What's the big deal about Conroe processors?

[Think]

Is it just a matter of price difference and less vcore implementation or are there any SIGNIFICANT architectural differences versus the Presler 955EE?

[pik-ard v1.1]

nope. they are completely different architectures. like going from AthlonXP to Athlon64. ;)

[Think]

nope. they are completely different architectures. like going from AthlonXP to Athlon64. ;)

Really? Well link would be nice to read the difference.

Thanks:)

[GoldenTiger]

Yep, and it is more efficient per clock than the Athlon64 FX60 dual-core 1MB cache CPU from what I've read by about 20%, too!

[Think]

Anyone got actual specs and details?

[pik-ard v1.1]

http://en.wikipedia.org/wiki/Intel_Core_Microarchitecture

look under the technology heading.

[Movieman]

Lets put it this way, at 54 it's got me feeling like a teenager again waiting to get my hands on one! The XE at 3333/4mb/1333...:slobber:
Now imagine that at 4000+...:rofl:

[mdzcpa]

Lets put it this way, at 54 it's got me feeling like a teenager again waiting to get my hands on one! The XE at 3333/4mb/1333...:slobber:


Same here. I'm looking forward to a new Intel rig. AMDs been awesome, but its time to flip back again.

P3
Athlon
P4
A64
Conroe

So it goes.

[Think]

Thanks for the link:).

[pik-ard v1.1]

Same here. I'm looking forward to a new Intel rig. AMDs been awesome, but its time to flip back again.

P3
Athlon
P4
A64
Conroe

So it goes.
hmm, my version:

P1
athlon (non-XP)
A64
Conroe (most likely)

and i've never really *used* an intel system as a gaming/benching system... should be fun.

[Think]

Damn. Any idea of the Intel P5WDG2 975x chipset will be compatible?

[n00b 0f l337]

No it wont be I beleive.

[Think]

No it wont be I beleive.

That's insane:nono: . What...a few friggin caps needed? Same socket for god sakes...geeeze that really eeks me.:mad:

[kyleslater]

Hmm. well I don't jump around so much..

P2 350
P3 450
P3 550
P3 750
Sempron 2400+
P4 630
Conroe... (maybe a E6400 or E6600 and an E4200).

[xombie2000]

Actually, some of the Intel and ASUS 975 are supposed to be compatible.

[Fred_Pohl]

nope. they are completely different architectures. like going from AthlonXP to Athlon64. ;)


Actually, Core differs much more from Netburst than A64 does from XP. Before some n00b even brings it up, NO, Core is not a rehash of the P3 any more than K8 is a rehash of the K6.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2748

[PkG.1337]

Intel P1 ???Mhz
Intel P3 400Mhz
Intel Celeron 800Mhz
Intel P4 1.5Ghz
Intel P4 1.8Ghz
Intel P4 3.2Ghz
Intel P-M 1.7GHz
Intel P4 3.2GHz
Intel Core Duo 2.16GHz

lol. i am such a intel fan boy

[Fuji]

nope. they are completely different architectures. like going from AthlonXP to Athlon64. ;)

The AthlonXP is not all that different from the Athlon64.

Conroe/Merom/Woodcrest can be related to the Athlon64 and Yonah to the AthlonXP although 80% of the circuitry had to be redone for Conroe.

But when you compare Presler to Conroe, you have a huge difference.

It's as big a jump as a Pentium 3 is to a Pentium 4 :)

[NinjaWreck]

Here's mine:

PII 266
PII 400
AMD 700
Athlon 1.4Ghz
Athlon XP 1800+
Athlon XP 2400+
P4 2.53c/P4 2.4c
P4 3.2c/P4 3.0e
AMD64 FX53 claw
AMD64 FX55 claw
AMD64 3000+ winnie
AMD64 3700+ SD
Opteron 170/P4 3.2c

[Teflon]

The AthlonXP is not all that different from the Athlon64.
)
:confused:

[Bugster]

Heres mine:

Intel P3 450mhz (slot 1)
xp 1600+
xp 1700+ DLT3C
Barton 2500+
P4 2.4 FSB800
A64 3200+ Winnie
Intel 630
Intel 920 (current one)

[Celeron Gamer]

Intel 80386
Intel 80486
Intel Pentium 133MHz
Intel Pentium 166MHz
VIA Cyrix 233MHz
Intel Celeron Mobile 550MHz
Intel Celeron 633MHz
Intel Celeron 667MHz
AMD Duron 1800+
Intel Pentium4 2.8GHz
AMD AthlonXP 2800+
Intel Celeron 2.7GHz
Intel CeleronD 315 2.26GHz
Intel Pentium4 2.4GHz

I'm much more of an Intel fanboy :P

[pik-ard v1.1]

The AthlonXP is not all that different from the Athlon64.

Conroe/Merom/Woodcrest can be related to the Athlon64 and Yonah to the AthlonXP although 80% of the circuitry had to be redone for Conroe.

But when you compare Presler to Conroe, you have a huge difference.

It's as big a jump as a Pentium 3 is to a Pentium 4 :)
um... actually... athlonXP was K7, and athlon 64 is K8. they're completely different cores, with probably nothing in common other than what K8 has in common with P7 (willy/northwood/prescott/whatever the crap).

where to you get the idea that they aren't that different?

[FunkyRider]

pik-ard v1.1, if you could grab yourself some info about K7/K8 arch,

you will see that K8's core (without IMC) is very likely to be a revised version of K7

K7 + AMD64 + IMC approximately equals -> K8

K7 to K8 is like intel Tualatin to Banias the only thing more is the IMC


Specs: http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2748&p=4

Just compare the changed between K7/K8 and NetBurst/Core, you will know the difference

[pik-ard v1.1]

hmm, i didn't know they had that much in common. but is that to say that they took a barton and said "how can we make this way better?", or something like "let's make something new, but we'll use the good points of the barton, just more refined."?

is what your're saying speculation, or has AMD said "we took the barton and heavily modified it to become K8"?

[adz]

i had 386
p1 mmx 200
p2 400
p3 450(laptop)
p4 1500
amd 64 3200+
p4 3200 (laptop)
amd64 3500+
core duo 1.66 (laptop)

[The Stilt]

That's insane:nono: . What...a few friggin caps needed? Same socket for god sakes...geeeze that really eeks me.:mad:

The problem doesn´t have anything to do with capacitors.
Asus P5WD2 / Prem. , P5WD2-E Prem and P5WD2G-WS motherboards all use ADP3181 PWM buck controller made by Analog Devices.

ADP3181 supports VR10 specification instead of required VR11.
Thats why these boards cannot support Conroe or any upcoming chips that require VR11 specification.

Asus should release P5WDxxx 1.03 revision that supports Conroe pretty soon.

[Fuji]

um... actually... athlonXP was K7, and athlon 64 is K8. they're completely different cores, with probably nothing in common other than what K8 has in common with P7 (willy/northwood/prescott/whatever the crap).

where to you get the idea that they aren't that different?
What i meant is that they are not all that different. They are not groundbreakingly different. That's not to say that they are the same, but it's not that different from Pentium 4 and Pentium 3. It's been said that the K8 is based on K7.

[Skyline GT-R]

Mine:

AMD K7 450MHz
P3 800MHz Coppermine
P4 2.4B
AMD Athlon 3400+
Conroe E6600 (my 3th 2.4GHz CPU in a row :P)

[JoeBar]

Asus should release P5WDxxx 1.03 revision that supports Conroe pretty soon.
I'm anxiously waiting for it! :)

[Think]

The problem doesn´t have anything to do with capacitors.
Asus P5WD2 / Prem. , P5WD2-E Prem and P5WD2G-WS motherboards all use ADP3181 PWM buck controller made by Analog Devices.

ADP3181 supports VR10 specification instead of required VR11.
Thats why these boards cannot support Conroe or any upcoming chips that require VR11 specification.

Asus should release P5WDxxx 1.03 revision that supports Conroe pretty soon.

Still eeks me. I had no idea that such a change in design was occurring so quickly. Oh well, I've been meaning to build a server in my home. I'll just use what I got for that and move on.

What a waste of money if you ask me. With such a significant change, my system has become a Dodo bird in less then 3 months.

[Smurfy]

Still eeks me. I had no idea that such a change in design was occurring so quickly. Oh well, I've been meaning to build a server in my home. I'll just use what I got for that and move on.

What a waste of money if you ask me. With such a significant change, my system has become a Dodo bird in less then 3 months.

Only in your head, it still does exacly the same as it did when you bought it.

[ahmad]

nope. they are completely different architectures. like going from AthlonXP to Athlon64. ;)

Is that right. Could you please tell me the differences between A64 and AXP?

Last I checked it was the "64bit"ness with an ondie memory controller and a die shrink (later on) that made the A64 so great.

[Think]

Only in your head, it still does exacly the same as it did when you bought it.

My understanding is that the new Conroe can make me a caffe late throught the Marvel USB ports:banana:

[Qba73]

My understanding is that the new Conroe can make me a caffe late throught the Marvel USB ports:banana:

and it can give you big mac's out of the CD drive..:D

Seriously i know how you feel think, im in the same boat your in. sucks doesnt it, socket stays the same, chipset supports it but we are @ss out by a controller.

how refreshing would it be if they came out with a socket adaptor like the ct-479 for these conroe's and the 975's that dont have the vr11
oh well i'll have to wait and see how the cards stack up and be ready for ebay for the 1.02 WS for the 1.03 WS.

Wonder if asus would honor a rma exchange being we paid top dollar for this board. to miss out on a new cpu by 5 months or so, then again 5 months in technology world is a lifetime..lol

[Thorburn]

Wonder if asus would honor a rma exchange being we paid top dollar for this board. to miss out on a new cpu by 5 months or so, then again 5 months in technology world is a lifetime..lol

They never claimed the board would support Core 2, unless the board is actually faulty they have no reason to replace it.

[Kjaks]

Intel has went from a cpu with long pipelines and netburst to a shorter pipelined one that is more like P3 or Dothan/Core Duo, but with a wider bus, higher speed and other improvements. Just google for Conroe specs..

[iterations]

Is it just a matter of price difference and less vcore implementation or are there any SIGNIFICANT architectural differences versus the Presler 955EE?
This is the big deal:

http://222.151.144.54/c-board/file/14.984s_T7400ES_3376.4.png

Cheers!

[pik-ard v1.1]

Is that right. Could you please tell me the differences between A64 and AXP?

Last I checked it was the "64bit"ness with an ondie memory controller and a die shrink (later on) that made the A64 so great.
all that was already brought to my attention, and i realized my error. nice try though. ;)

[Qba73]

They never claimed the board would support Core 2, unless the board is actually faulty they have no reason to replace it.


Kinda figured that but can't blame me for wishful thinking..lol

[Qba73]

Intel has went from a cpu with long pipelines and netburst to a shorter pipelined one that is more like P3 or Dothan/Core Duo, but with a wider bus, higher speed and other improvements. Just google for Conroe specs..

yeah the ol' p3 has gotten its revenge..i remember when the tualatin was whuppin the willamette's arse and then they went and killed the p3 and more importantly its architecture shelved. luckly it was revived for the mobile well its descendant at least, and now has come back to haunt the now defunct p4 and siblings. intel should of stayed with this approach rather than go speed crazy back then and who knows how the amd-intel back and forth would look like today, maybe intel would not have to play catch-up now...hopefully conroe will level they playing field if not tip it in its favor again.

[mzs_biteme]

OT but had to post my CPU legacy:

P 60@66:eek:
P 75@90
P 100@133
P 133@150
Cyrix 166+ (133MHz)
P 166MMX@200
P 233MMX@290 (I kid you not)
AMD K2 166@220
AMD K2 300@333
AMD K3 450@470:mad:
Celeron 266@400
Celeron 300@450
P3 550
Celeron 566@800
Duron 600@900
Duron 800@1050
Athlon 1200@1450
P4 1.6@2.1
P4 1.8@2.4
P4 2.6@2.9
P4 2.8@3.6
P4 3.0@3.9

Works for me...;)

[Maelstrom2160]

rm 80186, dos3
P60
386SX 16, co-processorless
P233 MMX
K6-3 450
1.2Ghz Thunderbird
XP2600
XP2500-M
Opteron 165

conroe :worship:


edit: ;)

[Qba73]

my best overclocker of all chips i ever owned was a p4 sl6z3 30 cap 2.4 northwood that did 3.6 on air (1200mhz oc, 300fsb 1200mhz effective fsb)

ahh i still remember that chip maybe because i retired her to my work pc that i am typing from now..:D

[charlie]

What's the big deal about CONROE?

Hmmm.... looks at my Celeron M notebook and then looks over at my FX-60/SLI rig...

Yeah, it's "different" from what's currently available.

[R101]

My PC machines - only good oc's noted :)

Intel 286-16MHz
AMD 386sx-40Mhz
Intel 486sx25MHz oc'd to 50Mhz :D
AMD 486DX133@180MHz (modded mobo, 4x45 :D)
Intel Pentium 90
Intel Pentium 166
Intel Celeron 300A
Intel Celeron 433@612MHz, 2.1V :)
Intel Celeron 566
Intel Coppermine 650@993 (ss only, kingmax ram was so cool back then)
AMD Duron 600@1100 (ss only)
AMD T-Bird 900
AMD XP1600
P4 NW 1600
P4 NW 2000
P4 NW 2.4@3.4 (oc that took most effort :) )
A64 s754 2800
P4 NW 2.4D@3.0
Laptop - Intel Celeron M 1.6 (still runnin' :) )

[R101]

OT but had to post my CPU legacy:

P 233MMX@290 (I kid you not)
AMD K2 166@220


Works for me...;)

Haha, just seen these.. A friend of mine first bought K2@233 which couldn't work at that frequency (225 was most it could go and be stable) and then he threw in a P166MMX which went to 270+.. Unbelievable speed for that time :).

[JoeBar]

Wonder if asus would honor a rma exchange being we paid top dollar for this board. to miss out on a new cpu by 5 months or so, then again 5 months in technology world is a lifetime..lol

The best after sales support... :p:

[Kaiser_Sose]

What are we going to have to spend to get a conroe? Are these cpus going to start out around $700+ or will we be able to get a $300 cpu and OC to a $1G cpu

Frpm the wiki ' Unfortunately, the FSB is the weak link in the new architecture, as it uses the infrastructure installed in the Pentium 4 era which cannot handle the full bandwidth of dual-channel DDR2 SDRAM, or the new memory architecture FB-DIMM. ' - anyone care to explain this and how it will hinder performance

[thunderstruck!]

My history is short. PIII, 2.4C, XP-M, A64, Conroe.

[Johnny Sack]

The deal is, a stock Conroe 2.13GHz, stock v-core and box cooler will destroy.


the purported quarters.
Intel Conroe 65nm Dual Core
E4200 2MB 1.60GHz 800MHz FSB Q4 $169. us
E6100 2MB 1.33GHz 1066MHz FSB Q1 2007 $149. us (35 Watts)*
E6200 2MB 1.60GHz 1066MHz FSB Q4 $179. us
E6300 2MB 1.86GHz 1066MHz FSB Q3 $209. us
E6400 2MB 2.13GHz 1066MHz FSB Q3 $239. us
E6500 2MB 2.40GHz 1066MHz FSB Q4 $269. us
E6600 4MB 2.40GHz 1066MHz FSB Q3 $309. us (65 Watts)
E6700 4MB 2.67GHz 1066MHz FSB Q3 $529. us
E6800 4MB 2.93GHz 1066MHz FSB Q4 $749. us
E6900 4MB 3.20GHz 1066MHz FSB Q4 $969. us
Intel Conroe XE 65nm Dual Core
E8000 4MB 3.33GHz 1333MHz FSB Q4 $1199. us (95 Watts)

[DilTech]

The AthlonXP is not all that different from the Athlon64.

Conroe/Merom/Woodcrest can be related to the Athlon64 and Yonah to the AthlonXP although 80% of the circuitry had to be redone for Conroe.

But when you compare Presler to Conroe, you have a huge difference.

It's as big a jump as a Pentium 3 is to a Pentium 4 :)

Actually, Coppermine(P3) was better than Williamette(first gen P4).


Finally, a man after my own heart that remembers that grand ole 286. Man, going to a COLOR screen was such an upgrade.

Ah, someone else remembers monochrome.

I still remember Deskmate, StereoShell(STS) file management, and PKZip......

[kyleslater]

Actually, Coppermine(P3) was better than Williamette(first gen P4).



Ah, someone else remembers monochrome.

I still remember Deskmate, StereoShell(STS) file management, and PKZip......
I had an old 286 with monochrome but did not think we were going back that far. Hey I even had a C64...

[Fred_Pohl]

The deal is, a stock Conroe 2.13GHz, stock v-core and box cooler will destroy.


the purported quarters.
Intel Conroe 65nm Dual Core
E4200 2MB 1.60GHz 800MHz FSB Q4 $169. us
E6100 2MB 1.33GHz 1066MHz FSB Q1 2007 $149. us (35 Watts)*
E6200 2MB 1.60GHz 1066MHz FSB Q4 $179. us
E6300 2MB 1.86GHz 1066MHz FSB Q3 $209. us
E6400 2MB 2.13GHz 1066MHz FSB Q3 $239. us
E6500 2MB 2.40GHz 1066MHz FSB Q4 $269. us
E6600 4MB 2.40GHz 1066MHz FSB Q3 $309. us (65 Watts)
E6700 4MB 2.67GHz 1066MHz FSB Q3 $529. us
E6800 4MB 2.93GHz 1066MHz FSB Q4 $749. us
E6900 4MB 3.20GHz 1066MHz FSB Q4 $969. us
Intel Conroe XE 65nm Dual Core
E8000 4MB 3.33GHz 1333MHz FSB Q4 $1199. us (95 Watts)

Just for comparison, here is what AMD plans to offer in Q3:

Athlon 64 FX-62 2MB 2.80GHz 1000MHz HTT $1,236 (???W)
Athlon 64 X2 5000+ 1MB 2.60GHz 1000MHz HTT $696 (95W)
Athlon 64 X2 4800+ 2MB 2.40GHz 1000MHz HTT $645
Athlon 64 X2 4600+ 1MB 2.40GHz 1000MHz HTT $558
Athlon 64 X2 4400+ 2MB 2.20GHz 1000MHz HTT $469
Athlon 64 X2 4200+ 1MB 2.20GHz 1000MHz HTT $365
Athlon 64 X2 4000+ 2MB 2.00GHz 1000MHz HTT $328
Athlon 64 X2 3800+ 1MB 2.00GHz 1000MHz HTT $303

To compare them simply subtract ~400MHz from AMD's clock speed to find a performance equivalent Conroe. :)

[andyisc00l]

OK this is starting to really bug me I posted in another topic and they didn't answer me really, but what is up with Pi? Why is it whenever I open a forum discussing the performance of a chip, Pi is brought in to the equation. Correct me if I'm wrong, but isn't it really, really pointless. Pointless meaning, if we actually looked at what it meant, it was nothing. I mean, an amd cpu getting 30 second pi times in comparison to a conroe chip getting 15.5 seconds means literally nothing? right? It just seems like taking Pi in to any consideration at all for the performance of a chip seems worthless. I feel like I'm taking crazy pills, lol.

[IvanAndreevich]

Just for comparison, here is what AMD plans to offer in Q3:

Athlon 64 FX-62 2MB 2.80GHz 1000MHz HTT $1,236 (???W)
Athlon 64 X2 5000+ 1MB 2.60GHz 1000MHz HTT $696 (95W)
Athlon 64 X2 4800+ 2MB 2.40GHz 1000MHz HTT $645
Athlon 64 X2 4600+ 1MB 2.40GHz 1000MHz HTT $558
Athlon 64 X2 4400+ 2MB 2.20GHz 1000MHz HTT $469
Athlon 64 X2 4200+ 1MB 2.20GHz 1000MHz HTT $365
Athlon 64 X2 4000+ 2MB 2.00GHz 1000MHz HTT $328
Athlon 64 X2 3800+ 1MB 2.00GHz 1000MHz HTT $303

To compare them simply subtract ~400MHz from AMD's clock speed to find a performance equivalent Conroe. :)
Agreed. AMD looks thoroughly OWNED. Possible worse than Intel was when A64 came out :eek:

[cupholder2.0]

Pentium 133 Mhz
Pentium 2 300Mhz
Celeron 400 Mhz
Celeron 1.8 Ghz NW
Pentium 4C NW-HT 2.4
Pentium 4C NW-HT 3.0
Pentium 4C NW-HT 3.4
Intel Pentium 4 670
Intel Pentium D 950
Soon: Conroe :banana:

[k0nsl]

OK this is starting to really bug me I posted in another topic and they didn't answer me really, but what is up with Pi? Why is it whenever I open a forum discussing the performance of a chip, Pi is brought in to the equation. Correct me if I'm wrong, but isn't it really, really pointless. Pointless meaning, if we actually looked at what it meant, it was nothing. I mean, an amd cpu getting 30 second pi times in comparison to a conroe chip getting 15.5 seconds means literally nothing? right? It just seems like taking Pi in to any consideration at all for the performance of a chip seems worthless. I feel like I'm taking crazy pills, lol.

Crazy pills? So a Williamette @ 2.1GHz which produces roughly 1.30m in SuperPI doesn't matter at all for the performance of a chip? The Williamette could still be good, because despite the fact that the Willamette did 1.30 (one minute and thirty seconds) it doesn't show the performance of it? :confused:

-k0nsl

[andyisc00l]

Crazy pills? So a Williamette @ 2.1GHz which produces roughly 1.30m in SuperPI doesn't matter at all for the performance of a chip? The Williamette could still be good, because despite the fact that the Willamette did 1.30 (one minute and thirty seconds) it doesn't show the performance of it? :confused:

-k0nsl
Well obviously when you compare old technology to new technology in this test, reguardless of how broad it is, it's going to show on the older stuff, but when comparing new technology, like where amd is getting upper 20 seconds lower 30s with their higher ocs, presler also upper 20s with high ocs, yonah getting the upper teens, and conroe getting lower teens...it doesn't seem like the test is describing an important characteristic of the cpu. It doesn't mean anything other then it can calculate pi faster which doesn't mean anything. Just because yonah @ 3.2ghz has an 18 second pi time, it doesn't mean amds 3.0ghz 29 second time is bad. Well its not as good of a pi time, but in no way does that reflect performance. In no way at all should there be any meaning behind conroe getting a 15 seconds pi time for its 3.2ghz clock in comparison to amds 29 second pi time at 3.0ghz...Yonah @ 2.4 ghz will probably beat amds time by a decent amount but is it better then a 3.0ghz amd cpu? Same with conroe..just cuz you have a 23 second pi time @ 2.1ghz, it doesn't mean that its in any way better then a 3.0ghz amd cpu...that's the point I was making. With that in mind it just seems like an unbiased test. Ya no? Who knows. And I don't mean anything by it...i was just throwing this up in the air cuz I never hear anyone say this test is stupid for showing off the performance of a chip...lol

[Fred_Pohl]

OK this is starting to really bug me I posted in another topic and they didn't answer me really, but what is up with Pi? Why is it whenever I open a forum discussing the performance of a chip, Pi is brought in to the equation. Correct me if I'm wrong, but isn't it really, really pointless. Pointless meaning, if we actually looked at what it meant, it was nothing. I mean, an amd cpu getting 30 second pi times in comparison to a conroe chip getting 15.5 seconds means literally nothing? right? It just seems like taking Pi in to any consideration at all for the performance of a chip seems worthless. I feel like I'm taking crazy pills, lol.

Pi calculation is basically a synthetic benchmark for these purposes but it does provide some insight into relative cpu/memory performance. Although Core may hold a 1000-1600MHz advantage over K8/X2/Opty in Pi while only holding a 300-500MHz advantage on average in more mainstream apps and games, Pi calculation has been and still is a very well accepted system benchmark for both performance and stability for several years now.

[Fred_Pohl]

Agreed. AMD looks thoroughly OWNED. Possible worse than Intel was when A64 came out :eek:

I haven't seen pwnage like this since K6-2s went up against C300As. As I recall, the first K8s weren't all that that impressive and killer 64 bit software is still elusive after how many years?

[Fred_Pohl]

Well obviously when you compare old technology to new technology in this test, reguardless of how broad it is, it's going to show on the older stuff, but when comparing new technology, like where amd is getting upper 20 seconds lower 30s with their higher ocs, presler also upper 20s with high ocs, yonah getting the upper teens, and conroe getting lower teens...it doesn't seem like the test is describing an important characteristic of the cpu. It doesn't mean anything other then it can calculate pi faster which doesn't mean anything. Just because yonah @ 3.2ghz has an 18 second pi time, it doesn't mean amds 3.0ghz 29 second time is bad. Well its not as good of a pi time, but in no way does that reflect performance. In no way at all should there be any meaning behind conroe getting a 15 seconds pi time for its 3.2ghz clock in comparison to amds 29 second pi time at 3.0ghz...Yonah @ 2.4 ghz will probably beat amds time by a decent amount but is it better then a 3.0ghz amd cpu? Same with conroe..just cuz you have a 23 second pi time @ 2.1ghz, it doesn't mean that its in any way better then a 3.0ghz amd cpu...that's the point I was making. With that in mind it just seems like an unbiased test. Ya no? Who knows. And I don't mean anything by it...i was just throwing this up in the air cuz I never hear anyone say this test is stupid for showing off the performance of a chip...lol

I hear ya! Just because Dothan/Yonah/Core rule SuperPi doesn't mean that they outperform k8/X2/Opty by the same huge margins in all apps and games. Just figure that Core outperforms X2 by ~400Mhz across the board and you will have found the truth, grasshopper.

[-.-PhanTom-.-]

Interesting perspective> http://sharikou.blogspot.com/2006/05/with-new-architecture-intel-will-be.html

[Thorburn]

Question the motives of a man who has an entire blog devoted to trashing Intel and saying how great AMD are.

[sealion]

That guy seems to know what he's talking about unlike my computer science teacher claiming a 3ghz celeron can beat a athlonFX 57.


Too many biased opinions in that blog.



The big deal a bout conroe is it is a step foward for intel. After a few years of useless side-steps; intel is moving foward.

[savantu]

Interesting perspective> http://sharikou.blogspot.com/2006/05/with-new-architecture-intel-will-be.html

When you post links to that guy's blog you lose credibility :stick: ...

[nn_step]

When you post links to that guy's blog you lose credibility :stick: ...
how so :stick:
Being open to different ideas doesn't mean that you are stupid.. it means you are considering more options

[G H Z]

Conroe is a joke, just a bunch of faked benchmarks & hype ;) :D

[Your_Boss]

how so :stick:
Being open to different ideas doesn't mean that you are stupid.. it means you are considering more options

True, but when you point to a guy who isn't open minded himself, kinda pointless. This guys "options" have already been considered and discounted in the same breath. I would like to see more objective people's opinions, not bias garbage. :rolleyes:

[nn_step]

True, but when you point to a guy who isn't open minded himself, kinda pointless. This guys "options" have already been considered and discounted in the same breath. I would like to see more objective people's opinions, not bias garbage. :rolleyes:
Although I agree that we should take all information with a Grain of salt..
but any company that is going to repeat the same mistake twice is not one I am likely to believe
http://img2.zol.com.cn/product/4_450x337/999/ceLj0DPtvAFgY.jpg
and legit questions that are posted on our Forums are attacked
http://www.xtremesystems.org/forums/showthread.php?t=95021&page=34
Don't you find that odd?

[Your_Boss]

I am all for grains of salt. ;)

I do give credit where credit is due...and Intel looks to have made some nice chips(Conroe/Memron/Yonah) Lotta ES results going around right now. And they look great..breaking WR and all. :slobber:

I do wish AMD would throw some ES chips out there if they have anything coming up soon. Intel is just getting a lot of play right now.

[savantu]

Although I agree that we should take all information with a Grain of salt..
but any company that is going to repeat the same mistake twice is not one I am likely to believe
http://img2.zol.com.cn/product/4_450x337/999/ceLj0DPtvAFgY.jpg
and legit questions that are posted on our Forums are attacked
http://www.xtremesystems.org/forums/showthread.php?t=95021&page=34
Don't you find that odd?

You don't understand the problem here : doing a multichip module isn't a mistake at all , it is a very smart strategy , sad part is few see it.

Conroe is 140mm^2.
Kentsfield should be 2x Conroe , 280mm^2. Numbers of possible defects increases at least 2^2 , 4 times.So with Kentsfield you might have abysimal yeilds.

Intel avoids this by using 2 separate Conroe dies in the same chip => it can offer a quad-core immediately with no impact on yields.

What's more , altough chips from the same part of the wafer are nearly identical , Intel can combine a "cold" core with a "hot" one while keeping the same TDP or 2 slower ones in the same TDP.AMD OTOH has none of these benefits

When chip size is small ( less than 150 mm^2 ) , either due to a new process or different architecture ( see Yonah ) you can build a single die multicore chip with good yields.

As a result Intel will offer a quad-core chip for desktops and 2S systems in january 2007 while the corresponding AMD chip will arrive guess when : in H1 2008 for crying out loud!

But you are free to let me quote you "being open to different ideas doesn't mean that you are stupid.. it means you are considering more options"

[k0nsl]

The only thing I am sceptical about at the moment is the fact that all results we have seen of Conroe has been with chips with 4MB L2 Cache and the ones we will be buying will only be 2MB. I try to convince myself that doubling the cache will not impact greatly the performance!

-k0nsl

[savantu]

The only thing I am sceptical about at the moment is the fact that all results we have seen of Conroe has been with chips with 4MB L2 Cache and the ones we will be buying will only be 2MB. I try to convince myself that doubling the cache will not impact greatly the performance!

-k0nsl

Sorry ? Conroe 2.4GHz (E6600 $316 ) and more are all 4MB L2.

[deltarealm]

The more cache the better. Imagine if you could have your entire system ram on the cpu.

[Thorburn]

The only thing I am sceptical about at the moment is the fact that all results we have seen of Conroe has been with chips with 4MB L2 Cache and the ones we will be buying will only be 2MB. I try to convince myself that doubling the cache will not impact greatly the performance!

-k0nsl

Just look at Yonah, 2MB L2 versions will be slower but the performance should still be good enough to take outpace the competition clock for clock.

There are still 4MB L2 cache versions available from launch, just at the higher clock speeds.

[k0nsl]

Sorry ? Conroe 2.4GHz (E6600 $316 ) and more are all 4MB L2.

LOL! Okay, I must have missunderstood something (glad I did). Thanks for clarifying that.

:toast:

-k0nsl

[Fred_Pohl]

Interesting perspective> http://sharikou.blogspot.com/2006/05/with-new-architecture-intel-will-be.html

Have you no shame? Aren't you embarrassed to link to that paid AMD shills FUD blog?

When this fanboy clown bashed the Core architecture because a dual Cloverton rig only scaled by 4.75x in 8-way Cinebench I emailed him evidence that an 8 way Opteron rig only scaled by 4.8x in the exact same benchmark. Needless to say he had no interest in the truth and neither retracted his erroneous blog FUD nor replied to my email.

BTW, I also posted the same 8-way Opteron scaling info to the fanboy's blog site but he censored it. I think it's crystal clear what his agenda is.

[Your_Boss]

True, when that guy can't even come up with a fabulous lie to answer your question, he'll just delete your question altogether. I've seen him do it. I was reading some guys reply and I was waiting for the blog guy to respond..I refreshed my browser..and the whole question is gone too! :nono: Then I knew this guy was FUD city.

On Topic:D

Conroes will be great..proof is already around the net.

BTW

2.4 GHz Conroe (and higher clocked) have 4MB L2
2.16 GHz Conroe (and lower clocked) have 2MB L2

I think this correct. I forget exactly where the cut-off is. ;)

[charlie]

You know... with the super LOOOOWWW prices on Cedar Mill/Presler and such... I was considering grabbing some cheap stuff for a new office rig. Then I realized :D Why? If I wait 2 months I can buy Conroe CPU @ $200 and some 965 Conroe board for <$200... sheesh! It's just NOT economical to touch ANYTHING til Conroe.

[automagic]

mine,

Kaypro (CPM)
intel 286
intel 386
pentium 100
celeron 400 @ 700
duron 700
thunderbird 1200 amd
northwood 2.4gig
athlon xp3200+
northwood 3.2

soon conroe

[amd4me]

Whats the price point for the Conroe?
And when is it set to drop?

[tdenton1138]

Let's see:

TRS-80 1 (http://www.obsoletecomputermuseum.org/trs80-m1/) (I only had the 'cassette deck' for storage..heh)
Apple // (http://apple2history.org/history/ah03.html)
Apple //+ (http://apple2history.org/history/ah06.html)
Apple //e (http://apple2history.org/history/ah07.html#x23) (with Transwarp (http://apple2history.org/museum/peripheralcards_nonapple/transwarp2.html) accelerator to finally break the 1Mhz barrier!) *
Amiga 1000 (http://www.amigau.com/aig/a1000.html) (68000 @ 7.16Mhz) * So far ahead of its time...
Amiga 3000 (http://www.amigau.com/aig/a3000.html) (68020 @ 25Mhz) *
P90 (stupidly bought a Falcon NW, but it sure was fast!)
P200MMX (Voodoo 2 SLI baby!) *
P3 600
Athlon 900
Athlon 1.33
Something else here I'm sure ...
XP2500 *
P4 2.53 *
Conroe

*still functional in my possession ;)

[Kjaks]

Whats the price point for the Conroe?

I'm not shure what you mean, but here's what the prices are going to be, like said earlier in the thread:

Intel:
E4200 2MB 1.60GHz 800MHz FSB Q4 $169. us
E6100 2MB 1.33GHz 1066MHz FSB Q1 2007 $149. us (35 Watts)
E6200 2MB 1.60GHz 1066MHz FSB Q4 $179. us
E6300 2MB 1.86GHz 1066MHz FSB Q3 $209. us
E6400 2MB 2.13GHz 1066MHz FSB Q3 $239. us
E6500 2MB 2.40GHz 1066MHz FSB Q4 $269. us
E6600 4MB 2.40GHz 1066MHz FSB Q3 $309. us (65 Watts)
E6700 4MB 2.67GHz 1066MHz FSB Q3 $529. us
E6800 4MB 2.93GHz 1066MHz FSB Q4 $749. us
E6900 4MB 3.20GHz 1066MHz FSB Q4 $969. us
E8000 4MB 3.33GHz 1333MHz FSB Q4 $1199. us (95 Watts)

AMD Q3:

Athlon 64 FX-62 2MB 2.80GHz 1000MHz HTT $1,236 (120W)
Athlon 64 X2 5000+ 1MB 2.60GHz 1000MHz HTT $696 (95W)
Athlon 64 X2 4800+ 2MB 2.40GHz 1000MHz HTT $645
Athlon 64 X2 4600+ 1MB 2.40GHz 1000MHz HTT $558
Athlon 64 X2 4400+ 2MB 2.20GHz 1000MHz HTT $469
Athlon 64 X2 4200+ 1MB 2.20GHz 1000MHz HTT $365
Athlon 64 X2 4000+ 2MB 2.00GHz 1000MHz HTT $328
Athlon 64 X2 3800+ 1MB 2.00GHz 1000MHz HTT $303

[Fred_Pohl]

True, when that guy can't even come up with a fabulous lie to answer your question, he'll just delete your question altogether. I've seen him do it. I was reading some guys reply and I was waiting for the blog guy to respond..I refreshed my browser..and the whole question is gone too! :nono: Then I knew this guy was FUD city.

On Topic:D

Conroes will be great..proof is already around the net.

BTW

2.4 GHz Conroe (and higher clocked) have 4MB L2
2.16 GHz Conroe (and lower clocked) have 2MB L2

I think this correct. I forget exactly where the cut-off is. ;)

IMO Sharikou holds a PhD in FUD. I wouldn't be a bit surprised to learn that he is being paid by AMD to post his nonsensical bull:banana::banana::banana::banana:. If not, he must be AMD's biggest delusional fanboy. He is careful to allow a few negative comments on his site but only when they don't contain any facts that contradict his FUD.

IIRC Intel plans to release a 2.4G model with 2MB L2 in Q4 and yes, there is more than enough evidence already available to show that Core 2 will be a resounding success. So far I've seen Core 2 BM results for the following tests and Core 2 trounces X2 by 15-50% at equal clock speed in all of them.

3DM01
3DM03
3DM05
3DM06
PCM05
FEAR
HL2
UT2004
Q4
Q4 SMP
SuperPi
Hexus PiFast
DivX 6.1
iTunes 6
WME 9

The average performance difference is between 20-25% in favor of Core 2. I researched almost 30 different benchmarks and established that X2 performance scales by 8-9% per 200MHz clock increase. This leads me to conclude that on average a dual-core AM2 cpu will need to run 400MHz faster to roughly equal the performance offered by Core 2.

[Salvador]

Have you no shame? Aren't you embarrassed to link to that paid AMD shills FUD blog?

When this fanboy clown bashed the Core architecture because a dual Cloverton rig only scaled by 4.75x in 8-way Cinebench I emailed him evidence that an 8 way Opteron rig only scaled by 4.8x in the exact same benchmark. Needless to say he had no interest in the truth and neither retracted his erroneous blog FUD nor replied to my email.

BTW, I also posted the same 8-way Opteron scaling info to the fanboy's blog site but he censored it. I think it's crystal clear what his agenda is.

Haha, Sharik brings so much laughter to our lives. It's so funny how he's making himself the laughing stock for the whole HW world.. If any, and I will be speachless, do believe this guy, they must be as fanatic as he too! :clap:
This is just writing down as much crap on a blog as humanly possible..

[amd4me]

My understanding is that the new Conroe can make me a caffe late throught the Marvel USB ports:banana:
Odds are I can make it faster.
I'm a professional Barista.

[-Sweeper_]

Damn. Any idea of the Intel P5WDG2 975x chipset will be compatible?

No. :(

[kyleslater]

You know what? almost any of those models (above 2.0Ghz) will be faster then what I have now... I get a maximum SuperPi1M of 27second @ 4.3Ghz but it's dreadfully unstable. The best stable is 28s. So that 19 seconds or whatever is plenty fast for me...

[Fred_Pohl]

Haha, Sharik brings so much laughter to our lives. It's so funny how he's making himself the laughing stock for the whole HW world.. If any, and I will be speachless, do believe this guy, they must be as fanatic as he too! :clap:
This is just writing down as much crap on a blog as humanly possible..

I must confess that as comical as Sharikou Fanboy is, I am still irritated by his lies. As unbelievable as it seems, there are people out there who actually believe his crap.

[kyleslater]

I must confess that as comical as Sharikou Fanboy is, I am still irritated by his lies. As unbelievable as it seems, there are people out there who actually believe his crap.
Well someone has to be a spokes person for the idiots :P

[gullf1sk]

I must confess that as comical as Sharikou Fanboy is, I am still irritated by his lies. As unbelievable as it seems, there are people out there who actually believe his crap.

Yeah i know what you mean. People like him bother me. Its even worse to read the comments of the people who agree with him.

[Fred_Pohl]

Well someone has to be a spokes person for the idiots :P


Sadly, that is true and Sarikou is just the deluded fanboy idiot for the job!

[Fred_Pohl]

Yeah i know what you mean. People like him bother me. Its even worse to read the comments of the people who agree with him.

It's really sad. I would like to enlighten them but many of them simply don't want to be enlightened and I really don't have the time anyway. A few months back I tried to interject some reality to the fanboys at AMDZone but most of them much prefer their fanboy deslusions. In reply to my factual info I heard fanboy fantasies ranging from AM2 showing 20% performance gains over S939 to AMD's 65nm yields being double Intel's to 4Ghz 65nm K8Ls launching in Q3 and spanking Core 2.

[savantu]

I am absolutely salivating over this. :slobber:

At 65W of heat production, you could almost afford to have a dual socket system running 2 x dual cores. Too bad no one has a chipset to support it.

All 6xxx Conroes are supposed to be 65w , it's a familly TDP.

Take with a gain of salt anything higher than E6800 2.93GHz which will appear in Q4.

As for the XE version rumor has it it will be here in July.If it is a 3.33GHz/4MB/1333 MHz part it's game over for you know who...

[Think]

Intel 8086 - 1979
Intel 955EE- now

I'm a patient person;)

[iterations]

I kinda doubt we will see the 3.33GHz part right off the bat. Most likely Intel will need some time to perfect the 65nm process. My gut tells me that is a pre-Xmas part.
I agree with you that it seems unlikely that the Extreme part will come out in July, but remember 65nm is not a new process for Intel, they have been shipping in volume since october 2005. Core Duo, and all of the latest Pentium Ds are all 65nm.

Core is a new microarchitecture, but I've seen data that shows that 65nm yields are the best Intel has ever seen, and they ramped to those yields faster then ever in their history.

So it is possible (though still unlikely) that the 3.33GHz part will come out imediately.

[ethernal]

Core is a new microarchitecture, but I've seen data that shows that 65nm yields are the best Intel has ever seen, and they ramped to those yields faster then ever in their history.

Sorry, this isn't really about yields or anything, but I'm just curious. You seem pretty knowledgeable about Intel, and I'm (obviously) not, so... my question is whether Intel has found a way around the exponentially increasing amount of leaking current (either by adding some form of high-k dielectrics or through something similar) or are simply reducing power consumption by increased power saving technology built into the chip.

I ask this because Intel seems confident in their ability to release chips at 45 nanometers in only two more years, and 32 in four. This is compared to the three years it took to retail chips from 90 nanometers down to 65nm. I would assume that based on this, they must have developed some method of at least limiting the problem of leaking current, especially since the 65nm chips seem to be as power efficient as ever with the Core Duo and Merom.

Thanks in advance if you have an answer. :)

[Dumo]

I dunno about all this tchnical things (ask Intel for answer):) ...But one thing I know that Yonah is speedy:D

[nn_step]

You don't understand the problem here : doing a multichip module isn't a mistake at all , it is a very smart strategy , sad part is few see it.

Conroe is 140mm^2.
Kentsfield should be 2x Conroe , 280mm^2. Numbers of possible defects increases at least 2^2 , 4 times.So with Kentsfield you might have abysimal yeilds.

Intel avoids this by using 2 separate Conroe dies in the same chip => it can offer a quad-core immediately with no impact on yields.

What's more , altough chips from the same part of the wafer are nearly identical , Intel can combine a "cold" core with a "hot" one while keeping the same TDP or 2 slower ones in the same TDP.AMD OTOH has none of these benefits

When chip size is small ( less than 150 mm^2 ) , either due to a new process or different architecture ( see Yonah ) you can build a single die multicore chip with good yields.

As a result Intel will offer a quad-core chip for desktops and 2S systems in january 2007 while the corresponding AMD chip will arrive guess when : in H1 2008 for crying out loud!

But you are free to let me quote you "being open to different ideas doesn't mean that you are stupid.. it means you are considering more options"
You fail to see the difference between a different idea and a Bad idea..
Dual cores fighting for the Bus choke teh performance even with a 1066Mhz bus.. imagine what 4 Cores would do..:fact:

[iterations]

Sorry, this isn't really about yields or anything, but I'm just curious. You seem pretty knowledgeable about Intel, and I'm (obviously) not, so... my question is whether Intel has found a way around the exponentially increasing amount of leaking current (either by adding some form of high-k dielectrics or through something similar) or are simply reducing power consumption by increased power saving technology built into the chip.

I ask this because Intel seems confident in their ability to release chips at 45 nanometers in only two more years, and 32 in four. This is compared to the three years it took to retail chips from 90 nanometers down to 65nm. I would assume that based on this, they must have developed some method of at least limiting the problem of leaking current, especially since the 65nm chips seem to be as power efficient as ever with the Core Duo and Merom.

Thanks in advance if you have an answer. :)
Well, the answer is definitely yes. There are many technologies coming out of the material science research from all semi conductor companies. In particular, high-k dielectrics that you mentioned, and pure metal gate are the most promsing ones.

However, there are literally hundreds of possibilties that are being analyzed, both on technical merit, and economic feasibility, such as using other materials (3-5 compounds, like InSb - indium antimonide) on top of the silicon layer, and building the transitor out of those materials, which exibit much higher electron mobility (and therefore lower dynamic capacitance, and hence lower leakage).

If you are interested in more detail, check out this page:

http://www.intel.com/idf/us/spring2006/webcast.htm and click the link at the bottom for the "high bandwidth" link for Paolo Gargini's talk for "Interactive webcast". Here Paolo goes into more detail for some of these technologies. I thought is was a great 40 minute talk.

By the way, unfortunately it appears that the link doesn't work in Firefox (at least for me) so you may need to hit it in IE.

[ethernal]

That's good to hear, not just for Intel but for the semiconductor industry as a whole.

I appreciate your time and insight. Gonna go watch the video now. :)

[goodcooper]

Sharikou's blog isn't that stupid... but he is always skirting the issue of single performance desktop vs desktop, which shows bits of bias...

BUT what he says about core 2's place in the server and enterprise market makes perfect sense...

the bus architecture of core2 in essence has gone back a gen (dunno where the guy gets 4 gens), but fact o the matter is, whatever is faster is faster, so who cares if its older...

but as soon as you get a couple gigabit cards in there, some hd controllers, performance is gonna hurt... but thankfully, we don't have to worry about that so much in the desktop market...

i think i'll buy a conroe for gaming, but if it came down to recommending/consulting/building a server for a customer, i'd have to stick with the opterons...

woodcrest has some 'splainin to do, in my eyes

don't know how they are going to make it work in the enterprise segment, can anyone refute what he says about ddr2 800 saturating that 'ol bus architecture? and what they might be doing is trying to snag the gamers who are now in the position of choosing which CPU goes in their server at work... "my A64 gets 185fps in cs source, so get an opteron!!!!", or "new conroes OWN A64s, i get 45fps more than my buddy incs source, so get a woodcrest" without taking a look at how different the architectures can be between a desktop and a server (but hey, sometimes they aren't much different at all) but ENTERPRISE SERVERS... woodcrest will not work, not with thier current model

[Salvador]

Sharikou's blog isn't that stupid...

blablablabla ...

... woodcrest has some 'splainin to do, in my eyes...

.... but ENTERPRISE SERVERS... woodcrest will not work, not with thier current model

Explain, fill out more.

Sharikou's blog is the biggest piece of :banana::banana::banana::banana: webpage ever created.. :rolleyes: Guys from XS.org should really know better.

My guess is that Woodcrest will work just perfectly. Shows that new Core possibly isn't that FSB limited as netburst was.

[Fred_Pohl]

You fail to see the difference between a different idea and a Bad idea..
Dual cores fighting for the Bus choke teh performance even with a 1066Mhz bus.. imagine what 4 Cores would do..:fact:

You're failing to take other factors into consideration. First, Conroe does not appear to be bandwidth bottlenecked by it's 1066fsb. If it is, it must be one helluva cpu to outperform X2 by ~25% clock for clock while bandwidth starved! Second, just because HTT provides enough bandwidth for 8 cores to a single K8 doesn't mean that there is any benefit from having 8 times the necessary bus bandwidth.

Single socket Woodcrest systems will have a more than adequate 1333fsb and 2-way Woodcrests will have 2x1333fsb. Early testing of a 2-way (8 core) Cloverton system showed that it scales by a factor of 4.75X in 8 way Cinebench vs 4.8X for an 8 core Opteron system. If this fsb bottleneck that everyone keeps on about doesn't even appear until you have eight 2GHz cores fighting for bandwidth from a 2x1333fsb, I don't see much of a problem for Woodcrest servers using 8 cores or less. Cloverton with a single 1333fsb could be a different story but if it is, I suspect that Intel will compensate by adding a second bus. Then in 2008 Intel has CSI coming at about the same time AMD hopes to release their first quad core cpus.

[Fred_Pohl]

Sharikou's blog isn't that stupid...

I read enough...

[savantu]

...
... Early testing of a 2-way (8 core) Cloverton system showed that it scales by a factor of 4.75X in 8 way Cinebench vs 4.8X for an 8 core Opteron system.

Dual Clovertown quad core 2.0 GHz
Cinebench 9.5 64-bit
1 core : 362 CB
8 cores : 1723 CB (4.76x)

8 x Opteron 854 (single core 2.8 GHz) 32Go
Cinebench 9.5 64-bit (wServer2k3 x64)
1 core : 448 CB
8 cores : 2154 CB (4.81x)

40% more GHz , assuming linear scaling for Cloverton , a 2.8GHz Cloverton should do 2412. ( 2200 more reasonable ) So clock/clock , Cloverton still has a little bit higher oomph... ;)

A 16 core Opteron scores about 2500 , that one hell of a scaling..:slapass:


Btw , the Opteron system has all the advantages , single cores , each one has 6.4GBs of BW.A dual core will cut that in half.

[crackhead2k]

It's really sad. I would like to enlighten them but many of them simply don't want to be enlightened and I really don't have the time anyway. A few months back I tried to interject some reality to the fanboys at AMDZone but most of them much prefer their fanboy deslusions. In reply to my factual info I heard fanboy fantasies ranging from AM2 showing 20% performance gains over S939 to AMD's 65nm yields being double Intel's to 4Ghz 65nm K8Ls launching in Q3 and spanking Core 2. Its kind of obvious whats going to happen if your post there... I wouldnt waste my time there :) Cuz its "AMDZONE" ;)

I like both AMD and Intel they will have there turn to be best anyways benifit for us right...?


You know... with the super LOOOOWWW prices on Cedar Mill/Presler and such... I was considering grabbing some cheap stuff for a new office rig. Then I realized Why? If I wait 2 months I can buy Conroe CPU @ $200 and some 965 Conroe board for <$200... sheesh! It's just NOT economical to touch ANYTHING til Conroe.

Im thinking the exact same thing:)

[JumpingJack]

I agree with you that it seems unlikely that the Extreme part will come out in July, but remember 65nm is not a new process for Intel, they have been shipping in volume since october 2005. Core Duo, and all of the latest Pentium Ds are all 65nm.

Core is a new microarchitecture, but I've seen data that shows that 65nm yields are the best Intel has ever seen, and they ramped to those yields faster then ever in their history.

So it is possible (though still unlikely) that the 3.33GHz part will come out imediately.

Actually, the jargon is pointing to a launch with the extreme edition:



Core 2 Extreme will actually become the first Conroe processor to launch, followed immediately by Core 2 Duo E6700.

Whether that is 2.92 or 3.33 remains to be seen. The intel 65 nm process, by the numbers reported so far (transistor metrics), is the best Intel has produced, and have record setting ring oscillator numbers. From what I can gather, 3.33 is more likely, you can also see it in the numbers from the Yohna over clocking -- 65 nm appears to ramp clocks very nicely -- again point to extremely healthy transistors.

[savantu]

Whether that is 2.92 or 3.33 remains to be seen. The intel 65 nm process, by the numbers reported so far (transistor metrics), is the best Intel has produced, and have record setting ring oscillator numbers. From what I can gather, 3.33 is more likely, you can also see it in the numbers from the Yohna over clocking -- 65 nm appears to ramp clocks very nicely -- again point to extremely healthy transistors.

It's 3 or 3.33GHz.I fully agree with the rest of what you've said.

2.93( Conroe E6800 Q4 266x11 ) is impossible to get with 333 (1333 FSB) base clock.

[iterations]

Actually, the jargon is pointing to a launch with the extreme edition:



Whether that is 2.92 or 3.33 remains to be seen. The intel 65 nm process, by the numbers reported so far (transistor metrics), is the best Intel has produced, and have record setting ring oscillator numbers. From what I can gather, 3.33 is more likely, you can also see it in the numbers from the Yohna over clocking -- 65 nm appears to ramp clocks very nicely -- again point to extremely healthy transistors.

Hehe, I completely agree, I'm just trying to temper (my own) expectations. A healthy launch with Core 2 Extreme being available would be awesome to say the least.

[FunkyRider]

Guess why Opteron scales badly in cinema rendering? Cuz they are NUMA - It's really hard to apply a single application to utilize NUMA. i.e. If the rendering engine is rendering a scene with a texture, that single texture must resides in some single memory system, which makes NUMA useless. Other cores just have to fetch the data from certain single memory, which will obviously saturate THAT bandwidth and idle other ones.

In order to utilize the NUMA's power, Opteron server is better off with something that each thread needs independent memory data to run, otherwise it will not be any better than intel old-skool FSB's UMA system.

In other aspect this shows different approaches of two companies:
AMD: Each core works independently. They have their own cache, their own memory.
Intel: Shared cache, shared memory.

These configurations will have strengths and weaknesses in one or the other applications. So, its just not an un-wise option to build a Woodcrest/Cloverton system for a render farm - they are not any slower than Opterons.


Some real life examples:
Op: each 2 cores has own IMC, bandwidth = 6.4GB/s, 2 way, so total bandwidth = 12.8GB/s
Wc: 2 cores, 2 way DIB, 4 ch FB-DIMM, bandwidth = 20GB/s, FSB bandwidth = 10.6GB/s

For above machines, single thread bandwidths are OP:6.4GB/s and WC:10.6GB/s
Two threads in one way: OP: 6.4GB/s (3.2GB/s each), WC: 10.6GB/s (5.4GB/s each)
Two threads in two way: OP: 12.8GB/s (6.4GB/s each), WC: 10.6GB/s (5.4GB/s each)
Two threads in two way but shared data*: OP: 6.4GB/s (3.2GB/s each), WC: 10.6GB/s (5.4GB/s each)
Four threads: OP: 12.8GB/s (3.2GB/s each), WC: 20GB/s (5GB/s each)
Four threads but shared data*: OP: 6.4GB/s (1.6GB/s each), WC: 20GB/s (5GB/s each)

* Shared data means even in NUMA system, data can only be hold in one CPU-Memory bank

only in third condition does Opteron outperform Woodcrest in terms of memory bandwidth.

[ethernal]

Actually, the jargon is pointing to a launch with the extreme edition:



I think DailyTech made a mistake. I know of the article you're talking about, and there is no other source that I can find that supports what they're saying. In addition, they reference a previous article to back up this claim and the article they reference has nothing to do with XE launching first. Most likely, it will come in late Q3/early Q4.

[goodcooper]

Explain, fill out more.

Sharikou's blog is the biggest piece of :banana::banana::banana::banana: webpage ever created.. :rolleyes: Guys from XS.org should really know better.

My guess is that Woodcrest will work just perfectly. Shows that new Core possibly isn't that FSB limited as netburst was.


well, first off, the old FSB architecture may be able to handle desktop I/O devices, but just take a look at AMDs architechture, they have a dedicated bus for memory, it is just so much more efficient once you start jamming that bus with I/O... also take into account HT compared to the FSB arch... HT is a 1ghz up and a 1ghz down (i must also note that, dual core Intel's, instead of using the bus to communicate with each other like before, now uses the cache, which runs at core, a VERY good development, which adds to it's current success), compare that to FSB... LET ALONE the fact that you no longer have to share memory bandwidth on that bus!

an ondie memory controller on new Intel chips would put them lightyears ahead,

when i'm talking about the types of servers that shouldn't be running woodcrest, i'm talking about I/O intensive servers, not necissarily render farms, render farms don't need to be pulling data off of a big pcix controller while feeding it out 3 gigabit nics while doing a scsi tape back up... you've got HOW many devices clogging up that bus now? PLUS the full bandwidth of DDR2 800?

please...

if i had a graphics card, some ram and a processor to make up my gaming machine, i'd choose conroe, but someone as yet to refute the bottlenecks that guy is bringing forward... you say its crap, but do you have any thing to back it up? cuz altho delivering his in a biased way, he has facts

i think some of you guys are letting the fanboy stuff get to you


EDIT: the fact of the matter is, this technology just doesn't scale well

[savantu]

well, first off, the old FSB architecture may be able to handle desktop I/O devices, but just take a look at AMDs architechture, they have a dedicated bus for memory, it is just so much more efficient once you start jamming that bus with I/O... also take into account HT compared to the FSB arch... HT is a 1ghz up and a 1ghz down (i must also note that, dual core Intel's, instead of using the bus to communicate with each other like before, now uses the cache, which runs at core, a VERY good development, which adds to it's current success), compare that to FSB... LET ALONE the fact that you no longer have to share memory bandwidth on that bus!

an ondie memory controller on new Intel chips would put them lightyears ahead,

when i'm talking about the types of servers that shouldn't be running woodcrest, i'm talking about I/O intensive servers, not necissarily render farms, render farms don't need to be pulling data off of a big pcix controller while feeding it out 3 gigabit nics while doing a scsi tape back up... you've got HOW many devices clogging up that bus now? PLUS the full bandwidth of DDR2 800?

please...


Aha ...answer me 2 questions :

1.what happens on a AMD server when I/O acceses memory ?For example the PCI-E videocard or GB NICs?Compare that with the Intel system.



i think some of you guys are letting the fanboy stuff get to you


EDIT: the fact of the matter is, this technology just doesn't scale well


2.How do you explain the lack of bigger than 4S Opteron servers while Xeon scales to 32S since , what , 4 years now ?

[Fred_Pohl]

EDIT: the fact of the matter is, this technology just doesn't scale well

Really?

http://www.theinquirer.net/?article=30963

"Now, the Intel show had a HP dual-socket Woodcrest at 3 GHz / FSB 1333 vs a Sun dual-socket Opteron 285 2.6 GHz / 1 GHz HT, both with 2 GB RAM etc, running SunGard credit analysis application. The Woodcrest completed the job 35% faster, while the power meter also showed roughly 6% less power consumed - 307 W vs 325 W."

"Performance-wise, after the mentioned adjustments, the 3GHz Woodcrest would still be about 25% faster than the 2.8GHz Socket F Opteron in the abovenamed application, and still with somewhat lower power consumption. Well OK, this is an Intel-chosen app. What about more neutral stuff? Here are the expectations based on the 'crystal ball' for, say, SPECint2000 per-core results, as well as SPECint200rate integer 2-core results:

Dempsey 3.73GHz-> 1,800 SPECint2000; 43 CINTrate-peak /May
Woodcrest 3GHz -> 2,400 SPECint2000; 59 CINTrate-peak /July
Opteron 2.8GHz Socket F -> 1,900 SPECint2000 ; 45 CINTrate-peak /September"

[nn_step]

Really?

http://www.theinquirer.net/?article=30963

"Now, the Intel show had a HP dual-socket Woodcrest at 3 GHz / FSB 1333 vs a Sun dual-socket Opteron 285 2.6 GHz / 1 GHz HT, both with 2 GB RAM etc, running SunGard credit analysis application. The Woodcrest completed the job 35% faster, while the power meter also showed roughly 6% less power consumed - 307 W vs 325 W."

"Performance-wise, after the mentioned adjustments, the 3GHz Woodcrest would still be about 25% faster than the 2.8GHz Socket F Opteron in the abovenamed application, and still with somewhat lower power consumption. Well OK, this is an Intel-chosen app. What about more neutral stuff? Here are the expectations based on the 'crystal ball' for, say, SPECint2000 per-core results, as well as SPECint200rate integer 2-core results:

Dempsey 3.73GHz-> 1,800 SPECint2000; 43 CINTrate-peak /May
Woodcrest 3GHz -> 2,400 SPECint2000; 59 CINTrate-peak /July
Opteron 2.8GHz Socket F -> 1,900 SPECint2000 ; 45 CINTrate-peak /September"

how about something current for spice
http://anandtech.com/IT/showdoc.aspx?i=2745
http://images.anandtech.com/graphs/paxvillempvsopt_031906110308/11288.png
http://images.anandtech.com/graphs/paxvillempvsopt_031906110308/11446.png
http://images.anandtech.com/graphs/paxvillempvsopt_031906110308/11447.png
http://images.anandtech.com/graphs/paxvillempvsopt_031906110308/11448.png
http://images.anandtech.com/graphs/paxvillempvsopt_031906110308/11449.png

Wow it looks like someone got thier Asses kicked in these benchmarks.. and it isn't AMD..:fact:
So even if the new Intel is 50% better perfromance than what it is replacing it can barely compete with CURRENT AMD offerings.

[Fred_Pohl]

Wow it looks like someone got thier Asses kicked in these benchmarks.. and it isn't AMD..:fact:
So even if the new Intel is 50% better perfromance than what it is replacing it can barely compete with CURRENT AMD offerings.

Thanks so much for adding that unrelated apples to oranges comparison. It tells us soooo much about what to expect from Woodcrest performance. :slap:

[cupholder2.0]

how so :stick:
Being open to different ideas doesn't mean that you are stupid.. it means you are considering more options

Um..you posted a link to some closed minded fanboy idiot. Im sure you knew it too...:slapass:

[cupholder2.0]

Wow it looks like someone got thier Asses kicked in these benchmarks.. and it isn't AMD..:fact:
So even if the new Intel is 50% better perfromance than what it is replacing it can barely compete with CURRENT AMD offerings.


LOL, I know your upset about Intel Core arcitecture coming, and the truth hurts :). :banana: :banana:

[JumpingJack]

Hehe, I completely agree, I'm just trying to temper (my own) expectations. A healthy launch with Core 2 Extreme being available would be awesome to say the least.

:toast: Makes sense.... better to be surprised than disappointed.

[JumpingJack]

Wow it looks like someone got thier Asses kicked in these benchmarks.. and it isn't AMD..:fact:
So even if the new Intel is 50% better perfromance than what it is replacing it can barely compete with CURRENT AMD offerings.

Could someone please pull out benches between the Northwood and K5??? :)

:p: Silly post due to a soon to be retired netburst... data is data, and the results are written on the wall, er, forum.

[iterations]

Wow it looks like someone got thier Asses kicked in these benchmarks.. and it isn't AMD..:fact:
So even if the new Intel is 50% better perfromance than what it is replacing it can barely compete with CURRENT AMD offerings.

http://www.xtremesystems.org/forums/showthread.php?t=97630

Please read this. This thread is degenerating into exactly what this warning was referencing.

[cupholder2.0]

Are there are boards out there other than the Aopen YDG? The Asus one for 145 supports merom but it sucks at OC :(.

[nn_step]

Um..you posted a link to some closed minded fanboy idiot. Im sure you knew it too...:slapass:
umm how his Anandtech a closed mind Idiot?:confused:
If I remember correctly he is the person you ALL point to for how Conroe is amazing :fact:

LOL, I know your upset about Intel Core arcitecture coming, and the truth hurts :). :banana: :banana:

nope.. I just love to see competition

[cupholder2.0]

umm how his Anandtech a closed mind Idiot?:confused:
If I remember correctly he is the person you ALL point to for how Conroe is amazing :fact:



What? I am referring to that blog..

http://sharikou.blogspot.com/2006/05/with-new-architecture-intel-will-be.html

Anyway I have neither the time nor the inclination to continue this pointless argument with you. So please refrain from violating the forum rules with your flame bait.

[perry_78]

This is pretty funny :) No-one knows anything about AM2 (practically and generally speaking), and then we have a number of benchmarks on a non-released platform.

Then the fight starts. Now why on earth would I think nn_step has the higher ground in this debate? :confused:

;)

[goodcooper]

ok, you guys have posted numbers for benchmarks that aren't refuting anything i'm saying...

i'm saying as the number of cores/devices/chipsets/memory controllers share that FSB, the lower your I/O performance (in latency more than raw mhz) is going to be, whether that has ANYTHING to do with the benchmarks you are posting or not... HT architecture IS more scalable, and that is a FACT... whether or not the numbers point to the new intels or not AT THIS MOMENT ON THIS BENCHMARK is not what i'm talking about, nobody will doubt each camp has thier own strengths... if they do, THAT is where you start pointing the fanboy finger

i mean i'm not saying this guy isn't an :banana::banana::banana:hole, but he presents SOME unrefutable numbers... its straight bits*cycles here folks, and it just doesn't work out in application

EDIT: i'd like to add my views on AM2 after perry's post.... in my eyes AM2 is nothing more than AMD trying to make it in better with the OEMs, allowing them to offer DDR2 in thier systems, we've seen the benchmarks, and beleive me, if the final product was going to be that far from what we see in preproduction benchmarks, they wouldn't even post the preproduction numbers... AM2 will not be that much faster than current chips at all... only its not really that big of a deal, becuase AM2 isn't going to be released with enthusiasts in mind... intel has the upper hand on the desktop market for the next gen (because they've always had it + now they finally have the performance to back up thier popularity), and i don't think AMD will be able to do anything about it, LET ALONE with something like AM2, something will HAVE to come out afterwards (better silicon/better manuf yeilds = more mhz, 65nm?)

[JumpingJack]

Sorry, this isn't really about yields or anything, but I'm just curious. You seem pretty knowledgeable about Intel, and I'm (obviously) not, so... my question is whether Intel has found a way around the exponentially increasing amount of leaking current (either by adding some form of high-k dielectrics or through something similar) or are simply reducing power consumption by increased power saving technology built into the chip.

I ask this because Intel seems confident in their ability to release chips at 45 nanometers in only two more years, and 32 in four. This is compared to the three years it took to retail chips from 90 nanometers down to 65nm. I would assume that based on this, they must have developed some method of at least limiting the problem of leaking current, especially since the 65nm chips seem to be as power efficient as ever with the Core Duo and Merom.

Thanks in advance if you have an answer. :)

Arrrrghhhh - leakage is a funny thing because of the high power of the P4 - Netburst architecture, people assume leakage is the primary culprit. It is not, clock a P4 down and decrease voltage to retain saturation and the power output characteristics are better then the competitions. There was nothing wrong with 90 nm process that caused excess leakage, it was more due to the fact that A) Netburst requires driving to higher clocks and B) Intel for odd reasons chose to double clock the ALU's requiring 2 PLL circuits (PLLs are the major power producers in the CPU).

Without getting into the physics, you can see this in the thermals between Dothan and Prescott, Dothan dissipates 27 watts at 1.8 - 2.0 GHz, but performs significantly better becuase of better IPC efficiency.

People inappropriately tag Intel's 90 nm process as a leaky process, it is not, leakage characteristics are actually quite good.


Jack

[accord99]

Without getting into the physics, you can see this in the thermals between Dothan and Prescott, Dothan dissipates 27 watts at 1.8 - 2.0 GHz, but performs significantly better becuase of better IPC efficiency.
It's even better than what its TDP suggest, Dothan's measured real world power usage (with CPU Burn) is <25W, including power loss due to the motherboard's voltage regulator. Less than the low-voltage Turion MT. Yonah is equally good, also beating the Turion MT.

http://www.silentpcreview.com/article313-page5.html

[Pandamonia]

every super pi score is amazing with conroe, but the 3dmark06 scores are the same as my 4800....

personally im a gamer and i love the latest stuff...

but so far conroe hasnt impressed me at all with 3dmark... also have u noticed how every person who tests these seems to not post 3dmark scores?

ive seen super pi 30 times... who cares how fast we can calculate pi?

i wana know how much better Ghost recon will play if i buy this chip!

3d head to head.... AMD FX vs Conroe EE

thats what im interested in....

i dont care who wins... ill buy who ever does.

[Fred_Pohl]

Now why on earth would I think nn_step has the higher ground in this debate? :confused:

;)

Good question.

[Fred_Pohl]

ok, you guys have posted numbers for benchmarks that aren't refuting anything i'm saying...

i'm saying as the number of cores/devices/chipsets/memory controllers share that FSB, the lower your I/O performance (in latency more than raw mhz) is going to be, whether that has ANYTHING to do with the benchmarks you are posting or not... HT architecture IS more scalable, and that is a FACT... whether or not the numbers point to the new intels or not AT THIS MOMENT ON THIS BENCHMARK is not what i'm talking about, nobody will doubt each camp has thier own strengths... if they do, THAT is where you start pointing the fanboy finger

You have a valid point there. HT is more scalable than Intel's GTL+ bus. However you started out by saying that you could not recommend Woodcrest for server/enterprise use over Opteron because Woodcrest will be I/O bottlenecked. What is missing from your argument is how great the impact of this theoretical bottleneck will be and at what point it will give Opteron a decisive advantage. Will it happen with four 3GHz cores sharing a dual 1333fsb or not until you have 16 cores sharing the same fsb bandwidth/latency? The evidence I've seen thus far suggests that <3GHz Woodcrest will have none of the I/O problems you speak of at the 4 core level and possibly not until more than 8 cores are sharing the same fsb.

Considering that the majority of server/enterprise systems use less than 8 cores per mobo, how can you dismiss Woodcrest as a viable alternative to Opteron based on scalability limitations that don't seem to appear until you have 8 or more cores sharing a single platform? IMO Woodcrest will be an excellent choice for server/enterprise systems in the mainstream enterprise market segment. HT may well give Opteron an advantage when 8 or more cores have to share a single platform but that remains a matter of speculation.

[Fred_Pohl]

i wana know how much better Ghost recon will play if i buy this chip!

3d head to head.... AMD FX vs Conroe EE

thats what im interested in....

i dont care who wins... ill buy who ever does.


http://images.anandtech.com/graphs/intel%20conroe%20idf%20preview_03070620313/11083.png

http://images.anandtech.com/graphs/intel%20conroe%20idf%20preview_03070620313/11084.png

http://images.anandtech.com/graphs/intel%20conroe%20idf%20preview_03070620313/11085.png

http://images.anandtech.com/graphs/intel%20conroe%20idf%20preview_03070620313/11086.png

http://images.anandtech.com/graphs/intel%20conroe%20idf%20preview_03070620313/11089.png

Enjoy your new Conroe and the money you save over FX62.

[nn_step]

ah yes the black box tests.. I was wondering when you would bring that up..
Considering the fact that Anand was not allowed to strip and fully test it.. I would take the scores with a grain of salt...

[goodcooper]

It's even better than what its TDP suggest, Dothan's measured real world power usage (with CPU Burn) is <25W, including power loss due to the motherboard's voltage regulator. Less than the low-voltage Turion MT. Yonah is equally good, also beating the Turion MT.

did this TDP also take into account Dothan's OFF DIE memory controller as well? there is more than just a cpu under the hood of that turion... once you factor that in, they become a lot more comparable (iirc, the dothan still wins, but it isn't by a landslide by any stretch of the imagination)



What is missing from your argument is how great the impact of this theoretical bottleneck

an EXCELLENT point... we really don't know how great the impact this bottleneck has (its not a theoritical one) with current chips, mainly because we don't have the chips, HT on the other hand... well... scalable...


The evidence I've seen thus far suggests that <3GHz Woodcrest will have none of the I/O problems you speak of at the 4 core level and possibly not until more than 8 cores are sharing the same fsb.

Ok, and does this evidence include other things on the bus besides said processors and memory? how about a scsi pcix controller, 2 or more gigabit cards on a pcix bus? shoot throw a pcie graphics card in there too

[Fred_Pohl]

3d head to head.... AMD FX vs Conroe EE

thats what im interested in....

i dont care who wins... ill buy who ever does.

BTW, the X2-4800+ is a damb fine cpu but do you really get these same scores in 3DM06?

http://coolaler.kj.idv.tw/merom/QLZT/XP64/5.gif

Unfortunately that score is for Merom @3.2G. Conroe will score score higher because of it's increased fsb speed.

[accord99]

did this TDP also take into account Dothan's OFF DIE memory controller as well? there is more than just a cpu under the hood of that turion... once you factor that in, they become a lot more comparable (iirc, the dothan still wins, but it isn't by a landslide by any stretch of the imagination)
The off-die memory controller accounts for 1-2W. And this is Dothan @1.3v vs the rather uncommon Turion MT at 1.2V. Versus the normal Turion ML, and it becomes a landslide similar to A64 vs Prescott.

[Fred_Pohl]

ah yes the black box tests.. I was wondering when you would bring that up..
Considering the fact that Anand was not allowed to strip and fully test it.. I would take the scores with a grain of salt...

I don't blame you for being skeptical about those results. At my regular forum I too was skeptical when many members predicted 20% performance gains from AM2, DDR3 support out of the box, 65nm K8Ls with 4MB L3 ZRAM cache shipping in June, etc, etc.

OTOH, when considering the numerous architectural advantages that Core 2 has over K8, I find it hard to be skeptical about those IDF results or the results being posted here daily by Coolaler and others lucky enough to get their hands on Core 2 ES chips.

[nn_step]

I don't blame you for being skeptical about those results. At my regular forum I too was skeptical when many members predicted 20% performance gains from AM2, DDR3 support out of the box, 65nm K8Ls with 4MB L3 ZRAM cache shipping in June, etc, etc.

OTOH, when considering the numerous architectural advantages that Core 2 has over K8, I find it hard to be skeptical about those IDF results or the results being posted here daily by Coolaler and others lucky enough to get their hands on Core 2 ES chips.
I admit Core Architecture is lightyears beyond Prescott but it is by nature a modified Pentium 3 Core never the less with some very nice advanced parts.
But like ALL Architectures they all have disadvantages..
Like the Prescott for all of its flaws.. till this day and probably for years to come... it is still the best for Photoshop..

[goodcooper]

The off-die memory controller accounts for 1-2W. And this is Dothan @1.3v vs the rather uncommon Turion MT at 1.2V. Versus the normal Turion ML, and it becomes a landslide similar to A64 vs Prescott.

the cheap compaq i use for work is a MT, and most of the time it stays at 1v..., and i remember seeing numbers on memory controllers using more like 5-6w for dual channel


anyway, this is OT

[Fred_Pohl]

an EXCELLENT point... we really don't know how great the impact this bottleneck has (its not a theoritical one) with current chips, mainly because we don't have the chips, HT on the other hand... well... scalable...



Ok, and does this evidence include other things on the bus besides said processors and memory? how about a scsi pcix controller, 2 or more gigabit cards on a pcix bus? shoot throw a pcie graphics card in there too

I believe we have arrived at an impasse. You apparently believe (correct me if I'm wrong here) that Woodcrest will be severely I/O bottlenecked by it's GTL+ fsb and I don't believe it will be a significant factor with 4 or even 8 cores at <3GHz sharing a dual 1333fsb. Neither of us has any hard evidence to support our positions so it appears that we will have to wait a few more weeks to find out.

BTW, I've seen Sharikou's blog and I do not consider his fanboy opinions to be hard evidence of anything other than the apparent fact that he will spin doctor anything he can find to make Intel look bad and AMD look good and censor those who present contradictory evidence.

[Fred_Pohl]

I admit Core Architecture is lightyears beyond Prescott but it is by nature a modified Pentium 3 Core never the less with some very nice advanced parts.


That is just so completely wrong. Core 2 is a modified P3 about as much as K8 is a modified K6. This is an excellent read:

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2748

[Thorburn]

ah yes the black box tests.. I was wondering when you would bring that up..
Considering the fact that Anand was not allowed to strip and fully test it.. I would take the scores with a grain of salt...

Trust me when I say these systems are all built to the highest quality, lies get found out eventually so whats the point in trying.
If you read the follow up articles on Anandtech then you'll know allowed to examine the system, update the BIOS, tweak memory timings, etc.
As for 'stripping down' the systems, you have maybe 20 or more publications all eager to get there own benchmark numbers, if you let them all pull the machines to pieces then you'd:
a) Need an event to last about 3 weeks, not 3 days
b) Have some idiot break the thing

At an event time is limited so you can't let every Tom, :banana::banana::banana::banana: and Harry dismantle the test machine just so they can prove a point to some fanboys.

[[XC] MarioMaster]

Meh, I've pretty much switched to AMD but I may go the way of the Conroe but still don't really feel the need to upgrade yet.
Intel 16mhz
Intel Celeron 433mhz
AMD Athlon XP 2500+
AMD Athlon 64 3000+ (laptop)

I have a 1.8 ghz celeron but it's not a gaming rig I got a bit ago, just for rosetta.

[Thorburn]

That is just so completely wrong. Core 2 is a modified P3 about as much as K8 is a modified K6. This is an excellent read:

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2748

Absolutely right, at a silicon level the Core 2 Duo is 80% different to the Core Duo, let alone the Pentium 3

[nn_step]

That is just so completely wrong. Core 2 is a modified P3 about as much as K8 is a modified K6. This is an excellent read:

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2748
Here is a better article for you
http://arstechnica.com/articles/paedia/cpu/core.ars

[goodcooper]

I believe we have arrived at an impasse. You apparently believe (correct me if I'm wrong here) that Woodcrest will be severely I/O bottlenecked by it's GTL+ fsb and I don't believe it will be a significant factor with 4 or even 8 cores at <3GHz sharing a dual 1333fsb. Neither of us has any hard evidence to support our positions so it appears that we will have to wait a few more weeks to find out.

BTW, I've seen Sharikou's blog and I do not consider his fanboy opinions to be hard evidence of anything other than the apparent fact that he will spin doctor anything he can find to make Intel look bad and AMD look good and censor those who present contradictory evidence.


severely, no not really, enough to make opteron a better choice? yes... not by the cpus and memory alone mind you, but once you factor in all the other I/O goodies...

my point is, we don't need to speculate about HT, because the numbers tell us there is no problem with HT, the simple fact that there are so many dedicated busses for things now instead of one clogged up and overused front side one

the numbers for woodcrest is making it look like we may need to see the real world performance, because like i said earlier, bits*cycles, it just doesn't add up, it is using new technology on an old platform, a platform that AMD has wisely decided to retire, and Intel hasn't


EDIT: also, who cares if the architecture of the core2 SEEMS like it took a step back? it is faster and that is the end of it, the way i see it, the discussion should be about the busses

[Fred_Pohl]

severely, no not really, enough to make opteron a better choice? yes... not by the cpus and memory alone mind you, but once you factor in all the other I/O goodies...

my point is, we don't need to speculate about HT, because the numbers tell us there is no problem with HT, the simple fact that there are so many dedicated busses for things now instead of one clogged up and overused front side one

the numbers for woodcrest is making it look like we may need to see the real world performance, because like i said earlier, bits*cycles, it just doesn't add up, it is using new technology on an old platform, a platform that AMD has wisely decided to retire, and Intel hasn't


EDIT: also, who cares if the architecture of the core2 SEEMS like it took a step back? it is faster and that is the end of it, the way i see it, the discussion should be about the busses

I dunno... if the I/O penalties are enough to make Opteron a better choice, despite ~20% lower performance per clock at ~20% higher power consumption, I think I would still call that "severe".

I won't argue that HT isn't far superior to GTL+ no matter how much Intel tweaks it. I just question how much of that bandwidth is actually needed and used at this time. IMO HT offers more bandwidth than necessary and Intel is offering just enough with their current fsb schemes. We'll see soon enough.

While I would like to see Intel abandon their NIH mentality regarding HT, at least until they get CSI finished, I don't foresee any significant benefits at the dual-core desktop level, which is where my interests lie.

Who on earth would think that the Core 2 architecture is a step back? Other than Sharikou and other like-minded AMD fanboys desperate to find any flaws, real or imagined, in Core 2, I can't imagine who would think that...

[Fred_Pohl]

Here is a better article for you
http://arstechnica.com/articles/paedia/cpu/core.ars

Thanks but I've already read that article and there is no mention in there of Core 2 Duo being a modified P3 either...

Although Core represents a very significant departure from Netburst and a return to a more traditional high IPC x86 design, that is where the similarities to the P3 end.

[Fred_Pohl]

Absolutely right, at a silicon level the Core 2 Duo is 80% different to the Core Duo, let alone the Pentium 3

Excellent point! Saying that Core 2 Duo is a modified P3 because it has little in common with Netburst is like saying that K8 is a modified 386 because it shares many of the same x86 instructions.

[iterations]

every super pi score is amazing with conroe, but the 3dmark06 scores are the same as my 4800....


Sorry, but I gotta call BS on that.

Clock for clock comparison for FX-60 and a Merom Engineering Sample:

3DMark06 CPU Test: http://coolaler.kj.idv.tw/merom/QLZT/XP64/5.gif

FX-60 @ 3.2GHz gets 2380 3DMark06 CPU score
Merom @ 3.2GHz gets 2833 3DMark06 CPU score

That is a 20% improvement clock for clock, and that Merom is air cooled.

[IntelUser]

Conroe is 140mm^2.
Kentsfield should be 2x Conroe , 280mm^2. Numbers of possible defects increases at least 2^2 , 4 times.So with Kentsfield you might have abysimal yeilds.

That's theory but considering how Prescott 1M is 112mm2 and Pentium D which supposed to be 224mm2 die is only 206mm2 die I disagree. The number of transistors on Prescott is 125 million but Pentium D is 230. Again not double.


Early testing of a 2-way (8 core) Cloverton system showed that it scales by a factor of 4.75X in 8 way Cinebench vs 4.8X for an 8 core Opteron system.
Wow it looks like someone got thier Asses kicked in these benchmarks.. and it isn't AMD.. FPRIVATE "TYPE=PICT;ALT="
So even if the new Intel is 50% better perfromance than what it is replacing it can barely compete with CURRENT AMD offerings.

Wow talk about being narrow minded. Intel's performance advantage with Woodcrest over Paxville is SpecInt_Rate, which scales almost perfectly with clock speeeds(~85%), and is almost not impacted by memory bandwidth.

First of all, Woodcrest platform will be 1333MHz FSB. Second of all, Woodcrest will have TWO 1333MHz FSB. Now compare that to Paxville, which has SINGLE 800MHz FSB.

One of the Intel presentation concerning Woodcrest says it will be 52% faster than Dempsey, which is a 2MB cache per core successor to Paxville, and has 2x1066MHz FSB. Now beat that. Dempsey is alone 20-50% faster than Paxville: http://www.realworldtech.com/page.cfm?ArticleID=RWT110805135916&p=6


every super pi score is amazing with conroe, but the 3dmark06 scores are the same as my 4800....

personally im a gamer and i love the latest stuff...

but so far conroe hasnt impressed me at all with 3dmark... also have u noticed how every person who tests these seems to not post 3dmark scores?

LOL, my friend wanted to submit his 3dmark score but realized that with SAME configuration, that he wasn't even close to the HDD scores posted on the internet. Funny how with same configuration, the differences in 3dmark score can be big as 30%. So yea you can't say it based on that.


ah yes the black box tests.. I was wondering when you would bring that up..
Considering the fact that Anand was not allowed to strip and fully test it.. I would take the scores with a grain of salt...

Pfft. Typical comments of a person who's always skeptical. How would you modify the system?? Do you know on the second revision of the benchmark that with THEIR OWN QUAKE 4 DEMO, it had greater differences between Conroe and A64 FX OC'ed?? How's that Intel unfairly modding Conroe to have advantage over A64, when Anand's own demo is EVEN FASTER!!??


Core has about as much in common with the P3 as the P3 does with the P2.

LOL. Then you have no more idea about computer chips than the average joe. Pentium III, all it brought is SSE instructions. Later Pentium III got faster bus and on-die L2 cache, but anyone who researched into Pentium II and Pentium III would know that mobile Pentium II's had on-die 256KB L2 cache, which is coincidentally same as Coppermine Pentium III's.

The differences between Core Duo and Core 2 Duo is like Pentium and Pentium II.


Early testing of a 2-way (8 core) Cloverton system showed that it scales by a factor of 4.75X in 8 way Cinebench vs 4.8X for an 8 core Opteron system.

That just tells that Cinebench isn't really a accurate benchmark for the entire CPU. Just look how Pentium EE is 15% faster than Pentium D, when Pentium D is half of the time faster.


The only thing I am sceptical about at the moment is the fact that all results we have seen of Conroe has been with chips with 4MB L2 Cache and the ones we will be buying will only be 2MB. I try to convince myself that doubling the cache will not impact greatly the performance!

You would know by looking at benchmarks the benefits of doubling the caches are less than 5% in most of the cases. Think the 2MB cache Core 2 Duo will be much slower?? Some people should get over the fact that Core 2 Duo isn't gonna bring mediocre 5-7% performance increase that previous Intel architectures brought.


I kinda doubt we will see the 3.33GHz part right off the bat. Most likely Intel will need some time to perfect the 65nm process. My gut tells me that is a pre-Xmas part.

You'll likely be right. The roadmaps indicating Q4 for EE should be very likely. Even though people say it should be taken with grain of salt, people also must have said that when Prescott is shown to be slower per clock than Northwood.

What I am saying here is if you can pick the right ones, the rumors are correct.


Sadly, that is true and Sarikou is just the deluded fanboy idiot for the job!

He has calculations saying Yonah will use 50W. He also says that the reason Yonah is used in desktop replacements are because its 50W. That's because initial designs for laptops for a fastest CPU will be desktop replacements. Look at now with T&L Core Duo laptops.

Enough said.


The average performance difference is between 20-25% in favor of Core 2. I researched almost 30 different benchmarks and established that X2 performance scales by 8-9% per 200MHz clock increase. This leads me to conclude that on average a dual-core AM2 cpu will need to run 400MHz faster to roughly equal the performance offered by Core 2.

400MHz difference when one is at 500MHz and 400Mhz difference when one is at 4GHz is a VAST difference.

Based on your numbers of 8-9% scaling, which isn't exactly true as games are generally 4-6% per 10%.

Core 2 Duo performance per clock will give at least 30% clock speed difference over the Athlon 64 AM2's. That's around 500-700Mhz difference.

[Direwolf]

OT- Quick stupid question- Will the soon to be released conroe motherboards all have a fsb wall at only 400fsb? Just a quickie :)...

[FunkyRider]

400 x 4 = 1600 that should be high enough for a stock 1066 processor IMO

[IntelUser]

Those skeptical about Conroe rumors should look back at Prescott. Couple of weeks before release, almost every news was negative. Prescott has 25 stage pipelines... Prescott has 30 stage pipelines, Prescott performs equal to Northwood per clock... Prescott performs lower than Northwood per clock... Prescott runs hot... Prescott is delayed... Prescott is at same clock speed as Northwood. And the WORST news all came true.

Look at Conroe. There isn't bad news except the sites that are made by AMD fanboys and the ones that are skeptical. When Core 2 Duo at 2.167GHz, 2MB cache, 1066MHz FSB, is competing for dominance against AM2 Athlon 64 FX, those peeps are gonna shut up, and lose credibility.

[Fred_Pohl]

400MHz difference when one is at 500MHz and 400Mhz difference when one is at 4GHz is a VAST difference.

Based on your numbers of 8-9% scaling, which isn't exactly true as games are generally 4-6% per 10%.

Core 2 Duo performance per clock will give at least 30% clock speed difference over the Athlon 64 AM2's. That's around 500-700Mhz difference.

That's not entirely fair as there are no 500MHz or 4GHz X2s. FWIW, all I did to arrive at that 8-9% per 200MHz figure was research the THG interactive CPU comparison charts and measure the performance differences between 1MB X2s at 2.4 and 2.6G and then spot checked various BMs at 2.0-2.2GHz. The difference was consistently about 8-9% even in most games. I didn't include any BMs that showed very little or no cpu scaling, such as memory bandwidth, etc. Not very scientific but I think it's a reasonable ballpark estimate for X2 scaling between 2-2.6GHz.

You may well be correct about your 30% (500-700MHz) figures but I prefer to remain more conservative until Conroe officially launches and most likely exceeds estimates that already give AMD fanboys serious heartburn. :D

[savantu]

That's theory but considering how Prescott 1M is 112mm2 and Pentium D which supposed to be 224mm2 die is only 206mm2 die I disagree. The number of transistors on Prescott is 125 million but Pentium D is 230. Again not double.



...

We're talking a multichip solution , PD 9xx not the single die PD8xx.Care to revise your statements now ?

[Moonraker™]

E6600 Can practically piss all over the AMD which is almost £500 more than it. (I think it's around £440 more right now)

[Meaker]

I cant see the Core duo 2 based chips scaling aswell as the opteron since even with two fsbs thats nothing compared to 8 and the hypertransport system connecting all of them.

[Lestat]

I cant see the Core duo 2 based chips scaling aswell as the opteron since even with two fsbs thats nothing compared to 8 and the hypertransport system connecting all of them.


have you really not been reading any of the conroe threads and the performance ?

its not an amd that break the Pi records, and is still breaking them.. it was a conroe.

its not an amd making and breaking all kinds of other records,, its a conroe.

i said it before and i'll say it again. the intel Core architecture is the next step in cpu's that we have been waiting for for a very long time.

[Pandamonia]

the intel core duo is a tech based on intel pentium 4 mobile Chips, it was a tech deveoloped paralell to netburst and when netburst showed its life was coming to an end they worked on the other tech and made it alot better.

there is nothing new about core duo. its just an improved paralell tech that was deveoloped cus netburst wasnt suitable for mobile.

for a while people have known this tech was better than netburst and had more potential but they didnt want the netburst to die so fast, intel wanted to rape your wallets for a few more years.

typical intel.

they finally have a product worth purchasing.

[XS Janus]

:rolleyes:
________
Black :banana::banana::banana::banana: (http://www.:banana::banana::banana::banana:tube.com/categories/11/black/videos/1)