Barcelona Opteron 2350(B1) arrived

Printable View

Show 100 post(s) from this thread on one page

10-26-2007, 12:54 AM
savantu

The L3 latency is 25 cycles , that's a lot.
10-26-2007, 12:58 AM
Lastviking

Some fast testing for showing K8 improvement over mhz

1 year old WinXp install for 24/7 os (thats why the times are slow and also 4x512mb and not tweaked)

Dfi Venus
Opty 165
4x512mb bh5

Used clockgen in windows to lower and higher the clocks"not same memmory speed and fsb, like the k10 scores"

54,907s 1600mhz 177fsb
43,641 2000 222
36,266 2400 267 ~7,375 improvement
34,843 2500 277
33,453 2600 289
32,203 2700 300
31,078 2800 311
30,000 2900 322
28,938 3000 333
28,047 3100 344 ~4,156 improvement from 2700mhz
10-26-2007, 01:34 AM
Lightman

Quote:

Originally Posted by kyosen

Please calm down about scaling for SuperPI time...
I'll try it again, with K8 Opteron 2212 and K10, with same OS:yepp:

Domo Arigato!
This will clearly show performance improvement between K8 and K10 in applications you will run! :up:
10-26-2007, 02:28 AM
JohnMike

Quote:

Originally Posted by kyosen

And these are latecy test results with Everest Ultimate v4.20:
at 2.0G=200x10: 76.0ns
at 2.2G=220x10: 69.0ns
at 2.4G=240x10: 63.4ns

That's not bad... I think the «2.4G=240x10: 63.4ns» is not far form Intel Qcore @2.4Ghz.

I would like to know how far RAM can run on a K10 system... If can catch Intel systems...

:shrug:
10-26-2007, 04:06 AM
PhilDoc

Thanks for all the hard work Kyosen. The L3 latency was pretty much expected. At this point no one seems to know how that will affect performance, but time will tell. Very supprised about the spi scores with gangled and ungangled. Could have had a better choice of terminology, lol, but maybe a repeat of the scores just to be sure. Looking forward to anything you'd like to run, looks great so far.
10-26-2007, 04:59 AM
hipro5

L3 Cache had a good impact back in the days of P4 Northwood (Gallatin) CPU....You could gain about 4% - 8% more depending on applications....So IF Barchy gets ONLY 4-8% more than K8, It's the L3 Cache, so AMD isn't going to be that fast....If it gains more, then it will be OK....;)
10-26-2007, 05:17 AM
Sparky

Quote:

Originally Posted by perkam

Then let me tell you in advance, it will :D

Perkam

I do hope you're right *crosses fingers*
10-26-2007, 07:03 AM
Blasar

Quote:

Originally Posted by SparkyJJO

I do hope you're right *crosses fingers*

I dont know if this is a stupid question that was answered before, if that´s the case excuse me, but how does Perkam knows

that they will be better??

Nevertheless, i´m eagerly waiting to see the Phenom results, and i hope Perkam is absolutely right..:yepp:
10-26-2007, 08:03 AM
cky2k6

Quote:

Originally Posted by perkam

Then let me tell you in advance, it will :D

Perkam

not trying to be a party pooper... but its a pretty bold statement, and from all that's been said, its just a higher clocking process tweak. let me tell you, i will be ecstatic if its more.
10-26-2007, 09:10 AM
freeloader

Any chance at some 2.6ghz love (or higher)? :D

BTW...good job! Finally nice to see some overclocking with these chips.
10-26-2007, 09:13 AM
PhilDoc

Quote:

Originally Posted by cky2k6

not trying to be a party pooper... but its a pretty bold statement, and from all that's been said, its just a higher clocking process tweak. let me tell you, i will be ecstatic if its more.

I'm not sure I like the way you opined that. :D
10-26-2007, 09:15 AM
BeardyMan

Hmm strange

Quad core

look at that bios shot

it says

L1 = 512Kb, 128X4
L2 = 2048kb , 512kb X4
L3 = 2048kb
is it me or isn't that wrong?
10-26-2007, 09:30 AM
Mad_Man

L3 is shared 2MB so its correct
10-26-2007, 09:30 AM
nesham

You are wrong. Barca have 4x128KB L1 + 4x512KB L2 + 2 MB L3 shared cache.
10-26-2007, 09:31 AM
s7e9h3n

Quote:

Originally Posted by BeardyMan

Hmm strange

Quad core

look at that bios shot

it says

L1 = 512Kb, 128X4
L2 = 2048kb , 512kb X4
L3 = 2048kb
is it me or isn't that wrong?

It's you ;) The l3 cache is shared....
10-26-2007, 09:32 AM
BeardyMan

Quote:

Originally Posted by nesham

You are wrong. Barca have 4x128KB L1 + 4x512KB L2 + 2 MB L3 shared cache.

LOL?:D
10-26-2007, 09:32 AM
BeardyMan

Quote:

Originally Posted by s7e9h3n

It's you ;) The l3 cache is shared....

No it's you :D

This forum requires that you wait 30 seconds between posts. Please try again in 4 seconds.
10-26-2007, 09:39 AM
akaBruno

Quote:

Originally Posted by PhilDoc

I'm not sure I like the way you opined that. :D

:clap:

o·pine (ō-pīn')

v., o·pined, o·pin·ing, o·pines.

v.tr.
To state as an opinion.

v.intr.
To express an opinion: opined on the defendant's testimony.

[Middle English opinen, from Old French opiner, from Latin opīnārī, to suppose.]
10-26-2007, 09:51 AM
cky2k6

Quote:

Originally Posted by akaBruno

:clap:

o·pine (ō-pīn')

v., o·pined, o·pin·ing, o·pines.

v.tr.
To state as an opinion.

v.intr.
To express an opinion: opined on the defendant's testimony.

[Middle English opinen, from Old French opiner, from Latin opīnārī, to suppose.]

his usage of the word was actually correct ;) thanks for the definition mr. chaucer
10-26-2007, 09:51 AM
tsuehpsyde

Thanks for the work Kyosen. :) Very much appreciated.
10-26-2007, 03:03 PM
tux08902

Does it clock any higher?
10-26-2007, 03:52 PM
Viper666

funny the L3 runs at the speed of the memory controller,on Barcy thats 1.8 @2Ghz on Phenom its alot higher ! I think I saw 4Ghz somewhere
10-26-2007, 05:35 PM
tictac

so we can overclock l3 cache? wow..
10-26-2007, 05:43 PM
akaBruno

Quote:

Originally Posted by tux08902

Does it clock any higher?

I doubt it will without some more juice pumpin. We're lookin at a NF4 here not a 790.

Kyosen and Stephen are owed a debt of gratitude. They've taken crap chips to a level we haven't seen yet. On crap boards, I might add...

As far as I know... these guys are the only ones on the internet that have given us a glimpse of the AMD future.

Thanks guys,

BRUNO
10-26-2007, 06:38 PM
PhilDoc

Imho, chipsets have always been a major problem for AMD and have hampered performance. Whether their own or other manufactures. The NF4 did happen to be one of the better though. Lets hope the new AMD 790 finally gives AMD users a great one.
10-26-2007, 07:23 PM
Viper666

Quote:

Originally Posted by akaBruno

I doubt it will without some more juice pumpin. We're lookin at a NF4 here not a 790.

Kyosen and Stephen are owed a debt of gratitude. They've taken crap chips to a level we haven't seen yet. On crap boards, I might add...

As far as I know... these guys are the only ones on the internet that have given us a glimpse of the AMD future.

Thanks guys,

BRUNO

Hum mine are running quite well TY on SUSE 10.3
10-26-2007, 07:25 PM
Gothic

Quote:

Originally Posted by hipro5

I think it only needs a single stage to find out that..... :D

Fresh water should do it. At least my opty coldbugs at 18c @ 363MHz HTT.
10-27-2007, 12:50 AM
KTE

Quote:

Originally Posted by Cronos

For dual-channel mode Spi 1M time was 38.968, and it is clear that overclocked results were taken with memory in dual-channel mode.
So, 38.968 / 32.672 = 1.1927 = 19.3% it is.

I don't want to dwell on this here especially out of respect for others, but I'd like to make it clear. My method was correct but I was in error when I forgot to normalize the values for decreasing number = increasing performance. The formula stays consistent but the math in this scenario becomes:-
((1/32.672 - 1/38.968) / 1/38.968) x 100
= ((0.0306072 - 0.0256621) / 0.0256621) x 100
= 19.27% performance gain with a 20% clock increase

Hope it helps. :)

========

Com'on Kyosen-san, we are all waiting for some usual magic :D

... And for Stephen :stick:
10-27-2007, 01:27 AM
SOLDNER-MOFO64

Quote:

Originally Posted by akaBruno

. Kyosen and Stephen are owed a debt of gratitude. They've taken crap chips to a level we haven't seen yet. On crap boards, I might add...

As far as I know... these guys are the only ones on the internet that have given us a glimpse of the AMD future.

Thanks guys,

BRUNO

Excuse me, "CRAP CHIPS" ? Oh you mean the first revision of new chips which is obviously labelled as CRAP by the hardcore uber elite clockers like you?

Quote:

Originally Posted by PhilDoc

Imho, chipsets have always been a major problem for AMD and have hampered performance. Whether their own or other manufactures. The NF4 did happen to be one of the better though. Lets hope the new AMD 790 finally gives AMD users a great one.

Explain how Nforce 2,3,4,500,550,570,590 or AMD 690G chipsets had MAJOR problems and hampered performance?
10-27-2007, 02:08 AM
kyosen

Today, I got my own Barcelona, Opteron 2346 HE:)

Bulk, but BA stepping:)
http://222.151.153.254/c-board/file/...6_GE_BA_x2.jpg

BIOS screen shot...NorthBridge clock is 1.60GHz
http://222.151.153.254/c-board/file/...BA_BIOS_NB.jpg

AMD Power Monitor, Everest, and CPU-Z
http://222.151.153.254/c-board/file/...rMon_Evrst.png

Memory clock is showed differently between CPU-Z and Everest...
I have no idea which is correct or not...

Now I'm thinking Vcore & Vnb mod,
but yeah I should run several benchmark programs
before I broke this board by mistake:D
10-27-2007, 02:32 AM
SOLDNER-MOFO64

best of luck m8
10-27-2007, 04:24 AM
Lightman

Quote:

Originally Posted by kyosen

.....
Now I'm thinking Vcore & Vnb mod,
but yeah I should run several benchmark programs
before I broke this board by mistake:D

You're very wise man :yepp:
10-27-2007, 05:23 AM
PhilDoc

We'd have to highjack the entire thread, but an example would be the NF3 and other early 939 chipsets. If you gave it too much voltage on vdimm you couldn't boot at 200 fsb. Most thought the boards were all buggered until Misteroadster and myself figure out you had to boot around 230 ish. There's and entire sticky about problems with the dfi nf4.

Anyway, too me this is a plus for AMD. It means there's extra performance that can be gained with an excellent chipset, which so far it appears they have with the 790
10-27-2007, 06:27 AM
Sparky

Quote:

Originally Posted by SOLDNER-MOFO64

Explain how Nforce 2,3,4,500,550,570,590 or AMD 690G chipsets had MAJOR problems and hampered performance?

Add ATI RD580 to the good list.
10-27-2007, 09:35 AM
quadnad

This is certainly a thread to watch. I'd love to see how the scaling shows up in real-world apps.
10-27-2007, 09:34 PM
Charles Wirth

Following the thread Kyosen-san, waiting for you to make those chips fly.
I hope you dont mind me doing a quick compare at your Mhz.
Vista 32bit, all services running.

http://222.151.147.26/c-board/file/C...t2350_2.4G.png
http://fugger.netfirms.com/c3.jpg
10-27-2007, 09:55 PM
s7e9h3n

Quote:

Originally Posted by FUGGER

Following the thread Kyosen-san, waiting for you to make those chips fly.
I hope you dont mind me doing a quick compare at your Mhz.
Vista 32bit, all services running.

Charles, it'd be nice if you would have done a comparison using hardware that's currently shipping :rolleyes: Last I checked, Newegg doesn't list any Yorkfield Qx9770's along with Corsair PC3-16000's (as well as an "unknown" mb) :p: Just wait a bit and a more fair comparison would be with b2 stepping barcelonas.

BTW, I'm surprised the Barcelona is even running that close to the Yorky @ 3.2....maybe there is hope for Phenom after all :yepp:
10-27-2007, 10:18 PM
Viper666

Lol

Quote:

Originally Posted by s7e9h3n

Charles, it'd be nice if you would have done a comparison using hardware that's currently shipping :rolleyes: Last I checked, Newegg doesn't list any Yorkfield Qx9770's along with Corsair PC3-16000's (as well as an "unknown" mb) :p: Just wait a bit and a more fair comparison would be with b2 stepping barcelonas.

BTW, I'm surprised the Barcelona is even running that close to the Yorky @ 3.2....maybe there is hope for Phenom after all :yepp:

"BTW, I'm surprised the Barcelona is even running that close to the Yorky @ 3.2....maybe there is hope for Phenom after all "

Come on S7 thats normal Intels $2500 processor always beats Amd's $400 one and its always the comparison that gets made. One thing for sure that Yorky cant beat Crays Red Storm(Powered BY K8 Opterons),Intel doesnt make anything that can Only IBM does

Yea I got this feeling in my gut that says,there gonna be a real big shocker comin soon

followed by major fanboy whinning and crying
10-27-2007, 10:23 PM
s7e9h3n

Oops...Or is the Yorky running @ 2.4? I say your cheating! :p: j/k
10-27-2007, 10:24 PM
mongoled

Quote:

Originally Posted by Viper666

"BTW, I'm surprised the Barcelona is even running that close to the Yorky @ 3.2....maybe there is hope for Phenom after all "

Come on S7 thats normal Intels $2500 processor always beats Amd's $400 one and its always the comparison that gets made. One thing for sure that Yorky cant beat Crays Red Storm(Powered BY K8 Opterons),Intel doesnt make anything that can Only IBM does

Yea I got this feeling in my gut that says,there gonna be a real big shocker comin soon

followed by major fanboy whinning and crying

Quote:

Originally Posted by s7e9h3n

Charles, it'd be nice if you would have done a comparison using hardware that's currently shipping :rolleyes: Last I checked, Newegg doesn't list any Yorkfield Qx9770's along with Corsair PC3-16000's (as well as an "unknown" mb) :p: Just wait a bit and a more fair comparison would be with b2 stepping barcelonas.

BTW, I'm surprised the Barcelona is even running that close to the Yorky @ 3.2....maybe there is hope for Phenom after all :yepp:

Well the Yorky is running at 2.4Ghz and looks like Charles is playing Devils advocate.......
10-27-2007, 10:40 PM
Charles Wirth

I clocked down, below the Q9450 with Vista 32bit.
Showing where things are now is interesting.
10-27-2007, 10:48 PM
hipro5

I think that what FUGGER is trying to say is that IF a person in the near future goes and buys TWO machines, one with Phenom at 2.4GHz and one with Yorkfield at 2.4GHz TOO (NOT expenssive at all) and compare them both, he'll get the above resaults...... Nothing more and nothing less....:)
10-27-2007, 10:51 PM
mongoled

Quote:

Originally Posted by hipro5

I think that what FUGGER is trying to say is that IF a person in the near future goes and buys TWO machines, one with Phenom at 2.4GHz and one with Yorkfield at 2.4GHz TOO (NOT expenssive at all) and compare them both, he'll get the above resaults...... Nothing more and nothing less....:)

Yes but this is flawed

Since when did the HE chips == Phenom

:P

oh an come sum1 put the 600 points gap into perspective.

i.e. using a baseline system, how many more mhz would be required to achieve those xtra points?

-edit-

700 points....
10-27-2007, 10:51 PM
Viper666

sorry I dont know him ,and no disrespect but Ive already seen benches of that proc. and it didnt beat barcy at 2.5 by that much !
10-27-2007, 10:57 PM
mongoled

Quote:

Originally Posted by Viper666

sorry I dont know him ,and no disrespect but Ive already seen benches of that proc. and it didnt beat barcy at 2.5 by that much !

Sorry to newer members of the forum

Charles == Fugger

:p:

Looking forward to the Yorkie-Phenoms head to head

:yepp:
10-27-2007, 11:21 PM
accord99

Quote:

Originally Posted by mongoled

i.e. using a baseline system, how many more mhz would be required to achieve those xtra points?

-edit-

700 points....

Since Cinebench scales almost perfectly with clockspeed on other processor architectures, it will probably take a 3.2GHz Barcelona to match the single-threaded performance of Yorkfield in 32-bit Cinebench R10 and 3GHz to match it at 4 threads.
10-28-2007, 12:00 AM
Mad_Man

FUGGER, to be a fair comparison, would you mind to run your bench again with only ONE stick of ram. or two in single channel for that matter?
and ddr2 at 400mhz 5-5-5 would equal to your 800mhz DDR3 at what? 9-9-9?

i would really like to see such a comparison
10-28-2007, 12:56 AM
mongoled

Quote:

Originally Posted by Mad_Man

FUGGER, to be a fair comparison, would you mind to run your bench again with only ONE stick of ram. or two in single channel for that matter?
and ddr2 at 400mhz 5-5-5 would equal to your 800mhz DDR3 at what? 9-9-9?

i would really like to see such a comparison

Looking forward to seeing this result

:D
10-28-2007, 12:56 AM
hipro5

Quote:

Originally Posted by Mad_Man

FUGGER, to be a fair comparison, would you mind to run your bench again with only ONE stick of ram. or two in single channel for that matter?
and ddr2 at 400mhz 5-5-5 would equal to your 800mhz DDR3 at what? 9-9-9?

i would really like to see such a comparison

We have to "match" on BOTH systems the Memory Bandwidth if we have to do it as you say....How much Everest Mem Bandwidth have the Barchies and how much the York?.....Latencies too.....Because due to internal Mem controler of the Barchie, the Mem Bandwidth is high......So to be more "accurate", BOTH Mem Bandwidths/Latencies MUST be "matched" somehow.....;)
10-28-2007, 12:59 AM
alayashu

a 'fair' comparison would be with a BA (wont exist afaik) or B2 Phenom:)
and i smell some Phenom benches at the horizon in the very near future:)
10-28-2007, 01:42 AM
Cronos

By the way , no application with 3.57 SMP speed up (like Cinebench R10) on Core2 Quad can depend on memory speed that much.
Memory-intensive apps. usually have 2-3 speedup on this processor.
10-28-2007, 03:04 AM
s7e9h3n

Quote:

Originally Posted by hipro5

We have to "match" on BOTH systems the Memory Bandwidth if we have to do it as you say....How much Everest Mem Bandwidth have the Barchies and how much the York?.....Latencies too.....Because due to internal Mem controler of the Barchie, the Mem Bandwidth is high......So to be more "accurate", BOTH Mem Bandwidths/Latencies MUST be "matched" somehow.....;)

Actually, the Yorky probably nets a much higher BW mark than the Barcelona in Everest. IIRC, I got something like ~5500 with the 2350's + 8gb DDR2 667, 5-5-5-x. Phenom will do much better since it's able to run regular DDR2 and @ a higher speed

Quote:

Originally Posted by alayashu

a 'fair' comparison would be with a BA (wont exist afaik) or B2 Phenom:)
and i smell some Phenom benches at the horizon in the very near future:)

BA's WON'T do any better in these benches than the B1's.
10-28-2007, 03:04 AM
cpuz

Quote:

Originally Posted by kyosen

Memory clock is showed differently between CPU-Z and Everest...
I have no idea which is correct or not...

Hi Kyosen,
I got your PM, we'll work on that.
Everest is assuming that memory clock is obtained from bus clock on K10, and CPU-Z computes it from CPU clock (as for K8).
IMO Everest is correct, but Tamas & I had no confirmation at the moment. We're expecting a benchmark can tell what the real clock is.

Nice rig, I'm impatient to see results on the new BA stepping :up:
10-28-2007, 03:53 AM
hipro5

Heh, heh, heh.......I was JUST informed that after the Phenom presentation in Poland, a mobo and a Phenom CPU will come into my place for testing..... :D :p:

EDIT: Of course I won't touch it with my own hands.......I'll bring esdee to do it.....He's a "painted" AMD guy.... :p:
10-28-2007, 07:04 AM
BeardyMan

Quote:

Originally Posted by hipro5

Heh, heh, heh.......I was JUST informed that after the Phenom presentation in Poland, a mobo and a Phenom CPU will come into my place for testing..... :D :p:

EDIT: Of course I won't touch it with my own hands.......I'll bring esdee to do it.....He's a "painted" AMD guy.... :p:

Overclocking is overclocking Hipro :D
you should know that ;)
10-28-2007, 07:05 AM
hipro5

Quote:

Originally Posted by BeardyMan

Overclocking is overclocking Hipro :D
you should know that ;)

Yes but I know that "AMD is AMD".....and I don't touch such things...... :D :p:
10-28-2007, 07:06 AM
mongoled

Quote:

Originally Posted by hipro5

Heh, heh, heh.......I was JUST informed that after the Phenom presentation in Poland, a mobo and a Phenom CPU will come into my place for testing..... :D :p:

EDIT: Of course I won't touch it with my own hands.......I'll bring esdee to do it.....He's a "painted" AMD guy.... :p:

Oh man I hope you are not kidding (se parakalo pes mou tiv alitheiaaaaaaa ):D

Any timeline on this, this is going to be good, ofcourse you are going to help our friend esdee ;)

:clap:
10-28-2007, 07:09 AM
BeardyMan

Quote:

Originally Posted by hipro5

Yes but I know that "AMD is AMD".....and I don't touch such things...... :D :p:

Oh if that's teh case, then you can send me you amd X2900Xt's :D

Live with it you are overclocking AMD...
do you feel it?

AMD :yepp: :fact: :poke: :buddies: :slobber: :wasntme:
10-28-2007, 07:17 AM
SOLDNER-MOFO64

Quote:

Originally Posted by hipro5

Yes but I know that "AMD is AMD".....and I don't touch such things...... :D :p:

:sick:

nevermind hipro, all good things must come to an end :(

soon you too will kneel and worship the true master :rofl:

What with your sig and all....free gifts already ;)

:welcome: to AMD :up:
10-28-2007, 07:40 AM
tictac

Quote:

Originally Posted by accord99

Since Cinebench scales almost perfectly with clockspeed on other processor architectures, it will probably take a 3.2GHz Barcelona to match the single-threaded performance of Yorkfield in 32-bit Cinebench R10 and 3GHz to match it at 4 threads.

yeah... probably 3.2Ghz barcelona to match up with 2.4Ghz Penryn DDR3... :eek:
10-28-2007, 07:43 AM
BeardyMan

Quote:

Originally Posted by tictac

yeah... probably 3.2Ghz barcelona to match up with 2.4Ghz Penryn DDR3... :eek:

why does K10 outform the penryn at speedup?
serious question don't what it means :shrug:

3.87 vs 3.55, while fugger is running high speed ddr3?
10-28-2007, 07:44 AM
tictac

Quote:

Originally Posted by BeardyMan

why does K10 outform the penryn at speedup?
serious question don't what it means :shrug:

3.87 vs 3.55, while fugger is running high speed ddr3?

native quadcore... :shrug:
10-28-2007, 08:32 AM
Viper666

:)

Quote:

Originally Posted by SOLDNER-MOFO64

:sick:

nevermind hipro, all good things must come to an end :(

soon you too will kneel and worship the true master :rofl:

What with your sig and all....free gifts already ;)

:welcome: to AMD :up:

Well I dont doute what you say at all based on understanding ,how this processor works,Its all about that memory controller speed removing latency from the L3 cache.I also think the Latency is there because the expected clock speed is very high.
10-28-2007, 08:51 AM
QuietIce

Quote:

Originally Posted by hipro5

Yes but I know that "AMD is AMD".....and I don't touch such things...... :D :p:

Ya', might get friction burns instead of heat burns ...! :rofl:

:p:
10-28-2007, 08:53 AM
Lightman

Quote:

Originally Posted by hipro5

Heh, heh, heh.......I was JUST informed that after the Phenom presentation in Poland, a mobo and a Phenom CPU will come into my place for testing..... :D :p:

EDIT: Of course I won't touch it with my own hands.......I'll bring esdee to do it.....He's a "painted" AMD guy.... :p:

Do you know who in Poland will test it?? Where the presentation will be held?? And most importantly WHEN? :p:
10-28-2007, 09:22 AM
BrowncoatGR

Quote:

Originally Posted by hipro5

Heh, heh, heh.......I was JUST informed that after the Phenom presentation in Poland, a mobo and a Phenom CPU will come into my place for testing..... :D :p:

EDIT: Of course I won't touch it with my own hands.......I'll bring esdee to do it.....He's a "painted" AMD guy.... :p:

Greek Power :up:
10-28-2007, 11:43 AM
Unoid

tictac you still making modified BIOS's? Or do I got the wrong guy.

I'm still rocking the DFI Ultra infinity NF2 mobo :)
10-28-2007, 02:22 PM
Cronos

Quote:

Originally Posted by BeardyMan

why does K10 outform the penryn at speedup?
serious question don't what it means :shrug:

3.87 vs 3.55, while fugger is running high speed ddr3?

Core2 shared L2 cache, mostly.
10-28-2007, 06:24 PM
KTE

Quote:

Originally Posted by s7e9h3n

Actually, the Yorky probably nets a much higher BW mark than the Barcelona in Everest. IIRC, I got something like ~5500 with the 2350's + 8gb DDR2 667, 5-5-5-x. Phenom will do much better since it's able to run regular DDR2 and @ a higher speed

AFAI tested Cinebench 10 scales very well with CPU and RAM speed, gaining as much as 40 CB marks from just 40MHz (DDR 80MHz) (all things kept constant).

Intel/DDR2 all the way up to 1000MHz doesn't get much in terms of benchmark bandwidth/latency compared to AMD K8/DDR/DDR2 but the new combo of Intel/DDR3 quite obviously gets higher bandwidth as you start going past DDR3-1333 at good timings. Charles' system was running 7-7-7-15 @ 1600MHz which is very close to your Barcelona timings but twice the RAM speed, and twice your bandwidth. His system was 693 CB ahead, which is within reach of high MHz/low latency RAM if the bandwidth is high.
10-28-2007, 08:12 PM
kyosen

Update: SuperPI1M&4M

Sorry for my slow working, and I'm happy to see Intel Guys also enjoying in this thread:D

I have three types of CPUs
K8 Opteron 2212HE = 2.0GHz = 200x10
K10 Opteron 2350(B1) = 2.0GHz = 200x10
K10 Opteron 2346(BA) = 1.8GHz = 200x9
We have no way to change multiplier of K10 so far,
I tested K8 Opteron at x9 with CrystalCPUID.
# KFSN4-DRE's BIOS has no multiplier option...
# but now Franck(as cpuz here) is working for K10 functions:up:

I used 1GB x2 memory modules.
They are Dual mode with K8 and Unganged mode with K10.
Memory timing was basically Auto, but I adjusted it in K8 case as same as possible in K10.

OS: Windows Server 2003, no-use services are set as manual starting.
MAXMEM, RAM disk, and copy-waza weren't applied.

SuperPI 1M comparison...screenshots are located on my BBS:
http://www.oohashi.jp/c-board/c-boar...ne;no=5229;id=

K8-1.8G=200x9: 46.468s
K10(BA)-1.8G=200x9: 43.172s -> 46.468/43.172=1.0763

K8-2.0G=222x9: 41.968s
K10(BA)-2.0G=222x9: 38.781s -> 41.968/38.781=1.0821

K8-2.0G=200x10: 41.828s
K10(B1)-2.0G=200x10: 39.125s -> 41.828/39.125=1.0690

K8-2.2G=220x10: 37.953s
K10(B1)-2.2G=220x10: 35.578s -> 37.953/35.578=1.0667

K8-2.4G=240x10: 34.734s
K10(B1)-2.4G=240x10: 32.484s -> 34.734/32.484=1.0692

SuperPI 4M comparison...screenshots are located on my BBS:
http://www.oohashi.jp/c-board/c-boar...ne;no=5230;id=

K8-1.8G=200x9: 3m59.359s
K10(BA)-1.8G=200x9: 3m47.125s -> 239.359/227.125=1.0538

K8-2.0G=222x9: 3m34.547s
K10(BA)-2.0G=222x9: 3m24.250s -> 214.547/204.250=1.0504

K8-2.0G=200x10: 3m35.047s
K10(B1)-2.0G=200x10: 3m26.953s -> 215.047/206.953=1.0391

K8-2.2G=220x10: 3m15.141s
K10(B1)-2.2G=220x10: 3m07.797s -> 195.141/187.797=1.0391

K8-2.4G=240x10: 2m59.297s
K10(B1)-2.4G=240x10: 2m51.906s -> 179.297/171.906=1.0429

Quick conclusion from above:
* There is no strange scaling behaviour within same CPU type.
* L3 contributes for shorter 1M calculation, and its effect reduces for longer 4M calculation.
* K10(BA) looks slightly...very slightly...faster than K10(B1).
Here, we need attention for NorthBridge multiplier...
K10 2346's NB is working at x8, and x9 in 2350's case.
So, if compared at same core&NB multiplier, BA's advance will increase a little more, I think.

Yeah K10 is the improved K8, indeed, so far, at least for SuperPI.
I wonder whether inmature BIOS may prevent the real K10 power or not,
but only AMD (and SOLDNER-MOFO64 too!?) knows:)

I'll try CineBench10 in 32bit & 64bit Windows...or 3DMarks?... tonight.
10-28-2007, 09:05 PM
Viper666

Hum

I was just looking in the SUSE 10.3 Hardware Info tool,under the processor,it shows prefetch for 3dnow(3dnowprefetch) but I dont see anything about SSE prefetch.ANyone know why?
10-28-2007, 10:49 PM
KTE

Very nice work Kyosen, thank you. :)

Can you compare one run of Barcelona at X CPU speed and run it twice at two different memory speeds (like 667 and 800) in Cinebench 10? Is that possible?

I'd like to see what effect RAM has on the score to see what we can expect from 1066 CAS4 RAM equipped Phenoms.

So;
@1.8G Increase between K8:K10(BA)-> 46.468/43.172=1.0763 - 7.6%
@2.0G Increase between K8:K10(BA)-> 41.968/38.781=1.0821 - 8.2%
@2.0G Increase between K8:K10(B1)-> 41.828/39.125=1.0690 - 6.9%
@2.2G Increase between K8:K10(B1)-> 37.953/35.578=1.0667 - 6.7%
@2.4G Increase between K8:K10(B1)-> 34.734/32.484=1.0692 - 6.9%

Thus including therein the RAM/NB speed gain;
K8 from 1.8GHz -> 2.4GHz = 33.33% Clock Increase - 33.78% Performance Increase
K10 (B1/BA) from 1.8GHz -> 2.4GHz = 33.33% Clock Increase - 32.90% Performance Increase

This is like Penryn over Core 2 so far although I'm not sure how the Intel chips scale on SPi.
10-28-2007, 11:35 PM
savantu

Quote:

Originally Posted by tictac

native quadcore... :shrug:

Huh ? IMC and HT are directly responsible for scaling.
K8 scales better than Core too..
10-29-2007, 04:36 AM
BeardyMan

3dmark06 would be good :D
10-29-2007, 04:44 AM
kyosen

Quote:

Originally Posted by BeardyMan

3dmark06 would be good :D

Now preparing...with 8800GT;)
10-29-2007, 05:06 AM
sponge bob

Quote:

Originally Posted by kyosen

Now preparing...with 8800GT;)

cant wait :d going to OC it 2,4 ghz + ?
10-29-2007, 11:01 AM
BeardyMan

Quote:

Originally Posted by kyosen

Now preparing...with 8800GT;)

Nice one Kyson , and somehow i feel you hava a suprise for us by running phenom at 3.6ghz or so :D
10-29-2007, 11:29 AM
gondezz

Excuse me for this n00b question:

If memory clock is derived from cpu clock divided by some number,
so what is northbridge clock for??

Thanks
10-29-2007, 11:38 AM
v0dka

I think it's not really a NB, this is just the IMC speed. On K8 it was always 1:1 with the CPU, apparantly now it's seperate.

This seems interesting to me as on K8 the IMC was suspected to play a major role in the max CPU clockspeed. A typical good K8 had mem revision B1 or BB as you will know, which would get you higher than a good week BW.
10-29-2007, 11:46 AM
gondezz

Thanks for reply..
So, IMC speed is 'internal speed' only?
To communicate with the core perhaps?
10-29-2007, 12:01 PM
kyosen

Update: 3Dmark06 with 8800GT

I could get 8800GT yesterday:)
http://222.151.153.254/c-board/file/...E_GF8800GT.jpg
...yes, I can't do 2 CPU operation with long PCIe board:D

I used 169.01 driver with inf file modified...it looks a bit faster than 167.26.
I tried 2.0G, 2.2G, 2.4G...and in 2.4G case, I tried 3core(1core disabled) and
2core(2core disabled) as emulation of Toliman and Kuma:)
I also tried on P5K3-D/Kentsfield with mild memory setting...

Results...screenshots are located on my BBS:
http://www.oohashi.jp/c-board/c-boar...ne;no=5231;id=

CPU clock: 3DMark Score / CPU Score

K10(BA) 2.0G=222x9: 9482 / 2999
K10(B1) 2.0G=200x10: 9353 / 2979
K10(B1) 2.2G=220x10: 10148 / 3267
K10(B1) 2.4G=240x10: 10796 / 3560
K10(B1, 3core) 2.4G=240x10: 10420 / 2830
K10(B1, 2core) 2.4G=240x10; 9423 / 1888

Kentsfield 2.4G=267x9: 11901 / 3845
Kentsfield 3.0G=334x9: 12965 / 4792

Quick conclusion from above:
*K10(BA) looks a bit faster than K10(B1), though multiplier aren't same.
*K10 score seems about 10% lower than Kentsfield at same clock.

So, K10 needs more clock!!!

I'll try CineBench10 in next time...
...now very sleeply...it's early morning here in Japan:)
10-29-2007, 12:01 PM
Calmatory

Hmm, so IF the ICM was the limiting factor when overclocking K8, now we can divide ICM speed with K10, will that mean that we can gain intel-like overclocks (+4.5GHz)? :O
10-29-2007, 12:19 PM
v0dka

Thanks Kyosen.

Too bad it lags 10% behind Kentsfield now. Maybe Phenom will close the gap to Kentsfield at least.

Quote:

Originally Posted by gondezz

Thanks for reply..
So, IMC speed is 'internal speed' only?
To communicate with the core perhaps?

Communication is done through the HT bus, which has a seperate speed. The Integrated Memory Controller communicates with the memory (read / write). I invite anyone with more technical background to fill in the details. :D

Quote:

Originally Posted by Calmatory

Hmm, so IF the ICM was the limiting factor when overclocking K8, now we can divide ICM speed with K10, will that mean that we can gain intel-like overclocks (+4.5GHz)? :O

An interesting theory is it not? But ofcourse there are other factors.
10-29-2007, 12:26 PM
Sparky

Quote:

Originally Posted by v0dka

Thanks Kyosen.

Too bad it lags 10% behind Kentsfield now. Maybe Phenom will close the gap to Kentsfield at least.

According to Perkam, B2 (which IIRC is what Phenom is going to be) is supposed to be much better. At this point I don't know what to believe, though I do hope he is right.
10-29-2007, 12:36 PM
grunge100

beardy,

How do you know the barc. proc. was in single channel mode?
10-29-2007, 01:15 PM
BeardyMan

Quote:

Originally Posted by MR_SmartAss

It is just BeardyMan having too much hopes about K10.
It was said 1000 times that CPUz displays it as single channel, while it is DUAL channel. It was also confirmed that the system is running in dual channel, comparing the scores of Super Pi with 1 and 2 memory modules. There was a response from Franck of CPUz too, in this thread.

Don't worry, at the end you will come and kneel for me :D
10-29-2007, 01:17 PM
grunge100

smartypants,

K10 is fine, I think its you that has a problem.:D
10-29-2007, 01:28 PM
BeardyMan

Quote:

Originally Posted by MR_SmartAss

Grunge100
OK, go ahead. What problem do you think I have?

have?:D
are:D
10-29-2007, 01:34 PM
LowRun

Quote:

Originally Posted by BeardyMan

Well i noticed Kyson's image says single channel on the AMD rig, perhaps someone chould run single and dual channel on his intel rig and se eteh performance drop...

Dude it's Kyosen not Kyson.
10-29-2007, 02:01 PM
BeardyMan

K10(B1) 2.4G=240x10: 10796 / 3560
Kentsfield 2.4G=267x9: 11901 / 3845

It would be interesting to see 267*9 on the AMD system or 240*10 on the intel and then to see how that cpu bench scales ;)
10-29-2007, 02:13 PM
knightwolf654

ok i am gonna come out and say it...
SHUT UP
no need for petty arguments.
now mabey i dident see it but is the bored ur running a 1207+? so it can utilize HT3? just curious.
10-29-2007, 02:17 PM
cky2k6

yes it would... hard one to predict *concentrating very very hard channelling my clairvoyant powers* oh my god, i got it! no the fsb wouldnt have any damn effect, when the hell has it ever??
10-29-2007, 02:20 PM
knightwolf654

i wasent talking the FSB, HT3 is on the die correct? and it would provide direct links to the ram and what not. instead of going through the FSB
10-29-2007, 02:23 PM
akaBruno

FSB x the multplier matters. MOBO fsb? Don't mean sh*t.

Things change everyday, though. :cool:
10-29-2007, 02:28 PM
mrcmtl

Quote:

Originally Posted by kyosen

FELIX kindly pointed out to me about K10's memory controller:)
According to AMD's document, BIOS and Kernel　Developer’s Guide,
K10's has 2 DCTs(DCT: Dram ConTroller!?), and it can work different two mode, i.e.,
(1) to behave as a single dual-channel DCT; this is called ganged mode;
(2) to behave as two single-channel DCTs; this is called unganged mode.

My understanding is:
Ganged mode reads or writes with 2 DCTs at the same time.
Unganged mode also can reads with 1 DCT and writes with another 1 DCT at the same time,
if the system requests as so.
Unganged mode is more flexible than ganged mode,
and ganged mode is equal to "conventional" dual channel mode.

)

I dont know how Ganged mode works...if it is a single read/write or two but Unganged mode is two different DCT, there part isn't shared...they both write or read...and just to add, K10 doesnt read/write anymore...it reads until the DCT buffer is out of space and bursts all the awaiting writes to avoid the constant read/write switch delay.
10-29-2007, 05:51 PM
oldblue

I'm curious about how the relative clock speed between the L3 and the cores affects the L3 cache latency. If I read this right, some combinations should give lower latency than others.

To support independent clocking and modular design, asynchronous dynamic FIFO buffers are used to communicate between different cores and the northbridge/L3 cache. These FIFOs absorb any global skew or clock rate variation, but the latency for passing through depends on the skew and frequency variance – which is why the L3 cache latency is variable.

Show 100 post(s) from this thread on one page

All times are GMT -8. The time now is 02:04 AM.

XtremeSystems