AMD 4x4-platform based on Socket 1207 / AMD roadmap ? [Archive]

turtle

09-22-2006, 01:26 AM

According to a source on our forum with access to roadmaps of AMD's 4x4-platform - also called ' Quad-Father ' - it is on Socket F. There are three new processors to be brought out especially for the platform; the FX-70 (2,6GHz), FX-72 (2,8GHz) and FX-74 (3,0GHz). These models will be sold, according to reports, only in sets of two. New FX-series are based on 90nm Windsor-core with 125W TDP and 2x1MB L2-cache. The marketing name ' FX ' will be no longer used in the future for AM2-processors, but that does not mean that faster chips will not come for the desktop: FX-74, for example, will also be brought out for Socket AM2, only under the name Athlon X2 6000+. Furthermore there is news concerning the first 65nm-chips of the company, which will appear towards the end of this year in stores. They all have 65W TDP but the maximum speed is higher than AMD indicated initially. The fastest chip based on 65nm Brisbane, as it happens, is not 2.4GHz but 2.6GHz. Still not as fast as 90nm cores, but nevertheless a good sign. Also, concerning cache quantity for 65nm-chips: instead of 3800 +.,4200 + and 4600+ (2x512kb) parts, 4000+,4400+ and 4800+ (2x1MB) chips are now expected. These chips are just as fast but have twice as much cache. Below is the complete overview of chips which AMD wants to bring out this year:

http://img240.imageshack.us/img240/2383/amdroadmapbj8.jpg
http://tweakers.net/nieuws/44504/AMD-4x4-platform-gebaseerd-op-Socket-1207.html

Please edit the title with a question mark...It's obviously still a rumor. If it's true it makes sense, but is very disappointing imho.

gOJDO

09-22-2006, 02:37 AM

if true, that means no cheap X2 for 4x4 like AMD fanboys were deraming and trolling about.

kemo

09-22-2006, 03:03 AM

ANY news about price ???
________
vaporizer (http://vaporizers.net/vaporizers)

Vapor

09-22-2006, 03:50 AM

'Voor kerst' = ??

And wow...what a bummer. One market they really could have taken from Intel is the budget-minded 'megatasker'. Now there's 250W of heatdump and it still might not give Intel's high-end market much of a challenge :( And it'll likely cost more, and it'll be harder to cool (at least Intel's two dies will be under the same IHS), and it definitely won't clock as well.

I really hope this is a rumor, a cheap 4-core system would have been very nice....

turtle

09-22-2006, 04:11 AM

LowRun

09-22-2006, 04:19 AM

Bad bad move if true, what a letdown :(

L&MFatty

09-22-2006, 04:51 AM

I'm more concerned about the prices ...

[XC] Teroedni

09-22-2006, 04:59 AM

http://img240.imageshack.us/img240/2383/amdroadmapbj8.jpg
http://tweakers.net/nieuws/44504/AMD-4x4-platform-gebaseerd-op-Socket-1207.html

Please edit the title with a question mark...It's obviously still a rumor. If it's true it makes sense, but is very disappointing imho.

Damn im so looking forward to the release of this.:D
And the fact that this system goes all the way down to x2 4400 means it will probaly be cheaper to build than a 1207 rig.

Now consider having 2x x2 4400 and ddr800 NUMA:woot:

Edit: Disregard this
I skimped without reading the text:slap:

The Ghost

09-22-2006, 05:10 AM

so is this suppose to use registered ram ?

there is a tyan mother board already that is set up just like amd said that 4x4 motherboard would be like , but it uses server cpu's

time will tell , maybe we will find more sources to agree or disagree

Vassili

09-22-2006, 05:19 AM

Damn im so looking forward to the release of this.:D
And the fact that this system goes all the way down to x2 4400 means it will probaly be cheaper to build than a 1207 rig.

Now consider having 2 x x2 4400 and ddr800 NUMA:woot:
LOL reading is hard isn't it.

Mr. Popo

09-22-2006, 05:28 AM

4 x 4 = 1207.
FX-70, FX-72 and FX-74 are 1207.

... I think.

[XC] Teroedni

09-22-2006, 05:30 AM

LOL reading is hard isn't it.
Damn I got it all wrong :(

So only socket 1207 is gonna be 4x4:stick: :mad:

turtle

09-22-2006, 05:39 AM

4 x 4 = 1207.
FX-70, FX-72 and FX-74 are 1207.

... I think.

That's what it looks like, just those three.

Well, and of course the spring refresh...Which could be what saves it from falling into the terrible abyss of bad ideas to market to the mainstream.

Mikesta

09-22-2006, 05:39 AM

I'm really starting to wonder whats going to happen with
AMD's sockets over the next year or so.....

Lightman

09-22-2006, 05:52 AM

;) I remember that AMD promised sets of 4x4 CPU's for under $1000, that means FX-74 for around $500 each:D pretty cool, but I will wait for K8L based FX for 4x4

gOJDO

09-22-2006, 06:04 AM

I dunno bout this, but even before it is confirmed I see Intel fanboys trolling about it... :stick:
Does it mean that the "intel fanboys trolls" told you the truth?
Is this the best of you LOE? :clap:

turtle

09-22-2006, 06:13 AM

Does it mean that the "intel fanboys trolls" told you the truth?
Is this the best of you LOE? :clap:

http://www.ctgilles.net/images/pictars/chill.jpg

alayashu

09-22-2006, 06:20 AM

there are allready "4x4"-like motherboards for socketF and 940... 4x4 will be AM2...

this is just a rumor, we'll have to wait and see.

[XC] leviathan18

09-22-2006, 06:23 AM

i think there is a rule that says you cant call other fanboys this is really getting out of control from certain guys here

Dimitriman

09-22-2006, 06:48 AM

I wouldn't be too concerned about the prices of these new FX.
Currently the X2 5000+, a 2.6Ghz 512Kb dual core is going for around 280 Euros/Dollars,
I serioulsy doubt AMD would be stupid enough to charge anything above 400 Euros/Dollars for the lower 2.6Ghz 1mb FX70 version.
They will most likely release the dual package for about 699 dollars which
would be considerably good for a quad core system, especially if thats less
than the Kentsfield when it comes out.

Sparky

09-22-2006, 06:52 AM

I don't quite understand it but that is ok. I'm holding out for K8L, I don't have any money anyway :p:

vitaminc

09-22-2006, 06:56 AM

there are allready "4x4"-like motherboards for socketF and 940... 4x4 will be AM2...

this is just a rumor, we'll have to wait and see.

I hope so too. It's a major letdown if 4x4 is on socket F. Pretty pointless to rebrand 2-way server technology as 4x4.

gOJDO

09-22-2006, 07:14 AM

http://www.ctgilles.net/images/pictars/chill.jpg
Damn :) you want me overdosed.

turtle

09-22-2006, 07:16 AM

Which brings back the conspiracy theory...

What if your new desktop architecture couldn't beat the competition and you planned on refreshing it shortly anyway with a new memory controller...How would you release it?

Personally, if I had a separate socket/platform for just servers/enthusiasts using at least dual socket-capable chips I'd keep it out of the single-socket limelight and push it as an at least dual cpu solution to consumers first...because I think (barring something absurd) 2xK8L on desktop or 4x/8x K8L on servers is gonna beat anything Intel has....and AM2 is still only 1-way last time I checked, and Socket F is at the LEAST 2-way.

@
Damn :) you want me overdosed.

Naw, only love amigo. You can never have too much chill :)

alayashu

09-22-2006, 07:32 AM

Which brings back the conspiracy theory...

What if your new desktop architecture couldn't beat the competition and you planned on refreshing it shortly anyway with a new memory controller...How would you release it?

Personally, if I had a separate socket/platform for just servers/enthusiasts using at least dual socket-capable chips I'd keep it out of the single-socket limelight and push it as an at least dual cpu solution to consumers first...because I think (barring something absurd) 2xK8L on desktop or 4x/8x K8L on servers is gonna beat anything Intel has....and AM2 is still only 1-way last time I checked, and Socket F is at the LEAST 2-way.

replace K8L in your post with K8 and it's all ok ;)

LexDiamonds

09-22-2006, 07:35 AM

[XC] Teroedni

09-22-2006, 07:47 AM

Now that i actually have read the text

So in October comes Stronger Sempron
Sempron 3800+ 2,2GHz 256KB 90nm 62W AM2 Oktober
Sempron 3500+ 2,0GHz 128KB 90nm 35W AM2 Oktober

Then In November Comes stronger X2
X2 6000+ 3,0GHz 2x1MB 90nm 125W AM2 November
X2 5600+ 2,8GHz 2x1MB 90nm 98W AM2 November
X2 5400+ 2,8GHz 2x512KB 90nm 98W AM2 November

Then in December comes 65 nm x2 (Cheaper to produce):D
X2 5000+ 2,6GHz 2x512KB 65nm 65W AM2 December
X2 4800+ 2,4GHz 2x1MB 65nm 65W AM2 December
X2 4400+ 2,2GHz 2x1MB 65nm 65W AM2 December
X2 4000+ 2,0GHz 2x1MB 65nm 65W AM2 December

Now this comes at the End on the Year when Intel is ramping up with Conroes
See a connection;)

AMD knows very well that Core dou 2 is a treat and are reoginasing their lineup to better compete.

No the big question is
Will we see Price Cuts?

I have a feeling we will:toast:

The Ghost

09-22-2006, 07:54 AM

The most dissapointing piece of information in this is the obvious truth that AMD is having problems scaling their 65nm chips.
i do not know if this is true , the best that i can tell is that this has been disproven

why is amd still using 90nm for it's fx cpu's , well i don't think that amd sells a whole lot of fx cpu's , and is probably making the fx chips in fab 30 , when fab 30 is converted to fab 38 , the fx cpu's will probably be 65nm

what do we think of amd being ableto take a 90nm cpu to the mhz that they are taking it to , this has to say something for amd and 90nm

Nedjo

09-22-2006, 08:01 AM

The most dissapointing piece of information in this is the obvious truth that AMD is having problems scaling their 65nm chips. If they were having good luck with speeds/yields it would make sense that they would move all their 2x1mb cache parts to 65nm asap.
Your conclusion is wrong!
You donâ€™t want to use volume production for parts that will sell in low quantities.
Beside that AMD has sharpened 90nm â€˜till the very edge of the tech, so itâ€™s natural that theyâ€™re using 90nm tech for the new FX CPUâ€™s. Those are really Opterons 2xxx series, and its common knowledge that Opterons are always produced on the highest quality wafers
In the long run, 4x4 on Socket F is really a way to go!

And of course it makes me smug seeing I was right long time ago ;)

http://www.xtremesystems.org/forums/showthread.php?t=111759&highlight=4x4+motherboards

I guess this will be the prime-time 4x4 mobo:

http://www.tyan.com/products/html/thundern3600b.html

uOpt

09-22-2006, 08:07 AM

I don't believe it.

The FX series is for enthusiasts, the major feature being the unlocked multiplier - which means overclocking friendly.

It doesn't make sense to put the enthusiast platform on registered RAM. There is no "overclocking" RAM available for it and it is more expensive, and right now all registered RAM has ECC which further limits speed scaling and increases cost.

For the enthusiast there is no benefit in registered RAM as they rarely use more than 2 or at most 4 sticks of memory, and they won't use ECC either.

The point about the original 4x4 design was to give people a dual-socket solution that they can use with "enthusiast"/overclocking memory.

Also, who's going to design an overclocking board for dual socket 1207? The board will all be server-style designs.

The above plan basically means "there is no 4x4". It goes back to what we had in first place - more than one CPU socket means moving to AMD's server platform, including the need for registered RAM.

Nedjo

09-22-2006, 08:19 AM

It's not a question of yours belives! That's AMD's plan, and me thinks it's a good one! They wnt to sell CPU's so they will give you two FX7x CPU's for the price of the one FX6x!

el rolio

09-22-2006, 08:42 AM

uOpt

09-22-2006, 09:35 AM

actually i got the impression that the 4x4 high end enthusiast platform was gonna not require registered ram at all.

Of course.

But socket 1207 does.

So that doesn't fit, something odd is going on here. Either the original report is wrong, or AMD effectively drops 4x4 (going back to "need server platform for > 1 CPU socket).

so what you will be seeing is asus, msi, dfi mobos, with 2 sockets, slots for your fav flavor of ram and then you buy a pair of processors initially and plop it into the LGA sockets. then clock the crap outta it and have an awesome system.

That was the original plan, unregistered RAM, 4x4 using socket AM2.

arisythila

09-22-2006, 10:05 AM

I wouldnt believe this stuff.. LOL The person that made that chart didnt even know how to spell Clock. They spelled it "Klok". AMD wouldnt do us like this. Atlease I hope not. Using ECC memory for thier ELITE line isnt really ELITE. I think this would be a bad mistake for them. I know they are smarter than this. I would be puzzled why they would bring a FX line to Socket F. Doesnt make sense to me. Which is why Im saying this is bunk.

~Mike

biohead

09-22-2006, 10:40 AM

I wouldnt believe this stuff.. LOL The person that made that chart didnt even know how to spell Clock. They spelled it "Klok".
How ignorant are you, it's dutch :stick:

nn_step

09-22-2006, 11:21 AM

Fud

afireinside

09-22-2006, 11:28 AM

So now 4x4 uses ECC, is on 90nm still, and costs more than most people will pay. Also they sell 90nm CPUs clocked higher than the 65nm? RIIIIIGHT

el rolio

09-22-2006, 11:58 AM

Of course.

But socket 1207 does.

So that doesn't fit, something odd is going on here. Either the original report is wrong, or AMD effectively drops 4x4 (going back to "need server platform for > 1 CPU socket).

That was the original plan, unregistered RAM, 4x4 using socket AM2.

ok, so just to make sure im not missing hte boat in my impression here:

is there anything about the socket and mobo that determines the ECC vs nonECC ram thing? i assumed that its built in as part of the ram controller on the cpu DIE. so along that lines, i assumed the 4x4 would be as you said in your orignal plan, save for the fact that it starts on the socket F, presumably cuz of hte extra HT connects blah blah blah.
so what we essentially get is:
budget line (sempron?)
consumer line (x2)
high end enthusiast line (4x4)
server line (opteron)

all that goes out the window if im wrong about the ECC requirement being based in the mem controller on die. if not, all of the info so far makes sense to me.

arisythila

09-22-2006, 12:07 PM

How ignorant are you, it's dutch :stick:

geen faecalien

~Mike

uOpt

09-22-2006, 01:19 PM

So now 4x4 uses ECC, is on 90nm still, and costs more than most people will pay. Also they sell 90nm CPUs clocked higher than the 65nm? RIIIIIGHT

Not ECC RAM but registered RAM. Big difference.

uOpt

09-22-2006, 01:24 PM

ok, so just to make sure im not missing hte boat in my impression here:

is there anything about the socket and mobo that determines the ECC vs nonECC ram thing? i assumed that its built in as part of the ram controller on the cpu DIE. so along that lines, i assumed the 4x4 would be as you said in your orignal plan, save for the fact that it starts on the socket F, presumably cuz of hte extra HT connects blah blah blah.
so what we essentially get is:
budget line (sempron?)
consumer line (x2)
high end enthusiast line (4x4)
server line (opteron)

all that goes out the window if im wrong about the ECC requirement being based in the mem controller on die. if not, all of the info so far makes sense to me.

You mix up ECC and registered/unbuffered. Different concepts. All AMD64s do optional ECC. Socket 754, 939 and AM2 always use unbuffered (unregistered). Socket 940 and 1207 always use registered.

I don't think it is possible to put a controller for unregistered RAM into a socket 1207. In any case, even if that's possible that would introduce a new processor line entirely incompatible with 1207 and all other sockets so that's not going to happen.

So it is still one of two:

Either 2+2 (4x4) is dead. And going 1207 means dead.
The report is wrong.

My money is on AMD not having found anybody to make a board for the original 2+2(4x4) concept with unbuffered RAM. There is just not enough money in that market, and it is a difficult board to make, and a difficult customer base (overclockers). Which mainboard maker would be so insane to do that when every overclocked kentsfield will be faster.

So I say 2+2(4x4) is dead and AMD just now markets it's server platform as "the new 4x4".

However, there is a chance we will get unlocked multipliers in 1207 FX chips, which would make me very happy.

el rolio

09-22-2006, 01:42 PM

ok, yea i meant registered, and ok thanks for breaking own your point of view and i definately see what you are saying now. so yea, i guess im with you there, either the report is wrong or 4x4 as we knew it is gone. im gonna hope for the possibility the report is wrong...of course.

biohead

09-22-2006, 01:59 PM

geen faecalien

~Mike
faecalien :confused:

SlicerSV

09-22-2006, 02:00 PM

why would the socket have anything to do with changing the ram controller?

i'm going to dismiss this entire thread and the original post as PURE AND STUPID SPECULATION, IF NOT SHEER FUD!

uOpt

09-22-2006, 02:06 PM

why would the socket have anything to do with changing the ram controller?

Since the memory controller is on the CPU, the socket in AMD systems is also fixed either registered or unbuffered. The additional pin in socket 940 over 939 was exactly for this.

The situation is different for Intel where the northbridge decides whether you go unbuffered or registered (or FB-DIMM) and the CPU socket could be the same.

Nedjo

09-22-2006, 02:06 PM

Fud
Well... not this time ;)

Fred_Pohl

09-22-2006, 06:53 PM

;) I remember that AMD promised sets of 4x4 CPU's for under $1000, that means FX-74 for around $500 each:D pretty cool, but I will wait for K8L based FX for 4x4

I think what AMD meant was a pair of FX70s for under $1000. :(

***Deimos***

09-22-2006, 11:01 PM

The most dissapointing piece of information in this is the obvious truth that AMD is having problems scaling their 65nm chips. If they were having good luck with speeds/yields it would make sense that they would move all their 2x1mb cache parts to 65nm asap.

They really need to be in the 3.6ghz range to successfully compete with existing and yet to be released conroes. This is a far cry from the 2.6ghz the 65 nanos are currently reaching (and yet to be released for a few more months). The only saving grace is that peak power consumption appears to be down on the new process.

For AMD, its K8L or bust. 4x4 is simply media hype to fight conroe.

I am a die hard AMD man but I cant for the life of me figure out how they let Intel sucker punch them for at least 12-18 months.

exisitng 90nm used for FX series, and used for highest clocked parts.
new 65nm used for volume and low power chips.

wait a minute,,,
hang on a second...
this sounds a tiny bit familiar.

its, EXACTLY what AMD did last time, with FX series made on 130nm and clocked higher, whereas 90nm winchesters used for volume 3000+, 3200+, etc parts. The pattern has repeated. Just as I said it probably would.

PS: if previous 130nm -> 90nm jump was any indication of power savings, then that 65W figure for the 2.6Ghz 65nm part is a cautious value. I look forwards to the return of P3 style 30W CPUs ;)

zakelwe

09-23-2006, 12:43 AM

The thing that puzzles me is why the midrange cpu's are going back to 1Mb cache ? I thought the whole idea for AMD was to reduce the cache size and therefore possibly make more money per cpu as the cache does not make too much difference ( a handful of percent ). With 65nm you'd think that they would still only have 512KB and make even more money.

So I am suspicious of that table. However that still begs why so little known at this point? I'm starting to think it is just a big marketing ploy. Hopefully I am wrong but if it is released and it is slower in most benchmarks than single Conroe then the Intel fanboys will be laughing their socks off and this will not be a safe place for the Amd people .. de de der :sofa:

I'm still hoping they will bring out a 3Ghz X2 on 939 for $100 so I can replace my ageing FX-55 ..but hope is fading :( sob sob.

Regards

Andy

nn_step

09-23-2006, 02:39 AM

Well... not this time ;)
Got undeniable proof?

The Ghost

09-23-2006, 03:54 AM

The thing that puzzles me is why the midrange cpu's are going back to 1Mb cache ? I thought the whole idea for AMD was to reduce the cache size and therefore possibly make more money per cpu as the cache does not make too much difference ( a handful of percent ). With 65nm you'd think that they would still only have 512KB and make even more money.

So I am suspicious of that table.
so am i , here is a link to a amd brisbane ES cpu

http://translate.google.com/translate?hl=en&sl=zh-CN&u=http://www.pcpop.com/doc/0/154/154075.shtml&sa=X&oi=translate&resnum=59&ct=result&prev=/search%3Fq%3Damd%2Bbrisbane%26num%3D100%26hl%3Den% 26lr%3D%26safe%3Doff%26rls%3Dcom.microsoft:en-US

Nedjo

09-23-2006, 05:31 AM

Got undeniable proof?
Yeah, but can’t publish it…

But use common sense! There is no way that mobo makes would make another AM2 line in their portfolio based on the niche market segment. With 4x4 Socket F mobos they are covering DP Opteron, and HighEnd Enthusiast market. That’s reasonable economics.

Second: imagine RMA risk with 1880 pins packed in the new PIB?! With SocketF FX’s AMD has near zero RMA risk.

Third: AMD is in the business of selling CPU’s, and If there is a chance for them to sell two CPU’s for 500-600 bucks, instead of one (for upgrading AM2 scenario), what do you think their strategy will be?

VulgarHandle

09-23-2006, 05:36 AM

except i remember a couple of articles in which many enthusiast mobo manuf. are quoted as "embracing the platform"....

arisythila

09-23-2006, 10:16 AM

I dont really care if they use a different CPU socket. I just hope I can use My GOOD DDR2 in it....

~Mike

cky2k6

09-23-2006, 10:30 AM

it makes sense for them to use socket f for the highend, once you also factor in the fact that torenza is coming... i actually think this is a very good idea, mobo makers dont really care what socket they solder on to the board, its not going to be a server board, just 2p enthusiast board. its really not that bad of a move, especially with the unified socket idea as well.

[cTx]Philosophy

09-23-2006, 05:12 PM

I just wanna see some performance numbers, to justify this conroe build, or kick myself in da bawls for jumping ship..

All this talk about architecture is just assumptions, unless theres bulet proof, then ill wait for facts..

arisythila

09-23-2006, 05:30 PM

Regardless of what you do you will be jumping ship... Go to Conroe now, 2 months later AMD has something better. Jump ship from Intel, and 2 months later they may have something better than 4x4. It sucks, but thats how this works.

~Mike

[cTx]Philosophy

09-23-2006, 05:37 PM

Well for me ive been AMD forever but Conroes performance was nothing to be left without...

Clovertowns benchies are ownage right now..

Hope AMD has their shi in shape for this one..

Fred_Pohl

09-23-2006, 07:46 PM

well IF amd OFFERS TWO FX70 chips AND a dual socket MOBO for 1000$ it is quite an offer

it will have no problems beating the s*it out of anything intel can offer in this price range

BUT I don't see any sense in having 4x4 s1207 - this is no 4x4 platform, it is a simple server board with server CPUs

I highly doubt that AMD is going to toss in a free 2x2 mobo with a pair of 2x2 FXs or that 2x2 will even come close to matching Kentsfield. The notion that 2x2 will beat the s*it out of Kentsfield is good for a laugh though. Thanks. :D

IIRC, AMD has said that 2x2 will use a 3rd HT link between the cpus. It seems likely that they need the extra pins for that. Not to mention that S1207 is physically superior to AM2.

***Deimos***

09-23-2006, 07:50 PM

Yeah, but canâ€™t publish itâ€¦

But use common sense! There is no way that mobo makes would make another AM2 line in their portfolio based on the niche market segment. With 4x4 Socket F mobos they are covering DP Opteron, and HighEnd Enthusiast market. Thatâ€™s reasonable economics.

Second: imagine RMA risk with 1880 pins packed in the new PIB?! With SocketF FXâ€™s AMD has near zero RMA risk.

Third: AMD is in the business of selling CPUâ€™s, and If there is a chance for them to sell two CPUâ€™s for 500-600 bucks, instead of one (for upgrading AM2 scenario), what do you think their strategy will be?

me thinks you are right.

[XC] hipno650

09-23-2006, 08:09 PM

[cTx]Philosophy

09-23-2006, 08:32 PM

i want to see how the performace increases on 4x4 cuz if two fx-70's are only slightly better than a single conroe x6800:nono: :nono: :nono: the conroe will will kill them when oced:nono: :nono: :nono: i sure hope amd gets this right!!!!!

Couldnt have said it better myself!!!

alayashu

09-24-2006, 02:45 AM

awdrifter

09-24-2006, 02:51 AM

nn_step

09-24-2006, 03:52 AM

Sorry for the noob question, but what's so bad about registered rams? Are they not as fast? Assuming the new 4x4 boards have oc options and we can run dividers on the rams, shouldn't we still be able to push the CPU to the max and just leaving the rams close to stock speed? But things really doesn't look good for AMD (at least for now). I hope this 4x4 will be cheaper than Intel's quad core cpu (Kentfield?), if not, I might have to switch and go Intel for my next build.
Registered Memory runs slower and has WORSE latency than unbuffered memory. Both of which seriously impact a latency sensitive AMD processor.
If you want MAX performance but don't expect MAX memory capacity.
Unbuffered is the way to go

The Ghost

09-24-2006, 04:20 AM

Are they not as fast?
you are correct , the fastest DDR2 registered memory is 667mhz , the uregistered memory is up and above the jedec standards , which is 800mhz

turtle

09-24-2006, 04:38 AM

If 4x4 is on 1207 and still requires buffered (registered) ram, what exactly is the point of trying to sell it over a dual-socket server board with 22xx Opterons which would be the exact same thing?

There's got to be something cooking on that front. Even if not, at least 1207 doesn't require ECC...I think a lot of people confuse the two.

Nedjo

09-24-2006, 10:33 AM

you are correct , the fastest DDR2 registered memory is 667mhz , the uregistered memory is up and above the jedec standards , which is 800mhz
Samsung has Registered DDR2-800:

http://www.samsung.com/products/semiconductor/DDR_DDR2/DDR2SDRAM/Module/RegisteredDIMM/M393T2950EZA/M393T2950EZA.htm

awdrifter

09-24-2006, 10:35 AM

Registered Memory runs slower and has WORSE latency than unbuffered memory. Both of which seriously impact a latency sensitive AMD processor.
If you want MAX performance but don't expect MAX memory capacity.
Unbuffered is the way to go
Then what's the advantage of registered ram? Why do servers use them? Thanks.

[XC] Teroedni

09-24-2006, 10:39 AM

Samsung has Registered DDR2-800:

http://www.samsung.com/products/semiconductor/DDR_DDR2/DDR2SDRAM/Module/RegisteredDIMM/M393T2950EZA/M393T2950EZA.htm

Yea but what about overclocking?

nn_step

09-24-2006, 12:02 PM

Then what's the advantage of registered ram? Why do servers use them? Thanks.
Registered memory is a specific form of memory designed to support a larger number of chips on a given memory module. To allow for all the memory chips on the module to properly function, it is necessary for the module to add a delay to all command requests to the memory. The delay is actually generated by an additional clock generator that resides on the module instead of relying on the motherboard. This will decrease the overall performance of a given memory module. The memory type is typically used in large server environments where large amounts of RAM are required. In these environments, the benefits of the increase in RAM outweigh the performance impacts caused by memory swapping.

Yea but what about overclocking?

:rofl: :lol: :rofl: :imo:

[XC] Teroedni

09-24-2006, 12:20 PM

Okey
It seems it is actually possible to overclock 1207 server boards if their bios supports it.
So a 4x4 certified board should be able to overclock well then:D

I guess Asus can just rebrand a server board like the Asus KFN32-D SLI and just add some bios options and voila 4x4;)

cky2k6

09-24-2006, 12:37 PM

Registered memory is a specific form of memory designed to support a larger number of chips on a given memory module. To allow for all the memory chips on the module to properly function, it is necessary for the module to add a delay to all command requests to the memory. The delay is actually generated by an additional clock generator that resides on the module instead of relying on the motherboard. This will decrease the overall performance of a given memory module. The memory type is typically used in large server environments where large amounts of RAM are required. In these environments, the benefits of the increase in RAM outweigh the performance impacts caused by memory swapping.
actually, registrered ram doesnt have too much of a penalty on latency, thats ecc, which basically just ads one more clock cycle, so cas 2 ram is cas 3 ecc. registered non-ecc ram will actually function at the same latency as it was non-buffered, its just that nobody has really cared enough to make non-ecc ram registered, as 32 bit os's cant use registered ram anyways, as you can have up to 16gb of ram as u-buffered. registered ram will be a little slower, but not a whole clock cycle slower like ecc.

arisythila

09-24-2006, 02:09 PM

I believe what he was meaning. Is there is no decent Registered memory out.. Fastest we've seen in 1 gb x 2 is DDR2-667. Thats crap in my opinion. I want to see some nice DDR2-1000 that is overclockable to DDR2-1200, like we can with regular DDR2 memory. This is what I want.. I just hope if they take the route of going to Scoket 1207.. They make it so we can use normal DDR2 memory.

~Mike

nn_step

09-24-2006, 03:47 PM

actually, registrered ram doesnt have too much of a penalty on latency, thats ecc, which basically just ads one more clock cycle, so cas 2 ram is cas 3 ecc. registered non-ecc ram will actually function at the same latency as it was non-buffered, its just that nobody has really cared enough to make non-ecc ram registered, as 32 bit os's cant use registered ram anyways, as you can have up to 16gb of ram as u-buffered. registered ram will be a little slower, but not a whole clock cycle slower like ecc.
What makes you think you can have non-ECC registrered ram?

***Deimos***

09-24-2006, 04:13 PM

What makes you think you can have non-ECC registrered ram?

good point

registered/buffered memory primarily serves the purpose of reducing the huge capacitative load on the memory controller allowing more memory to be installed. (for effect of capacitative load on memory controller, compare overclock/timing of single sided DIMM vs double sided, and 1 DIMM/channel vs 2 DIMMs (ie 4 total).

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

The Ghost

09-24-2006, 04:58 PM

Samsung has Registered DDR2-800:

http://www.samsung.com/products/semiconductor/DDR_DDR2/DDR2SDRAM/Module/RegisteredDIMM/M393T2950EZA/M393T2950EZA.htm
nice find , the first that i have seen :)

cky2k6

09-24-2006, 05:20 PM

What makes you think you can have non-ECC registrered ram?
because they two techs are not related... ecc is for error correction, registered ram is for capacity. theyre only found together since theyre both critical on servers, and useless so far for enthusiasts... there has been non-ecc registered ram in the past, not common, but it exists...

oldblue

09-24-2006, 05:20 PM

This is what Rahul Sood, founder and CTO of VoodooPC, had to say:

With the help of Nvidia, and others, AMD is creating an entirely new motherboard concept for 4x4. It's not your standard-issue workstation motherboard; rather, it's an enthusiast-class multiprocessor motherboard with support for unbuffered non-ECC enthusiast-class memory. AMD also promises to release three tiers of 4x4 in the box where processor kits will start well under $1,000 and go up from there.

http://voodoopc.blogspot.com/2006/09/amd-creating-new-category-of.html

Fred_Pohl

09-24-2006, 06:34 PM

Just because 4x4 will be using socket 1207 doesn't mean that 4x4 will require registered ram.

***Deimos***

09-24-2006, 08:34 PM

I guess people are being haunted by memories of the initial FX series (ie registered RAM, socket 940)

Personally, I can't wait and see what AMD delivers with "socket" 1207. Reserved pins for more HT links? Reserved pins for other memory types? Very curious to see how many DIMMs new Opterons will be able to push. I bet that memory controller is built like a tank.

nn_step

09-24-2006, 09:03 PM

Not to mention the native DDR3 support :D

Nedjo

09-25-2006, 02:12 AM

HKEPC view on the subject (they are claming that the source is mobo manufacturers...):

http://www.hkepc.com/bbs/itnews.php?tid=674691&starttime=0&endtime=0

el rolio

09-25-2006, 04:20 AM

uOpt

09-27-2006, 07:19 AM

Sorry for the noob question, but what's so bad about registered rams? Are they not as fast?

There is only a minimal overhead for the registered and at this time almost nonexistent for the ECC. You can easily test the latter by turning on and off ECC in a board with unregistered RAM. I might have some numbers somewhere.

The problem is overclocking RAM. Overclocking RAM is a product made from selected chips, or at least selected modules.

The enthusiast RAM vendors would hate to, and probably will not, offer an overclocking registered product. That didn't even happen when the first Athlon FXes were the craze and required registered DDR. OCZ and Mushkin made some low-latency but that was it.

The sad thing is that it is all rubbish anyway. RAM speed only has a minimal impact on real-world performance and that "run 1:1 for best overclock" rule seems to be a myth, too (at least my max dual mprime stabe clocks don't drop when I take down the RAM).

But the overclocking crowd doesn't care. They want fast RAM anyway.

If 4x4 is on 1207 and still requires buffered (registered) ram, what exactly is the point of trying to sell it over a dual-socket server board with 22xx Opterons which would be the exact same thing?

The only point would be the unlocked multiplier and the faint hope that board vendors offer overclocking boards for dual 1207.

There's got to be something cooking on that front. Even if not, at least 1207 doesn't require ECC...I think a lot of people confuse the two.

All current processors support but no not require ECC, including AMD64, Core2, PowerPC and Itanium.

It would in fact be possible for the enthusiast RAM vendors to make registered non-ECC to cut down cost and maximize overclock potential for socket 1207 if they think there's a market of people buying it (there isn't).

Then what's the advantage of registered ram? Why do servers use them? Thanks.

You reduce the load on the control lines so that the same processor can drive more chips. As a result the same mircoprocess can drive twice as many modules and each module can have twice as many chips, resulting in a 4x increase of memory capacity. It also reduces the change of RAM error from random electrical sources since the signal is now stronger.

Keep in mind that in registered RAM as in for socket603/ 604, 940 and 1207 only the control lanes are buffered, the data lanes are not. FB-Dimms buffer the data, too, but FB also changes the whole setup to a cascading one.

Just because 4x4 will be using socket 1207 doesn't mean that 4x4 will require registered ram.

Read the disription of the pin layout. The socket 1207 platform is for registered RAM.

The memory controller is on the CPU for AMD64s, so there no switching between registered and unregistered with the same CPU and/or socket.

In the Intel world that would work, because the CPU doesn't have the memory controller. The same Intel CPU can theoretically use registered and unregistered RAM in different boads. But in practice they are locked, too, because Intel usually only relases one chipset, and then the chipset fixes what kind of RAM you need.

Or in other words, you could run a Woodcrest with DDR2 unbuffered or non-FB registered, as in the CPU can, but you cannot buy a chipset to do so.

Poodle

09-28-2006, 05:37 AM

What's all this chatter about ECC/reg memory? :stick:

Hey guys. It would be pointless of introducing 4X4 with anything other than normal unreg/Non ECC memory modules because if they wouldn't it would be just a normal Socket F system.

Last year I believe that Abit showed of an Opteron socket 940 (e.g. not 939) board that didn't make it to the shops but seemed to work, that could handle unreg/non ECC. So why wouldn't all the big guys over at Amd not be able to get this to work?

Amd wouldn't sell any 4x4 if they wouldn't. We know it and they must surely know it! If i'm wrong the 4x4 was just a media stunt to get people not to jump on Kentsfield.

lapdog

09-28-2006, 05:54 AM

These FX chips must be different than Socket F. If they are the same they will sell like hotcakes to potential Opteron buyers as a single 2.6GHz Opteron 285 is over $900. I am sure the memory controller is the only difference.

Lightman

09-28-2006, 07:46 AM

so all you smart people, can someone explain to me what is the point of 4x4 if it is going to require server mobo, server chip and server memory?

it wont require server memory. Any of you can use some logic and deduct that 4x4 on server board wouldn't take any time to go to market. But designing 4x4 took some time which means that AMD was changing, developing something different than normal 2p server. It certainly will be on 1207 but it won't be re branded 2p workstation board with pcie 16x.

nn_step

09-28-2006, 09:00 AM

so all you smart people, can someone explain to me what is the point of 4x4 if it is going to require server mobo, server chip and server memory?
I think we reasonably deduced that it isn't really a server Mobo, it doesn't require an exotic server chip or Server ram. So one can assume saying it is on one socket or another is speculation

uOpt

09-28-2006, 09:02 AM

Hey guys. It would be pointless of introducing 4X4 with anything other than normal unreg/Non ECC memory modules because if they wouldn't it would be just a normal Socket F system.

Yeah, but apparently AMD tried to do that initially, when they announced 2+2(4x4) to be AM2 based and failed to secure mainboard makers for what is essentially an entirely new platform.

So now that have to share with an existing platform.

Last year I believe that Abit showed of an Opteron socket 940 (e.g. not 939) board that didn't make it to the shops but seemed to work, that could handle unreg/non ECC. So why wouldn't all the big guys over at Amd not be able to get this to work?

Oh my, not that rubbish again. You are not the troll over from 2cpu.com, are you?

That board never made retail and I know nobody who actually ever used one.

Technically, it might be possible. ECC is never required and you can move the buffer chips from the RAM modules to the board. Of course, that would be pointless since although you can now use unregistered RAM you still have the register chip limiting overclocks, not to mention it is now more expensive (since you pay for 4 register chips even if you only use 2 modules).

Amd wouldn't sell any 4x4 if they wouldn't. We know it and they must surely know it! If i'm wrong the 4x4 was just a media stunt to get people not to jump on Kentsfield.

The cause of events makes it most likely that they tried to establish an AM2 based 4x4 platform with unregistered RAM and failed, and now instead of cancelling 4x4 they just "rebase" it. That means the only technical effect of 4x4 will be that we get 1207 chips with unlocked multipliers and maybe a few boards with overvolting options.

Now, whether anybody produces 4x4 boards (no matter which socket) with serious overvolting options is an entirely different matter. The length of lines on the PCB board and the fact that you have to shuffe around twice the power makes that a very serious risk for the vendor. If you can sell only a couple hundred boards and half of them come back with blown up mosfets you won't make any profit.

In any case, socket 1207 is a registered RAM platform and nobody can do anything about that.

Poodle

09-28-2006, 09:27 AM

Yeah, but apparently AMD tried to do that initially, when they announced 2+2(4x4) to be AM2 based and failed to secure mainboard makers for what is essentially an entirely new platform.

So now that have to share with an existing platform.

Oh my, not that rubbish again. You are not the troll over from 2cpu.com, are you?

That board never made retail and I know nobody who actually ever used one.

Technically, it might be possible. ECC is never required and you can move the buffer chips from the RAM modules to the board. Of course, that would be pointless since although you can now use unregistered RAM you still have the register chip limiting overclocks, not to mention it is now more expensive (since you pay for 4 register chips even if you only use 2 modules).

The cause of events makes it most likely that they tried to establish an AM2 based 4x4 platform with unregistered RAM and failed, and now instead of cancelling 4x4 they just "rebase" it. That means the only technical effect of 4x4 will be that we get 1207 chips with unlocked multipliers and maybe a few boards with overvolting options.

Now, whether anybody produces 4x4 boards (no matter which socket) with serious overvolting options is an entirely different matter. The length of lines on the PCB board and the fact that you have to shuffe around twice the power makes that a very serious risk for the vendor. If you can sell only a couple hundred boards and half of them come back with blown up mosfets you won't make any profit.

In any case, socket 1207 is a registered RAM platform and nobody can do anything about that.

1: First I don't hang out much over at 2cpu.com more than to :slobber: when I've heard something new is going on. I don't know what "troll" you're talking about. I don't even got an account there if you want to know...

2. You seem to be sure on your argument so I have to believe you. This said I can't say that 4x4 feels that hot to me anymore . Non ecc/reg was half the point with it. :( I was first glad about it ending up on 1207 but now I don't know...

3. What made me sure of my argument was that the new M2R32-MVP can handle ECC/reg on Am2 so I guessed that the otherway around also was possible.

oldblue

09-28-2006, 09:42 AM

Read the disription of the pin layout. The socket 1207 platform is for registered RAM.

Could you point me to a description of the socket F (1207) pin layout? I haven't been able to find one.

uOpt

09-28-2006, 10:16 AM

3. What made me sure of my argument was that the new M2R32-MVP can handle ECC/reg on Am2 so I guessed that the otherway around also was possible.

Where's that info coming from? Any pointers?

%%

oldblue, since the memory controller is on the chip the lanes must match the memory access interface. Since we already know that 1207 is for registered with the normal Opterons, there is no way to use the same pin layout for unregistered RAM.

Unless they are willing to re-define what each pin does but then the result would be a new socket, even though it might have the same number of pins. That would instantly kill mainboard maker's interest since you couldn't exchange processors and boards with the other socket 120, that's out of question. Intel got away with doing that with the two 478 sockets, but both of them were major platforms so mainboard vendors could afford it.

Poodle

09-28-2006, 10:21 AM

Where's that info coming from? Any pointers?

%%

oldblue, since the memory controller is on the chip the lanes must match the memory access interface. Since we already know that 1207 is for registered with the normal Opterons, there is no way to use the same pin layout for unregistered RAM.

Unless they are willing to re-define what each pin does but then the result would be a new socket, even though it might have the same number of pins. That would instantly kill mainboard maker's interest since you couldn't exchange processors and boards with the other socket 120, that's out of question. Intel got away with doing that with the two 478 sockets, but both of them were major platforms so mainboard vendors could afford it.

Well here: http://www.sweclockers.com/nyhet/?id=4693 in Swedish I'm arfaid. Now I'm not sure anymore if the author was completly sober when he wrote it though. :)

Edit: Maybe he wasn't drunk after all: http://www.neoseeker.com/Hardware/Products/asus_m2r32_mvp/

uOpt

09-28-2006, 10:27 AM

I'm quite sure I read it in the press release for the board.

That would be a typo. Wouldn't be the first time this happens to Abit, I remember vividly how one of their 939 boards web pages stated max RAM 8 GB. Right.

While it is theoretically possible (but utterly pointless) to run unregistered RAM from a registered memory controller by moving the register chips from the RAM modules to the board, there is nothing you can do to access registered RAM from an unregistered memory controller. And running both registered and unregistered from the same slots is out of question, too, and they certainly wouldn't lok themself into registered RAM.

%%

Those who want 4x4 with unregistered memory have only one hope: that the original report of the socket 1207 rebasing turns out wrong, or that the decision gets reverted when they magically find a mainboard vendor for AM2 4x4.

Poodle

09-28-2006, 10:45 AM

That would be a typo. Wouldn't be the first time this happens to Abit, I remember vividly how one of their 939 boards web pages stated max RAM 8 GB. Right.

While it is theoretically possible (but utterly pointless) to run unregistered RAM from a registered memory controller by moving the register chips from the RAM modules to the board, there is nothing you can do to access registered RAM from an unregistered memory controller. And running both registered and unregistered from the same slots is out of question, too, and they certainly wouldn't lok themself into registered RAM.

%%

Those who want 4x4 with unregistered memory have only one hope: that the original report of the socket 1207 rebasing turns out wrong, or that the decision gets reverted when they magically find a mainboard vendor for AM2 4x4.

Well it's sad: I'd like to have one but now I think I'll probably stick to my E6600 for a long time and save the money for once. :)

Maybe new hopes around Am3 and K8L. We'll see.

oldblue

09-28-2006, 11:42 AM

While it is theoretically possible (but utterly pointless) to run unregistered RAM from a registered memory controller by moving the register chips from the RAM modules to the board, there is nothing you can do to access registered RAM from an unregistered memory controller.

What if the memory controller supports both registered and unregistered RAM? Could there then be two different types of socket F motherboards, one that uses registered RAM and another that uses unregistered RAM?

[XC] leviathan18

09-28-2006, 11:49 AM

mmm i guess im the only one here that will say this the mem controller is in the cpu so depends on the cpu if accepts ecc and registered ram

uOpt

09-28-2006, 11:49 AM

What if the memory controller supports both registered and unregistered RAM?

There's an extra pin for registered memory, it's not just "software".

Could there then be two different types of socket F motherboards, one that uses registered RAM and another that uses unregistered RAM?

That would make no sense. That would be a like a new incompaible socket, just happens to have 1207 pin. Nobody would make a board for that given that only rich enthusiasts would buy it (most enthusiasts are poor).

if AMD could force a new platform then would force an AM2 4x4 but apparently they tried and failed on that one.

awdrifter

09-28-2006, 12:03 PM

If the new 4x4 does turns out to be just a rebadged 2 socket server board with OC'ing options, it actually doesn't seem so bad. Cuz that means we can use the 22xx series server cpu on it right? So we could possibly builld a 4 core system for less than $1k.

oldblue

09-28-2006, 12:13 PM

There's an extra pin for registered memory, it's not just "software".

I understand. I was wondering whether it would be possible that the registered memory pin could be active in the "registered memory" motherboard but not the "unregistered memory" board.

That would make no sense. That would be a like a new incompatible socket, just happens to have 1207 pin.

Not entirely. You would still be able to use the same processors in either motherboard, as long as the memory controller on the processor is capable of supporting either registered or unregistered memory.

Nobody would make a board for that given that only rich enthusiasts would buy it (most enthusiasts are poor).

if AMD could force a new platform then would force an AM2 4x4 but apparently they tried and failed on that one.

I understand this argument, and I tend to agree with it. But I'm trying to better understand the technical problems of offering socket F motherboards with unregistered RAM.

lapdog

09-28-2006, 07:58 PM

I see no reason AMD can't make this work on only a few mobos. They can't have the new FX line work on all Socket F mobos or they will loose Opteron sales--or have to half the price of the top of the line Opterons. Not happening. It will be a new memory controller that works on only a few mobos. After all this is just a ploy to lessen attention from the real quad cores, just like the photos at the INQ (http://www.theinquirer.net/default.aspx?article=34742).

uOpt

09-28-2006, 08:45 PM

I see no reason AMD can't make this work on only a few mobos.

No way they are going to find a mainboard maker for this plan. Nobody's going to design new boards that only work for 4x4 and only with special 4x4 CPUs (the original plan would have normal AM2 FXes).

lapdog

09-30-2006, 05:49 AM

No way they are going to find a mainboard maker for this plan.. With only one or two supporting mobos the volume may be high enough for a high price enthusiast product. The extreme high end is always fairly low volume anyway.

nn_step

09-30-2006, 01:13 PM

With only one or two supporting mobos the volume may be high enough for a high price enthusiast product. The extreme high end is always fairly low volume anyway.
At the same time DFI made their Venus board and not very of them were sold, SOO maybe some could do something similiar for 4x4 and sell even more for similiar profit

zakelwe

09-30-2006, 02:12 PM

If anyone will do it then Asus will, just look at their 7800GT x2 vga card.

Considering that this solution is seemingly more and more bespoke why is the enthusiast who needs 4 cores not going to be swayed by popping a Kentsfield into a 975 motherboard ?

Regards

Andy

VulgarHandle

09-30-2006, 03:08 PM

Considering that this solution is seemingly more and more bespoke why is the enthusiast who needs 4 cores not going to be swayed by popping a Kentsfield into a 975 motherboard ?

cause he can make it 8 cores couple months later...

***Deimos***

09-30-2006, 04:19 PM

["Technically, it might be possible. ECC is never required and you can move the buffer chips from the RAM modules to the board. Of course, that would be pointless since although you can now use unregistered RAM you still have the register chip limiting overclocks, not to mention it is now more expensive (since you pay for 4 register chips even if you only use 2 modules)."

ok, I dont really know that much about pin layout of registered memory (or ECC), but here is my take:

buffered/registered: have extra logic to buffer the command/address signals, introducing extra latency, but lower capacitative load on memory controller. I think a memory controller that supports registered, would also be able to use non registered, although not simulatenously... the timings would need to be adjusted.

ECC: If I recall correctly, that's one extra memory chip (per 8) to check if data is correct. So, output is 72 bits not 64 bits, per DIMM. I think the DIMMs are notched differently and wouldn't fit, so I guess you wouldn't be able to mix with regular un-buffered.

shimq1

10-01-2006, 11:07 AM

At the same time DFI made their Venus board and not very of them were sold, SOO maybe some could do something similiar for 4x4 and sell even more for similiar profit
There were only 1,000 DFI Venus boards made, and they were all sold.

nn_step

10-01-2006, 11:23 AM

There were only 1,000 DFI Venus boards made, and they were all sold.
And I highly doubt no less than 1,000 4x4 boards would be sold.
Hell for Budget servers, it would sell like wild fire

***Deimos***

10-01-2006, 03:03 PM

And I highly doubt no less than 1,000 4x4 boards would be sold.
Hell for Budget servers, it would sell like wild fire

budget server is almost an oxymoron.

3NZ0

10-01-2006, 03:34 PM

budget server is almost an oxymoron.
I disagree, both my servers (my old one and my new one have been rather cheap, also delivering a good whack of power too.

(first was my p4 3.2c main rig with a few changes to make it a sever, second is a pentium 930d at 4ghz)

shimq1

10-01-2006, 03:57 PM

And I highly doubt no less than 1,000 4x4 boards would be sold.
Hell for Budget servers, it would sell like wild fire
You're wording is kinda confusing, do you mean you doubt less than 1,000 4x4 boards will be sold.

or

You doubt more than 1,000 4x4 boards will be sold.

ozzimark

10-01-2006, 04:02 PM

ECC: If I recall correctly, that's one extra memory chip (per 8) to check if data is correct. So, output is 72 bits not 64 bits, per DIMM. I think the DIMMs are notched differently and wouldn't fit, so I guess you wouldn't be able to mix with regular un-buffered.
ecc is interchangable with non-ecc from what i understand ;)

nn_step

10-01-2006, 06:01 PM

You're wording is kinda confusing, do you mean you doubt less than 1,000 4x4 boards will be sold.

or

You doubt more than 1,000 4x4 boards will be sold.
I doubt that less than 1,000 4x4 boards would be sold (should someone make them) But in all probability if AM2 supports quadcore AM3 Processors. I AM very sure LOTs and LOTs of these boards will sell

uOpt

10-02-2006, 05:19 AM

ECC: If I recall correctly, that's one extra memory chip (per 8) to check if data is correct. So, output is 72 bits not 64 bits, per DIMM. I think the DIMMs are notched differently and wouldn't fit, so I guess you wouldn't be able to mix with regular un-buffered.

I mix them all the time. Of course the extra ECC chip is then ignored even on the modules that have it.

ECC is really not the problem here, it is fully ignoreable.

uOpt

10-02-2006, 05:26 AM

There were 1000 Venus boards sold, but it was a $300 board taking any 939 CPU and taking any unregistered DDR RAM.

Let's assume it is technically possible and 4x4 is socket 1207 with the CPU using unregistered RAM, then you can exchange those CPUs neither with the 1207 server platform, nor with AM2.

And to sell these 4x4 board they need to overclock high, have high-quality capacitors etc etc. And it needs to be dual, so it almost certainly needs to be 6-layer.

There is no way way that such a board could cost below $500, and then you add two CPUs that don't run anywhere else. You really cannot compare that to the Venus board situation.

I don't think 1000 board of "1207 special unbuffered" can be sold.

Think back about the original plan: they didn't find anybody to make that board even when it would use normal AM2 CPUs.

***Deimos***

10-02-2006, 08:25 AM

uOpt

10-02-2006, 10:09 AM

woah.. at least $500?
a motherboard hardly costs that much.. maybe more like $50 (and thats including all the in-box accessories). With more expensive chipset and other chips on the motherboard up to $100.. maybe even $120.

If they tried to sell for $500, nobody would buy it. Even $300 is really pushing it/

I don't think you looked at high-quality dual boards lately, or any 6-layer board for that matter. Or any socket F board. Then add some overclocking potential which at least means more engineering and stronger power distribution.

nn_step

10-02-2006, 03:32 PM

I don't think you looked at high-quality dual boards lately, or any 6-layer board for that matter. Or any socket F board. Then add some overclocking potential which at least means more engineering and stronger power distribution.
What if it is Socket AM2 not Socket F

emboss

10-03-2006, 09:55 PM

Being socket 1207 doesn't stop it from supporting unbuffered RAM. Remember the MP/MPX chipsets? It could run two unbuffered or 4 registered sticks.

The same capability apears to be built into the K8 memory controllers as well (see the K8 BIOS Programming Guide). Which is not that surprising really - the only electrical difference between registered and unbuffered DDR2 is a difference in the capacitive loading (ignoring the two address parity pins which would be ignored and unaffected by an unbuffered module anyhow). The only other difference is in the timings which are easily configurable.

The question then comes down to whether the motherboard manufacturers have done something in hardware to break this ability, or whether all that is required is a BIOS change to allow the board to start with unbuffered DIMMS and configure the CPU appropriately.

cky2k6

10-04-2006, 05:43 AM

they are not going to use current s1207 boards for 4x4 for petes sake. why would they use a server board for enthusiasts. its going to be the same board as any enthusiast board, just with two processors and using s1207. im sure the chipset will be similar to the server boards, theyre probably going to use nforce pro, but thats where the similarities end.

Dimitriman

10-04-2006, 07:17 AM

A small detail to add is that most server boards don't aim at having SLI or Crossfire enabled. There are a few boards from each brand with high end workstation capability which have SLI capability but there aren't many. This is probably where the 4x4 scheme will have the biggest impact, will most likely require all the boards to have dual x16 PCI express and SLI/Crossfire support. One can probably expect the price of these mobos to be lower than that of server boards (which in general vary between 250$ and 500$ for dual sockets), all in all i think the 4x4 should be an attractive design for gamers and enthusiasts even with ECC ram.

nn_step

10-04-2006, 09:08 AM

el rolio

10-04-2006, 11:19 AM

yea nn step i wa wondering about that too what with the recent info beiing speculated at diff news sites across the web. what of this am2+ stuff etc. gos hi guess lots of thigns will be revealed come nov/dec

nn_step

10-04-2006, 11:33 AM

uOpt

10-04-2006, 01:03 PM

I'm just going to throw this idea out to see if anyone might view this idea as reasonable.
AM2+ supports HTT 3.0 and the ability to split HTT lanes. Only a 8bit coherent HTT link is needed for 4x4
Thus AM2+ boards could support 4x4 and it is waiting for that change before such a setup happens. Just a thought.

I don't think that the lack of a coherent HTT has anything to do with the decision (if there was one) to drop 2+2 on AM2/unbuffered.

Really, the three hyperlinks are on the AMD64 die and they get coded to be coherent or not (I/O) based on what market the processor is selected for.

It is highly unlikely that random yield problems are the limit for the number of coherent hyperlinks on a given CPU.

Or in other words: if AMD wants to give AM2 CPUs a coherent hyperlink there is is nothing keeping them from doing so.

nn_step

10-04-2006, 01:06 PM

I don't think that the lack of a coherent HTT has anything to do with the decision (if there was one) to drop 2+2 on AM2/unbuffered.

Really, the three hyperlinks are on the AMD64 die and they get coded to be coherent or not (I/O) based on what market the processor is selected for.

It is highly unlikely that random yield problems are the limit for the number of coherent hyperlinks on a given CPU.

Or in other words: if AMD wants to give AM2 CPUs a coherent hyperlink there is is nothing keeping them from doing so.
umm I believe we have already covered this.
ALL AM2 cpus DO have a coherent Hypertransport link
The main debate was if they had two like 2P setups and their associated processors

uOpt

10-06-2006, 10:01 AM

I don't know whether you guys spotted that:
http://badhardware.blogspot.com/2006/10/amds-k8l-revealed-in-cray-rainier.html

http://photos1.blogger.com/blogger/7288/2341/400/k8l_yjM7OCMYeXam.jpg

So, that FX is Socket F. Not some new socket happening to have 1207 pins, and not AM2. So it's straight server platform, and registered RAM.

nn_step

10-06-2006, 10:22 AM

somehow I question a technical site that has the phrase "* Mel Gibson * for President !"

Poodle

10-06-2006, 12:29 PM

somehow I question a technical site that has the phrase "* Mel Gibson * for President !"

:toast:

Will the Antares X2s have L3 as well? :confused: Well good stuff...

informal

10-07-2006, 04:12 AM

http://www.theinquirer.net/default.aspx?article=34918

Two FX-70 CPU pack matches pricing, clocks of Core 2 QC6700

By Theo Valich: Saturday 07 October 2006, 07:09
WITH THE IMMINENT launch of Core 2 QC6700 (codename: Kentsfield), AMD will respond by introducing the QuadFather platform, marketing dubbed the 4x4. The platform consists of Opteron workstation motherboards using Socket 1207 and equipped with two PCIe x16 slots for graphics cards and retail packaged CPUs.

We have learned that not many motherboard vendors will support the platform, so it is up to the highest-end manufacturers such as Tyan, SuperMicro, or Iwill to offer 4x4 motherboards. This will hardly differ from current 1207-pin-socketed Opteron motherboards. At the same time, it seems that AMD will finally remove the need for Registered ECC memory for Athlons, but another source is countering this info with the claim that FX-70 series will be nothing more but the Opterons and that the processors will require usage of Registered DIMM memory.

The retail boxes are coming in a bit larger size than current packaging and feature an open look at two CPUs, but for the first time, AMD will be launching not one, but three FX CPU packages. All of the CPUs will feature 256KB of L1 and 2MB of L2 cache memory (each core has 64+64KB L1 and 1MB L2 cache), and be placed in three speed grades.

FX-70 CPUs will be clocked at 2.6 GHz, with the price of $999 for a dual-processor retail box.

FX-72 will be clocked at 2.8 GHz, and with the price of $1132, it presents an interesting choice, since it outclocks Kentsfield.

FX-74 is clocked at 3.0 GHz and finally breaks the 3 GHz clock barrier for desktop and dual-core CPUs in general. However, the price is currently planned at sky-high $1500, so who ever spends that amount of money in the year of K8L (Barcelona) arrival, it should better have another $1,500 in the bank - for quad-core Barcelona babies, of course.

All in all, it seems that there isn't much sense for AMD to release QuadFather platform in Q4'06, but this was probably directed by marketing, which needs strong answer to Kentsfield. However AMD, are you guys too little, too late when it comes to enthusiasts, which you kept with your doors shut while everything was la-la in Opteron land? µ

[XC] Teroedni

10-08-2006, 02:23 AM

somehow I question a technical site that has the phrase "* Mel Gibson * for President !"

what?
You mean you have more confidence in Bush:D

I can see one good thing about 4x4 being socket 1207
NUMA:toast: :D

uOpt

10-09-2006, 08:37 AM

Only dual CPU packages?

What if you damage one chip while putting it in?

That's the stupidist thing I heard today.

Prices are good, though. $1500 for two 3.0 dual-cores would be interesting - right now.

el rolio

10-26-2006, 08:11 AM

http://www.theinquirer.net/default.aspx?article=35334

snippet:

Here is the answer to a big question which has bothered many of our readers. Although 4x4 uses the Opteron processor socket "F", it will not use registered ECC memory. Athlon 64 FX-70/72/74 will ask for regular DDR2 memory modules, with Corsair, Geil and OCZ offering blazingly fast memory for the system. Corsair will get some preference in reviews due to the fact that it sports the so-called "Enhanced Performance Profile", also known as SLI Memory. This was co-developed by Corsair and Nvidia. However, there are now over five different memory manufacturers offering a combined total of 17 different memory kits.

uOpt

10-26-2006, 08:20 AM

Lightman

10-26-2006, 08:23 AM

http://theinquirer.net/default.aspx?article=35333

he launch line up was revealed here, but the pricing has been "re-constructed" severely downwards, with the fastest CPU pack now matching the price of Intel's Kentsfield.

So I'm expecting pair of FX-74 for under $1000USD :D
Wonder what prices will be on FX-70 and FX-72 CPU's.

Good for us

Lightman

10-26-2006, 08:27 AM

So they actually made a whole new platform.

New CPUs, new chipsets and new motherboards, none of them are compatible with either AM2 or Socket F. At all. And CPUs sold in pairs so if you blow one up then what?

That's insane. I'm lucky I didn't buy their stock when I wanted to.

It's still S1207 and updated NF590 x2, so may by some bright person will manage to 'mod' Opteron and fit it in 4x4 mobo. Everything is possible. Remember Athlon XP was moddable to Athlon MP only connecting one bridge on CPU. I know that this time it will be much more complicated, but who knows??

The Ghost

10-26-2006, 08:27 AM

it seems like amd and intel are releasing their new cpu's on november 14

that should be a interesting day

perkam

10-26-2006, 08:28 AM

http://www.theinquirer.net/default.aspx?article=35334

snippet:SLI memory can go lick a fire hydrant.

I don't see what the excitement is over 4x4. It's basically a socket f server comp without the glory of having to say you have a dual socket server comp. It's the biggest obstacle to anyone getting one.

You buy a socket f server and you take to WCG forums and they'll start :slobber: ing all over the place....4x4 will "meh"ify the dp consumer market...I see no good coming out of this. The cheaper ram is a good touch but will leave no resale value to the comp as most demand for it is with ECC ram.

It's basically you go to a computing event and go to Tyan's exhibit and say

You: "Dude i got 4x4"
Tyan rep: "Yea...you got crippled dual processor server that will be obsolete when AMD K8L Launches"

:p

Perkam

Turtle 1

10-26-2006, 08:44 AM

it seems like amd and intel are releasing their new cpu's on november 14

that should be a interesting day

I think you better check when Intel is releasing Kentsfield. AS for 4x4 it looks very interesting!

The Ghost

10-26-2006, 08:46 AM

It's basically you go to a computing event and go to Tyan's exhibit and say

You: "Dude i got 4x4"
Tyan rep: "Yea...you got crippled dual processor server that will be obsolete when AMD K8L Launches"

Perkam
i thought that amd has said that they would make a x4 cpu to drop in the 4x4 ? maybe this won't be out dated so fast

this could be interesting

i did wish that amd would have made the 4x4 on the am2 socket

The Ghost

10-26-2006, 08:50 AM

I think you better check when Intel is releasing Kentsfield.
AS for 4x4 it looks very interesting!
i didn't make this up

http://www.theinquirer.net/default.aspx?article=35346
INTEL decided to paper launch its upcoming CPU, but the space between launch and availability isn't that long.

The company told the world+dog that reviews of Intel Core 2 Extreme Quad Core Processor QX6700 will hit the wires on November 2nd 6AM time in continental Europe. This long name is the official one that Intel picked for this dual core baby.

These CPUs will be available on the November 14th and this includes retail and etail for systems based on it.

The Xeon 5300 reviews should arrive on November the 14th UK time and available that day too.

Turtle 1

10-26-2006, 09:11 AM

i didn't make this up

http://www.theinquirer.net/default.aspx?article=35346
INTEL decided to paper launch its upcoming CPU, but the space between launch and availability isn't that long.

The company told the world+dog that reviews of Intel Core 2 Extreme Quad Core Processor QX6700 will hit the wires on November 2nd 6AM time in continental Europe. This long name is the official one that Intel picked for this dual core baby.

These CPUs will be available on the November 14th and this includes retail and etail for systems based on it.

The Xeon 5300 reviews should arrive on November the 14th UK time and available that day too.

Short ans, is nov 2. release. Forget the double talk. People in these forums have already tested these cpu's. But the Nov 14 release of 4x4 will be an interesting day of reviews. I personally don't care how it turns out. Trueth is I hope 4x4 owns Kentsfield. Intel needs a kick in the pants.
I personnally am moving away from the desktop cpu & m/b 's . My last build well probably be a 2 socket [B]TIGERTON using FBdimms . So it won't be a true gaming machine.

My son is waiting to see the 4x4 results and K8L . If its as good as many believe thats what he wants me to build for him . So I will.

ozzimark

10-26-2006, 09:11 AM

New CPUs, new chipsets and new motherboards, none of them are compatible with either AM2 or Socket F.
hm? it IS socket F ;)

and just curious, when was the last time you actually killed a cpu?

VulgarHandle

10-26-2006, 09:14 AM

I don't see what the excitement is over 4x4.
FX74 for $1000 or less(that's for both 3.0ghz dual-core cpu's)
cheaper unbuffered ram

as opposed to $1600 for 2 Opteron 2218(dual-core at only 2.6ghz)
more expensive registered ram

not to mention overclocking and 4 pci express slots....

and later upgrade to dual quad-cores, then later upgrade the mobo and get ddr3

nn_step

10-26-2006, 09:26 AM

VulgarHandle

10-26-2006, 09:44 AM

perkam

10-26-2006, 10:10 AM

true, but do you really need the fastest ram possible when dealing with 4gigs?

or, 4 512kb sticks for 2 gigs of faster, and still not gouge yourself

EDIT: i do have 1 question, since the CPU's carry the 'FX' moniker, will they be unlocked?You can learn a lot about servers JUST by going through Dell's AMD server configs.

Due to individual IMCs on each proc on a dp server, 4x512mb will be more beneficial than 2x1gb for dp setups. (Basically 2x512mb dual channel for each proc).

Perkam

Turtle 1

10-26-2006, 10:15 AM

FX74 for $1000 or less(that's for both 3.0ghz dual-core cpu's)
cheaper unbuffered ram

as opposed to $1600 for 2 Opteron 2218(dual-core at only 2.6ghz)
more expensive registered ram

not to mention overclocking and 4 pci express slots....

and later upgrade to dual quad-cores, then later upgrade the mobo and get ddr3

Hay thats great news . 2 x 3.ghz FX 74 cpu's for $1000 or less thats wonderful . I thought 2 of these were like $1500. The M/B I suppose you guys would say they cost like what. $150 . Thats sounds cheap enough.
4 pci-e slots for quad slew. Buy the NV G80 . 4x650= $2600 A bargain by anyone's standard.
Of course Vista $400 and 4x 1g ram = $500. In a system like this a 1000 watt PC power and cooling = $400. Top of the line sound card in a system like this is a must have $300. AT least 4 raptors anything less would be stupid for a system like this= $1400. I would imagine a bigger case is a must have= $150. If you run quad sli . Water is a must have or this will sound like a jet. = Min. $500. So lets see what we have for Cost of this in the maxed out version.
2x3.ghz FX74's $1000 ???????? price check.

M/B $150 ????????? price check

4x G80's $2600

Vista $400 a must have for DX10

Sound card $300 With a system like this a must have.

4X raptors $1400 Same as above.

Larger case $150

Watercooling $500 min. Must have or noise be a huge problem

PS $ 400 maybe use 2 ps.

total $6900

This is actually cheap for a system like this, Than of course there is the hidden cost savings. In the winter time forget about heating your PC room this baby will heat your house.

Power draw at the wall socket only 1200 watts adds to cost.

So there you have it the ultimate Gaming pc. :fact:

perkam

10-26-2006, 10:26 AM

Hay thats great news . 2 x 3.ghz FX 74 cpu's for $1000 or less thats wonderful . I thought 2 of these were like $1500. The M/B I suppose you guys would say they cost like what. $150 . Thats sounds cheap enough.
4 pci-e slots for quad slew. Buy the NV G80 . 4x650= $2600 A bargain by anyone's standard.
Of course Vista $400 and 4x 1g ram = $500. In a system like this a 1000 watt PC power and cooling = $400. Top of the line sound card in a system like this is a must have $300. AT least 4 raptors anything less would be stupid for a system like this= $1400. I would imagine a bigger case is a must have= $150. If you run quad sli . Water is a must have or this will sound like a jet. = Min. $500. So lets see what we have for Cost of this in the maxed out version.
2x3.ghz FX74's $1000 ???????? price check.

M/B $150 ????????? price check

4x G80's $2600

Vista $400 a must have for DX10

Sound card $300 With a system like this a must have.

4X raptors $1400 Same as above.

Larger case $150

Watercooling $500 min. Must have or noise be a huge problem

PS $ 400 maybe use 2 ps.

total $6900

This is actually cheap for a system like this, Than of course there is the hidden cost savings. In the winter time forget about heating your PC room this baby will heat your house.

Power draw at the wall socket only 1200 watts adds to cost.

So there you have it the ultimate Gaming pc. :fact:Leave the hw configs with people with a tad bit more experience pls :p

Perkam

Poodle

10-26-2006, 10:33 AM

@Turtle... Well the Kenty isn't El Mejor Cheapo either.... :stick: And consumes alot of power aswell. And you are forgeting that unlocked cpus always cost alot. Pentum EE costed alot while a X2 at half the price outperformed it as well as a E6600 outperforms a fx62 easy at half the price 2day.

As for cooling, nothing beats air (peformance/ease+dollar). You must have missed the P190 which will cool any 4x4 with a couple of Ninjas, no sweat.
I'm buying another Raptor to my rig which already has two and it's a C2D. So the harddrive thing doesn't hold either.
Everyone knows that 4x4 won't be cheap. But you will get a serversystem for less money that can do more. And Intel won't have anything in the same price that has anything near the BW.

IMHO. Socket F and unreg ram is sexy as hell. :)

Turtle 1

10-26-2006, 10:49 AM

Leave the hw configs with people with a tad bit more experience pls :p

Perkam

Please . I said plainly that this was a top of the line configuration. You no nothing about my experience. If I have something configured wrong for the top of the line configs. Please feel free to point them out. Other wise take your superior experience and shove were sun doesn't shine.

The above configs. Is based on what A top of the line 4x4 should look like.

I think its a dam cheap price for the hardware involved . I admit last year I didn't sell to many gamers because of Conroe's release. I told my customers why and most waited. Some I lost. But I sleep well when I sleep. This year I have already built and delivered 27 conroe based gamers. Not 1 was under $5000. So stick to post and stay out of peoples private lives please.

perkam

10-26-2006, 10:53 AM

Please . I said plainly that this was a top of the line configuration. You no nothing about my experirnce. If I have something configured wrong for the top of the line configs. Please feel free to point them out. Other wise take your superior experience and shove were sun doesn't shine.

The above configs. Is based on what A top of the line 4x4 should look like.

I think its a dam cheap price for the hardware involved . I admit last year I didn't sell to many gamers because of Conroe's release. I told my customers why and most waited. Some I lost. But I sleep well when I sleep. This year I have already built and delivered 27 conroe based gamers. Not 1 was under $5000. So stick to post and stay out of peoples private lives please.Ha lol...don't need to pick up those boxing mits for that lol...I was talking about hw config exp not computing exp :p

My point was that I wouldn't do hw configs on platforms that have less than 50% of the required hardware to build a comp.

I've used estimated conroe prices for the intel guide in the sig...but that was when there was some certainty as to price, availability, support and performance.

Perkam

Turtle 1

10-26-2006, 10:54 AM

@Turtle... Well the Kenty isn't El Mejor Cheapo either.... :stick: And consumes alot of power aswell.

As for cooling, nothing beats air (peformance/ease+dollar). You must have missed the P190 which will cool any 4x4 with a couple of Ninjas, no sweat.
I'm buying another Raptor to my rig which already has two and it's a C2D. So the harddrive thing doesn't hold either.
Everyone knows that 4x4 won't be cheap. But you will get a serversystem for less money that can do more. And Intel won't have anything in the same price that has anything near the BW.

IMHO. Socket F and unreg ram is sexy as hell. :)

Ya air is fine . I never said this was the cheapest way to build a 4x4 I plainly stated top of the line. Price out a Falcon with those spec's your looking at $12000+. As for the raptor thing why you adding another one if you all ready have two?

Like I said the price I gave above is cheap for the hardware involved for a self build.

Turtle 1

10-26-2006, 10:57 AM

Ha lol...don't need to pick up those boxing mits for that lol...I was talking about hw config exp not computing exp :p

My point was that I wouldn't do hw configs on platforms that have less than 50% of the required hardware to build a comp.

I've used estimated conroe prices for the intel guide in the sig...but that was when there was some certainty as to price, availability, support and performance.

Perkam

Thats better. Yes I understand what your saying. But if you look at the prices I listed I was conserative. I won't do a kentsfield configuration in this thread as its an AMD thread. But your right a top of the line Kentfield = $$$$$$

uOpt

10-26-2006, 11:30 AM

hm? it IS socket F ;)

It might be socket F but the 4x4 FX chips have a memory controller for unregistered RAM, and the Socket F Opterons for registered. The boards have memory sockets wired for unregistered RAM. So no exchange of CPUs or boards with Socket F.

It would be first after a long time that anybody releases a memory controller that can do both registered and unbuffered at the same time, and even if they do then the board probably would need separate sockets for both kinds of RAM.

and just curious, when was the last time you actually killed a cpu?

None yet, actually, but Murphy's law says I will have I ruin $1600 :)

nn_step

10-26-2006, 11:35 AM

It would be first after a long time that anybody releases a memory controller that can do both registered and unbuffered at the same time, and even if they do then the board probably would need separate sockets for both kinds of RAM.
Or just one board for Registered and another for Unbuffered. and the Proc can use which ever they want.
And if anyone could do such a thing, it would have to BE AMD, since they have the world's best DDR Controller and soon the world's best DDR2 controller :D

Turtle 1

10-26-2006, 12:06 PM

Or just one board for Registered and another for Unbuffered. and the Proc can use which ever they want.
And if anyone could do such a thing, it would have to BE AMD, since they have the world's best DDR Controller and soon the world's best DDR2 controller :D

Nice. But didn't AMD license Rambus tech . To do there memory controller. You make it sound as if AMD invented it. Whats the bus speed of AMD2? Is it like 200. or is it 400. I believe its like 200 but I have been wrong many times in this long life. Quad pumped= DDR2 800 if cpu freq. is 200.

ozzimark

10-26-2006, 12:20 PM

nn_step

10-26-2006, 12:25 PM

Nice. But didn't AMD license Rambus tech . To do there memory controller. You make it sound as if AMD invented it. Whats the bus speed of AMD2? Is it like 200. or is it 400. I believe its like 200 but I have been wrong many times in this long life. Quad pumped= DDR2 800 if cpu freq. is 200.
AMD licensed Rambus' FB-DIMM tech
Not DDR2

Lightman

10-26-2006, 12:47 PM

i was under the impression that registered ram was a bios/MC thing, not a slot/board layout thing.

You're right! I've got reg. DDR-I and they fit perfectly to DIMM slots on normal board, simply system will not boot :D

to uOpt: I was once building 3 (three) workstations based on MSI K7D-Master (762MPX) and Athlons MP(XP mod). Because in that time in Poland it was extremely hard to buy Registered RAM I went with unRegistered ones. This board/chipset allowed to install 2 unreg. DIMMS in first two banks (up to 1GB max), only when you needed more memory space the only way was to buy Reg. DIMMS. Think about this... ;) In this world nothing is certain :D

flippin_waffles

10-26-2006, 02:42 PM

For whatever it's worth, i've a pair of registered 512mB corsair modules that i have in my s940 fx51 system. I thought i'd try and add them to my s939 system and see, just fot the hell of it, if i could use them for a total of 2gB in this 4400 system. However it would not boot with the 2 combined. But i installed the 2 registered modules by themselves, and the system worked perfectly fine.

Turtle 1

10-26-2006, 02:54 PM

AMD licensed Rambus' FB-DIMM tech
Not DDR2

Thanks NN here I thought that FB-dimms was an intel type project. and had nothing to do with Rambus. Other than of course the patients rambus holds on DDR/DDR2 and DDR3 tech.

Here's an interesting link . Not because its about FBD's but what it says at the bottom/ Nehalem totally modular

Whoever pushed the whole concept of making Nehalem totally modular, pat yourself on the back, you probably saved the project. When that chip does roll out, look for FBDs to be more or less off the table.

http://www.theinquirer.net/default.aspx?article=34220

emboss

10-26-2006, 05:52 PM

i was under the impression that registered ram was a bios/MC thing, not a slot/board layout thing.

It's a bit of both. The only difference between registered and unbuffered DIMMs are
1) Registered DIMMs have buffers on the control lines which results in lower loading of the control lines.
2) There's a couple more lines for registered dimms to do with control signal error detection.

The lower loading of the control lines simply means that you can pile more DIMMs onto a single channel, and doesn't change anything on the board. For the extra control lines, obviously the board needs to route these from the CPU socket (assuming an IMC) to the DIMMs. Additionally, the memory controller itself (and the socket) needs to support these extra lines.

After that, it's simply timing, which in the case of the K8 is all handled by the BIOS.

If you look in the K8 BIOS programmers guide, you can see that any K8 should, in theory, support both registered and unregistered RAM. Which isn't all that surprising given that the same design is used on pretty much every socket and there's so little difference between unbuffered and registered RAM. S939/754 (and presumably AM2, though AMD hasn't released the pinout on that yet) can't support registered RAM because the socket doesn't have the extra control lines. S940/F do support registered RAM, and should therefore be able to also support unbuffered RAM if the BIOS will allow it and configure the IMC accordingly.

However, this doesn't mean that it makes sense to use Opterons on a 4x4 board. I haven't seen the updated pricing on the lower-end FX's, but the only ones that come out to be less than a grand for the pair are the 2212 (2GHz, $754/pair) and the 2210 (1.8 GHz, $510/pair) and their HE variants ($900/pair and $632/pair respectively). It would really only make sense to get a 2210, and that would require a decent overclock to compete with a single dual-core ~3GHz chip.

It would make more sense to use FX's on a server board, but I'm guessing AMD knows this and will stop them working either though hardware (force-disabling registered support on the IMC) or software (tell the BIOS guys "don't support FX's").

informal

10-26-2006, 06:59 PM

Turtle 1

10-26-2006, 07:49 PM

There is no FSB in AMD chips(except mem. controler which works at CPU speed)... :rolleyes:
Read the DC specs

Ya I was told that before. But I just can't grasp the idea. This is what confuses me. I will use intel . fsb of cpu = 266 quad pumped DDR2 533 = fsb of 1066.

now on AMD2 DDR2 800 would = what 200 frequency quad pumped =800 frequency running at cpu speed. Is that correct? So amd has less latency . but the ram is running at lower frequency. This is were i get confused. Because DDR2 533 is slower than DDR2 800. SO if using DDR2 800 with a 2.8ghz AMd= 2400 frequency or what ?

Poodle

10-26-2006, 11:12 PM

Ya air is fine . I never said this was the cheapest way to build a 4x4 I plainly stated top of the line. Price out a Falcon with those spec's your looking at $12000+. As for the raptor thing why you adding another one if you all ready have two?

Like I said the price I gave above is cheap for the hardware involved for a self build.

Ok. Maybe I over reacted but your tone was something like: "you have to get raptors for a 4x4..."

To your question: i think I'm gonna put the game drive on a raid-0.

Turtle 1

10-26-2006, 11:28 PM

Ok. Maybe I over reacted but your tone was something like: "you have to get raptors for a 4x4..."

To your question: i think I'm gonna put the game drive on a raid-o.

Hay its alright. I didn't mind. I didn't use caps.;) I said several times top of the line system. Its cool.

Poodle

10-27-2006, 12:29 AM

Hay its alright. I didn't mind. I didn't use caps.;) I said several times top of the line system. Its cool.

I've got to cool down. I think it's the long waiting for the Dfi rd600 that makes nervous and tense :brick: (and half the forum with me).

Turtle 1

10-27-2006, 01:09 AM

Ya it seems like it could a nice card. But If its not all you think it should be .

The Asus workstation 64 M/B is a nice board.

Poodle

10-27-2006, 02:15 AM

Ya it seems like it could a nice card. But If its not all you think it should be .

The Asus workstation 64 M/B is a nice board.

Which board ever is? ;) Ok, enough Off T.

perkam

10-27-2006, 07:23 AM

Hay its alright. I didn't mind. I didn't use caps.;) I said several times top of the line system. Its cool.yup...people over react all the time

Turtle 1 :slap:

Perkam

informal

10-27-2006, 08:12 AM

Turtle 1

10-27-2006, 05:39 PM

Turtle 1 ,you are confusing many things when AMD is in question.
First of all,mem. controler works at the speed of the CPU.No more and no less!
Second of all,AMD uses the whole numbers for the memory dividers for forming the frequency of the RAM.In other words,2.8 GHz FX62 chips when paired with DDR2-800 memory sets the clock for the memory to 400 MHz(divider 7).Nothing in this chip works at 2400Mhz like you said in some post of yours...
Other example,the 5000+(2.6GHz chip) when paired with DDR2-800 memory sets the memory in the following way :2600/400=6.5 which gets rounded to the next whole number which is 7.So the actual freq. of the memory is 2600/7=371 Mhz.Memory controler still works at the 2.6Ghz(full CPU freq.,and this is true for any part of the CPU like caches et al).And you need tho read the specs for DDR2 memory and about the base clock for instance.I can't correct you all the time and write the basic stuff...

I was asking about how you guys figured it. I should go get the thread because your making it sound like I was declaring something when in fact I was Asking . I was tring to figure it the wayI would Intel . I Didn't understand. I see now that the memory is running at cpu freq. But I still dont get this part of it. memory freq is = to cpu freq 2600/400 What does the 400freq represent. and what does Quad pump memory effect in as far what numder is quad pumped and were did that number come from.

ozzimark

10-27-2006, 05:47 PM

I was asking about how you guys figured it. I should go get the thread because your making it sound like I was declaring something when in fact I was Asking . I was tring to figure it the wayI would Intel . I Didn't understand. I see now that the memory is running at cpu freq. But I still dont get this part of it. memory freq is = to cpu freq 2600/400 What does the 400freq represent. and what does Quad pump memory effect as in as far what numder is quad pumped and were did that number come from.
yeah, you definitely seem really lost when it comes to amd's memory stuff. here goes.. only cause i've explained it like 1396726 times already to other people :D

htt is the base clock. many call it fsb. it's not.
cpu mhz = htt * cpu multiplier
ht link speed = htt * ht link multi (sometimes called ldt multi)
memory speed = cpu/x

x = ceiling ( cpu multi / memory ratio )
where the memory ratio is...
400mhz (ddr2-800) = 2/1 (2:1)
333mhz = 5/3
266mhz = 4/3
200mhz = 1/1
lowest x value is 5

for an example, if i set 340mhz htt, 9x cpu multi, 3x ht link multi, 400mhz memory ratio
cpu mhz = 340*9 = 3060mhz
ht link = 1020mhz
x = ceiling ( 9 / 2/1 ) = ceiling (4.5) = 5
memory speed is 3060mhz / 5 = 612mhz

understand now? :toast:

Turtle 1

10-28-2006, 10:36 AM

yeah, you definitely seem really lost when it comes to amd's memory stuff. here goes.. only cause i've explained it like 1396726 times already to other people :D

htt is the base clock. many call it fsb. it's not.
cpu mhz = htt * cpu multiplier
ht link speed = htt * ht link multi (sometimes called ldt multi)
memory speed = cpu/x

x = ceiling ( cpu multi / memory ratio )
where the memory ratio is...
400mhz (ddr2-800) = 2/1 (2:1)
333mhz = 5/3
266mhz = 4/3
200mhz = 1/1
lowest x value is 5

for an example, if i set 340mhz htt, 9x cpu multi, 3x ht link multi, 400mhz memory ratio
cpu mhz = 340*9 = 3060mhz
ht link = 1020mhz
x = ceiling ( 9 / 2/1 ) = ceiling (4.5) = 5
memory speed is 3060mhz / 5 = 612mhz

understand now? :toast:

Thanks A ton . I got it finely. This is the part I was really struggling with

x = ceiling ( cpu multi / memory ratio )
where the memory ratio is...
400mhz (ddr2-800) = 2/1 (2:1)
333mhz = 5/3
266mhz = 4/3
200mhz = 1/1 <- this is what had me really puzzled now that I see 400 is 2/1( quad pumped 200mhz 1:1 DDR2 800)
lowest x value is 5