9800 GX2 Cebit Pics.

barkey52 · 4 Mar 2008 at 15:46

well we all know the greatest dual chip card was the XGI Volari V8 Duo

Makunouchi · 4 Mar 2008 at 16:15

Wish OCUK would put this on pre-order already ^^

jigger · 4 Mar 2008 at 16:23

easyrider said:
Saying it is 512bit is wrong?

No?

Yep, it's wrong if it's 256 bit x2.

Willis · 4 Mar 2008 at 16:42

jigger said:
Yep, it's wrong if it's 256 bit x2.

Are we 100% sure thats its 2 x 256Bit or is there a chance it could be 512 Bit x 2 :confused:

Just so many people say its this and that and as yet no hard facts :rolleyes:

Gashman · 4 Mar 2008 at 16:49

Willis said:
Are we 100% sure thats its 2 x 256Bit or is there a chance it could be 512 Bit x 2

Just so many people say its this and that and as yet no hard facts

there not 100% sure, just guessing i assume since as far as i know the technical data hasn't been released for these cards yet, there still under NDA right? and too all the 'X2 is technically better' crap, what brings you to that conclusion, just because they put two GPUs on the same PCB doesn't make it any more of an engineering masterpiece than two PCBs together, at the end of the day its the exact same thing just executed differently, i would be willing to bet good money that if they wanted to NVIDIA could make a dual-GPU, one PCB 8800 series card, but why bother, whats easier developing an entirely new PCB layout and heatsink from scratch or just bolting two existing PCBs together and doing it that way

Gerard · 4 Mar 2008 at 16:50

Gashman said:
what brings you to that conclusion, just because they put two GPUs on the same PCB doesn't make it any more of an engineering masterpiece than two PCBs together,

Yes it does.

Bonjour · 4 Mar 2008 at 16:53

Gashman said:
there not 100% sure, just guessing i assume since as far as i know the technical data hasn't been released for these cards yet, there still under NDA right? and too all the 'X2 is technically better' crap, what brings you to that conclusion, just because they put two GPUs on the same PCB doesn't make it any more of an engineering masterpiece than two PCBs together, at the end of the day its the exact same thing just executed differently, i would be willing to bet good money that if they wanted to NVIDIA could make a dual-GPU, one PCB 8800 series card, but why bother, whats easier developing an entirely new PCB layout and heatsink from scratch or just bolting two existing PCBs together and doing it that way

Slightly weird, in that you first say that the X2 is not technically better and then go on to admit that it's harder to develop an entirely new pcb and heatsink than just bolt two cards together.

I'm just wondering which is the case..

Lightnix · 4 Mar 2008 at 16:53

It's certainly 100% bonafida not 2x512-bit - the cards would be damn enormous (twice as many memory chips), much more so than they are now... Then there's the extra heat it'd output... Then the extra logic they'd have to add to the cores. It'd all get very hot and hairy...

Also, more factual proof:

Memory bus-width can be worked out with the formula: 64/2*c where c is the number of memory chips because GDDR3 works in 64-bit channels comprised of two chips ...

As you can see in this shot:

There would appear to be 8 chips, so: 64/2*8 = 256.

If anyone wants to correct my maths, feel free.

fornowagain · 4 Mar 2008 at 17:02

Lightnix said:
If anyone wants to correct my maths, feel free.

Trouble is maybe they used double density chips, or more could be mounted on the back. Not forgetting that a 256bit or 512bit card can use 512Mb of vram. I'd say its 256bit because its a G92 with 16 ROP's 4x64bit channels, going by their architecture 512bit would be 32 ROP's 8x64bit.

Nearly 200 Euro for the block, ouch.

Flanno · 4 Mar 2008 at 17:09

Quick question...a bit off topic.
But I am running 2 OCUK GTX's in SLI on a 780i mobo right now (and thinking of adding a third for tri-sli since I got my 30" screen). They don't clock past 630/1000, and 1530 shaders. This I assume will run quieter, cooler and faster then a 9800GX2 ? Upgrading is a waste of cash..correct ?

Skeeter · 4 Mar 2008 at 17:30

To all those arguing about the 'engineering feat' of 2 PCB vs one... im sure im near correct in thinking that its just a case of plugging either a 12"x4" PCB area or 2x 12"x4" PCB area into a CAD program, entering all the componenets that need to go on, and clicking the 'Make PCB Layout to fit all this' button.... which probublly takes the software about 2 seconds....

Being able to re-arange things to make them fit in a smaller space isnt exactly something to shout about?

Flanno said:
Upgrading is a waste of cash..correct ?

If it does turn out to be just 2 8800GTS's in a single card for £400 then this statement makes sence for any situation

helmutcheese · 4 Mar 2008 at 17:32

3dFX did it in 1998 with the Vodoo not 2000 so were before ATI.

Either way the comments about "Feats of Engineering" etc dont seem to mean much as I would buy neither of these cards.

Bonjour · 4 Mar 2008 at 17:36

helmutcheese said:
3dFX did it in 1998 with the Vodoo not 2000 so were before ATI.

Either way the comments about "Feats of Engineering" etc dont seem to mean much as I would buy neither of these cards.

The 3dfx solution was two PCI cards SLI'd together with a cable.

The Rage Fury Maxx was the first single card SLI-style solution AFAIK.

helmutcheese · 4 Mar 2008 at 17:40

Yes 1 range of card was indeed SLI and the other was 2 Cores, Google and WiKi show both.

Before my PC use but just posted orginally as comments above making out the current ATI X2 was 1st when its been done in recent years with 6000 series (I knew this) and 7000 series (new to me).

Bizarre · 4 Mar 2008 at 17:49

Remember when Nvidia pimped their 7950gx2 as having a 512 bit bus when infact we all knew it was just 2x256.

http://www.nvidia.co.uk/page/geforce_7950.html

bluestreak101 · 4 Mar 2008 at 17:52

Who really gives a monkeys about who did it first, the fact is if we all had cash to burn, we'd probably buy one for sh*** and giggles anyway...

It looks sexy as hell and is probably powerful enough to run anything we can throw at it.

Willis · 4 Mar 2008 at 17:54

bluestreak101 said:
Who really gives a monkeys about who did it first, the fact is if we all had cash to burn, we'd probably buy one for sh*** and giggles anyway...

It looks sexy as hell and is probably powerful enough to run anything we can throw at it.

here here

ten80p · 4 Mar 2008 at 17:59

Im curious as to why some think it's important as to which implementation is the best or correct way of doing it from a technical point of view
Surely once you've installed the card in your rig, the only factor you'd be interested in is the performance, not what it looks like or how the cards been designed/built etc

I can see why things like sound and heat are important to some, and I understand the width of the cards are also important in terms of single/dual slot design and space within your case but all talk about ATI 2 cores on one PCB as opposed to Nvidia 2 cards welded together..........why is it relevant ?

Final8y · 4 Mar 2008 at 18:10

ten80p said:
Im curious as to why some think it's important as to which implementation is the best or correct way of doing it from a technical point of view
Surely once you've installed the card in your rig, the only factor you'd be interested in is the performance, not what it looks like or how the cards been designed/built etc

I can see why things like sound and heat are important to some, and I understand the width of the cards are also important in terms of single/dual slot design and space within your case but all talk about ATI 2 cores on one PCB as opposed to Nvidia 2 cards welded together..........why is it relevant ?

I can use Maze 5 blocks on ATI cards that i know of from 1800xt upto the 3870x2 & what a cost saving.

helmutcheese · 4 Mar 2008 at 18:13

Opps, thats 9800GTX not X2. sorry

IGNORE