GeForce GTX295 Card Exposed

[jasonelmore]

It should put up some really nice folding numbers.

[mikeyakame]

OBR.

If time permits would you mind checking the memory IC part numbers on the 295.

Last I recall NV were using Samsung ICs which if correct NV using a 1242Mhz clock, would probably be using K4J52324QH-HJ08 which is a 1200MHz 0.8ns 2.05v bin. If so they won't have all that much headroom for overclocking.
Otherwise the only other alternative would be K4J52324QH-HJ7A, 1300MHz 0.77ns 2.05v bin. Curiosity has the best of me

Just a small request if permits.

Thanks in advance.

[jas420221]

But in SLI with the 280's, one card's memory and bandwith is used and the other card's GPU is Processing also, but not memory..

Also, for all intents and purposes, the 9800gx2 IS SLI. There is a ribbon that connects the two, just like SLI. You can also disable SLI in the control panel with it.

Now dont misunderstand me, I dont hate this implementation, but lets call a Spade a Spade...

[NapalmV5]

No there's no new tech. You were basicly saying all this card is was 2 280's in SLI on one card. But in SLI with the 280's, one card's memory and bandwith is used and the other card's GPU is Processing also, but not memory. With this solution (295) the bandwidth, memory, and GPU is all shared.

http://www.techpowerup.com/reviews/Z...front_full.jpg

[Blacky]

this card will make GT300 card look bad just as the 9800GX2 did for GT200 at launch :/




soo what is this? less specification than previous announced? is this fake or not its really getting confusing. Can someone confirm this (points at OBR )?

http://forums.vr-zone.com/showthread.php?t=366972

[jas420221]

If those specs are true, the 295 is nothing but 2 GTX260 216s with a die shrink slammed together...

You can use the EDIT feature instead of double posting.

[jasonelmore]

If those specs are true, the 295 is nothing but 2 GTX260 216s with a die shrink slammed together...

You can use the EDIT feature instead of double posting.

Those specs are not right. It has already been confirmed that it will definately have 240 shaders X 2. not 216. and from what i'm hearing the stock clock's are not right either.

[jas420221]

So then its a GTX280 with a die shrink slammed together...

Can you please elaborate on your point about the 2nd cards' memory not being used in SLI... I know its not shared, but AFAIK, it more certainly is USED.

[[XC] Synthetickiller]

well.... i wont step up if it aint worth it lets see what the numbers are before we jump to conclusions but im more than happy with my gtx260-216 also depends very much on price to. when these cards out ?

Which card do you plan on getting if you do upgrade?

[Blacky]

Those specs are not right. It has already been confirmed that it will definately have 240 shaders X 2. not 216. and from what i'm hearing the stock clock's are not right either.

where ?

[trinibwoy]

So then its a GTX280 with a die shrink slammed together...

Oh no! You mean it's just like every other dual-GPU card ever made?

[jas420221]

Oh no! You mean it's just like every other dual-GPU card ever made?

From Nvidia...yes.

[jasonelmore]

So then its a GTX280 with a die shrink slammed together...

Can you please elaborate on your point about the 2nd cards' memory not being used in SLI... I know its not shared, but AFAIK, it more certainly is USED.

No not quite. It uses the 260's architechture. It still uses a 448 Bus instead of a 280's 512 interface. Bandwidth and the amount of memory is relative to 2X260's thats why you see the memory at 1.792GB instead of a solid 2GB.
BUT.. and this is a big but.. It has 240 Shader Processors per core instead of 192 or 216.

What i mean was the GPU's shared memory and bandwidth on both cards was scaled through the PCI Express Slot and not the SLI connector. Just the GPU is scaled through the SLI connector. Now that both cards are on one single card they are scaled togather through one single bus.

Ya got all of that trem?

[jas420221]

Thanks for the explanation...you should type what you mean next time.

Trem... LOL

I sure do jasonelmore, unberninja, Proamplitude, and w/e other names you used over at EOCF via Russian Proxy after you were permanently banned for being a complete idiot...

[jasonelmore]

Yeah Tremglox formally from EOCF, you can drop the act. Rev busted you out.

[trinibwoy]

From Nvidia...yes.

Nope, not just Nvidia. There is no theoretical or measured benefit to AMD's approach. It still takes up two slots and is still based on crossfire and still draws boatloads of power. It's just as slammed together as anything Nvidia ever did with their GX2's. Convincing yourself otherwise is just delusional.

[jas420221]

If you insist Trini...

Once sideport is needed and enabled, then it will have its architectural advantages, yes? I know that has little to do with 2 vs 1pcb but...

[ToTTenTranz]

Doesn't make any sense to launch another high-end single gpu so soon. With GTX295, the GT300 would be kind of useless.. Especially in that timeframe.

Then again, nVidia already made that mistake, with GTX280..

[LowRun]

Nope, not just Nvidia. There is no theoretical or measured benefit to AMD's approach. It still takes up two slots and is still based on crossfire and still draws boatloads of power. It's just as slammed together as anything Nvidia ever did with their GX2's. Convincing yourself otherwise is just delusional.

After market cooling solutions is almost a no go with GX2 design, seems like a really good reason to prefer the AMD's approach to me and nothing delusional at all

[trinibwoy]

After market cooling solutions is almost a no go with GX2 design, seems like a really good reason to prefer the AMD's approach to me and nothing delusional at all

Yeah you're right, if you're looking at after-market cooling solutions the single-PCB design will be easier to work with. But we were talking about the relative technical (in)elegance of the two approaches. Neither provides any tangible advantage to scalability or performance of multi-GPU.

And jas, sideport was promising on the X2 but it turned out just to be another high-bandwidth link. And when they realized that PCIE 2.0 provided enough bandwidth already sideport became redundant and was disabled. There needs to be more integration before multi-GPU moves to the next level. It's a moving target though since on-die bandwidth and processing needs will continue to outstrip any sort of inter-die communication or data-sharing approach.

[DeanZ]

dont know if anyone saw this yet from expreview

http://en.expreview.com/2008/12/16/g...html#more-1646

shows some real potential

[jas420221]

And jas, sideport was promising on the X2 but it turned out just to be another high-bandwidth link. And when they realized that PCIE 2.0 provided enough bandwidth already sideport became redundant and was disabled. There needs to be more integration before multi-GPU moves to the next level. It's a moving target though since on-die bandwidth and processing needs will continue to outstrip any sort of inter-die communication or data-sharing approach.

Tell me something I dont know..

[dan7777]

dont know if anyone saw this yet from expreview

http://en.expreview.com/2008/12/16/g...html#more-1646

shows some real potential

performce looking good all depends now on price realy.

[Sampsa]

review coming tomorrow 18th of December?

http://www.guru3d.com/news/tomorrow-18-december-/

[Bo_Fox]

yeah baby

i like it, a lot.

ROFL!!!