NVIDIA GF100 & Friends speculation

2ZEGW7.png
 
Gigabyte's ultra durable 2oz copper pcb

I hope you do know that 2oz copper is very common in high-end PCB's right?

And the x16 bridge chip looks like ye olde PLX chip which is branded accordingly (like on 6990.)

If the availability reports and Sampsa's preliminary benchmarks are anything to go by (SLI is faster than SLI on a stick) we might be in for an "ouch"
 
I hope you do know that 2oz copper is very common in high-end PCB's right?

And the x16 bridge chip looks like ye olde PLX chip which is branded accordingly (like on 6990.)

If the availability reports and Sampsa's preliminary benchmarks are anything to go by (SLI is faster than SLI on a stick) we might be in for an "ouch"

You mean the performance drop from having only 1 x16 wide PCIE bus is enough to hurt the performance enough to classify it as "slower than SLI"?

Anyway, I don't think it's the PLX chip, at least on AMD/ATI the "core" area of heatspreader is closer to chip borders than on that image.
 
You mean the performance drop from having only 1 x16 wide PCIE bus is enough to hurt the performance enough to classify it as "slower than SLI"?

Anyway, I don't think it's the PLX chip, at least on AMD/ATI the "core" area of heatspreader is closer to chip borders than on that image.

Every chip adds latency, no matter if it's PLX, NF200, Pericom or Hydra. that's why the most simple solutions are the fastest 2x16 > 2x8(Pericom) > 1x16+PLX

But this isn't the cause of the loss in those benchmarks. I actually thought they would beat the 6990 in almost every benchmark.
 
Last edited by a moderator:
Kinda doubt those are completely real, but if they are, kudos for the word 'Horsepower'. It is awesome.
 
They're speccing power consumption lower than the 6990? I'll believe that when I see it. Looks like a clean solution though and they'll probably sell all that they can (choose to) make.

Have you forgotten NV's "Average" power consumption already?


Kinda doubt those are completely real, but if they are, kudos for the word 'Horsepower'. It is awesome.
Yeah, at least they won't lose on horsepower this way ;)
 
Horsepower seems much more honest (or at least with no real reference) than abusing 'TDP' for something meaning average power.
 
Fx57 from DonanımHaber has received an info about one of the AIBs roadmap papers mention GTX 590 is "World's Fastest Single Graphic Card"..

http://translate.google.com/translate?hl=tr&sl=tr&tl=en&u=http%3A%2F%2Fwww.donanimhaber.com%2Ftm.asp%3Fm%3D47469003

For which it's enough that it's fastest in some specific test, for example batman: arkham asylum physx enabled and there's no contest, and they can print it's the fastest (as long as in very, very small print somewhere in slide that probably never even gets sent anywhere it's mentioned that it's referring only to this specific case)
 
For which it's enough that it's fastest in some specific test, for example batman: arkham asylum physx enabled and there's no contest, and they can print it's the fastest (as long as in very, very small print somewhere in slide that probably never even gets sent anywhere it's mentioned that it's referring only to this specific case)
They wouldn't need to resort to cheap tricks (physx) for that. Just take something like HAWX2 (with or without tesselation) and no doubt it will be faster.
Overall though I think it could be quite close.
Also, for TDP, imho due to low voltage (and clocks) it should draw a bit less than 2 x GTX 570, despite using fully enabled chips. So for games, not exceeding 375W seems doable (2 x GTX 570 would be around 400W). Quite sure though for unthrottled furmark this value would be exceeded (2 x GTX 570 is around 500W there - so maybe around 450W).
 
If they release a dual card AND it's still slower in most cases than 6990 that seems like an epic fail. An almost why bother releasing it fail.

I suppose, it gives them a single slot solution that wins more benchmarks than any previous single slot solution they had, or something, if this indeed turns out true.

Anyways as an aside I have noticed AMD is really turning up the heat with driver performance improvements and I think that's a reason 6990 benched as such a powerhouse.
 
Last edited by a moderator:
They should cancel a product because the competition released something faster two weeks earlier? Are you being serious?
 
People who purchase these cards should also be quite educated on overclocking. You can raise clocks and voltages atleast on Evgas cards without voiding the warranty. GTX 590 should have great OC headroom, if the 60x mhz clocks are true.
 
:runaway:

This is based on what, exactly?

If it head great headroom, wouldn't they just go fro 792MHz per GPU and be done with it?

The power consumption will be through the roof with those clocks and voltages that are needed to reach them, but the cores should be cabable to do about the same as the ones in GTX 580 when given enough juice. It is castrated card at stock settings to keep within "reasonable" power consumption, but its user enabled nVidias "AUSUM equivalent" should provide a huge boost.

Look how much headroom GTX 460 had for example, why didn't it launch with 750-800mhz? AMDs cards are clocked closer to their limits at stock. Pretty much all GTX 5xx gpus will do atleast 850mhz at 1.1v for the core, and you can set that voltage through software without any bios changes, I don't think these will be any different, actually I think they have chosen pretty good chips for these limited number cards.

edit2

by this "nVidias "AUSUM equivalent"" I'm not saying that it'll have a switch or anything, but just regular OC by the end user, but who knows maybe it'll have a switch...Also I'm not saying that the OC headroom will be awesome, if the stock voltages are low, but once the voltages are cranked up to regular GTX 5xx levels, it should fly.

Also based only on my gut :smile:
 
Last edited by a moderator:
Back
Top