NVIDIA GF100 & Friends speculation

Maybe the Chinese were lucky in guessing correct numbers then, or actually have valid numbers. Doesn't make any difference. Heise did actually benchmark a card.

 

Disagree. The Unigine bench of 470 at 4xAA matched the leaked result.. It seems 470 would be on par with 5870 at 8xAA.

It is quite curious what is the driver used in the leaked bench. The driver sent out to AIC must be the beta version, as the latest Geforce driver 196.75 WHQL actually doesn't support the upcoming GF100. Is it possible that NV just let the beta driver hide the performance??

 

http://translate.google.com/transla...p://www.hardware-infos.com/news.php?news=3450

 

Where did they say that ?
The translation may be borked, but all I'm reading is that they saw it @ CeBIT in Hanover. They didn't actually test the card themselves.

Still, they say they saw it in action, so I guess that's different.

 

What do you disagree with？I'm confused....

 

What they hint to, but couldn't confirm, about the GTX 470 sitting in between the GTX 285 and GTX 295, is what I would expect (close to the GTX 295 in fact).

But if the numbers posted earlier are true, the GTX 470 is barely faster than the GTX 285...which is a big disappointment.

 

The result is 'genuine' based on the beta driver they had.

 

Looks like they had to slow down the gddr5 on the 470 to give the 480 an ample lead. I bet that 800mhz ram is specd at 5gbps

 

That sounds like wishful thinking. If they wanted to blindside ATi on launch day why are they hyping geometry and tessellation performance so much? Granted we are at a point where new features are more effective differentiators than raw FPS but intentionally hobbling early drivers doesn't make sense in the current scenario.

 

This is definately wrong, and means 800 (3200) instead of 1600 (6400), using 1600MHz chips would raise costs by at least a leg, if not arm and leg

 

Some people report the base clock (800) and some report the DDR clock (1600) and some report the effective data transfer (3200). It's not a big deal.

 

It's a big deal when it can create confusion.

 

I have higher expectations of my B3D brethren

 

Well, it would still fit within PCIe specs and thus get the sticker

And AMD at least kept their TDPs pretty much even with last gen, real consumption being even lower in many cases (comparing HD4800 to 5800, of course)

 

I remembered some nv PR was quite confident amd was still in the dark of searching the GF100 performance. Does they just pretend to be another hyping?

The final specification of both 470 and 480 haven't released to AIC yet. It is quite strange for a soon-to-released product line within 22 days. Remember what happened to rv770? One month is not enough time for amd to react efficiently even if they know the specification in advance. I think doing this may imply: 1) they struggle to balance the poor yield, clock and performance; or 2) they want to hide something. Surely I admit I have no idea for the production procedure of graphic card. My estimate may be wrong...

 

Is it really too hard to tell?

 

Yes...it's just hyping by your smile. But I can't understand why AIC still mouth tied with the clock of a short-to release product. It is just strange in the pre launch period.

 

The first gtx480,gtx470 cards will have pencil writen specs on retail boxes it seems

 

and don't forget to line across the ATI 5000 on the back...

 

Perhaps they don't want AMD to find out too much and spoil their launch with the release of an HD 5890...