NVIDIA GF100 & Friends speculation

Let me guess:

Crysis Warhead (DX10) @ 1920x1200

GF100: 134 fps
HD 5970: 50 fps

 

Your 1tri/clock theory plus trisetup running at hot clock? Please don't hurt me it says speculation up there in the title (oh and where's Arun when you need him?)

I don't know if it can be relevant to GF100, but G80 was capable of 4 pixels/clock address and setup and 8 INT8 bilerps/clock.

 

Please tell me the above is tongue-in-cheek. Please.

Ailuros: you don't want to know what Arun believes about Fermi's setup rate...really!

 

"Delivery", "9", and "3".

Spoiler

The topic of the graph was how many months behind schedule the product launch is.

 

You're such a fun spoiler

 

I would say that those numbers are fake until proven otherwise. Here is where I got them from:

http://www.youtube.com/watch?v=xXETJLWTeV8

Edit: The HD 5970 number actually seems plausible based on this:

http://www.anandtech.com/video/showdoc.aspx?i=3679&p=8

Now, could GF100 be >4x faster than GTX 285 on this benchmark? I would doubt it, but if true, holy hell

 

Yeahrightsureok and it makes the GF100 somewhat over 3x times faster than a GTX295 based on those results..... That silly video you're refering to uses Rys' GF100 block diagram. Don't you think that NVIDIA would have their own diagram?

 

Only 3x GTX295? I'm kind of disappointed. *SCNR*

 

http://www.digitimes.com/news/a20100114PD202.html


Nvidia is expected to see its share of the global discrete graphics chip market drop from 65% in 2009 to 60% or even lower due to strong competition from AMD, according to sources from graphics card makers.

...

Although Nvidia plans to launch its 40nm Fermi-GF100 graphics chip in March 2010, mass shipments are unlikely to start until April, the sources noted. Nvidia responded saying its launch schedule remains unchanged.

...

 

That's expected. With competition's products already out, a delay will always cause a drop in market share.

 

While I'd love to have a card with that kind of headroom, the chances of such an ameteurish slide coming from big green are zero.

Next leak please.

 

GF100 isn't all that relevant to marketshare, since this metric is driven by low-end and mainstream parts. I'm not sure when Fermi derivatives meant to address these segments are expected, but for now (well, starting early February, at least) I can't really think of an Nvidia card worth buying at any price point. So Digitimes's report isn't very surprising.

 

At least a convincing enthusiast part - if any - provides a nice halo effect for lower-end solutions.

 

Which is exactly what I said isn't it ?...

The delay I mentioned involves all parts, high/mid/low end, which is what the competition has already out (or will have soon...I think only Cedar is missing) even before Fermi is out.

 

That's right. It doesn't capture "marketshare", but rather "mindshare" of "who has the fastest" on the market.

 

Less is more™

 

"NVIDIA GF100 pre-release speculation"

My bets for performance:
best case scenario: GTX 285 x 1.8 (1400m transistors vs. 3000m + small bottleneck)
worst case scenario: GTX 285 x 1.3 (shit happened)

 

Doesn't 48xx use 16 ROPs and yet competitive vs GTX28x with it's 32 ROPs. I mean less yet still competitive?

 

yep

It's just a shame they weren't planning on releasing it with Less™ though