NVIDIA GF100 & Friends speculation

[Vincent]

NV won't drop CUDA. It will just get deprecated or de-emphasized. They have to support the legacy.

DK

Without mass market share as the compensation for research and development, Nvidia is doomed.

 

[KimB]

Without mass market share as the compensation for research and development, Nvidia is doomed.

What are you talking about?

 

[Jawed]

Gtx480m

http://www.nvidia.com/object/product-geforce-gtx-480m-us.html

 

[no-X]

352 SPs at 850 MHz? Quite conservative approach...

 

[Jawed]

It states 897 GFLOPS. That 850MHz is just the base clock.

 

[Psycho]

No, but suddently they multiply by 3 again (850*352*3= 897M)... I thought it could only do 2 ops (ie one MAD) per clock now, but maybe marketing doesn't want to know.
Baseclock is 425: http://www.tcmagazine.com/tcm/news/hardware/28292/nvidia-reveals-its-first-directx-11-mobile-gpu

 

[Jawed]

Damn I didn't realise how stupid a base clock of 850MHz is when I posted earlier

 

[Vincent]

What are you talking about?

The current condition seems to be more complicated for Nvidia to design chips with features targeted for either Quadro/Tesla or Geforce product lines. I believe that all GF10X series will be remodified architecture based on GF100. and most of them ( except low-end ) may contain different streaming multiprocessor (SM) architecture as compared to old GF100 series.

 

[Vincent]

GTX465 Review from Xbitlabs

http://www.xbitlabs.com/articles/video/display/geforce-gtx-465.html

 

[BRiT]

GTX465 Review from Xbitlabs

Already linked from in this earlier post: http://forum.beyond3d.com/showpost.php?p=1433735&postcount=5834

 

[Tchock]

Hmm, the GTX480 M is a 104 or 100?

If it's an 104 then it indeed has 384 ALUs at full capacity, and 352 being the top bin. And the 465 might be here to stay with the GF104 slotting in somehow (either that or as the 460)

EDIT: Wow, TechReport says it's a GF100.

 

[Alexko]

GF104 is supposed to have 256 SPs, not 384...

 

[Tchock]

GF104 is supposed to have 256 SPs, not 384...

Definitely sure? Thought 384 was actually on the line, especially with a size of Cypress, a much smaller memory interface, and "cut cache" (ugh).

 

[Alexko]

Definitely sure? Thought 384 was actually on the line, especially with a size of Cypress, a much smaller memory interface, and "cut cache" (ugh).

Not definitely, no, but most rumors point to that figure. Plus, the GTX 465 is GF100-based, and has 352 SPs, this much we know. And it seems very likely that the GTX 460 is GF104-based, so it should have less SPs than the GTX 465, especially since it's probably clocked higher.

 

[Jawed]

Nvidia has 'absolutely no plans' to kill the GeForce GTX 470

Wow, kitguru actually has some use, by stirring the pot:

http://www.techreport.com/discussions.x/18975

"We feel that there's some room below [the GTX 470 and GTX 480] for a reasonably priced version of our GF100 [graphics processor]."

NVidia goes on record "pre-announcing" another card based on GF100

 

[mczak]

EDIT: Wow, TechReport says it's a GF100.

Yeah, nvidia must be smoking serious
Though with those clocks (gddr5 with a base clock of 600Mhz!!!) I'm not sure how much faster this actually is compared to their current fastest offering (gtx285m). I would guess probably not that much...

 

[Mindfury]

http://translate.google.com.hk/tran...ttp://we.pcinlife.com/thread-1431238-1-1.html

GTX460 fan design is too fragile Actually a small fan protruding outside of the radiator cover, and radiator cover easily deformation of material, accidentally touched the little card, run the fan on the lead contact with metal heat sinks, to interrupt a fan Now turn the fans up there to fly out feeling great centrifugal force

Fortunately, not heat, I do not know that the fan in question out of power is also measured whether the Right

 

[Mindfury]

GTX460(GF104) PCB layout:

http://we.pcinlife.com/thread-1431292-1-1.html

 

[Kaotik]

Is it just me, or is the GPU-package rectangle opposed to square ones we're used to?

 

[Vincent]

GTX460(GF104) PCB layout:

http://we.pcinlife.com/thread-1431292-1-1.html

If this leaked picture is true beyond doubt, the die size of GF104 could be over 330mm^2.