NVIDIA GF100 & Friends speculation

[Kaotik]

Aren't all those specs on the Newegg site as well.. where you can actually order them?

Not to mention the XBit Labs review from nearly a week ago
http://www.xbitlabs.com/articles/video/display/geforce-gtx-465.html

 

[Ailuros]

Sure, but try taking that desktop with you on your next trans-pacific business trip so you can kill down time with some gaming.

My business trips aren't that boring, that I need to waste my spare time in front of any sort of PC LOL

 

[CarstenS]

Since we're still on "prerelease" in this thread:
http://hothardware.com/News/Computex-2010-Dual-GPU-GTX-470-Videocard-Spotted/

Galaxy shows a massive GF100-card with two chips, 2x 8-pin connectors and 2x 256 Bit memory - so probably a Dual-GTX465?

 

[no-X]

The sticker says "2x GTX470 prototype"... but it looks like a dummy board.

 

[CarstenS]

Yeah, I've read the sticker too. But tell me - how will you populate a 320 Bit bus with only 8 32-bit memories? Of course it's a prototype and all, but if it's really that dum(b, my), they could've added the missing memory just for fun's sake - the board already has provisions for it.

2x8+PEG -> 375 watts, seems a little low for two chips with 215 watts each (including memory of course).

 

[no-X]

The other modules can be located at the PCB's back.

 

[CarstenS]

Then you'd have quite possibly 2x 256 Bit for each GPU. Assuming that there's not two single RAM chips on the back of the PCB while two solder pads in the front are not populated.

 

[Erinyes]

Well the real question is not if they can build it or supply power to it, but if they can actually cool it? Dissipating >300W of heat is not an easy task, and this card looks like its closer to 400W and probably over 400W if they use a full GTX 470

 

[LordEC911]

well the real question is not if they can build it or supply power to it, but if they can actually cool it? Dissipating >300w of heat is not an easy task, and this card looks like its closer to 400w and probably over 400w if they use a full gtx 470

~350w tdp

 

[Mindfury]

GTX470m?

 

[Erinyes]

GTX470m?

Seems quite possible that its a GF 104 based part, where does it say GTX 470M? The GDDR5 clock speed does point to a mobile part though, even GTX 480 has a GDDR5 clock of only 2400 Mhz and the 1.5GB VRAM size fits with the 192 bit bus. Core and shader clocks are rather high at 675 mhz and 1350 mhz

 

[neliz]

GTX 460M on a MSI model:

http://www.engadget.com/2010/06/02/nvidia-shows-16-new-optimus-laptops-at-computex-teases-460m-gpu/

Link to pic.

 

[CarstenS]

~350w tdp

HD 5970's cooling solution is supposedly capable of getting rid of about 400 watts.

 

[Bouncing Zabaglione Bros.]

~350w tdp

Is that a Nvidia 350 watts, ie only under "normal conditions"? If you actually stress the hardware, will it be significantly more than that as per the other Fermi chips..?

 

[fellix]

 

[Unknown Soldier]

That GT240 is for PhysX, it was announced some time ago that one of the companies in this case Galaxy, might go down that route.

 

[Tchock]

Seems quite possible that its a GF 104 based part, where does it say GTX 470M? The GDDR5 clock speed does point to a mobile part though, even GTX 480 has a GDDR5 clock of only 2400 Mhz and the 1.5GB VRAM size fits with the 192 bit bus. Core and shader clocks are rather high at 675 mhz and 1350 mhz

N11E -GS

480M should be N11E-GTX, 470M - N11E-GT or GS, and 460M should be N11E-GS or N11P.

I see the 3 mentioned together, but that is rarely indicative of what market segments they are targeting...

So if 480M is indeed a GF100, impressive. GF104 seemed to have shrunk a lot compared to what they touted though, or we're going through this yields ordeal again.

 

[mczak]

N11E -GS

480M should be N11E-GTX, 470M - N11E-GT or GS, and 460M should be N11E-GS or N11P.

I see the 3 mentioned together, but that is rarely indicative of what market segments they are targeting...

So if 480M is indeed a GF100, impressive. GF104 seemed to have shrunk a lot compared to what they touted though, or we're going through this yields ordeal again.

I don't quite get it. With the suggested clocks, 480M only looks to have about 12% or so more shader power than 460M. That doesn't make a whole lot of sense, as the 470M (assuming that's a less disabled GF104 compared to 460M with similar clocks) would hence possibly beat the 480M.

 

[Tchock]

TTM?

Differences in the bandwidth though. Oh, and the "GPC"s. Tesselation perf and fillrate halved for 104s?

As is now, the chip displayed above should fit in the 50-60W TDP threshold.

 

[mczak]

TTM?

Ok that's a valid reason.

Differences in the bandwidth though. Oh, and the "GPC"s. Tesselation perf and fillrate halved for 104s?

Well, unless GF104 now only has 192bit bus suddenly, I can't see why a supposed 470M wouldn't have the same (in fact probably more) bandwidth than the 480M. Same goes for tesselation perf and fillrate - supposedly 480M has 3 active GPCs (and 32 ROPs) at a very low clock, even the 460M 2 GPC (presumably) could already be just as fast (true for the 24 ROPs too).

As is now, the chip displayed above should fit in the 50-60W TDP threshold.

I'm not sure how you arrived at that, but I guess anything fitting into the former do-not-exceed limit of 75W for mobile parts would be progress...