NVIDIA: Beyond G80...

http://www.vr-zone.com/?i=4690

We told you earlier about NVIDIA skipping G81 or GeForce 8900 to go straight for something more powerful. It seems like INQ has insider scoop that NVIDIA mulling the successor to G80 but still would like to produce GX2 if they managed to solve the thermal and power issues. After that it will be time for G9x flavour, codenamed G90. NVIDIA is planning to produce 65nm G9x chips based on G80 architecture with much higher frequencies and powered with GDDR4. If NVIDIA makes G80 smaller there may also be just enough space for 512Mbit controller. If all goes as planned, the chip is scheduled for second part of 2007.

Thats quite a jump there. 90nm to 65nm, 512mbit memory interface, GDDR4, could be quite a chip.
Although Vr-zones source was the INQ.
 
Some more news about the upcoming midrange G8x. (G86/G84)

http://www.vr-zone.com/?i=4704

According to our sources, there will not be any AGP versions for GeForce 8 series because simply it can't support. Therefore all DX10 G80, G84 and G86 cards will be PCI Express based. Also NVIDIA has not revealed the official marketing name for G84 and G86 yet but there will be 2 SKUs for G84 and one SKU for G86 at launch. G84GTS (G84-300) and G84GT (G84-400) are slated for launch in April while G86 (G86-300) will launch later in May. Interesting, G86-300 is 128-bit like the G84 while the rest of the G86 variants are 64-bit.
 
Has nVIDIA implemented xS hybrid modes back into the G80? Ive heard that it works using the latest Vista drivers from a few people, but im not too sure.

Why has nVIDIA decided to release the 8500 series? Im kind of confused at this launch.
 
Has nVIDIA implemented xS hybrid modes back into the G80? Ive heard that it works using the latest Vista drivers from a few people, but im not too sure.

http://www.vr-zone.com/?i=4727

w0mbat in our forums has found the following information in the latest Forceware drivers for Vista :
NVIDIA_G80.DEV_0191.1 = "NVIDIA GeForce 8800 GTX"
NVIDIA_G80.DEV_0193.1 = "NVIDIA GeForce 8800 GTS"
NVIDIA_G80.DEV_0194.1 = "NVIDIA GeForce 8800 Ultra"
NVIDIA_G80.DEV_019E.1 = "NVIDIA Quadro FX 4600"
NVIDIA_G84.DEV_0400.1 = "NVIDIA GeForce 8600 GTS"
NVIDIA_G84.DEV_0402.1 = "NVIDIA GeForce 8600 GT"
NVIDIA_G86.DEV_0421.1 = "NVIDIA GeForce 8500 GT"
NVIDIA_G86.DEV_0422.1 = "NVIDIA GeForce 8400 GS"
NVIDIA_G86.DEV_0423.1 = "NVIDIA GeForce 8300 GS"
It seems that NVIDIA is planning for a higher clocked GeForce 8800 Ultra and a cheaper GeForce 8400GS card.

Sounds like a higher OCed 8800GTX with samsung 1.0ns GDDR3 memory or GDDR4?

IMO this indicates that nVIDIA knows that they dont need a G80 refresh to fight against R600 when its launched.
 
IMO this indicates that nVIDIA knows that they dont need a G80 refresh to fight against R600 when its launched.

Or they just don't have one ;)

I was expecting them to introduce something between the 8800GTS and 8800GTX cause there's currently a big gap between those two. If the Ultra opens up a similiar gap on the GTX then it's going to be a very fast card.
 
Or they just don't have one ;)

I was expecting them to introduce something between the 8800GTS and 8800GTX cause there's currently a big gap between those two. If the Ultra opens up a similiar gap on the GTX then it's going to be a very fast card.

A GTX 384MB could be helpful there. The question might be do they have the yields to provide those chips at higher volumes. And it could be questionable if they'd be at a disadvantage in recent games with a 38468MB framebuffer performance-wise. Obviously they released GTS as it exists today for a reason, and that reason was yields. Interestingly, the rumours of AMD's XT is that it is just down clocked 10-15% (see HKEPC), not down clocked and disabled units too (GTS).
 
Last edited by a moderator:
Sounds like a higher OCed 8800GTX with samsung 1.0ns GDDR3 memory or GDDR4?

Has NV ever changed process without changing model number at the high end? Did the fabled 80nm G80 just recede into the mists of might-have-beens, or is this it?
 
A GTX 368MB could be helpful there. The question might be do they have the yields to provide those chips at higher volumes. And it could be questionable if they'd be at a disadvantage in recent games with a 368MB framebuffer performance-wise. Obviously they released GTS as it exists today for a reason, and that reason was yields. Interestingly, the rumours of AMD's XT is that it is just down clocked 10-15% (see HKEPC), not down clocked and disabled units too (GTS).

You mean 8800GTX 384?
 
Has NV ever changed process without changing model number at the high end? Did the fabled 80nm G80 just recede into the mists of might-have-beens, or is this it?


Its possible its not 80 nm, I was wondering when the Ultra moniker would come back
 
Okay, so this thread is now reserved to discuss NVIDIA's high-end roadmap, including the GeForce 8800 Ultra. There's a new thread for G84/G86.

As for the codename - I'd suspect if it's a 'true' optical shrink (same metal masks), then NVIDIA would seriously consider keeping the same chip codename. We'll see.
 
Interesting.
Contrary to the previous 8600/8300 vr-zone report, the 8500 GT appears to be based on the G86, not on a crippled G84, like they said it would be...
 
Interesting.
Contrary to the previous 8600/8300 vr-zone report, the 8500 GT appears to be based on the G86, not on a crippled G84, like they said it would be...
Might have been a good idea to check the post just above you to know where to post that, hehe ;)
 
Back
Top