True, but he seems pretty confident about some specifics both here and over at 3dc. As far as samples being in the hands of devs, it seems kind of early for that. Are samples usually out this far ahead of commercial launch (assuming said launch is in late Q3/early Q4 09)
Nvidia GT300 core: Speculation
- Thread starter Shtal
- Start date
- Status
True, but he seems pretty confident about some specifics both here and over at 3dc. As far as samples being in the hands of devs, it seems kind of early for that. Are samples usually out this far ahead of commercial launch (assuming said launch is in late Q3/early Q4 09)
Yeah, G80 samples were out in january 2006 (if memory serves right) also - albeit a bit cloaked to not spoil the unified-party too much
Interesting. It seems to have some serious shader power but the rest is only a modest upgrade from GT200. I wonder if the clock speeds will be much faster.
What i would like to know is if each of those hypothetical 384 scalar processors is merely a "copy" of the ones in GT200's/GT200b's (which, in turn, weren't that much different from G80's/G92's), or if it has significant additional capabilities that make 1:1 shader clock/processing potential difficult to compare against their older siblings.
Not merely in supporting DX11 (if there's support at all), but especially in any new methods to improve per-sp/per-clock current DX10/DX9 code processing efficiency.
So, no DX11 or even DX10.1, then ?
Mildly disappointing, but if performance and price are truly competitive, it doesn't matter much anyway.
Last edited by a moderator:
Right:
"[...]wird der Chip ab dem zweiten Quartal dieses Jahres mit 384 Shadereinheiten an den Start gehen, die zum GT200 unangetastet geblieben sind, sprich, es können pro Takt weiterhin zwei MADD und ein MUL berechnet werden. "
-> They say, GT212 will, as GT200, be capable of two (sic!) MADD plus one MUL per cycle per ALU.
Those specs make sense which means they're probably wrong
But f they're accurate and it really is only just under 300mm^2 with a 256-bit bus it sounds like Nvidia didn't do much to increase density.
For the people, who cannot speak german. Here the english version of the article:
http://www.hardware-infos.com/news.php?news=2629&sprache=1
http://www.hardware-infos.com/news.php?news=2629&sprache=1
256-bit MC means only 16 ROPs so what about AA performance in higher resolutions (1920x1200)? I think more probably it will have 512-bit MC. Moreover is there any possibility to make chip smaller than 300mm2 with increasing ALUs more than 50% and even increasing TMUs?
IF those specs are true then GT212 will be extremely fast (30-40% faster than GTX285 IMO) and not expensive to produce.
IF those specs are true then GT212 will be extremely fast (30-40% faster than GTX285 IMO) and not expensive to produce.
Not to mention that a "GT212 X2" could probably hold off the high-end competition from AMD, at least until "GT300" later this year.
This chip must not have 16 ROPs by 256-bit MC. Read the article I have posted:
"Because Nvidia's memory interface refers to the ROPs, these would be cut in half from 32 to 16, too. But we assume that Nvidia either installs eight instead of four grid units per ROP partition, or alternatively uses 32 instead of 64 Bit wide controllers, which would both steady the pixel performance of the chip."
Yes, there is a possibility to make chip smaller than 300mm2 with increasing ALUs more than 50%, because at GT200 240 SPs uses only 1/4 of the whole die size.
- Status