NVIDIA Corporation on Wednesday finally rolled out its new highly-acclaimed graphics processor formerly-known under the code-name NV40. The new chip that got GeForce 6800 brand-name will again bring the boundaries of programmability forward and will deliver performance not achievable by the previous generations of graphics products.
The high-end of the new GeForce 6-series line will be the chip called GeForce 6800 Ultra operating at 400MHz and containing approximately 220 million of transistors that form 16 pixel pipelines, 6 vertex processors, a new programmable video processor as well as other important integrated circuits of the chip. The performance-mainstream version of the GeForce 6-series will be called GeForce 6800 and will have only 12 pixel pipelines, according to some reports.

The GeForce 6800 Ultra will feature 256MB of GDDR3 memory operating at 1100MHz and will retail for $499. The GeForce 6800 is expected to have 128MB of DDR memory onboard and retail for $299.
NVIDIA makes the GeForce 6800 Ultra processors in IBM’s foundry in East-Fishkill, New York.

“As Doom III development winds to a close, my work has turned to development of the next generation rendering technology. The NV40 is my platform of choice due to its support of very long fragment programs, generalized floating point blending and filtering, and the extremely high performance,” said John Carmack, president and technical director of id Software.
The GeForce 6-series graphics processors from NVIDIA feature some improvements in the programmability of their pixel shader and vertex shader engines. Featuring some additional capabilities, the chips correspond to Microsoft’s DirectX Shader Model 3.0 that evolves from the Model 2.0 available on GeForce FX as well as RADEON 9500 and higher graphics chips.
For Carmack, the graphics rendering development process takes time. He first spoke about the Doom III back in 1999 saying that the original GeForce 256 was his platform of choice. The Doom III game requires a GeForce FX or a GeForce 6-series graphics card to demonstrate all its glory, according to performance estimations and Carmack himself.
X-bit labs’ performance tests show that NVIDIA GeForce 6800 Ultra is about two times faster than the GeForce FX 5950 Ultra and RADEON 9800 XT in benchmarks that require powerful pixel shader calculations or with “eye-candy” features, such as full-scene antialiasing and anisotropic filtering enabled.
Numerous PC-makers are expected to integrate the new GeForce 6800 and GeForce 6800 Ultra graphics cards into their personal computers coming out this Spring and Summer. Add-in graphics cards makers, such as Aopen, ASUS Computer, BFG Technologies, Chaintech Computer, eVGA.com, Gainward, Gigabyte Technology, Leadtek Research, MSI Computer Corporation, Palit Microsystems, PNY Technologies and XFX. Retail graphics cards based on the GeForce 6800 models are slated for release in the next 45 days – sometime in late May or early June.
To learn more information about the architecture of the NVIDIA GeForce 6800 Ultra and its performance check out our article entitled “NVIDIA GeForce 6800 Ultra and GeForce 6800: NV40 Enters the Scene”.
Comments currently:
3
Discussion started: 04/15/04 04:27:39 AM
Latest comment: 04/16/04 04:23:23 AM
Expand all threads |
Collapse all threads
[1-2]
1.
A certain win for NVIDIA.
The power requirements are set too high, but in such high price cards, i think it's no problem to pay more for a good PSU or to pay the electricity bill at the end of each month.
If ATI chip requires less power and have similar performance ATI wins again.
The Geforce fx5xxx was a complete failure, while NVIDIA have those nice "paper specs" or marketing specs like I like to call it (30Gb/s, 475/500Mhz clock), the chip never made good use of the bandwidth available.
I am starting to think that this is all about memory bandwidth and the use you give to it and the tech behind to save it. ATI proved that with their mediocre 300Mhz chips vs NVIDIA 500Mhz, and ATI win. AMD and Intel have the same issue, who can make better use of the memory bandwidth available? We all know that this isn’t about MHz. Intel must stop fooling us and they will with the model number think at least I hope that. I'm starting to think that chips aren’t making good use of the memory available, specially the dual channel ones.
ATI should start thinking now what to do, since their position is good, but could be better if weren’t HL2. I think they should process these guys since they have lost a lot of market, this was the "game" that was going to show the huge difference between one architecture and the other, or you could say the game that could have killed NVIDIA Geforce fx5xxx chips. But now NVIDIA is now superior so if the game comes out now NVIDIA already have a chip that plays with it good to.
FSAA was a bit disappointing since I was expecting better image quality than ATI and it’s not the case, games that doesn’t work with FSAA like HALO, Splitter cell I was also expecting that they could have some way of made it possible, but no, and it’s still only 4X FSAA since a card with that speed I was think 6X and 8X was already possible. ATI have already 6X and ATI 4X is superior to NVIDIA 4X I don’t know what FSAA ATI will have but if maintain the 6X and add the 8X even with a “nice” performance hit ATI should always have this advantage and if it bring new image quality levels then we certainly will have a race at least at new levels because NVIDIA FX5xxx was far from being a competitor to ATI in all segments.
[
Posted by: I

|
Date: 04/15/04 04:27:39 AM]
2.
This test is inconclusive.
If NV40 did NOT beat the Radeon9800XT, then this would have been truely pathetic.
At least now we know that this card is NOT a dud.
However, it's too early to draw any conclusions. Wait until the RV420 comes out before you decide what to get.
[
Posted by: Dan

|
Date: 04/15/04 02:38:06 PM]
+ expand thread (1 answer)
- collapse thread
Well it's not a dud, but i bet there are still the same optimizations that have pushed the 5xxx series scores up or if you want the scores are high (double)because besides the double all (vertex/tmu/etc...) there are other optimizations.
ATI should have put otimizations when nvidia started to put them to in their drivers, instead of forget them (removed), and used some kind of enable/disable option.
Some one said that ATI is going down because the R420 is based on the R3xx and that's bad? So what if they double all like nvidia did i bet that we will have at least a 2x performance increase to.
Where is the 128bit image quality in here?
http://www.tomshardware.com/graphic/20040414/geforce_6800- 46.html
I never see a proff in any game that nvidia 128bit shader quality is better then the Ati 96bits.
I think Nvidia is still using 12 and 16 bit precision, dont forged to add that to the performance numbers.
[
Posted by: I

|
Date: 04/16/04 04:23:23 AM]
[1-2]