Weitere nv30 Infos vom Reaktor

5 August - Nvidia clarifies the situation with NV30: timeframes, architecture, performance estimates


On one of the recent conferences, Nvidia`s CEO, Jen-Hsen Huang, told analysts that the upcoming graphics processor code-named NV30 is not yet taped-out. Most of the industry observers started at once claim that NV30 will be late as it is not ready. We decided to ask our sources at the company about what is happening and, in fact, they told us some hot`n`juicy details about the current condition of NV30 and its architecture. They even were kind enough to whisper us some performance estimates.

The tape-out of this Cinematic Shading GPU was made back in May. But we should understand that developing so complex processors is extremely hard and there may be some erratas with the real chip. Modern GPUs consist of a number of blocks that are put together at the final stage of the design. The problem with the final stage is that there are used a lot of applications and emulators. Developers have to write some additional scripts so that different programs could work in collaboration. Sometimes those patches are a bit incorrect and the silicon version functions not the way it should have been. It happed with 3dfx`s Napalm and now it seems to detach Nvidia a bit. We should figure out that the latter has a very big designer`s team and we believe that an ordinary errata cannot set the GPU aside for three or four months some sources claim.

According to the original schedule that was observed by us earlier this year, NV30 was planned to hit the shops in October, starting volume production of the boards in September and sampling this month. As the final design is not yet taped-out and approved, according to the CEO himself, we believe that the sampling will begin a couple of weeks later and it will cause the boards to be out a bit later. Provided that there are no problems with TSMC`s 0.13 microns fabs, Nvidia has almost all the chances to meet its own schedule. However, Mr. Huang said that the boards will be available by Holiday, it means Christmas. Is it official change in Nvidia`s plans?

The architecture of NV30 in general is more flexible than that of RADEON 9700. Basically, NV30 ideology represents the approach of Renderman and certain other algorithms used in cinematography. Although NV30 cannot really replace Renderman, it can make the life of game-developers, 3D-artists and directors significantly easier. NV30 and Cg support large number of syntactical constructions, including loops and branching. The total number of vertex instructions in 65536 what is a lot and also NV30 supports 16 or 32-bit floating point components.

Frankly speaking, there is too lot to discuss about the newcomer and its ability to support all those functions in real-life, so, we will leave it for a separate article with analysis. However, our friends also told us some real estimates of NV30 raw-performance which we are going to share with you:
In Quake III Arena in 1280x1024 with 4x FSAA enabled, NV30 is going to be 2.5 times faster than the GeForce4 Ti4600.
In The Next Doom the board based on NV30 will be able to show 3.5 times or or even more of the performance the current Nvidia`s flagman has to offer us there.
NV30 will score three times more than the GeForce4 Ti4600 in 3D Mark 2001.
Effective HQ Pixel Fillrate (2x anisotropic filtering enabled) of the newcomer with will be about 2.7 times more than that of the fastest NV25.
As for pixel-shading speed, it will be 4 times of the NV25.
Nvidia claims that its upcoming GPU is capable of processing 200 millions triangles per second.
We have no idea if these estimates are accurate as there are no details about them and how they were achieved.

What you really-really want are the actual specifications of the GPU and the boards based on it. In fact, we have already published some interesting things a number of weeks ago in our news-story called “Another blast from the future: NV30”. We were correct in almost all our predictions except the memory speed. Nvidia wants it to be about 1 GHz delivering amazing 48 GB/s bandwidth when accompanied by the 3rd generation of their LightSpeed Memory Architecture. We are not sure that Samsung will be able to deliver them 1 GHz DDR-II memory by September.

We have just presented some brief ideas about NV30. We can believe Nvidia claiming that NV30 based products will appear by October, we can believe analysts and observers who say that Cinematic Shading GPU will not be on the shelves until Christmas or even January-February. Basically, those, who know when NV30 will definitely be available, will not tell you and us anything, those, who do not, will present you the launch date and time with the venue NV30 will be announced in. Also we have some doubts about Renderman ideology on PC just now, but that is another story...

_______________________-

hmm, interessant