Geforce 4 on the spot

LIKE MOST OTHERS IN THIS BIZ we're always a-browsing on the Web searching for the interesting thoughts of our colleagues who are also fighting this hard battle of news survival.

If you publish something that's controversial, it might affect the share price of firms and then all hell can break loose.

We were interested to see news on NV News concerning the NV25 chip and that caused us to think a little.

According to this, information we can expect six pixel pipelines that may bring some performance gain in the four-pixel pipelines market.

This is certainly the way future graphics technology is going - more and more pipelines. Perhaps eight would be good but this may just be too much for the current technology.

Nvnews also said that the chip will be manufactured in 0.13µ (micron) technology which is the logical step forward.

But that depends on the people in the foundries and if, as Eva Glass reported over the weekend, there are problems with that process at TSMC, then Nvidia must cast around in a different direction.

Certainly, Nvidias need to shrink technology. The 57 million trannies in the Geforce 3 TI make it really hot, and needs a lot of cooling.

As for the additional vertex unit we reported months ago and for which we were accused of rumour-mongering, the news that it is now becoming reality comes as no surprise to me.

All you have to do is think about NV2A or the Xbox chip that has two vertex units and it will not be hard for Nvidia to make a similar marchitecture available on the PC platform too.

One possible contradiction is that even the one-vertex unit is not exactly overused, so what need for a second one? It will again take at least six months from its introduction until we see some serious usage of these nice features.

The 300 MHz core and 600-660 MHz memory is not a surprise since even Columbia from Via was suppose to run this frequency, but as we've said, it will be late. Even with this memory bandwidth, Nvidia is likely to fall behind ATi which uses better optimisation on the Z buffer. It's the difference between brute force and smart technology, I think.

One of the things that doesn't seem logical to me, but which NV News says is "that we can expect Voodoo5 5500-esque Anti-Aliasing feature. The presumption is that the NV25 will bring a Rotated-Grid AA implementation to the table."

After many interviews with Nvidia, I learned that NV30 will be the first chip that will have ex-3DFX engineers involved.

So it would be irrational to expect Voodoo technology in this chip, as it is not so easy to team up engineers and make them work together. Developing chips takes a long time and the 3DFX acquisition was just a year ago.

Nvidia knows that I already asked for TwinView high end cards with TI generation since we know that TI is not NV17 or NV25 it is just NV20 ultra. This move is logical since ATi has both high end and mainstream cards with this option. This can be considered as the biggest advantage of the ATi Radeon 8500.

Last but not least is the statement that Nvidia will support hardware iDCT. This is one of the things where Nvidia is still lagging behind ATi.

Nvidia will again be the fastest player in town and they will be on the market before ATi, but maybe the latter will surprise us with the R300. This is a tough market.

Truform is a nice feature but we don't see Nvidia going in that direction. This, I believe, is a matter of face and I still do see any use or even announcements that some developers will use Nvidia implementation on the higher order surfaces called polynomial surfaces. Some improvements in T&L unit will appear but still there is one question I have.

IS NV25, GeForce 4, or whatever it's going to be called, DirectX 8, 8.1 or DirectX 9 hardware? I hope to answer that soon. µ