Almost all the boxes here are Dell boxes. The exception being my personal custom PC:

Athlon XP 2100+
EPOX 8KHA+
3COM OfficeConnect SOHO/100
512MB Mushkin PC2100 CAS2.5 DDR
GeForce 3 Ti 500 64MB
SoundBlaster Audigy
WD200GB-ATA100
Maxtor 40GB-ATA100
Maxtor 20GB-ATA66
Yamaha-24x10x40-ATA33
…etc

Actually, you’ll note that he states that the precision mode that ATi offers is somewhere between the ones NVidia offers. He explicitly states that NVidia has the higher precision:

[q]To make it even more complicated, the exact
precision that ATI uses is in between the floating point precisions offered by Nvidia, so when Nvidia runs fragment programs, they are at a higher precision than ATI’s, which is some justification for the slower speed. Nvidia assures me that there is a lot of room for improving the fragment program performance with improved driver compiler technology.[/q]

Of course I knew that ATi was lower precision than NVidia a while ago. They only do 96-bit shader operations while NVidia supports 16, 32, 64, 128 bit precisions (I’m not sure about the 16bit one, or if there’s an 8bit one). So as far as quality goes, NVidia’s is better because of their TRUE full 128-bit floating point capability. ATi doesn’t have this *yet*. I sure hope they do on the R350.

Personally, as far as drivers go, I doubt we’ll be seeing any decent “open source” drivers for the 9×00 series. ATi has made it abundantly clear to certain individuals (regardless of that one unofficial posting by an ATi person on the DRI list) that they don’t feel the need to disclose “confidential” hardware information anymore since they have their own Linux drivers.

So no one can play the “open specs” advantage of ATi anymore. And currently NVidia has BSD, Win9X, Win2k, WinXP, MacOSX, and Linux drivers. That’s more than ATi can boast….

And as a BSD user I feel rather put out in the cold by ATi over that. Until they say otherwise, NVidia has my vote in $$$.

It sounds like from his post that when he uses the ARB code path that he doesn’t have control over the precision that’s used (which can give a BIG speed boost) whereas when using the NVidia specific optimized path he has full control and can get a big speed boost. In comparison, ATi has no way of controlling the precision and it’s high precision all the time (well, 96-bit for most things anyway).

The thing I was extremely disheartened to hear was the lack of blending operations in floating point modes, this is shocking to me really.

Anyway, overall the view of the card I was looking for from a developer standpoint. From my own personal development experience (which is far limited in comparison to his), his statements about NVidia’s drivers being somewhat better I can echo fully. ATi is getting there, but we’ll see what happens…

And don’t give me that old song and dance of “NVidia crashes my X11 like crazy!” I’m sick of hearing it, the office I work at is FILLED with NVidia cards (Nothing else, ok, there is one or two Radeon 7000 boxes but we don’t speak of them). I have never had a crash in X due to my drivers. I’ll take Carmack’s line on that and say: “There are many factors that can influence stability that are beyond the control of the driver” in so many words. In fact, I’m trying to remember when X last crashed…

Competition is always good for my pocketbook.