As a designer of Weitek's VGA core, this is a very interesting read. I had no idea how valuable the core was to nVidia. As Weitek was going under, I also remember interviewing with 3dfx and thinking how arrogant they were. I'm not surprised they eventually lost
>Roach: ... we did make some efforts to get a VGA core. We actually licensed it from Unisys, who had it
internally, so they didn't mind licensing it to a merchant chip vendor, and we did that, and I think Barry
deserves credit for doing the best he could to make something out of all of this, and he gave it kind of one
more generation, one and a half generations, but the fundamental approach was still, coming from a high
cost core, trying to go down, whereas the other guys had it much better figured out from low cost to
coming up.
Whole panel is depressing to read. They had two strong products, x86 copro and SUN design win copro. Rest was losing money or not selling well. As soon as SUN moved on company started having serious problems. They were convinced Weitek was in high end graphics, but every Weitek P9000/9100 benchmark I can find puts them in the middle of common PC vendor pack losing to much cheaper chips/cards. 5186/5286 vga cores performed pretty terribly (8bit video ram access when combined with P9000?), and so did the one integrated into P9100 delivering half the performance of low end Trident PCI (not liking Doom 8bit writes pattern?). Stupid marketing gimmicks, feng shui, wasting money on bad calls like speech recognition and chasing high margin products in a market racing to commoditize everything. It looks like they might have lost best people early on with Edmund Sun starting C-Cube going into video acceleration, sold for >$2 Billion in 2000, and Chi-Shin Wang 8x8 doing conferencing and still holding on with ~2000 employees and almost a $1B revenue.
Weitek dominated PC 2D graphics accelerator performance for a while in the 90s, but they also suffered from 3dfx's problem of poor VGA compatibility. So they hired me to help design a new VGA core from scratch. Apparently, we did that well, but it was too late, since 3D came and they completely missed that boat because of financial problems preventing them from hiring 3D savvy talent.
They were also very well known for having the fastest math co-processors in the 386 PC days. The 486 killed that business though.
Another thing they made was a faster (2x speed) Sun Sparc processor that they sold successfully as an upgrade for a while.
The company's last profitable quarter was when they did the nVidia license deal.
Oh, and we invented unified memory architecture when we designed a PC chipset that included onboard 2d graphics and VGA. It was a cool product, but maybe ahead of its time. It was also hard to sell that not being Intel. The architect of that product went to nVidia
I was very convinced at that time that 3dfx didn't have a good roadmap and Nvidia would prevail based on their professionalism and superior ability to design silicon.
They already had what they needed from me :). I do think it's cool that every pixel they displayed for approximately the next decade went through my display pipeline
I remember in the interview mentioning they should also license Weitek's VGA core, but they said VGA was dying and they didn't want to dedicate silicon area to it. Besides they had an external chip to do it "for now" - i.e until it died. They were also very contemptuous of nVidia and completely convinced of their own superiority
Even crazier is that someone made an adapter that converts the TPM header found on some modern motherboards into an ISA slot because apparently that has all the necessary bus lines exposed. So that, you know, you could plug a sound blaster in there to play your DOS games with sound.
