More Stardent Junk
To follow up that previous post, here’s some more Stardent Vistra 800 stuff from that box in the attic.
After we finished getting the first prototype built, it posed for its media photos. Here are a couple of outtakes. Cute little machine, wasn’t it?
In that picture, you can see AVS displaying a PDB file. In the full resolution version you can see the neat Phong sphere rendering algorithm we developed on the GS1000. It was a really clever algorithm which gave us full phong shading cheaply, but it got a little hairy when you had a perspective projection matrix. A perspective projection doesn’t map spheres to circles, and the algorithm depends on rasterizing a circle. We ended up fitting a circle to the projection of the sphere in that case.
This one shows Tom Van Sant’s GeoSphere image. He used a Stellar GS1000 to assemble this cloudless view of the earth from a large collection satellite images. It was a really amazing dataset. I remember when he stopped by to show it to us after he first finished it. I’m afraid I made him pretty mad. While he was a lunch, I took a program I had that drew texture mapped spheres (this was before we’d written AVS) and I put his dataset into it. It was the first time it had been mapped onto a sphere and he was out of the room at the time. Sorry about that Tom! Later they made some really cool 2 meter spheres with this image on them. It was sort of the earliest precursor of Google Earth.
I also found a copy of the review of the Vistra in the June 1991 issue of Digital Review magazine. They said:
At 31.53 MicroVAX II units of processing (MVUPs), the CPU performance of the Vistra … certainly isn’t shabby, but overwhelming CPU power is not what the Vistra will be best known for. The Vistra 800 is ultimately a graphics workstation, designed to put 3-D graphics processing power on the desktop at minimal cost.
Here’s a copy of their benchmark data:
The Stardent 3020 and the Silicon Graphics Power Iris 4D/320 were much bigger machines with multiple processors, but that HP 9000-720 was comparable. The PA-RISC in it was certainly the hot chip that year.
The list prices for the Vistra were:
- $27,995 for the base 16bpp 800
- $31,995 for the 24bpp 800e (with the AGC)
- $40,995 for the 800ex with the 2 extra i860’s (the AGM)
Of course those prices included 32MB of DRAM, a 400MB disk, and a 19” monitor. Those were pretty impressive numbers in 1991.
That summer we took a couple of the machines to SIGGRAPH in Las Vegas. For some reason they were really unhappy there. Something about the heat or the humidity or something. The night before the show opened, I was up all night trying to get them to go with half the company standing around watching and saying:
What’s wrong Mike? When’s it going to work Mike? The show opens tomorrow Mike!
Oh well, that kind of thing was half of the fun of startups. Someday I’ll tell you about the time Jeff and I were hacking executables with a binary editor at 2:00 AM because we didn’t have a compiler with us or the time when the shipping company dropped the MicroVax unloading it from the plane and cracked the motherboard the day before the show.
Great pair of posts, thanks for sharing! I actually just got ahold of an i860 programming manual yesterday and started reading it. I wish I could’ve programmed one of these machines, but I was in high school at the time. By the time I got out of college, a lot of the interesting RISC architectures were fading from the scene; many I wasn’t even aware of until years later.