CompuServe Messages

I have what I thought was a unique problem, and I've been afraid to come out of the closet with it, BUT NOW I KNOW I'M NOT ALONE! The problem is with 3D Studio and any make of Gateway (confirmed on all 486 DX2-66's through P-100's). Regardless of which video card is in use, regardless of how much RAM you have (this problem has been verified with between 16 and 32 Meg of RAM), what version of DOS (demonstrated with 5.0 thru 6.22), and whether a PCI or VLB is present. It seems to be a CPU/FPU thing in conjunction with the Micronics boards that Gateway uses. The use of any memory management beyond Phar Lap contributes to the severity of the problem, as does the number and type of drivers loaded in config.sys and autoexec.bat. (CD ROM drivers seem to be especially onerous). The big revelations – I haven't heard of it happening to any other brand name PCs., and the problem can be LESSENED (though not minimized or eliminated) somewhat by continuously pressing the PF11 key, so that the keyboard buffer fills up. In fact, a rendering frame that might take an hour without special handling, drops to 5 to 10 minutes. I learned recently that a company in San Francisco using Gateways in a production animation environment were having the same problem, and had also discovered the partial solution of depressing the PF11 key. So, what IS the problem? In scenes where either/both face and vertex counts exceed 100,000, OR ipas routines such as Skin, Flame, Bones, etc are in use, rendering can slow to a crawl, swap files will be created even when sufficient RAM is present to process, etc. The telltale sign is that the fuel gauge stops moving while transforming objects, while procedural objects are being created, or while normals are being prepared. How do I know this is happening, and that the rendering times are out in left field, you ask? On a recent project, I had to lease 5 Pentium 90s, 1 Pentium 75, and a Pentium 60. These clones all outperformed my Gateway P-90 in rendering by factors of 2 to 4, depending on whether not I effectively taped down the PF11 key. If the tape worked loose, and somehow the key was not depressed, the next frame rendering time could increase by a factor of 30. At first I thought the files were causing my machines (Pentium and DX2-66) to lock up, since they had only 32 Meg of RAM versus the 48 on the leased machines. Not so – after a few reboots, I inadvertently let my Pentium continue rendering, even though the gas gauge did not move for nearly an hour and a half, and there was no disk activity. The gauge suddently began moving again after the hour and a half, and the frame finished rendering within the next few minutes. A 94 minute render, versus 3 1/2 minutes on the clone Pentium 50. Sad stuff! Today, when I tried to recreate the problem to make sure I got the various messages right, I got a fatal one (for the render) that I hadn't seen before – "EXP Mesh Op Failure – no object ready" while the last round of normal preparation was occuring. The message occurred five minutes into the render of the first frame. This particular file has thirteen skin objects, two flame .ixp boxes, and consists of around 20,000 vertices and 36,000 faces. Anyone have a similar experience, and some type of long term solution? I'm very happy with Gateway quality and service, but my next PC will be a clone if I can't fix this problem with my current machines. I may even arrange an across-the-board trade and keep my RAM. If their are no solutions, are there any takers for a new Pentium 90?? Obviously, the move to NT, and a different memory management strategy for 3D Studio may make all this a non-issue with release 5. Thanks in advance. I'm posting copy of this on the Gateway forum, with some additional background for the "uninitiated" there. Bob Weil