CompuServe Thread

My main work horse is a 12 slot MB DX2-66 with 32 meg ram. I had so many device drivers and TSR's that it really started to get cramped. One day, recently, after upgrading one of my rendering slaves to DX2-66, I noticed it was running rendering in 2:35 what my main machine was doing in 5:12. Both were DX2 with 32 meg ram, no paging. The difference was I had only 431K of base ram on the slow machine and 631K on the fast machine. Also, the cache was 256k on the fast machine and 128K on the slow machine. I jumped to the conclusion thinking it was the cache ram that was slowing things down. A friend advised me to settle the argument by testing the rendering with the same base ram on each machine. Surprise! When I increased the base ram on my slow machine to 621K by eliminating all tsr's and devices unnecessary to slave rendering, the slow machine did the same project in the same rendering time as the original fast machine. Consequently, I built a pair of boot startup files to optimize speed for all slave renderings. When I go into slave mode on all machines I just reboot to optimize speed. I do have an exception on one machine I load a DOS TSR for receiving background FAXES. It slows that machine by a half a second per frame and goes to complete stop while accepting an incoming fax , then starts rendering automaticly. I did some limited benchmarking and determined the threshhold for base ram was 550K, below which things really start to crawl. This threshhold might be also dependent on the file being rendered as I only did the study with that one file I was working on at the time. Note: I use QEMM 7 and lantastic 5.0 loaded high to achieve these base ram figures.