CompuServe Thread

#Render speed

5 messages in this thread
#193929From: Stephen C. LevyOct 4, 1995 1:44 AM
John, Thanks for your numbers. How much RAM do you have, and how much and what type of L2 cache? Thanx again, SCL
#193976From: John StetzerOct 4, 1995 11:15 AM
>>How much RAM do you have 64 megs. >>how much and what type of L2 cache 256k. John Stetzer JWS Sorry, just installed it yesterday and wanted to pass along the results as quickly as possible. Re. the Display/No Display figures; indeed, go figure, but there you have it (happened the same way 3 times in a row for each).
#194024From: Stephen C. LevyOct 4, 1995 3:32 PM
John, Thanks for the add'l info. Main point of curiosity now is about cache — pipelined burst or not? Thanx again, SCL
#194043From: John StetzerOct 4, 1995 5:13 PM
>>pipelined burst or not? Yep (or so I was just told). Think the extra 256k cache contributed towards the slight speed gap? John Stetzer JWS
#194074From: Stephen C. LevyOct 4, 1995 7:17 PM
John, We're on the same wavelength. L2 size difference, and we're ignoring (1) the effects of HD subsystems, (2) impact of memory managers, config.sys, autoexec and 3DS.set differences, and (3) theoretically identical machines will, due to manufacturing tolerances, vary slightly in speed. Each of these variables could conceivably account for the difference by itself, but likeliest is a combination of them. If you're interested in trying to pick up that last few seconds, you could try a few things like setting up a separate boot menuitem that loads a VERY minimal configuration (no EMM386, no HIMEM.SYS, no memory manager at all) of just the barest drivers necessary to run 3DS. If you still have 250K of low DOS available, it may help. (I gather that a side effect of this is that you "lose" all memory over 64 megs, so those with more may actually be going backwards by trying this). On a DX2/66, I gained 10 secs/frame this way, but on two different P90s, the effect was only a third or so. Let me know if you try it. Thanks again, SCL