#Possible bug
11-Jul-94 09:22:22
Sb: #112109-#Possible bug
Fm: Jonas Ruikis [ADESK] 73172,1351
To: Ian Kelsall 100141,435
Ian,
<< Pentium problems.. >>
repost…
"Hello All,
Well, the good news is, my rendering time on the P5/90 has dropped from 90
minutes to 3 minutes! Follow these simple instructions to get the same out of
your machine. Put your hand on the Pentium's ZIF socket, and say "I'll never
buy a Power PC" three times. Poof! Instant performance increase, and this
week's winning lotto numbers appear embedded in the BIOS serial number on your
next boot. Serves me right for reading all those Mac and Toaster magazines
when I should have been memorizing my device driver memory allocation tables.
Here's what worked.
CONFIG.SYS before loaded HIMEM, EMM386, SETVER, POWER, IFSHLP, DOS HIGH, DOS
UMB, and CD-ROM drivers.
CONFIG.SYS after loads HIMEM, EMM386, SETVER, DOS HIGH, DOS UMB. I also bumped
FILES up to 100 and BUFFERS to 40.
AUTOEXEC before loaded the NETWORK, CD-ROM, SMARTDRV, MOUSE, VMODE, and 3DS.
AUTOEXEC after loads the MOUSE, NETWORK, VMODE, and 3DS. FYI, VMODE is for the
Gateway monitor, and I moved the MOUSE driver before the network driver since
it takes more memory.
Also, I significantly reduced the amount of memory used by the network startup
files by setting Lantastic's default parameters to realistic (I hope) minimums.
Time will tell.
Here's what didn't work. Setting FILES and BUFFERS too low added to rendering
time. Disabling HIMEM and EMM386 didn't help, but it didn't hurt either.
Removing the NOEMS parameter from EMM386 did slow rendering time (logically,
Cap'n Kirk). Including the old DOS parameter FCBS added to rendering time.
Playing with the BIOS setup didn't help. Gateway (or more properly, AMI) only
lets one ENABLE or DISABLE cache memory, and as you can imagine, disabling it
sends rendering time into the stratos"fear". The only other CMOS setup
parameters deal with the speed of the ISA bus and IDE transfers. I didn't play
with them. If anyone out there thinks disabling the shadow memory, slowing the
ISA bus speed, or choosing a specific IDE transfer mode will help, drop me a
line and I'll look under the hood again before I have to. Using all local
materials instead of hitting materials off of a network server's hard drive did
not improve rendering time. Running MEMMAKER after revising CONFIG.SYS and
AUTOEXEC.BAT didn't help. Blowing cayenne pepper into the vents in the front
didn't help, but it made me feel better.
I'm still a little worried about the whole deal, though. Trimming the fat from
the network configuration cut rendering time to 21 minutes. Available
memory=61688, Used memory=61688, Swap=128, and Page faults=188. I didn't write
down the exact parameters for the 90 minute render, but Swap wasn't much more
than 128 (maybe 500?). The current setup performs the same rendering task in 3
minutes, with 63736 available and 61864 used, 0 swap and 0 page faults.
If 3DS could talk, I think it would be saying, "Hey Mac, you're 1872 steps away
from the quicksand."