First post, by Skyscraper
I use an old socket 7 Soyo Sy-5eas motherboard and a pentium 166mmx in my retro build.
It uses the Via VPX chipset and can cache 64mb of memory.
The board has one single slot for PC66 sdram which according to the manual can handle up to a 32mb module.
I knew for a fact that it can handle a 128mb module since I sucessfully used a single sided 64mb module.
I thought that I might aswell see how much of a performance hit I would suffer if I installed a larger module.
I searched the net and much as I thought the memory selling sites stated that the board could take a double sided 128mb module.
I also found forum posts written by someone that tried a double sided 256mb module but only got 128mb just as you would have thought.
I tried a 512mb module, and got 256mb! which made me very satisfied, the strange thing is that I diddnt notice much of a performance hit at all.
I lost 2% mips in sisoft sandra cpu test but the mflops stayed the same, the multimedia benchmark the same and the memory benchmark numbers the same.
All this was with identical bios settings using the bios that was on the board when it was sold in 1997.
But since I im thinking of trying a k6-3 if I can find one I wanted to update the bios.
I found a bios from year 2000 which I presume have support for k6-2/3 altough limited to the chipsets officially max supported fsb of 75mhz (83mhz works just fine).
The thing is the new bios made as much of a performance inpact as going over the cachable range. I lost another 2% mips in sisoft sandra, again the mflops, multimedia and memory bench stayed the same.
I also tried PC Mark 2002 and I lost the exact same 4% of the CPU score from the memory upgrade and bios update combined. I also lost 4% of the memory score.
I did only run pc mark with the old bios with 64mb memory and with the new bios and 256mb so I cant be sure if its the same 2% + 2% or if the performance hit came from one or the other.
Perhaps sisoft sandra and pc mark isnt the best tools to measure the performance in windows 98?
I did also run speedsys and there you dont see a performance hit at all, neither from the memory upgrade nor from the bios update, but DOS does not fill the memory top down so I can understand that
Is the performance inpact of going over the cachable range in windows overrated?
If a bios update can have just as much as a performace inpact as going over the cacheable range then perhaps lots of people run their systems with unnecessary small ammounts of memory?
Running Windows 98 SE with 64mb memory does not leave much memory free at all but with 256mb there is plenty of breathing room.
I would like some input on these matters, whats your experience?
New PC: i9 12900K @5GHz all cores @1.2v. MSI PRO Z690-A. 32GB DDR4 3600 CL14. 3070Ti.
Old PC: Dual Xeon X5690@4.6GHz, EVGA SR-2, 48GB DDR3R@2000MHz, Intel X25-M. GTX 980ti.
Older PC: K6-3+ 400@600MHz, PC-Chips M577, 256MB SDRAM, AWE64, Voodoo Banshee.