very impressive score and getting closer to the cyrix.

What cache are running in the vlb system? 10ns?
Do you have support for WB on L2?

I tried my x5 on 1:1 pci divider and 1:2/3
Also L1 on WB and WT and always end up on 16.6

So I am thinking this is the limit of my setup unless I get faster cache chips and find a bios if one exhists with WB for L2

I am quiet happy and its a dos only machine but benchmarking is a game in itself.

I might try even lower timings but as I recall from months ago system was not stable anything under 2 2 2 2

guys how the heck can the AMD 166 MHz be that slower than the Cyrix 133 MHz???? 😲

amd 160 is a clocked 486 pretty much and the cyrix is more then just a 486 a cut down 686 in a socket 3 there is alot of material about it, also the cyrix is running 2x66fsb vs 4 x 40 amd

My SRAM is of the counterfeit chinese variety and is supposedly rated for 10 ns. I do not beleive the SRAM speed is increasing the performance compared to 15 ns SRAM. The Asus VL/I-486SV2GX4 motherboard w/updated BIOS are probably most of the reason for the speed. This is probably one of the most favoured socket 3 VLB/ISA boards.

I see, I thought that might of been the case. I have this board but the cut down 5v version only perhaps I will start a project sometime in the future to mod it to 3.3V (I skimmed over some guides) and update the bios but not atm.

In the meantime I was also skimming over some old posts from you feipoa and from the quick reading I was doing I understand there is a moded bios for the MB-8433UUD-A which has WB L2 cache enabled, did I understand it right?

You think there is any bios past 3/26/96 that can give a performance boost?

There is a 5/20/96 BIOS for this board, but there isn't a speed boost because of it. I modified the 5/20/96 BIOS to alter some default BIOS settings, and more importantly, to unhide the write-through L2 cache type feature. This board is permanently set for write-back operation of the L2 cache. So if you install 64 MB of RAM and are using 256K of double-banked cache (recommended), then only half your RAM will be cached. If you change the L2 cache type to write-through, you can cache all 64 MB of RAM.

Intel486dx33 wrote:

These are my settings for AMD 5x86-133-P75 over clocked to 160mhz. I am having ram and cache performace problems. I am using ben […]
Show full quote

These are my settings for AMD 5x86-133-P75 over clocked to 160mhz.
I am having ram and cache performace problems.
I am using benchmark tools from Philscomputerlab.com
Computer works fine just problem playing MP3’s.
I am not sure if its the bios, CPU, or ram , or audio driver ?
How can I check ?

First, what audio driver? Are you running these tests on Windows 95/98? Even a high-end 486 machine like that AMD will struggle with mp3 files unless the player/decoder was optimized for slower CPUs. Most software quickly gave up on this and instead demanded a Pentium (preferably MMX or above) and focused on various GUI things to show during playback.

Also, a 3-1-1-1 cache setting is pretty good I'd say unless you want a 2-1-1-1. Well that's not going to happen without 12ns or 10ns cache chips I think.

feipoa wrote:

There is a 5/20/96 BIOS for this board, but there isn't a speed boost because of it. I modified the 5/20/96 BIOS to alter some default BIOS settings, and more importantly, to unhide the write-through L2 cache type feature. This board is permanently set for write-back operation of the L2 cache. So if you install 64 MB of RAM and are using 256K of double-banked cache (recommended), then only half your RAM will be cached. If you change the L2 cache type to write-through, you can cache all 64 MB of RAM.

Ok thanx for the good explanation

Deunan wrote:

Intel486dx33 wrote:

These are my settings for AMD 5x86-133-P75 over clocked to 160mhz. I am having ram and cache performace problems. I am using ben […]
Show full quote

These are my settings for AMD 5x86-133-P75 over clocked to 160mhz.
I am having ram and cache performace problems.
I am using benchmark tools from Philscomputerlab.com
Computer works fine just problem playing MP3’s.
I am not sure if its the bios, CPU, or ram , or audio driver ?
How can I check ?

First, what audio driver? Are you running these tests on Windows 95/98? Even a high-end 486 machine like that AMD will struggle with mp3 files unless the player/decoder was optimized for slower CPUs. Most software quickly gave up on this and instead demanded a Pentium (preferably MMX or above) and focused on various GUI things to show during playback.

Also, a 3-1-1-1 cache setting is pretty good I'd say unless you want a 2-1-1-1. Well that's not going to happen without 12ns or 10ns cache chips I think.

1) Use Winamp 2.05
2) Disable autoscroll for the song title (this has a huge impact)
3) Disable the spectrum analyser
4) In the mp3 decoder settings, set the decode type to 486
5) Use the wav output option rather than direct sound. IIRC, direct sound requires a few more % CPU.
6) Reference: Post your 486 Winamp benchmarks
7) Try Winplay3 as an alternative player
8) Still can't get it to play without skipping? Disable stereo output from Nullsoft Module Decoder and decrease mixing rate, e.g. to 22 KHz.

From memory, I think even an Am5x86-133 can play mp3s without skipping.

After all this talk about quake I had to go run it. My fastest VLB board (Genoa X4) with AMD 5x86 at 133 does 14.0fps with status bar, 12.4fps full screen. At 120MHz it does 13.2fps with status bar, 11.7fps full screen. Sadly, this one doesn't run at 150 or 160. And the board doesn't boot with a Cyrix 5x86.

Meanwhile the HOT-433 with Cyrix 5x86 is terrible at Quake for some reason. Only gets 12fps.

One interesting tidbit that I discovered: changing the clock multiplier jumper on the Cyrix 5x86 takes effect immediately. I removed and replaced the jumper a few times watching it get faster and then slower, but I guess I did it one time too many as the program finally crashed with a stack overflow.

So I got out a riva tnt 1 not 2 as recommended and still 16.6 fps on quake, s3 trio 64 next 16.6 still, looks like its maxed out at that. Cache chips are 15ns but I don't think it will make a difference.

Next I got out a atx1415, pretty much same setup as the biostar with the s3 trio64 and 16.6 fps .. magic number both these motherboards use the same umb chipset 8886bf / 8881f so yeah it makes sense. I ran the atc with cache off in the bios and didn't seem to make any difference

Im going to get my lonestar pci board out next

What would be the best PCI video card to use on these 5x86 builds ?
Win95b compatible.

The TNT will only show some performance benefit if you are using GLQuake.

Has anyone run the Cyrix 120 with optimal settings in a VLB board with a good video card?

I presume 40mhz vs 33mhz in video benchmarks is not going to give much benefit?

Anonymous Coward has, but I recall he couldn't run with an optimised LSSER.
EDIT: I'm fairly certain there was a user on the forum who had a VLB/ISA system with the Cyrix 5x86's BTB, FP_FAST, and LSSER feature set accordingly, but I forget who it was.

I went back for the final round:
1. Luckystar lucky Star LS486-e c1 17.1 fps
2. Atc 1415 16.9 fps ( .3 fps improvent after turning on pci/memory burst mode bios settings)
3. Biostar mb8433-uud 16.6 fps (no improvement with pci/memory burst options on)

I used a riva tnt 16mb pci and S3 trio 64 1mb and both cards got exactly the same scores.

I wasn't able to beat Feipoa's vlb setup but breaking the 17fps barrier I feel I will be able to sleep good tonight and even my speedsys scores on my LS486-e are slightly better vs feipowa's Asus VL/I-486SV2GX4 his quake score is very impressive and unbeatable. I wonder if anybody on a pci/vlb system running amd160 can beat 17.6.. probably not

Turning on the below settings on the ls486-e were the final straw that took me just over 17fps

I'm curious as to why the Biostar is underperforming on your Am5x86-160 system. Do you have the PCI divisor set to 1:1? Is your cache at 2-1-1-1 and memory read/write at 0ws/0ws?

2 1 1 1 crashes on all my systems, command.com error system frozen, so 2 2 2 2 0 0 and pci 1:1 is the best I can squeeze. I am using 15ns cache.

I just found this piece of info however.

atc1415 actually stops showing cache on 2 2 2 2(shows on post screen as 256) but speedsys and cache check don't see L2 going back to 3 1 1 1 shows cache in speedsys again and gives me 16.6 fps in quake, curious the same score as biostar at 2 2 2 2 with cache showing in speedsys.

Perhaps there is some mechanism to throttle the speed down on the biostar while having cache work?

Im going to check later biostar on 2 2 2 2 with cache disabled in bios, maybye this will be a similar condition to the atc1415

Also I did a few more runs on the atc1415 before packing it up and managed to get 17.0 fps with the riva tnt, beginners luck maybye.

I have five MB-8433UUD boards with differing versions. original version, v2, v3.0, and v3.1. All of them can handle 40 MHz with 2-1-1-1 with 15 ns cache. Are you using 256K of double-banked cache? Don't use single-banked cache. Are you using the BIOS I posted for this board so that you can use L2 cache in WT mode? How much RAM are you using and how many sticks? Are you using EDO or FPM? Do not use EDO - it is not faster on this board and requires slower timings at 40 MHz.

ahh so 2 1 1 1 never worked for me on the biostar so I took it as a benchmark and did all the testing on 2 2 2 2. Just tried the atc1415 at 2 1 1 1 and got 17.3

Going back to the biostar I still get invalid command.com or sometimes boots but quake quits with a page error.

By double banked do you mean using 1 cache chip per 2 banks? bottom pins of 1 bank and top pins if second?
If so then no.

I never used the bios you told me about because you said WT is enabled by default in stock bios and I am only using 32mb so there is no need to disable it hence no performance boost from the modded bios.

I am using 2 matched 16mb fpm 60ns single sided ram chips I know its not ideal because 2 chips but next best thing I got is a double sided 32mb chip but it is 70ns apparently (I tried both combinations of rams on the biostar and they both get 16.6 on 2 2 2 2

One time I was lucky to get a benchmark with 2 1 1 1 and 1 x 32 fpm 70ns chip and got 17.0 fps so still lagging behind atc1415

the act1415 and luckystar I used the same umc cache

The biostar is in a case so it was too hard to swap cache and this is the cache biostar has

This is a biostar v3 I do have a v2 also maybye ill try that for comparison too