Ahh...so you have it running 😀

My PIO-2 only accept FPM RAM too but 128mb max according to the manual...I bet it has the same VIA chipset as your mobo?

Rob some cache chips from your crappiest 486 mobo so you have at least 256k cache and stuff an AMD 5x86 and run it at 40 or 50mhz FSB since you have a PCI divider jumper as I do on my PIO-2

btw. please do a pcpbench benchmark with a lets say Diamond Stealth 3d card

Amigaz wrote:

My PIO-2 only accept FPM RAM too but 128mb max according to the manual...I bet it has the same VIA chipset as your mobo?

The 2 VIA chips have this info written on them -

VT82C505
VT82C496G

Amigaz wrote:

Rob some cache chips from your crappiest 486 mobo so you have at least 256k cache ...

I will do this very soon ...

Amigaz wrote:

... and stuff an AMD 5x86 and run it at 40 or 50mhz FSB since you have a PCI divider jumper as I do on my PIO-2

That sounds like fun. 😀 I think I have an AMD 5x86 CPU somewhere. However, the Cyrix CPU I have seems OK - please see the 45.xx score on the speedsys screenshot - it's as good as the AMD 133 CPU. Also, here is some info about this Cyrix CPU -

http://www.cpu-world.com/CPUs/5x86/Cyrix-5x86-100GP.html

It says that its Bus speed is 50. I will double-check the mobo's jumper settings, to see if the board is running at 50...

Amigaz wrote:

btw. please do a pcpbench benchmark with a lets say Diamond Stealth 3d card

Unfortunately, I don't know where my DS3D card is ATM. Please could you recommend another card(s) for me to try? Thanks.

retro games 100 wrote:

The 2 VIA chips have this info written on them - […]
Show full quote

Amigaz wrote:

My PIO-2 only accept FPM RAM too but 128mb max according to the manual...I bet it has the same VIA chipset as your mobo?

The 2 VIA chips have this info written on them -

VT82C505
VT82C496G

Amigaz wrote:

Rob some cache chips from your crappiest 486 mobo so you have at least 256k cache ...

I will do this very soon ...

Amigaz wrote:

... and stuff an AMD 5x86 and run it at 40 or 50mhz FSB since you have a PCI divider jumper as I do on my PIO-2

That sounds like fun. 😀 I think I have an AMD 5x86 CPU somewhere. However, the Cyrix CPU I have seems OK - please see the 45.xx score on the speedsys screenshot - it's as good as the AMD 133 CPU. Also, here is some info about this Cyrix CPU -

http://www.cpu-world.com/CPUs/5x86/Cyrix-5x86-100GP.html

It says that its Bus speed is 50. I will double-check the mobo's jumper settings, to see if the board is running at 50...

Amigaz wrote:

btw. please do a pcpbench benchmark with a lets say Diamond Stealth 3d card

Unfortunately, I don't know where my DS3D card is ATM. Please could you recommend another card(s) for me to try? Thanks.

Exactly the same chipset I have on my PIO-2
I wonder what differs between the PIO-2 and PIO-3 except the smaller size of the PIO-3 and it can take less RAM

Your Cyrix is probably running 3x33mhz = 100mhz
It's a very powerful CPU..I'm using it on my Asus PVI 486SP3 😀

Do you have a matrox millenium or Mystique instead to try with?

Amigaz wrote:

Your Cyrix is probably running 3x33mhz = 100mhz

I've just checked the mobo's "PCI BUS CLOCK SELECTION" jumper. It's set to its default setting, which is "<= 33MHz (PCI clock = CPU clock)". I have now changed this jumper configuration, and set it to: ">33 MHz (PCI clock = CPU clock/2)". I haven't switched on power yet. Will this run the CPU @ 50 bus speed?

Amigaz wrote:

Do you have a matrox millenium or Mystique instead to try with?

I just realised something - I have a matrox millennium 2 graphics card in a drawer. I will use that very soon. Before I do, I'd like to double-check with you that my Bus Clock jumper change indicated above is OK..

retro games 100 wrote:

I've just checked the mobo's "PCI BUS CLOCK SELECTION" jumper. It's set to its default setting, which is "<= 33MHz (PCI clock = […]
Show full quote

Amigaz wrote:

Your Cyrix is probably running 3x33mhz = 100mhz

I've just checked the mobo's "PCI BUS CLOCK SELECTION" jumper. It's set to its default setting, which is "<= 33MHz (PCI clock = CPU clock)". I have now changed this jumper configuration, and set it to: ">33 MHz (PCI clock = CPU clock/2)". I haven't switched on power yet. Will this run the CPU @ 50 bus speed?

Amigaz wrote:

Do you have a matrox millenium or Mystique instead to try with?

I just realised something - I have a matrox millennium 2 graphics card in a drawer. I will use that very soon. Before I do, I'd like to double-check with you that my Bus Clock jumper change indicated above is OK..

With that jumper setting you will now run the PCI bus at 16mhz instead of the same speed as the CPU "FSB" which is 33mhz
It's just a 1/2 divider so whatever the FSB the CPU is running at you get 50% of that speed on the PCI bus if you jumper it to 1/2...this jumper should be used if you use 40mhz FSB or higher but it's not nessecary for 40mhz FSB...try and see 😀

I've got a problem with the mobo's cache. The board was supplied with 128K of cache, and a TAG chip. That configuration works. Now, I have added 128K more cache, to make 256K. That configuration makes the mobo very unstable. Here's the technical info -

Bank 0:

4 chips, all marked:

W24257AK-15 (14 metal legs)
95451

TAG:

W24129AK-15
95220

This 128K config works OK.

Now, I added 4 chips to Bank 1:

4 chips, all marked:

W24257AK-15 (14 metal legs. Also, same code as above.)
94490 (Different code to above.)

This 256K config is not stable. Please note - I changed the mobo's cache jumper settings from 128k to 256k. Do I need to change the TAG chip? Also, I have some more W24257AK-15 (14 metal legs) chips I could try. Also, I have 4 larger cache chips, with 16 metal legs. These 4 chips have these markings:

W24M512AK-15
94521

Thanks a lot for any advice.

Edit: Solved! I removed the existing TAG chip marked:

W24129AK-15
95220

And replaced it with a (same sized: 14 legs) chip marked:

W24257AK-15
94490

In fact, this replacement TAG chip is the same chip as the 4 new chips used for Bank 1, in order to increase the cache from 128K to 256K.

Benchies will follow soon...

Edit 2:

I have temporarily misplaced the Matrox card. I will definitely find it, and rerun speedsys. For now, I have rerun speedsys with the extra mobo cache of 256k. Also, I upgraded the RAM from 16MB to 32MB. I don't know what type of RAM it is. It probably isn't Fast Page. Also, I maxed out the advanced chipset BIOS settings. So, I set cache to "turbo", and RAM to "fast", etc.

I wonder how much performance you'd loose if you stick the 120mhz version of the cx5x86 I use here socket 3 system on stereoids *Updated* use the PCI divider....
Planning to test that cpu on my PIO-2 in the future

btw. glad you got the cache working 😀

I noticed in your "120 mhz thread" that you are using a "Diamond Stealth 3D 2000 2mb 50ns EDO DRAM". I had a look in my box of junk, and found this, a S3 Trio64V+ card, but it's got the 2mb upgrade, and these 2 upgrade chips are marked with "bla bla...-50" That looks a bit like 50ns EDO to me. (Just guessing. Anyway, in it goes...)

Also, some of the mobo's BIOS settings are as follows:

Memory Parity Check = Disabled
Video BIOS Shadow = Enbled
C8000-CBFFF Shadow = Enabled. I set this to Enabled. Also, I set the other 5 similar settings to Enabled. Does this speed up the graphics?

Decoupled Refresh = Enabled
Video BIOS Cacheable = Enabled
System BIOS Cacheable = Enabled
Combine Alter & Tab Bits = Disabled
CHRDY for ISA Master = Enabled
Mempry Hole At 15Mb Addr. = Disabled
Cache Timing Control = Turbo
DRAM Timing Control = Fast
Fast DRAM = Enabled

CPU to PCI Write Buffer = Enabled
PCI Master Write Buffer = Enabled
PCI Master Prefetch = Enabled
PCI Master 1 WS Write = Disabled

BTW, the 2 x 16MB = 32MB RAM sticks I am using have got the number 60 written on them. Maybe they are FPM RAM?

I run some Pcpbench marks, using DOS 6.22 installed on a compact flash device.

Pcpbench mode 100 (640x400x8) not LFB = 5.1
Pcpbench mode 101 (640x480x8) not LFB = 4.5
Pcpbench mode 103 (800x600x8) not LFB = 3.4

Unfortunately, I still haven't attempted to run fastvid, in order to enable LFB. Also, both my auto + config files are empty. Do I need to run emm386 and himem? I'm sorry that some of this technical stuff still hasn't sunk in yet, but I've got a "mental block" about some of it. I don't know what emm386 or himem do, or if/when I need them.

retro games 100 wrote:

I noticed in your "120 mhz thread" that you are using a "Diamond Stealth 3D 2000 2mb 50ns EDO DRAM". I had a look in my box of […]
Show full quote

I noticed in your "120 mhz thread" that you are using a "Diamond Stealth 3D 2000 2mb 50ns EDO DRAM". I had a look in my box of junk, and found this, a S3 Trio64V+ card, but it's got the 2mb upgrade, and these 2 upgrade chips are marked with "bla bla...-50" That looks a bit like 50ns EDO to me. (Just guessing. Anyway, in it goes...)

Also, some of the mobo's BIOS settings are as follows:

Memory Parity Check = Disabled
Video BIOS Shadow = Enbled
C8000-CBFFF Shadow = Enabled. I set this to Enabled. Also, I set the other 5 similar settings to Enabled. Does this speed up the graphics?

Decoupled Refresh = Enabled
Video BIOS Cacheable = Enabled
System BIOS Cacheable = Enabled
Combine Alter & Tab Bits = Disabled
CHRDY for ISA Master = Enabled
Mempry Hole At 15Mb Addr. = Disabled
Cache Timing Control = Turbo
DRAM Timing Control = Fast
Fast DRAM = Enabled

CPU to PCI Write Buffer = Enabled
PCI Master Write Buffer = Enabled
PCI Master Prefetch = Enabled
PCI Master 1 WS Write = Disabled

BTW, the 2 x 16MB = 32MB RAM sticks I am using have got the number 60 written on them. Maybe they are FPM RAM?

I run some Pcpbench marks, using DOS 6.22 installed on a compact flash device.

Pcpbench mode 100 (640x400x8) not LFB = 5.1
Pcpbench mode 101 (640x480x8) not LFB = 4.5
Pcpbench mode 103 (800x600x8) not LFB = 3.4

Unfortunately, I still haven't attempted to run fastvid, in order to enable LFB. Also, both my auto + config files are empty. Do I need to run emm386 and himem? I'm sorry that some of this technical stuff still hasn't sunk in yet, but I've got a "mental block" about some of it. I don't know what emm386 or himem do, or if/when I need them.

Try univbe and run the benchmark (univbe 6.7)
http://www.mediafire.com/?4mznuyeymnj

it's always good to have at least himem.sys for the app's

Thanks a lot. I have now got univbe installed.

Also, I put these lines in to my config.sys file:

DEVICE=C:\Dos\HIMEM.SYS
DOS=HIGH,UMB
DEVICE=C:\Dos\EMM386.EXE RAM
FILES=30
STACKS=0,0
BUFFERS=20

Pcpbench mode 100 (640x400x8) LFB = 5.6
Pcpbench mode 101 (640x480x8) LFB = 4.9
Pcpbench mode 103 (800x600x8) LFB = 3.8

retro games 100 wrote:

Thanks a lot. I have now got univbe installed. […]
Show full quote

Thanks a lot. I have now got univbe installed.

Also, I put these lines in to my config.sys file:

DEVICE=C:\Dos\HIMEM.SYS
DOS=HIGH,UMB
DEVICE=C:\Dos\EMM386.EXE RAM
FILES=30
STACKS=0,0
BUFFERS=20

Pcpbench mode 100 (640x400x8) LFB = 5.6
Pcpbench mode 101 (640x480x8) LFB = 4.9
Pcpbench mode 103 (800x600x8) LFB = 3.8

Those benchmarks look real crappy to me...try another gfx card 😜

I removed the 2mb S3 Trio64V+ card, and replaced it with a 4mb Diamond / nVidia Riva 128. When I run Pcpbench, it runs in "LFB mode" automatically - I don't have to run univbe beforehand.

Pcpbench mode 100 (640x400x8) LFB = 5.8 (was 5.6)

This is odd. Why are the pcpbench scores so bad on this mobo? I wonder if the advanced chipset BIOS settings are to blame? This thread seems interesting -

Finally!! A stable Socket 3 System!!

On page 2, Vogons user "2Mourty" says that if he changes the L2 cache setting, then the memory bandwidth increases. Amigaz, you then reply with "Weird...welcome to the world of buggy'n'crappy socket 3 mobo's" Hehe!

If you look at my speedsys screenshots above, the memory scores are poor. I might mess about with the advanced chipset BIOS options, and see if I can improve things by decreasing/lowering the current "aggressive" settings...

I have just run Cache Check version 7, by Ray Van Tassle. The command line used was:

cachchk7 -4

I used the -4 switch, because this makes Cache Check run in "486 mode", and also makes Cache Check skip the CPU check. Cache Check reports this information:

This machine seems to have both L1 and L2 cache. [reading]
L1 cache is 16KB -- 205.1 MB/s 5.1 ns/byte (1654%) (257%) 1.7 clks
L2 cache is 256KB -- 79.5 MB/s 13.2 ns/byte (641%) (100%) 4.5 clks
Main memory speed -- 12.4 MB/s 84.9 ns/byte (100%) [reading] 28.7 clks
Effective RAM access time (read ) is 339ns (a RAM bank is 4 bytes wide).
Effective RAM access time (write) is 92ns (a RAM bank is 4 bytes wide).
486 Clocked at 88.9 Mhz. Cache ENABLED.

Don't know what's wrong but you should at least have a score of 7-8 in pcpbench

Something doesn't feel right. I changed the SIMM RAM. I found a pair of single sided 16MB SIMMs, total RAM = 32MB. The chips are marked with "bla...60". I think they are Fast Page SIMMs. They look quite "modern". The chips are thin, like RAM seen on more modern graphics cards. The advanced chipset BIOS settings are maxed out, and the mobo seems OK and stable. Yet Pcpbench produces low scores of approx 5.8, with a Riva 128. Strange.

I ran this, on the command line: doom -timedemo3 demo3

Comparison benchmarks (and timedemo instructions) can be found here:

http://www.complang.tuwien.ac.at/misc/doombench.html

My "realtics" score is 2087. By comparison, someone else scored 2086, and their system is:

AMD486-100, 8KWT, Hippo DCA 2, 82C802G 33, Quadtel PS805, S3c805 VLB, 2M DRAM, MB uses EDRAM

I had one of these boards back in the day (1997) when I didn't really know nearly as much about computers, and as I recall it was dreadfully slow no matter what I tried with it (Intel DX4 100, AMD 5x86 133, different memory, cache amounts) . I had a friend who really knew what he was doing who tried tweaking my system for better perfromance and couldn't get it to really perform. Was this VIA chipset just solid and reliable but not very fast perhaps? I wish I had that board now to mess with! I replaced it with a board that had a SIS chipset on advice of my friend, and it was much faster.

Have you ran CTCM.exe and CTCM7.exe to see what those results are? CTCM7 will tell you weather all the RAM is being cached properly by L2 cache. CTCM7 will probably only go as far as telling you about the CPU and cacheable areas before it freezes on a 486 system (it works properly with a Pentium though). Use CTCM.exe to see some detailed bandwidth readings, which might give you additional info on top of your other results.

I'm currently messing with a Chaintech 4SPI 486 PCI board and in Speedsys I'm getting nearly 40MB/s main memory throughput from it, and 53MB/s L2 throughput using a 5x86 133MHz. I'm working on getting some faster FPM memory to see if I can improve those results. Will post some info soon!

@Retro Games 100: Which program did you use for your Sys-info pic in the initial post? Is it speedsys?