Pretty specced up system you've got there.
Love the pics of the pc out in the snow 🤣

Hi and welcome! 😀

Nice rig you got there, I see you use one of favourite tower's...Vobis Highscreen Colani designed 😉

udam_u wrote:

I don't like 3dbench, because this is very capricious program, for example it works bad with my cyrix 5x86 processor (strange result 0 fps!).

Firstly: Cool project! Love the picture in the snow. Makes me miss european xmas / winter / snow a lot 😢

Secondly: There are two versions of 3dbench!

Version 1.0 which works up to 99 fps and which I recommend for max. 486DX4 machines.

And there is version 1.0c which works up to 999 fps and which I recommend for Pentium / K6 / 5x86 and higher...

Both version can be downloaded at my first post here: 3DBENCH CPU Benchmark Database

Using 1.0 on fast machines and 1.0c on slow machines will result in odd / wrong / no results.

In a nutshell: If you run 3dbench version 1.0c you will very likely record the highest 486 score in our 3dbench database 🤣

udam_u wrote:

I love old computers (especially 486) and i think that this is very good place for me.

Be gone, spawn of Satan! Your kind aren't welcome around here! *Only joking!*

Excellent and fun project you have there! 😀 I am curious about something - Speedsys says that your 586 chip is a 486, and it's 0.50 um.

I tried to overclock an AMD 5x86 133MHz ADZ to 3x60=180MHz, but I could not get DOS to boot up. I used a UMC chipset PCI mobo. I tried all CPU voltage settings. What is your CPU voltage setting, for your CPU? Thanks.

Thank You for replies. (:

Nice rig you got there, I see you use one of favourite tower's...Vobis Highscreen Colani designed.

Yes, this is really case with Colani sign. You are very perceptive!
I love this tower because of its nuclear-like power switch.
Case photos:


Have you got Colani too?

@Mau1wurf1977 You are right. I have run 3dbench 1.0c and results seem to be correct now. I obtained 102,1fps.

During my work on optimizing bios settings i realized that mdk dos benchmark is very bios settings sensitive (cache tymings, wait states, etc.).
(link to benchmark: mdk_bench.zip)
Have you ever run it? If so, what do you think about it?

I am curious about something - Speedsys says that your 586 chip is a 486, and it's 0.50 um.

I think that this is because I have set 3x FSB multiplier. Maybe those old CPUs don't have cpuid instruction like pentium and later processors. Don't be puzzled if programs like speedsys will show wrong results when you use non standard CPU settings. If I change multiplier to 2x (internal 4x), bios, chkcpu and speedsys recognize my CPU as AMD 5x86 P-75 (0,35um).

CPU AMD 5x86 133MHz ADZ@180MHz(3*60MHz)

CPU AMD 5x86 133MHz ADZ, non overclocked, standard settings

What is your CPU voltage setting, for your CPU?

It works fine with voltage set both 4 or 5V. What model of motherboard have you used?

udam_u wrote:

I love old computers (especially 486) and i think that this is very good place for me.

Then you've come to the right place, mate!

I've wandered around the net myself for a while before I found a (couple) forum(s) that had people that shared my interest (namely 386 and up, using windows and my main reason for using old rigs is because it's fun trying out stuff with old computers and because I love gaming 😉 ).

I'm contemplating creating a 486 hot rod myself. Theres some very interesting info about hardware that isn't easily found elsewhere (like having a Pentium 2-400Mhz that has a flexible multiplier).

Part of the fun is in finding out ways to run hardware in a way that wasn't intended.

Same with software, kinda

udam_u wrote:

It works fine with voltage set both 4 or 5V. What model of motherboard have you used?

It's an "SYL8884PCI-EIO", and I talk about it on Vogons here. I think your Shuttle HOT-433 mobo must be one of the finest 486 mobos ever made. 😀

11.5fps timedemo demo1 seems low. My 486 gets 14fps average there. Poor PCI performance?

Also, try running the benchmark in ModeX 320x200 and VESA 320x200 (use vid_describemodes and vid_mode to figure out what you have and use it to set it as the video modes menu doesn't cover it all)

udam_u wrote:

@Mau1wurf1977 You are right. I have run 3dbench 1.0c and results seem to be correct now. I obtained 102,1fps.

Impressive!

Beats a Pentium 100!

@retro games 100 What a pity I haven't got SYL8884PCI-EIO (I have: Shuttle Hot-433, Elitegroup UM8810P-AIO, Gigabyte GA - 486AM). I suppose that this is processor related problem. You have to find better one. I have got five 5x86 133MHz ADZ but only two work fine at 180MHz.

@leileilol Can you list your 486 specification please? I don't think that this is PCI performance problem. Those old machines have slow memory so you can't reach full PCI speed.
Maybe you ran timedemo at vid_mode 0 (320x200)? I have checked my results again. This is what I obtained:
5x86 133MHz ADZ @ 180MHz full stable bios settings:
vid_mode 17 (320x240): 11,6fps
When I set more agressive cache timings (dos still stable but BSOD when booting to WIN95) I obtained:
vid_mode 17 12,3fps

vid_mode 0 (320x200) 14,3fps

@Tetrium This forum is the biggest repository of knowledge i have ever seen. My first computer was 5x86 133MHz but i hadn't got knowledge to overclock it. Now i want to catch up. (;

I have made some progress in Peltier.
First of all i had to make insulating frame for CPU and Peltier module. This is very important because condensed water can damage computer (this is active heat pump so temperature on cold side can drop below ambient temperature).

Peltier in frame

Peltier, CPU bounded together in frame

Photo above shows my first mistake. I placed Peltier hot side to CPU. I realized that something is wrong when computer hang up at 133MHz. Fortunately I used Peltier in this position for a short time and CPU still work correctly.
Because of power lack standard power supply 12V rail (only 9A) i had to use additional power supply for Peltier.

This allowed me to reach full stable 5x86 133MHz@180MHz_4V in home.

Warning
[boring part]
I use Peltier TEC1-1270 (89W) in my project. You can read its datasheet here:
http://www.hebeiltd.com.cn/peltier.datasheet/TEC1-12710.pdf

From this datasheet you can evaluate expected temperature on Peltier cold side.

Sketch of the calculations:
Up-voltage applied to Peltier
Rp-Peltier resistance
Ic-Peltier current
Qj-Joule heat (Qj=Ic^2 *Rp)
Th-Peltier hot side temperature
Tc-Peltier cold side temperature
dT-temperature diference between cold and hot side of Peltier
Ta-ambient temperature
Qcpu-heat produced by overclocked CPU
TDP-heat produced by non overclocked CPU
fs-non overclocked CPU frequency
fo-overclocked CPU frequency
Us-non overclocked CPU voltage
Uo-overclocked CPU voltage
Qh-heat transferred to cooler
Rw-cooler thermal resistance [C/W]

Ic=Up/Rp=12/1,24=9.68A (near optimal Imax=10.5A)
Qj=9.68^2*1.24=116.44W

TDP=3.7W, fo=180Mhz, fs=133MHz, Uo=4V, Us=3.45V
Qcpu=TDP*(fo\fs)*(Uo\Us)^2
Qcpu@180MHz_4V=6.74W

Qh=Qj+Qcpu
Qh=123.18W

ambient temperature in my house Ta =18C, Rw=0.5[C/W]
Th=Qh*Rw+Ta=79.59C
Tc=Th-dT (dT=70C from datasheet)
Tc=79.59-70=9,58C, so my CPU should have temperature below ambient temperature.

I had problem finding small, cheap and eficient power supply. Finally I decided to try 5V form internal power supply (full satisfactory 22A).
Additional calculations proved that this is better way.
Ic=4.63A
Qj=23.16W
Qh=29.9W
Th=32.95C dT=35C
Tc=32.95-35=-2,05C (28.31F)!
[/boring part]

Peltier connected to 5V internal power supply rail.

Now my CPU is stable in home at 180MHz and 3.6V! I can also close my case because i don't use external power suply any more. I haven't tested my system outside yet, because i'm waiting for lower temperature (current temperature is above 0C). In home 200Mhz still doesn't work. I suppose i need better CPU.
I'm going to make some control logic (control current by PWM, measure CPU temperature etc.) for my Peltier module. This will allow me to get real fedback.

HNY Vogons! (:

udam_u wrote:

I have got five 5x86 133MHz ADZ but only two work fine at 180MHz.

Ah, I see! 😀 Two of your five ADZs are "magic"! 😉

I have had bad day today. It seems that my PSU in Colani Tower is dead. I think that Peltier killed it. I took out motherboard and I'm going to make some investigation. (;

I'm still looking for good sound card for my new rig. I think that SB AWE32 is the best, because this was high-end sound card with wide support in games. What is the difference between various AWE32's? For example this contains all chips on PCB:
http://upload.wikimedia.org/wikipedia/commons … _PNP_CT3990.jpg
and this one haven't got one chip:
http://www.lauppert.ws/screen1/sc-awe32.jpg
Both cards are CT3990.

That's a bummer for you. I hope you get your tower back up and running soon. From the photos, that computer looks nothing short of intimidating.

As for the two different AWE 32s, my only guess can be changes to another chip or two made the extra chip useless. Or maybe they sold different versions under the same part number. I don't remember the exact part number for the Vibra 16 in my Gateway 2000 system but I remember seeing cards with the same part number that had CD ROM interface; mine doesn't.

The missing chip is the CT1748 CSP/ASP. No big loss... 😉

udam_u wrote:

Maybe you ran timedemo at vid_mode 0 (320x200)?

Yep and that is how it should be done. vid_mode 0 only maintains consistency and it's the fastest of the 3 types of modes supported anyway (ModeX 320x200 is slower and VESA 320x200 is even slower)

Thank you for replies. (:

That's a bummer for you. I hope you get your tower back up and running soon. From the photos, that computer looks nothing short of intimidating.

Yeah i was a little bit confused. Today I swapped orginal PSU with reserve one and tested all components. Happily all parts work fine! (:
I haven't observed condensed water during test. I think that Orginal PSU is to weak for this system (2xHDD, 3.5FDD, 5.25FDD, CDROM, Ethernet card, Voodoo 1, TsengET6000 and finally Peltier). I want to repair this PSU by myself. I need some time to acquire necessary knowledge...

As for the two different AWE 32s, my only guess can be changes to another chip or two made the extra chip useless. Or maybe they sold different versions under the same part number. I don't remember the exact part number for the Vibra 16 in my Gateway 2000 system but I remember seeing cards with the same part number that had CD ROM interface; mine doesn't.

It sounds very logic. Maybe software developers didn't use this chip so Creative decided to remove it.

The missing chip is the CT1748 CSP/ASP. No big loss... Wink

Yesterday I tried to find information about this chip. I red this thread (saw your post (; ):
http://queststudios.com/smf/index.php?PHPSESS … dc&topic=2421.0
Do you know any application which use CSP?

@leileilol Yes but I wrote that I had run it at 320x240. (; Everyone can set resolution to 320x240 and compare with my result.

Hey I was just browsing around looking at things related to 486s, Quake, and other such things I'm interested in... and I came across this thread. My main 486 box is a Compaq Presario 425, and I'm afraid that with weaker specs I get another 2FPS in Quake than you.. I'd like to know why.

Specs:
Evergreen 586 (Am5x86-133)
33Mhz FSB
20MB RAM
Cirrus Logic GD-5420 GPU, 512K RAM (iirc, need to check that still)
(I think that the GPU is VLB, but it's onboard so hard to tell, could be ISA)
FreeDOS w/ JemmEx

I get 12.78fps if I do a timerefresh standing just behind the entry pad at the start of the second level of the game. It's a spot where looking one direction there's smooth framerates and the other choppy, so a good spot to do FPS tests.

Perhaps you're just testing in a more intensive spot, I don't know - could you take a measurement there?

If you still get ~11FPS there, then I'm curious why my system performs better with lower specs. I'd think that the 180Mhz chip with higher FSB would provide enough of a boost to run at 17FPS or so all by itself - not to mention the video card you're using should be quite superior..

Does my Presario just have magical powers? xD

Try timedemo instead? When I benchmark Quake, I do timedemo demo1 at the console. I usually do full screen, and the default video resolution, which is I think 320x200.

Yushatak wrote:

Perhaps you're just testing in a more intensive spot, I don't know - could you take a measurement there?

He insists on testing in the non-standard non-default resolution of 320x240 which I don't even like. I'm not changing my methods for one guy. The additional 40 rows of pixels make all the difference, but not only that, his viewsize is maximum (no inventory bar)!

A 13fps avg timerefresh on start.bsp for that chip is as normal as it gets