Thanks a lot people! This was one of my most interesting tests. I have removed the big copper cooler heatsink on top of this ADW CPU, and moped up the Artic Silver 5 compound on it, to reveal a production date inscription of: A 9630DPE. Is this a good/"magic" date? BTW, I'm sure the heatsink cooler and paste helped a little bit, to maintain stability.

Very unfortunately, right ATM, I don't have a Voodoo2 card to hand. Sorry.

Hehe, regarding the compact flash aparatus. I'm not absolutely certain if this makes a difference. I have done various tests however. Sometimes I use a CF device attached to the mobo's IDE pins. Sometimes I use an IDE cable with the CF attached to the end of it. I have found that the CF+cable is less reliable than a CF attached to the mobo's IDE pins, when doing huge overclocks. Perhaps it is because the data has to travel along a cable, back and forth, and when you do overclocks, things can go wrong?

I have removed the big copper cooler heatsink on top of this ADW CPU, and moped up the Artic Silver 5 compound on it, to reveal a production date inscription of: A 9630DPE. Is this a good/"magic" date?

Thank you for your sacrifice. (: Your CPU is very good! Mines don't work stable at 180MHz without peltier. Maybe ADW series takes more adventage from increasing voltage. I realized that it doesn't make any difference if I set voltage to 4 or 5 volts. Are you using the same cooling as on biostar?

If you've got a similar set up, please experiment with the I/O recovery time in the BIOS. I changed this setting from 2 to 8, and that allowed me to do this outrageous overclock.

This is a very important discovery. Unfortunately, currently I haven't got CPU working at 200MHz in home environment. I will be looking for ADW. (:

I have found that the CF+cable is less reliable than a CF attached to the mobo's IDE pins, when doing huge overclocks.

It seems funny but devil is in the details. I always thought that this only applies to internal motherboard wires (external digital intputs use "synchronizers" to prevent encountering metastable voltages) but you overclocked CPU to 200MHz so you are right.

Thank You for your experiment, now I know that it is possible to overclock this CPU to 200MHz. (:

udam_u, you're welcome! 😀 Previously, I had poor results with the ADZ variant. I couldn't achieve either 180MHz or 200MHz. However, the ADW variant works well at both 180 and 200 MHz. Sorry I can't remember, what variant are you using?

I'm using a slightly different cooler, compared to the one used on the 486 Biostar mobo from the other thread about "UMC 486 PCI". It's a bit heavier, and I'm also using a bit more thermal paste. The paste covers all of the CPU, and also some of it is smeared over the underside of the cooler. It can be seen here. Please scroll down that webpage, and look at the 3rd photo. The one I am currently using is the one on the right - with the heatpipes. The one I used on the Biostar 486 mobo can be seen on the left - this one is still in its shrinkwrap.

I just tried another ADW chip I have. That works too, at 200 MHz. All of the BIOS timings were set to "maximum performance", except for one crucial setting, and that was the cache timings. They were set to "slowest performance". Ah, it's a pity this mobo's cache chips are 15ns. If I had some 12 or 10ns chips, I could get faster benchies!

My 3DBench 1.0c score is 105.6. I seem to remember my 3x, 60MHz = 180 MHz 3DBench 1.0c score was ~107. (Begin edit: It was 106.7. End edit.) However, the 200 MHz configuration is very useful, because it provides extra CPU power, and benefits games such as Quake. It was also fun just messing about with this stuff.

Also, I set all of the BIOS timings to "slow", and then attempted to boot up using a bus speed of 66 MHz, with the multi set to 3x. I got no POST.

Btw, can you also make it run @ 3x66Mhz instead of 4x50Mhz?

Tetrium wrote:

Btw, can you also make it run @ 3x66Mhz instead of 4x50Mhz?

The answer to your question can be seen immediately above it.

How are you mounting the heatsink/fan?
I have a few spare coolers (LGA775, Socket A) and was just given a DX4-100 a week ago.

Not going to go insane but it would be interesting to see if it's prime stable at 120MHz.
Well, that's if Prime95 runs on a 486...

TheLazy1 wrote:

How are you mounting the heatsink/fan? I have a few spare coolers (LGA775, Socket A) and was just given a DX4-100 a week ago. […]
Show full quote

How are you mounting the heatsink/fan?
I have a few spare coolers (LGA775, Socket A) and was just given a DX4-100 a week ago.

Not going to go insane but it would be interesting to see if it's prime stable at 120MHz.
Well, that's if Prime95 runs on a 486...

Unfortunately, I'm not actually mounting the heatsink. ATM, it's just resting on top of the CPU. When I want to install this mobo in to a case, I'll have to be creative, and think of a plan of action to see how this can be achieved. It might not be easy...

That's a really good idea - to see if your DX4-100 can be OC'd from a bus speed of 33 to 40 MHz and run stable with Prime95. I cannot see any reason why you won't be able to test this. Prime95 must surely run in Windows 95, and a 486 can definitely host such an operating system. I did a search for Prime95, and every link mentioned Windows 95.

retro games 100 wrote:

Tetrium wrote:

Btw, can you also make it run @ 3x66Mhz instead of 4x50Mhz?

The answer to your question can be seen immediately above it.

doh... 😊

Strange though, it must be the motherboard then.

Edit:though the motherboard can in fact run your SX chip @ 66Mhz.
What could cause the SX chip to boot at 66Mhz fsb and the 5x86 (which should be able to run @ 200Mhz) to fail at 66x3 but work at 50x4?

CPU stability. Higher FSB puts more stress on the CPU.The best way to find out would be an AMD 486DX4 100 enhanced chip that could go above 120MHz and set the multi to 2x.

retro games 100 wrote:

Unfortunately, I'm not actually mounting the heatsink. ATM, it's just resting on top of the CPU. When I want to install this mobo in to a case, I'll have to be creative, and think of a plan of action to see how this can be achieved. It might not be easy...

There should be glue like TIMs, but I don't think those can be undone.
I have a 486 chip here with glued on heatsink so I cannot get any model #s off of it.

Even more insane:
Lapping the heatsink to be used on a 486, but that's just crazy...

TheLazy1 wrote:

Even more insane:
Lapping the heatsink to be used on a 486, but that's just crazy...

I actually considered that...a little 😜
Though I never tried, I'll probably practice on a couple heatsinks first

The problem is you need a good mounting mechanism to provide enough even pressure, otherwise it's probably no better than the 3 cent coolers around now.
Engineers: Here is your mission.

udam_u, you're welcome! Happy Previously, I had poor results with the ADZ variant. I couldn't achieve either 180MHz or 200MHz. However, the ADW variant works well at both 180 and 200 MHz. Sorry I can't remember, what variant are you using?

Yesterday I tried to boot up my AM5x86 (I'm using ADZ series) with FSB set to 66MHz. I was surprised because it worked without any special cooling system (just simply pentium cooler). Unfortunatelly it crashed after bios entering screen. I tried changing io recovery times, cache timings, FSB/PCI divider but without positive feedback...

I'm using standard old HDD (I haven't got CF card). Do you think it is important to use flash drive? Have you ever tested your 200MHz 486 with standard HDD?

Thanks! (:

udam_u wrote:

I'm using standard old HDD (I haven't got CF card). Do you think it is important to use flash drive? Have you ever tested your 200MHz 486 with standard HDD?

Unfortunately, I have never tested my 200 MHz 486 with an old HDD. However, I will retest this system again, and use an 80 GB "real" hard disk drive - that's the closest I can get to using an old "real" HDD. I will try to do this when I have time...

I think that using a flash drive, especially one that is not on the end of cable, provides useful stability when doing huge overclocks. The CF unit, and the CF card itself, can be bought fairly cheaply, and IMHO, it's a worthwhile investment. There's one thing that I have found useful, when experimenting with old hardware - it's a really good idea to have a "tool kit" containing lots of different retro items - different things you can try out, which may yield different and sometimes unexpectedly good results. I would recommend that you get some CF hardware for this reason.

Also, my favourite overclocks are in this order: 4x / 50 MHz, then 3x / 60 MHz, then my least favourite is 3x / 66 MHz, because it is the most unstable, and you also have to set the BIOS timings to slow values. I would put all of your efforts in to achieving 4x / 50 MHz. Good luck! 😀

Thanks!
I will buy CF kit in the future. Currently I need to save money. I found CPU collector who offered me selling five AM5x86 ADW but the same money problem bother me in the completion of transaction. [:

I use an 1GB hard drive for my testing purpose. If you encounter a problem with hard drive detection when trying to connect an 80GB HDD the reason will lie in 8GB BIOS limitation.

At this moment my computing area is occupied by another person from my family but tomorrow I'm going to reach stable 4x50Mhz!

Regards! (:

EDIT1:
I made additional tests today and discovered that my AM5x86 overclocked to 200MHz or 198MHz boot to DOS when L1 cache is disabled. I achieved this on ga-486am/s motherboard. One thing is dissapointing - those additional efforts allowed me to obtain ~9 points in speedsys! [:

486AM/S motherboard doesn't support EDO memory so I had to use FPM module. However I think that it behaves better than um8810 in spite of lacking EDO memory support. um8810 contain very strange IDE controller (based on cmd chip). I encountered some hdd read problem when using it even with standard CPU settings. Also I successful overclocked Cyrix 5x86 100MHz to 133MHz on ga-486am/s motherboard (2x66MHz, cache enabled).

During tests I checked all important bios configurations (I hope) with 3 graphics cards so the only way to obtain better results is buying mentioned above ADW processors. (:

EDIT2:
I forgot to mention I found a mistake in ga486am/s jumper configuration description (from stason and th99). Correct should be: configuration for 256KB (A) -> BANK1 and configuration for 256KB (B) -> BANK0.

udam_u wrote:

... so the only way to obtain better results is buying mentioned above ADW processors. (:

Yes, and also if your ADW CPU tests do not provide completely satisfactory results, please do consider getting a CF. BTW, I would recommend getting something like a 512MB device. I've tested about 10 different CF cards, and I don't like a lot of them, for example the faster ones seem to have compatibility problems.

Also, I think there are a lot of fakes being sold on ebay. I have 3 Transcend 133x 2GB CF cards, and they all behave differently when used on the same IDE controller. Perhaps invest a bit more, and get one from a well known online retailer, rather than ebay.

I think a stock Cyrix 5x86 - 120 Mhz will likely outperform this AMD at 200 Mhz. I've been running mine for years in a Biostar MB-8433UUD and a PC Chips M919. They run stable on the fastest RAM and cache settings. I've been able to turn on all enhanced features of the Cyrix 5x86 in Windows NT4.0 and Windows 98SE except for branch prediction.

Cachechk with Cyrix 5x86-120 in the M919 reports:

L1 cache is 16KB -- 246.3 MB/s
L2 cache is 256KB -- 95.8 MB/s
Main memory speed -- 55.1 MB/s
Effective RAM access time (read) is 76ns
Effective RAM access time (write) is 51ns
Clocked at 486 120.2 MHz. Cache ENABLED.

The cachechk results of my AMD 586-160 in the same motherboards at the fastest stable settings does not even come close.

Cachechk with AMD 586 in the M919 reports:

L1 cache is 16KB -- 164.9 MB/s
L2 cache is 256KB -- 74.5 MB/s
Main memory speed -- 47.4 MB/s
Effective RAM access time (read) is 88ns
Effective RAM access time (write) is 51ns
"AuthenticAMD" 486 Clocked at 486 160.6 MHz. Cache ENABLED.

The memory performance of your AMD 200 Mhz is similar to that of a Cyrix 5x86 - 100 Mhz. Here is what I get,

Cachechk with Cyrix 5x86-100 in the M919 reports:

L1 cache is 16KB -- 205.2 MB/s
L2 cache is 256KB -- 79.8 MB/s
Main memory speed -- 45.8 MB/s
Effective RAM access time (read) is 91ns
Effective RAM access time (write) is 61ns
Clocked at 486 100.2 MHz. Cache ENABLED.

I have reported only stable configurations such that MemTest does not show any errors and conditions where Windows NT4.0 has been run for at least 1 year without any BSOD's with continual use. I generally use the Biostar now as it has 512KB L2 cache and addresses 256MB FPM ram.

Awesome! 😀 Some screenshots of Speedsys would be nice, and would make reporting your scores much easier for you. Also, can you please run Quake 1 (shareware 1.06) full screen, timedemo demo1, and tell us what FPS you get? Thanks a lot.

Are Cyrix 5x86-120 chips rare these days? I had a look on ebay, and couldn't find any.

retro games 100 wrote:

Awesome! 😀 Some screenshots of Speedsys would be nice, and would make reporting your scores much easier for you. Also, can you please run Quake 1 (shareware 1.06) full screen, timedemo demo1, and tell us what FPS you get? Thanks a lot.

Are Cyrix 5x86-120 chips rare these days? I had a look on ebay, and couldn't find any.

Yes. The Cyrix chips are hard to find. That's why I imagine the group of collectors chipped in for that large job lot someone mentioned not too long ago. The 100mhz and slower 486/5x86 Cyrix chips are everywhere but not the 120's for some reason. The Cyrix 5x86 actually had a lot in common with the 6x86 chip internally so it was pretty damn fast for a chip that fit in a 486 socket and used the 486 instruction set. It was damn close to a Pentium in architecture but the instruction set it used was mostly 486 instructions. There was only a handful of Pentium instructions that were supported.