Yah, it showed 133 for 2x66, so IDK if 125 was weird quantising, just guessing or what, plainly not the highest number it could show, but all the benches I ran against it at that time, which were not many, probably sysinfo, Veldthuis's snooper, maybe 3Dbench and a doom timedemo, were showing it level or a bit behind the 40Mhz x 3 setting for 120.. since we didn't have softfsb, cpuid etc to read clocks direct then, I didn't know what else to think. I thought I saw some IBM based upgrade chips with funky multis sometime though, but none of them seem to be cropping up here.

Edit: Amptron and PCChips.. the models seldom ever quite matched up, the PCChips might have 3VLB installed, the Amptron would leave one empty, the PCChips might have soldered cache, the Amptron might have sockets or vice versa. So seemed like Amptron was getting special orders from PCChips or something.

Cyrix and IBM didn't have fractional multipliers until the 6x86MX. Even the Cyrix MediaGXm at 300 MHz was still using integer multipliers.

BitWrangler wrote on 2021-05-16, 01:46:

Edit: Amptron and PCChips.. the models seldom ever quite matched up, the PCChips might have 3VLB installed, the Amptron would leave one empty, the PCChips might have soldered cache, the Amptron might have sockets or vice versa. So seemed like Amptron was getting special orders from PCChips or something.

Amptron, PCCHIPS and ECS had their own rev numbers and yes I can say with a great deal of certainty that they were rarely optioned the same as a vanilla PCCHIPS

The Amptron version I had of the m571 had a different rev number, bios and # of pci slots.
Memory compatibility was different between the 2 as well.
The compatible BIOs for the Amptron rev were older as well though I found I could use the wrong bios from a completely different model of board.

Interesting times.

douglar wrote on 2021-05-11, 19:45:

CPU Galaxy is raising the bar -- 200 Mhz 486, 66Mhz PCI, Peltier cooling, 21.6 FPS […]
Show full quote

CPU Galaxy is raising the bar -- 200 Mhz 486, 66Mhz PCI, Peltier cooling, 21.6 FPS

https://www.youtube.com/watch?v=LI1_RlVLhu8&t=1333s

Then they do a 29.3 FPS GL Quake @ 640 x 480 just because they can.

Untitled.png

Peltier cooling is the PC world equivalent of NOS. I guess it's great to show off, but you're just going to end up destroying your hardware if you run it like that constantly.
I consider 180MHz to be the upper limit of what am5x86 can do without burning out, and even that is really stretching it. If you want to blow up Cyrix 5x86s at 150 or POD83s at 125, be my guest.

What we really need to push the 486 higher is a newly developed upgrade using something smaller than .35micron.

Guessing SOCs are out of the question. You want a new CPU that can physically plug into existing Socket 3 motherboards, right?

What about an FPGA solution? There are some cheapish FPGAs out there that should be more than powerful enough to be set up as a 486 compatible CPU. The tricky part would be the interface between it and the motherboard.

Anonymous Coward wrote on 2021-05-16, 05:51:

What we really need to push the 486 higher is a newly developed upgrade using something smaller than .35micron.

The Media GXm uses 0.35 um and it can go up to 300 MHz, so I suspect the transistor gate length, or process size, isn't the only factor limiting the Cyrix 5x86 chips.

I think CPU Galaxy's objective is just to see how fast a 486-class CPU can go in Quake, irregardless of the practicality of having such a system in a case with all other supporting hardware. I'd be willing to run a single test of a Cyrix 5x86 at 150 MHz in Quake 1.06 using a peltier, but only if it counts as a "486" for this comparison.

I would like to see a more rigorous analysis of the Am5x86 at 200 MHz, e.g.

1. Is such a large heatsink fan necessary? Would a socket 370 heatsink/fan suffice?
2. How much benefit performance benefit did running the graphics at 66 MHz award, that is, if the 1/2 FSB-to-PCI multiplier was enabled, how much lower would the score be?
3. What benchmark score is observed when the FSB is run at 50 MHz x 4 instead of 66 MHz x 3 ?
4. What is the minimum wattage peltier needed for 200 MHz when using a socket 370 heatsink/fan?
5. Are you able to install Windows NT4 and Windows 98SE without the system crashing when clocked to 200 MHz?
6. When the motherboard is inside a case, is there any condensation build-up on or around the CPU which could cause a short if it were to drip.
7. Does the ISA bus speed run at a fraction of the PCI bus clock or the FSB? If the PCI bus clock, what is the highest integer divisor? On my MB-8433UUD it is only 4, meaning the ISA bus is running at 16.67 MHz, and there is no 7.14 MHz option on the MB-8433UUD.

I think if the optimal peltier wattage is determined and there isn't significant condensation, 180 MHz at 4 V or less might be practical for a cased system. However, I have head that these cheap peltiers aren't the most reliable, so you would want to find a way to integrate an Arduino-based thermistor LED display on the case bezel. Or possibly leave a thermocouple wire under the socket and plug in your DMM whenever you use the system.

I have a peltier I bought way back in the day that I used when trying to overclock the snot out of K6-2 CPUs. With the peltier cooled with a super janky watercooling setup incorporating a homemade waterblock, a bucket, and a fish tank filter pump. it would ice the CPU and socket within a few minutes, even when the system was running full-tilt.

That cooler being used needs to be able to handle the wattage from the CPU + peltier in order to get the lowest temps possible.

Have the water chilled in a water cooling setup and you get even lower temps.

How much benefit performance benefit did running the graphics at 66 MHz award, that is, if the 1/2 FSB-to-PCI multiplier was enabled, how much lower would the score be?

Depends in which test. Quake in VGA mode has very small increase, scoring only 0.1-0.2 fps more with 66 Mhz PCI bus. It should score slightly more in VESA modes and probably 1+ fps more in Mode X, compared to normal 33 MHz PCI bus results.

When the motherboard is inside a case, is there any condensation build-up on or around the CPU which could cause a short if it were to drip.

I doubt that such experiment was for daily use. But you can insulate the area around CPU socket with plasticine. Which is a common practice during below zero Celsius runs in overclocking community.

Does the ISA bus speed run at a fraction of the PCI bus clock or the FSB?

He ran it at minimal possible value on LuckyStar board, which still should be somewhat overclocked at 66 Mhz. Maximum possible value is 22 Mhz or 1/3 of PCI clock. It works perfectly fine with some sound cards, video cards or controllers.

douglar wrote on 2021-05-11, 19:45:

CPU Galaxy is raising the bar -- 200 Mhz 486, 66Mhz PCI, Peltier cooling, 21.6 FPS […]
Show full quote

CPU Galaxy is raising the bar -- 200 Mhz 486, 66Mhz PCI, Peltier cooling, 21.6 FPS

https://www.youtube.com/watch?v=LI1_RlVLhu8&t=1333s

Then they do a 29.3 FPS GL Quake @ 640 x 480 just because they can.

Untitled.png

Just saw the new "486 world record" info. Thanks for the heads up. Great stuff.

Coincidentally today I was testing some am5x86-133 cpus, to see which ones have to go.
Two of them seem to be long term stable at 180mhz, at least with the usual tests we use here.
Didn't put them through the grinder of offline computation yet.
I actually didn't overclock for top perf, more like which processor lights-up at 180, but the used settings were well optimized.
So, for example - i get 71.7 fps in Doom and something like over 30 fps in PC Player benchmark.
Can provide videos for legitimacy.

While this is pretty cool by itself, the more interesting thing is that at some point I hit 84.5 fps in Doom, which was surprising.
Will try to replicate soon in front of camera.

Anonymous Coward wrote on 2021-05-16, 05:51:

Peltier cooling is the PC world equivalent of NOS. I guess it's great to show off, but you're just going to end up destroying your hardware if you run it like that constantly.
I consider 180MHz to be the upper limit of what am5x86 can do without burning out, and even that is really stretching it. If you want to blow up Cyrix 5x86s at 150 or POD83s at 125, be my guest.

What we really need to push the 486 higher is a newly developed upgrade using something smaller than .35micron.

Cpu galaxy's video looked pretty convincing in my opinion.
But I agree that putting together a system like this will require some good preventive healthcare effort to avoid unexpected fatalities.
I had somewhere drafts for custom 3d printed case aimed at exactly that. One day ...

pshipkov wrote on 2021-05-16, 09:57:

Coincidentally today I was testing some am5x86-133 cpus, to see which ones have to go.
Two of them seem to be long term stable at 180mhz, at least with the usual tests we use here.
Didn't put them through the grinder of offline computation yet.

What voltage, which motherboard, what cooling? Did it finish Quake 1.06 timedemo and boot Windows 9x?

I have at least two motherboards that can handle 180MHz (and potentially more - need to actually investigate the 200 option really):
Biostar MB-843UUD-A rev.2
LuckyStar LS486E rev.C2

One of the two CPUs (ADZ, BGC) work well at 3.45V.
The other one needs 4V.
The ADZ feels more flaky. Unconfirmed really.

Without peltier element Quake 1 is often problematic. With it - quake, win95, etc - are fine.
Yesterday i used 10x15x2.5 cm heatsink but it is way too big for the need. It never gets even warm.
Need to test it of course, but 5x5x3 heatsink with decent fan will be more than enough.

Yesterday i tested with the 843UUD motherboard, since i stabilized it for good based on your feedback recently, if you remember.
Just checked - all BIOS settings on max, except two:
- DRAM write wait-state = 1
- DEVSEL decode = medium

Video showing quake 1 (19.6 fps), doom (72.23 fps), pc player benchmark (29 fps).
I am not sure if these numbers can be further improved on the Biostar board, but i didn't try for real, just a feeling.
Early cache write mode can get +1 fps in quake 1, but then Doom gets funky, etc.

LuckyStar LS486E is SIS based. SiS496/7 is slightly clock-to-clock faster than UMC8886BF/1F, so that is one potential path forward.

Wrapping up few remaining retro things here and may give the fastest-486 business a try.

I have just tried one of my LuckyStar LS486E rev.C2 with (fake?) AMD 5x86 160ADZ at 200 MHz (3x66) with peltier element at slowest settings in BIOS. PCI runing 1/2 FSB. Matrox GPU.
1. Computer starts when CPU is 8C
2. It boots to DOS when CPU is 5C
I can't get temperature lower, so only run Speedsys, 3D bench (high) and Quake. Doom freezes at 5C.
3D bench 95,5
Quake 19.9 fps

lukas12p, if you play around with different graphics cards and use 66 MHz FSB, can you get close to CPU Galaxy's results? 20.8 fps for 180 MHz and 21.6 fps for 200 MHz. What CPU voltage are you using?

Classic Matrox cards, on my experience, can do 66 Mhz just fine. Obviously Millennium II can do it, since it's native AGP card.

feipoa wrote on 2021-05-16, 22:58:

lukas12p, if you play around with different graphics cards and use 66 MHz FSB, can you get close to CPU Galaxy's results? 20.8 fps for 180 MHz and 21.6 fps for 200 MHz. What CPU voltage are you using?

I have noticed that PCI divider was set to 1:1, so Matrox works at 66 MHz and I have just checked Voodoo3 3000 PCi and it also works. Both cards gave the same results in Quake1 and PCI clock doesn't matter.
CPU @200 MHz works at 4V, now getting temp. 4C under load and 20.6 fps in Quake1, Doom freezes.
What voltage would be safe? I have more of this motherboards and two with volt.mod so can adjust any voltage (used for IBM 5x86 100@133)

4C under load and 20.6 fps in Quake1

Try VESA 320x200.

What voltage would be safe?

4V should be safe for prolonged use.

Ark2000PV lights-up at 66MHz and is a hair faster than Matrox Millennium in DOS.
If lukas12p has one of these - it may be better for the DOS tests.

Soooo anyone got a UMC U5S DX2 engineering sample they wanna try running at 100Mhz? 😁

Only got an SL myself, I was burning my fingers on the sink it when I had it at 60Mhz 1:1 and it was keeping up with my P60.

On cooling, I'd tend to favor the waterloop to a cooler full of salted ice (Look up ice cream bath) approach rather than pelts.

pshipkov wrote on 2021-05-17, 02:47:

Ark2000PV lights-up at 66MHz and is a hair faster than Matrox Millennium in DOS.
If lukas12p has one of these - it may be better for the DOS tests.

I had good luck with my ark1000vl at 66MHz on VL-bus as well.