Seems like you all were better off than I was in 1996/1997. In Jan 1997, the most I could afford was the cheapest entry-level machine available from some no-name whole in the wall computer builder. It was a PC Chips based, cacheless, M919 v3.4 with an Am5x86-133 and 8 MB of RAM. Win95 diskette edition choked on it until I eventually upgraded the RAM to 32 MB. When people asked me what computer I had, I was embarrassed to answer, so I would reply "133" and hope they wouldn't ask further questions. Even the Am5x86-133 was far faster than what I had been using, which was some 286 upgraded to 386 with 4 MB of RAM. That system broke, so I had actually been using one of those Compaq briefcase 8088 machines before getting the PC Chips 486. This transition in my computer history is probably what sparked my retro interest for proper upgrades to 286, 386, and 486 class hardware.

I also felt that this is more of a socket 3 competition than 486 one.
And since @mpe shared his POD100 Quake3 achievement, i felt compelled to follow his example.

Quick boot-to-Quake1-roll on a fully stable system with full gear:
https://www.petershipkov.com/temp/retro_pc_im … quake1_27.3.mp4

Asus VLI-486SV2GX4, POD100, Ark1000VL, 32Mb RAM, 1024Kb SRAM, all settings on max.

That's one beefy hybrid.

Crazy that these numbers are actually faster than a regular P100. Unless using Triton chipset with PB cache (which can push it to 2-3 fps higher).

Currently L1 is in WT mode.
Noticed some instabilities with WB that i will look at soon.
When in WB mode, Quake 1 hits 27+ fps.

Not much into hybrids myself.
I use this PC with 160mhz 486 AMD ADZ CPU, but since it has very fast DOS graphics, it was a good fit for the conversation here, with bolted POD100 on it.

feipoa wrote on 2021-02-08, 23:34:

Seems like you all were better off than I was in 1996/1997. In Jan 1997, the most I could afford was the cheapest entry-level machine available from some no-name whole in the wall computer builder. It was a PC Chips based, cacheless, M919 v3.4 with an Am5x86-133 and 8 MB of RAM. Win95 diskette edition choked on it until I eventually upgraded the RAM to 32 MB. When people asked me what computer I had, I was embarrassed to answer, so I would reply "133" and hope they wouldn't ask further questions. Even the Am5x86-133 was far faster than what I had been using, which was some 286 upgraded to 386 with 4 MB of RAM. That system broke, so I had actually been using one of those Compaq briefcase 8088 machines before getting the PC Chips 486. This transition in my computer history is probably what sparked my retro interest for proper upgrades to 286, 386, and 486 class hardware.

Feipoa, you wrote me a message, but for some reasons I am not allowed to answer here. Just contact me at cpugalaxy@gmx.at

BR

Congratulations for your last video, CPU Galaxy! Your video are always so interesting!
I wonder how the pci bus can run (stable) at 60 mhz, not only the video card, but also the eide controller. Usually the pci bus gets tricky after 40mhz, it would be possible that the chipset use some kind of wait states for the pci?
Or the eide controller could use the isa clock instead of the pci?

On faster boards with fsb 60, 66 MHz You have PCI divider (0,5*fsb) You can turn on.
But still all system gets unstable, depends on components, like I can't run any ISA card when use 66 MHz fsb on LuckyStar 486 MB.

The Serpent Rider wrote on 2019-12-26, 14:43:

So I've been doing all sorts of crazy stuff with my high-end 486 setup... […]
Show full quote

So I've been doing all sorts of crazy stuff with my high-end 486 setup...

CPU: AMD 5x86-133 ADZ @ 180Mhz
Motherboard: LuckyStar LS-486E Rev. C2 - 60Mhz FSB with 1:1 PCI divider
L2 Cache: Magic 15ns 256Kb
RAM: 2x 8Mb EDO SIMM 60ns (tmtech chips)
Video: 3dfx Voodoo 3 3000 PCI
HDD: Transcend IDE DOM 1Gb
Additional hardware: PS/2 Mouse ISA card adapter

===================================================================
Settings:
LUCKYSTAR LS-486E SETTINGS.png

CacheCheck:
CACHECHEK 486 180Mhz.png

SpeedSys:
SPEEDSYS.gif
___________________Read___________Write____________Move__
Cache Level 1_________170.89 MB/s________76.61 MB/s________220.15 MB/s
Cache Level 2__________83.67 MB/s________76.41 MB/s_________38.34 MB/s
Memory _____________57.34 MB/s________76.58 MB/s_________20.45 MB/s

3D games and benchmarks:
Doom Shareware v1.09 timedemo - 1098 ticks / 68.0 fps
Quake Shareware v1.06 timedemo - 19.4 fps
PCPlayer Benchmark 320x200 - 28.0 fps
PCPlayer Benchmark 640x480 - 12.5 fps
Chris's 3D Benchmark 320x200 - 76.0 fps
Chris's 3D Benchmark 640x480 - 23.3 fps
Duke Nukem 3D observable framerate (VESA 320x200) - 70+ fps
Duke Nukem 3D observable framerate (VESA 640x480) - 28+ fps

CPU-Z VINTAGE:
AMD 5x86 ADZ 180Mhz CPU-Z.png
Interesting sidenote: According to CPU-Z, my setup is very close to the infamous IDT WinChip C6 200Mhz. So it's really a 486 class CPU on steroids.

P.S.
Now try to beat my scores =P

My new video is online and I did beat your scores. 😀

https://youtu.be/qPRSHI4JoWI

CPU Galaxy wrote on 2021-03-27, 14:44:

My new video is online and I did beat your scores. 😀

Contrats! But you've forgot to push it a bit further with VESA modes. By using VESA 320x200 I've gained +0.8 fps without doing anything else. Unfortunately Permedia cards won't do, due to buggy support in VESA modes. More info about it here - PCI Video cards benchmark: Electric Boogaloo Edition

Not bad at all, but I think that you might easily beat your own record with that fine CPU, a 20.8 FPS score in Quake can't be the limit for a 486 machine. If that CPU can complete all the benchmarks at 180 MHz with a passive cooler at room temperature, then, theoretically, it might operate stably at 200 MHz by maintaining its temperature at about 0*C. Such temperature could be maintained for a while by utilizing a freezer trick.

If you'll manage to pull a stunt like such, then Quake frame rate might go above 22 FPS and THAT would be pretty darn impressive 😀

a 20.8 FPS score in Quake can't be the limit for a 486 machin

It's already not limit, since with VESA it will perform even better. And to be fair, Permedia 2 aren't absolute fastest PCI cards for VGA or Mode X. Definitely could squeeze a bit more at least in Doom. And yes, that difference on much more powerful system (link above) slightly scales even on appropriate to DOS era hardware - Re: PCI Video cards benchmark: Electric Boogaloo Edition

ARK2000 will probably do wonders for Doom score, but I doubt it's PCI66 friendly. Worth a try though.

P.S.
BTW I vaguely remember there was a separate thread on vogons with AMD 5x86 / TNT2 setup which scored 71-72 fps in Doom, but I think it was 200 Mhz, i.e. 66 Mhz FSB with more relaxed timings.

Great stuff @CPU Galaxy.

I also think that the shown in the video system is somewhat clock-to-clock slow.
Have here rigs that run at 160mhz and are only frame apart from your numbers and actually lead in some cases.

Also, how real the rig is ?
Meaning - is it completely and long term stable ?
The simple DOS tests are ok, but not indicative enough in my experience, especially if you test without full gear - sound, lan, win, more involving computation, etc.
Did you check more in-depth, or the exercise was for checking the peak perf only ?

yes, I know that I have still space for improvements. The 200 Mhz setup is already tested and running stable at 0 degrees with active cooling. But before that I will make a follow up on the 180 MHz setup under windows.

I found curious that your M919 is implementing 1/2 * FSB = PCI, when FSB >= 40mhz. On my M919 v3.4 B/F with the same BIOS date, it would implement a 2/3 * FSB when FSB >= 40mhz. So at 40 MHz, the PCI bus would run at 26.7 MHz, not 20 MHz like on your board.

CPU Galaxy wrote on 2021-03-27, 22:53:

yes, I know that I have still space for improvements. The 200 Mhz setup is already tested and running stable at 0 degrees with active cooling. But before that I will make a follow up on the 180 MHz setup under windows.

So now you tested both the fastest and the slowest 486.
How about a few charts showing the enormous gap between them? =)

I would be itching to drop that pathetic 20MHz SX in that impressive motherboard and see how much performance you could recoup with some OC =)

Just watched the CPU Galaxy video - this is all great stuff, thank you folks 😀

bofh.fromhell wrote on 2021-03-28, 14:03:

So now you tested both the fastest and the slowest 486. How about a few charts showing the enormous gap between them? =) […]
Show full quote

CPU Galaxy wrote on 2021-03-27, 22:53:

yes, I know that I have still space for improvements. The 200 Mhz setup is already tested and running stable at 0 degrees with active cooling. But before that I will make a follow up on the 180 MHz setup under windows.

So now you tested both the fastest and the slowest 486.
How about a few charts showing the enormous gap between them? =)

I would be itching to drop that pathetic 20MHz SX in that impressive motherboard and see how much performance you could recoup with some OC =)

Slowest? I have a board that can run certain socket 3 CPUs at 8mhz.

CPUGalaxy, would you be willing to let me know the result you get when you run Quake 1.06 shareware on its own, that is, not from Phils benchmark pack? Most of the result displayed here are done not with Phil's bench pack. I get different values when using his pack and not using his pack. I recall him writing once that he made some changes to some configuration settings/files, so it would be ideal to install Quake 1.06 SW to a new folder. Then run Quake.exe -nosound -nomouse -nonet -nocdaudio. Once quake loads, press ` to bring down the console, then -timedemo demo1

This is what I looked at recently while playing with pod100 cpus but there was not difference in perf.