It's probably like where a GF2MX400 will do better on a K6-III than a FX5950 will.

@Feipoa

Updated the charts with your numbers.
Linked your post in the directory on the first page.
You have the highest Doom score at 180MHz currently. Congrats !

To my best knowledge Doom is tested the way it is in Phil's benchmarks package - max details, full screen with status bar.
One more note for VLB/PCI motherboards - there is no best video card. Usually it is Ark1000VL or S3 Trio64 VLB for DOS and MGA/Voodoo3/Permedia2 PCI for Windows. Considering period correct-ish components.

@CoffeeOne
Updated the 160MHz charts with your top Wolf3D score. Congrats !

Where did those GLQuake results with MGA come from? What resolution and driver version are being run? Matrox results for GLQuake should not be besting or equating to a Voodoo3. Something's funny with those results.

The default for DOOM is with the utility/weapon bar at the bottom. I feel this is the most natural configuration to benchmark DOOM. Quake also has the utility bar at the bottom. I don't know why Phil's bench pack is benching without that utility bar, but it is not how we were benchmarking DOOM before phil's compilation came about.

Unless stated otherwise, all my DOOM benchmark results posted on this forum have been with this view:

it contains a maximised screen with the the utility/weapon/health bar.

I bench everything to "phil standard" because frankly it is the most accessible benchmark pack that everyone uses. Expecting someone to change default settings vs extract and go is ill-advised

maxtherabbit wrote on 2024-03-09, 02:30:

I bench everything to "phil standard" because frankly it is the most accessible benchmark pack that everyone uses. Expecting someone to change default settings vs extract and go is ill-advised

As far as I recall, it is extract and go from the DOOM_SW zip, as it is from the Quake_SW zip. As you are a late comer to the scene, I can see how you would have started using the Phil's pack. My 486 and 686 benchmarks were done before Phil decided to stick all the standard shareware demos into a single zip and change defaults.

I've been messing with retro hardware since I got an XT in the late 90s. When Phil's pack came out I immediately switched to it. For reasons which should be obvious. Convenience > all, especially when it comes to repetitive tasks

maxtherabbit wrote on 2024-03-09, 02:59:

I've been messing with retro hardware since I got an XT in the late 90s. When Phil's pack came out I immediately switched to it. For reasons which should be obvious. Convenience > all, especially when it comes to repetitive tasks

I sense a defensive posture. I mean, a late comer to the scene on this forum. I think most of use were playing with computers since 1990; that is mostly expected.

When Phil released his benchmark zip, I didn't have much any use for it. All the benchmarks were already readily available and I did not see the need to deviate from the defaults that we had been using on this forum from years prior. I have over ten thousand data points from before Phil's benchmark pack, am I supposed to adjust them retroactively because someone made a zip compilation? Or keep using what I/we had been doing so that my numbers are consistent? I picked the later.

You can do what you like. I am not asking anybody to change what they want to do, as you alluded to here: "Expecting someone to change default settings vs extract and go is ill-advised." I was reminding everyone that all my DOOM benchmark results on this forum, whether from 2011 or 2024 are done the same way, that is, maximised with the utility bar.

feipoa wrote on 2024-03-09, 03:37:

I have over ten thousand data points from before Phil's benchmark pack, am I supposed to adjust them retroactively because someone made a zip compilation? Or keep using what I/we had been doing so that my numbers are consistent? I picked the later.

That makes good sense, I don't archive my results so a change was trivial.

My main point was that when there is a single file available that is accessible to everyone with ease, it would be expected for it to become the standard. Why deal with transferring and extracting multiple things when you don't have to? Rational self-interest remains the best predictor of human behaviour.

maxtherabbit wrote on 2024-03-09, 02:59:

For reasons which should be obvious. Convenience > all, especially when it comes to repetitive tasks

so laziness 😀 Phil made a mistake destroying comparability with decades worth of old benchmarks.

rasz_pl wrote on 2024-03-09, 04:19:

maxtherabbit wrote on 2024-03-09, 02:59:

For reasons which should be obvious. Convenience > all, especially when it comes to repetitive tasks

so laziness 😀 Phil made a mistake destroying comparability with decades worth of old benchmarks.

yes perhaps he did, but it is done now

and you can call it what you like, but I prefer to say that the hallmark of the advanced mind is finding creative ways to avoid work 😉

The GLQuake/MGA numbers come from @Chadti99 - from his MS-4144 and Biostar UUD boards runing properly.
He built a fully stable 180/200mhz PC around the UUD.
I will let him clarify here.

---

as for phil's benchmarks:
they provide minimally automated minimal set of DOS benchmarks. nothing interesting on its own, but were the first consistent thing around.
i guess we collectively pushed them as the standard over the last few years, out of lazyness.
i actually have a disk image that has few more things in it but never bothered to promote it. few members here use it. some of the tests there are few more clicks than the single hotkey in phils tests, which is already too much i suspect - C compiler tests, complex offline graphics tests, couple of games to test audio and networking, etc.

---

wintune2 in win31 and win95 - this test always fluctuates between runs, but results are fairly consistent between win31 and 95, so i use it on both OSes with high confidency.

Keyboard controllers matter.

Realized that i never touched on an important detail that i bet many overlook during their 286/386/486 overclocking activities - the keyboard controllers.
For reaching and surpassing
- 25MHz at 0 wait states for 286
- 50MHz at tight timings for 386
- 180/200MHz for 486
it is important to have the right keyboard controller inserted in the motherboard.
The 486 case is less affected by them but they still matter.

The side effect of incapable chips will result in all sorts of seemingly unrelated issues. For example:
- flaky IDE controllers and local storage devices, including not being recognized entirely during POST
- if lucky to boot to DOS or other OS - random hangs, probably key press spamming, or invalid signals
- memory parity check screens out of the blue
and so on.

Often the difference between overclocking success and failure hinges on these chips.

The solution ?
Use AMI's KB-BIOS-VER-F.
They are the best and make a very real difference.

Probably most IC's have an impact on 286 and 386 systems which are being pushed way beyond their operating parameters.

Unfortunately for me, AMI's KB-BIOS-VER-F controller, if I recall right, lacks PS/2 functionality. All my systems are setup with PS/2 mice. Do you have a comprehensive list of all the keyboard controller modules you have tried; what notes accompany each controller? Maybe there are some with PS/2 which meet your requirements.

Right, you have a thing about ps/2 mice.
The answer to your question is quite simple - flip through the pics of the mobos in the thread.
The KBD controllers on them, are more or less representative of what was tested.
But cannot guarantee that each end everyone of them was inspected - didn’t keep track.

I can attest that the 'F' AMI KBC does in fact support ps/2 mice. I have several of them in my interposers

My memory is certainly not infallible, in fact, it has become rather error prone. Thus, I went to check, and I do have an AMI KB-BIOS-VER-F that is sitting in my bin of KBC-PS2 rejects. If you say it works with PS/2 mice, then it could be that some boards are acceptable with the KB-BIOS-VER-F, while others are not. I've run into this situation a few times. Or, it could also be that I removed the VER-F from a motherboard because my PS/2 adaptor already had a Holtek on it, and I didn't bother to test VER-F w/PS2. Nonetheless, as I am in the mist of assembling several PS/2 KBC adaptors, I will test VER-F again. If the VER-F works with PS/2 mice, and pshipkov says it is a possible solution to tight timings, all the better! I will use it on my SXL2-90 build, which is already at the limit of my motherboard.

EDIT: I can confirm that the KB-BIOS-VER-F also works in PS/2 mouse mode.

Epson Lynx (Epson E01161NA, E01165EB) used in Epson Equity IIe line of PCs

Bought this motherboard by accident.
Once it arrived i didn't actually know what to do with it but since it was here decided to examine it.
Was not able to find anything online about it, including TheWayBackMachine.
By the look of it is obvious that the assembly was built for some specific purpose.
Was able to scrape some vague notes about office scanner or perhaps a printer.
That was incorrect presumption, it is just another proprietary contraption - the ones i try to keep away from.

The board came with 12MHz Intel processor, 2x1Mb 100ns memory modules, and a 48MHz crystal oscillator.
I expected the base frequency to be 12MHz (48/4), but the divisor is actually 6, so base frequency was mere 8MHz.
With de-turbo it drops further. At the same time the provided manual in the following post talks about 8/12MHz, so not sure what is going on.

Has 1 Mb on-board memory. Slow 100ns rated chips.
That first megabyte cannot be bypassed. No jumpers, etc.
Tried even removing the chips but then only 64Kb RAM get recognized, despite the SIMM slots being populated with 4x1Mb modules.
Maximum supported memory is 5Mb.

Instead of the standard AT DIN-5 keyboard connector there is a 5-pin header in the lower right corner.
Used that with some loose wires going to the keyboard DIN-5 pins.

No BIOS program.
During POST a message is displayed about using some SETUP diskette but that is nowhere to be found.
Because of that i was not able to configure HDD/CF geometry for IDE extension card. Had to work off Gotek floppy emulator.

The onboard FDD controller works fine. If IDE extension card is inserted its FDD controller is used instead of the onboard one.

The power connector had non-standard AT plastic pieces, but the pins are AT-standard.
Had to remove the plastic pieces to be able to insert the two common power plugs.

The most i was able to get out of the motherboard is 11.11MHz.
Didn't run most of the tests - the board is clearly not build for performance.
For reference - Superscape produces meager 5.4 fps which is only ~1 fps higher than the best 8086 score.
Yet, i am not complaining. The board is actually fun contraption to thinker with.

For completeness listed it in the combined benchmark results.

Seems to belong in an Epson Equity IIe see pics in this fleabay listing https://www.ebay.ca/itm/354945317803
Apparently there's a serial board goes in the left side riser plug and hdd interface in the left side one.
like that https://sierra-indigo.com/epson-lynx-sp-board … d-y18620210000/
and that https://sierra-indigo.com/epson-lynx-hi-board … d-y18620310000/

edit:
short tech ref here
https://files.support.epson.com/pdf/e2e___/e2e___pg.pdf
No sign of setup disk, but DOSdays report that generic setup utility works on other models in the line.

Okay. My search-fu failed.
No idea how you related the Lynx board to Epson Equity line of computers, but i appreciate it. Especially the scanned manual.
I am going to touch-up what i wrote early.
Thanks.

You are welcome... me and google have this understanding... they give me what I want quickly, and I don't spam their search for 72 hours using ever more abstract and ridiculous search strings still trying to find it... some of it is being devious with the query, some of it is doing an end run around their text search "thesaurus attack" by browsing the images.