pshipkov wrote on 2023-01-21, 18:56:

Decided to consolidate the captured performance metrics for all tested motherboards. Goal is to improve consistency and readabil […]
Show full quote

Decided to consolidate the captured performance metrics for all tested motherboards.
Goal is to improve consistency and readability.
Data will be split in few separate posts based on hardware class - 286, 386, 486 (160/180/200MHz).
The original posts will be edited accordingly.
Starting with 486 class motherboards at 160MHz.
(charts will be updated if new items get examined later)

486 class motherboards with AMD Am5x86 processor running at 160MHz (4x40)

Hi again,
I must say I am really surprised a little bit by your values.
I always have lower values, also when I seem to use identical hardware
for example
Asus SVGOX4 Rev.1.2 160MHz 1MB L2, all timinings best except DRAM burst disabled.
VLB Ark 1000VL 1MB

some dos bench marks:
Wolf 3D: 142.4
Doom max: 65.1
PCP 640x480: 10.9
Quake 640x480: 18.1 (Here you have 18.8 just as an example).

So not even in DOS I come close 😁. Does the DOS version play a role??? I used the one from Windows 98SE.
I keep trying, will upgrade the ARK to 2MB (but I don't think that makes a difference for DOS)
Then I will compare the valued with the ones of a SV2GX4 Rev.2.0 (where this "DRAM write burst" can be also enabled, but again, I also doubt that makes a difference)
Then I will try different CPUs. Can this make a difference? I also doubt it.

Before post some consolidated benchmarks for 386 class hardware will test some of the motherboards with IBM BL3 CPU - for completeness.
First in line is one of my favorite 386 mobos - JUKO 386DX W/B.
Two new sections added to the original post.

---

@CoffeeOne

If we assume that SVGOX4 does not have some built-in wait states and is equal to SV2GX4, which i think is the case, and if you use the same video card, then it must be the software configuration.
DOS setup is very compact and simple, so we can quickly find if something is off there.
The first apparent difference is that i use DOS versions in the range of 5/6/6.2/6.22 for this class of hardware.
Maybe something in DOS 7 (Win98) is affecting you ?

Also, can you post the contents of your AUTOEXEC.BAT and CONFIG.SYS ?

I noticed that @PC-Engineer posted same/similar performance numbers to mine with his SV2GX4.
This is a confirmation that i am not doing anything unique here.

pshipkov wrote on 2023-01-30, 03:39:

Before post some consolidated benchmarks for 386 class hardware will test some of the motherboards with IBM BL3 CPU - for completeness.
First in line is one of my favorite 386 mobos - JUKO 386DX W/B.
Two new sections added to the original post.

When noting specs and benchmarks, e.g. IBM BL3 at 100MHz, could you also specify FSB and maybe ISA freq as well? This way, when others try to replicate your settings and results, they can do so without wondering, if you are running the BL3 at 3x33 or 2x50 and ISA at 12 or 16 MHz. So it would look something like, IBM BL3 at 100 MHz (2x50) & ISA at 16 MHz At minimum, it would be good to note the ISA freq. in with the BIOS settings.

pshipkov wrote on 2023-01-30, 03:39:

Before post some consolidated benchmarks for 386 class hardware will test some of the motherboards with IBM BL3 CPU - for comple […]
Show full quote

Before post some consolidated benchmarks for 386 class hardware will test some of the motherboards with IBM BL3 CPU - for completeness.
First in line is one of my favorite 386 mobos - JUKO 386DX W/B.
Two new sections added to the original post.

---

@CoffeeOne

If we assume that SVGOX4 does not have some built-in wait states and is equal to SV2GX4, which i think is the case, and if you use the same video card, then it must be the software configuration.
DOS setup is very compact and simple, so we can quickly find if something is off there.
The first apparent difference is that i use DOS versions in the range of 5/6/6.2/6.22 for this class of hardware.
Maybe something in DOS 7 (Win98) is affecting you ?

Also, can you post the contents of your AUTOEXEC.BAT and CONFIG.SYS ?

I noticed that @PC-Engineer posted same/similar performance numbers to mine with his SV2GX4.
This is a confirmation that i am not doing anything unique here.

Hello,

I quickly retested with the same dos config from harddisk with the SV2GX4 Rev.2.0, same CPU, every setting on max, 64MB FPM (SVGOX4 had 32MB, but that should be really not matter. I hope at least)
OK, while PCP and Wolf3D are exactly the same, interestingly Doom and Quake got a tiny bit faster.
Quake now 18.3 consistently, before 18.1. But still off from your 18.8.

Let's check the DOS config.
Unfortantely I trashed the Windows 98SE installation, so only the dos part of it works 😁
so I don't have an easy copy / paste.

But it is like that (himem.sys, NO EMS, dos=high, german keyboard driver)
config.sys:
device = ..... himem.sys
device = ..... display.sys
country = ...... , country.sys
dos = high
lastdrive = f

autoexec.bat:
mode con codepage prepare ....
mode con codepage select=850
path = ....
keyb gr

Basically the only things that consumes dos memory is the german keyboard driver.
Something like 595000 bytes free.

Do you think it is the amount of free dos memory?
How do you run the dos bench tests?
himem.sys noems? no himem.sys at all? alternative himem.sys?

UPDATE: I found the values from PC-Engineer: SV2GX4, 160MHz ARK1000:
He had PCPBench 640x480: 10.9: That is exactly the value that I have. You posted somewhere 11.1
Quake: 18.4, I have 18.3, you have 18.8
Doom 67.4, I have 65.5, you have 68.

So I am somehow in between. My doom value is too low for some reason???

And still your 18.8 fps in Quake is outstanding 😁

Converted that PC to POD100, but the software setup is still the same.
DOS 6.22

config.sys:
DEVICE=C:\...\HIMEM.SYS /TESTMEM:OFF /M:1
DOS=HIGH,UMB (optional)
DEVICEHIGH=C:\...\EIDE2300.SYS /T /M0:8

autoexec.bat - nothing special/relevant to this:
PATH C:\...

I can spin it again in one of the next days to double check.
I remember that user Chadti99 mentioned something about the Quake 1 score too - cannot remember the details anymore.

Siemens ROLM 98d5192
Late industrial 386DX motherboard based on UMC's UM82C481AF UM82C482AF UM82C206F chipset.

Unusual beast in truly pristine condition. An evidence that it was used in clean room environment.
72-pin SIMM memory slots.
64Kb of soldered on-board Mosel Vitelic level 2 cache. No option to upgrade (jumpers, hardware path, etc)
External battery only.
No FPU socket.
A "Siemens certified" sticker from May 2001.

More capable boards based on this chipset were examined before, so didn't expect anything special from it and it didn't surprise either.

Supports TI/Cyrix 486DLC, TI 486SXL2 and IBM BL3 processors in addition to the standard 386DX ones.
Not fully stable past 40MHz base frequency, but decently ok at up to 50MHz.
Tested at the peak stability point (40MHz).
All BIOS settings on max except DRAM WAIT STATE = 1 (best is 0).

Didn't bother with full set of tests.
Some of them require FPU which requires a socket which i am not going to bother installing.

Used STB Nitro ISA (Cirrus Logic GD-5434), 2Mb video card.

--- IBM BL3 at 80MHz (2x40), ISA bus at 10MHz

Speedsys:

Wolf3D: 45.5 fps
Superscape: 33.3 fps
PC Player: 7.7 fps (without FPU this test is skewed)
Doom: 18.4 fps
WinTune 2: 2423 kpixels/sec

--- TI 486SXL2 at 40MHz, ISA bus at 10MHz

Wolf3D: 38.0 fps
Superscape: 23.2 fps
PC Player: 6.1 fps (without FPU this test is skewed)
Doom: 12.4 fps
WinTune 2: 1266 kpixels/sec

--- AMD 386DX at 40MHz, ISA bus at 10MHz

Surprisingly the board is unstable in this configuration.
Tried hard to overcome that but without success.
Had to step down to 33MHz for complete stability.
Lost interest right there.

If somebody is interested in scores of a patched 3DMARK99MAX for non-pentium-CPUs at a 180-MHz-AMD-486 and a Voodoo 3,
they can be found here:
Re: Modern graphics on a 486

Continuing with the reorganization of this thread by consolidating all collected performance metrics into single chart with references to it from the individual posts.

386 class motherboards with AMD 386DX, TI 486SXL2, IBM BL2/BL3 processors

How to read the data ?

This "study" took place over a time span of several years.
The intention was to identify the best motherboard and built the fastest PC around it.
All tests were carried out from the point of view of peak stable overclock instead of finding the relative clock-to-clock performance on a common basis (CPU/BUS frequencies).

Initially i had only AMD 386DX processors.
Later on managed to obtain SXL2 chips.
Last year switched to BL3 silicon.
At the same time i don't collect hardware, so throughout this CPU progression many of the previously examined boards were not with me anymore.
As a result quite a few boards were not 3-way examined.

The offline graphics tests (3D Studio R3 and LightWave3D) are very sensitive for complete system stability, but also - very time consuming - from tens of minutes to several hours.
Because of that only the most promising motherboards were grinded through them - the ones i considered building that mythical PC around.
So, in a way, these two test (especially the LW3D one and especially the BL3/SXL2 sections) are telling what the best stuff is.

With that said, i am going to fill some of the gaps for completeness.
All items listed as "to be tested" will be updated in the coming days.

Finally, this information covers many (most) prominent brands/models/chipsets from that time.

Gray bars indicate peak performance at compromised system stability.

Woof, massive data compilation. This must have taken some time. Some comments follow.

Not sure if this was an accident, but the Alaris Cougar 3x33 results are missing from the DOOM chart.

It is too bad you did not note the ISA speed. There are those who think running the ISA over 11 MHz is undesirable. From my understanding of your work, is it safe to assume the ISA speed is in the 13 - 20 MHz range, depending on the board and FSB?

To put the numbers into greater perspective, I was thinking it might be good to add results for an i486DX33 or maybe DX2-66.

If you wanted to add my SXL2-80 DOOM results on HaydnII, they can be found here: Re: Creating a voltage regulated 386 BL3 module from existing parts
Quick copy paste:

1DTK PEM-4030Y [Symphony Haydn II, GD5434, 256K]
2BL2: 80 MHz at 2x40 and ISA at 13.3 MHz, 3.6 V, then DOOM = 3654 = 20.44 fps
3SXL2: 80 MHz at 2x40 and ISA at 13.3 MHz, 4.1 V, then DOOM = 3406 = 21.93 fps
4
5Chaintech 340SCD [SiS Rabbit, GD5434, 256K]
6BL2: 80 MHz at 2x40 and ISA at 13.3 MHz, 3.6 V, then DOOM = 3562 = 20.97 fps
7SXL2: 80 MHz at 2x40 and ISA at 13.3 MHz, 4.1 V, then DOOM = 3801 = 19.79 fps

Took a moment to put everything in one place, but it will be a time saver going forward.

I left few gray bars in the charts as an indication of top performance achieved regardless of complete system stability.
In the case of Alaris Cougar 3x33 for Doom - it is slower than DTK PEM-4036Y with BL3 processor running at 110MHz, so it was not added there.
Only fully stable systems and their results are listed (well, with few small exceptions).

I started adding ISA bus speed to the individual posts.
The left side of the charts (the row description) gets too long if ISA bus frequency is added.
Excel does not allow me to stretch that area further and part of the string gets lost.
But from what i have seen so far - 10 to 14 MHz is working pretty well almost everywhere.
Many boards and expansion cards can handle 15-20MHz just fine.

Added your numbers.
If you can provide the rest of the simple DOS interactive graphics benchmark results at least, that will be great.
They are easy to achieve and quick to test.

Thanks, that's an awesome comparison. As helpful as the 486 586 stickies.

I was daydreaming about BL3s last week. Think it was because I got a sniff of a color screen 20mhz portable.

BitWrangler wrote on 2023-02-13, 02:17:

Thanks, that's an awesome comparison. As helpful as the 486 586 stickies.

I was daydreaming about BL3s last week. Think it was because I got a sniff of a color screen 20mhz portable.

Which are unfortunately not sticky per se...

Thanks BitWrangler.
---
I feel i am missing the joke with the "sniff of color screen 20mhz portable" is ?

I meant I saw a larger format old 386 laptop for sale some distance away with a color LCD, i.e. I caught the scent of it as a hunting hound would, but that it's likely to get away. But that the possibility though slim of getting it, had caused me to wonder if it could be upgraded with a BL3 and whether that would make a nice machine.

One time I seen a no-name DTR-like notebook that used 386DX, 3.5" hard drive and floppy in a 2 1/2" thick notebook long ago, opened up for service. Had not seen one again ever since.

It is not luggable form factor, really looks like very thick notebook computer.

And I was very sure since this has ceramic 386DX. Not the jelly bean 386SX stuff.

Cheers,

I'm pretty impressed with the performance of the Alaris Leopard. It's a gimped 16-bit motherboard, but it can still compete with 32-bit boards.

pentiumspeed wrote on 2023-02-13, 21:58:

One time I seen a no-name DTR-like notebook that used 386DX, 3.5" hard drive and floppy in a 2 1/2" thick notebook long ago, ope […]
Show full quote

One time I seen a no-name DTR-like notebook that used 386DX, 3.5" hard drive and floppy in a 2 1/2" thick notebook long ago, opened up for service. Had not seen one again ever since.

It is not luggable form factor, really looks like very thick notebook computer.

And I was very sure since this has ceramic 386DX. Not the jelly bean 386SX stuff.

Cheers,

This one had that kind of look like a portable CAD/Scientific machine they would have been at the time, power user only for the price, but it may have just been a lower end SX machine made chunky to take cheap desktop size parts.

Guys, please, does anybody know if this Abit AB-PM4 support AMD 5x86-133MHz? Bios shows for every settings max AMD DX4-120MHz.

@Anonymous Coward

Agreed.
Alaris Leopard is neat, but only the C variant - the one with VLB bus and L2 cache.
Alaris Cougar passes for the premium version, but if you look at the benchmarks - Cougar (3x25) is slower than Leopard in offline compute and FPU heavy tasks, which makes sense.

@barry07
If the board functions with Am5x86 CPU then you are good. The fact that BIOS does not list it properly during POST is not a problem.
How do you like this board ?
It will be great if you share some info about it in 4x40MHz or faster configuration.

I've never seen a Leopard rev A or B. I assume they exist...but wouldn't they also be VLB?
I think the board you're thinking of is the Leopard LX (sometimes called Opal LX). That's a different product.