@BastlerMike

When I did some 486 PCI mobo benchmarking with PCI videocard performance I came to the conclusion that Intel and UMC chipset mobo's had the worst performance so I would do a different chipset rating that yours

I got the best score out of an MSI MS-4143 VLB/PCI mobo with ALi chipset

Hmmm, isn't the MS-4143 a PCI-only board ?
I remember it had ALI M1439/M1445 chipset and the performance wasn't very good.
I think you mean MS-4134 !?

BastlerMike wrote:

Hmmm, isn't the MS-4143 a PCI-only board ?
I remember it had ALI M1439/M1445 chipset and the performance wasn't very good.
I think you mean MS-4134 !?

You are correct, my mistake

What program and what graphics card(s) did you use ?

BastlerMike wrote:

What program and what graphics card(s) did you use ?

Hercules Dynamite 128(Tseng ET6000), Hercules Stingray (ARK 1000)
Matrox Mystique, Diamond Stealth 3D 2000 (S3 Virge DX)

I used pcpbench for the benchmarks

When I looked for a suitable board for my 486 machine years ago, I got three boards donated from friends and acquaintances:

FIC 486-VIP-IO2 (VIA 82C486A/82C505)
Asus PVI-486SP3 (SIS 496/497)
Asus PCI-I 486SP3G (Intel Saturn)

Old articles from C't magazine suggested that the Intel chipset board would leave the others in the dust, but in reality, it turned out to be the slowest. Besides, it doesn't support 40 MHz bus speed, which is very nice to have when you want to run a fast 486.

The FIC board had a lot of features, but the BIOS was ancient and turned out to be non-flashable, so I went with the SIS 496.

5u3 wrote:

FIC 486-VIP-IO2 (VIA 82C486A/82C505)

FIC 486-VIP-IO --> VIA 82C486A/82C505
FIC 486-VIP-IO2 --> VIA 82C496G/82C505

😉

Today I found an old papersheet with several bus performance measurements (via 32-bit STOSD) I made some time ago.

Regarding these, I have to revise my order I posted earlier.
I will make a table and present it here. Some results are quite interesting.

Let's kick off some Pentium II results already (btw where did my 486 go?)

Intel-Pentium II-66-3.5-233-umm-iforgot2000-i440LX-512kb-AGP-PowerVR KYRO II-64MB-leilei-195.3

Sheet has been updated! Post #2

The following table shows the maximum bus transfer rates (MB/s) of different boards/cards combinations. The graphics cards Hercules Dynamite 128 (Tseng ET6000), Matrox Millennium (MGA2064) and Elsa Winner 1000 (S3 Trio64V+) were used.
They were first tested on a Pentium mainboard (MS-5120) to determine the card's max. data transfer rates when no bus limitations are present. You can see that in this case the rate is limited by the cards and not by the bus.
All other cases show that the transfer rates are not limited by the cards but only by the bus.

Hmm, i don't know why the SIS is so slow and I don't see any "suckage" with the UMC Chipsets 😒

retro games 100 wrote:

Generally speaking, is a socket 3 PCI board perfectly satisfactory, or would you consider a decent VLB/ISA board to be superior to all PCI boards?

Regarding the results, I would say that the PCI bus yields no significant higher transfer rates on 486 systems compared to VLB designs.
Every VLB board should be capable of transferring >35 MB/s over the bus, even more when clocked higher than 33 MHz.
The major benefit of having PCI is the higher flexibility because of the huge assortment of card alternatives.
-only an opinion-

@bastlermike

Did you use 3dbenchmark for your test? I believe it mostly stresses the CPU

I don't remember the name of the tool, i was something with "transfer..." or something 😕
These measurements are only bus- not CPU-dependent. In all cases the bus frequency was 33 MHz. The tools writes data to the video memory and so determines the maximum bus transfer rates. Tests were executed with DX2/66 or Am5x86- 133 (P5-66 and P55C-166 on the Pentium boards).

I have been inspired by the info in this thread, and decided to benchmark a 486 mobo using 3DBench. I used only one CPU for every test, an Intel SX-33, and set the mobo's bus speed to 25, 33, and also to 40. I used an S3 PCI video card for all tests. After the tests were done, I sorted the results table in 3DBench score order. Looking at these results, I am guessing that this 486 mobo can practically cover all ground from an early IBM PC, up to the fastest 386. And beyond. 😉

O M G what a herculean effort! Must have taken you quite some time to compile this list!

Very nice scaling, quite impressive...

Mau1wurf1977 wrote:

O M G what a herculean effort! Must have taken you quite some time to compile this list!

Very nice scaling, quite impressive...

Thanks! 😁 Hehe, I should have made a note of what time I started today. I think it's taken me about 7 hours.

I would pack up all the parts you used for this test, put it in a box and slap the benchmark results on the box. Because this will come in handy one day...

Actually, I reckon these test results will be approximately the same, if you used most PCI 486 boards. I used an Asus PVI-486SP3, with 32 MB of FPM RAM, and a Number Nine S3 video card. I think if you were to use roughly equivalent parts, you would get roughly equivalent results. I think the most encouraging thing about these test results is that using a combination of L1 / L2 / System speed / Turbo / BIOS timings, you can achieve flexible fine tuning, in order to achieve sub 486 speeds. 😀

I think that most 486 BIOSes have L1 and L2 disable options, and probably every 486 board has a turbo button. I expect that most if not all mobos have a system FSB speed that you can adjust. The BIOS timings only made a small impact, so they are not important, but I think that most 486 BIOSes have some memory timing options. I enjoyed doing those tests, and making that table. I like lots of numbers! 🤣

I don't know where you guys get the patience to compile all these titanic benchmarks, but you guys just ROCK!!! So much of informations!! (Btw, my head's already spinning looking at the list! 😁)

Malik, I was really impressed with your socket 5 classic Pentium slow down project. I forgot to mention above that it was your thread alongside Mau1wurf1977's original post that inspired me to do the 486 speed testing project. 😀

This post is an excellent resource- good work to all who contributed.

I'm wondering if anything similar has been done for PCPBench? Main reason I ask is because my 486 is pretty sluggish in SVGA modes and I was looking to replace the S3 805 VLB card with something a bit snappier. I have an old LPX motherboard with an integrated S3 Trio64 and with the same processor in this board, the system is significantly faster/smoother in hi res modes.
PCPBENCH tests at the higher resolutions but unfortunately I can't find many PCPBENCH results on the net.