As I am covering more S5 chipsets I have to deal with some really quirky ones.

Right now I am testing the Opti Viper-M 82C556/557/558. This is the most problematic chipset so far.

At least the motherboard I got iis not compatible with DOS4GW extender (just hangs). First time I experienced that . Everything else works fine. I had to patch binaries to replace DOS4GW for DOS32A and then I can run DOOM and other DOS4GW stuff. However, not sure if the results are comparable now...

A 1995 motherboard not compatible with DOOM sounds like users must had a plenty of fun with it!

mpe wrote on 2020-06-02, 11:16:

As I am covering more S5 chipsets I have to deal with some really quirky ones.

Right now I am testing the Opti Viper-M 82C556/557/558. This is the most problematic chipset so far.
...

But quality of the board is really good. The manufacture used solid-state capacitors that is very unusual.

mpe wrote on 2020-06-02, 11:16:

As I am covering more S5 chipsets I have to deal with some really quirky ones. […]
Show full quote

As I am covering more S5 chipsets I have to deal with some really quirky ones.

Right now I am testing the Opti Viper-M 82C556/557/558. This is the most problematic chipset so far.

At least the motherboard I got iis not compatible with DOS4GW extender (just hangs). First time I experienced that . Everything else works fine. I had to patch binaries to replace DOS4GW for DOS32A and then I can run DOOM and other DOS4GW stuff. However, not sure if the results are comparable now...

A 1995 motherboard not compatible with DOOM sounds like users must had a plenty of fun with it!

That Board is 100% DOOM compatible it is just slow.
You have some defective / misconfigured part.
i also own OPTI Boards and they all work fine.

Where are the Cache Chips ? Do you run EDO or FPM ? did you set jumpers according ?
Do you run VLB or PCI Card ?

As for the L2 - I am testing all combinations - 256kB / 512kB / 1024 kB async as well as 256kB and 512kB PB COAST modules. PCI VGA and EDO RAM. Configuration options are sparse on this board. Here is the full board:

I had another Opti Viper board which worked fine although that one used a newer version of OPTi Viper and did not have any async SRAMs sockets - just the CELP slot

I suspect this could be some BIOS issue. Also investigation chipset register if I can tweak it a bit as usual utilities don't work. Will have to spent some time on this one. So I guess the results could change a bit.

Your board seems like it might have come out of an AcerPower system. Do you know the model?

No. But perhaps it can be established from the POST string:

Acer AN5C

Need to identify these two old motherboards

Thanks at least I know what it is. I could find any useful stuff about the board (BIOS, manual).

However, I found datasheet for the chipset with register info. So perhaps I can write a tool to compensate lack of BIOS options with some programming...

I have a feeling you will probably find the manual one day. Judging by the connectors on it, it looks to have come out of an OEM system (probably at least Acer, maybe Ambra, Fujitsu and others)...so whatever documentation or BIOS you find will be the OEM version. Since your splash screen says Acer, I would put my money on the AcerPower series, but maybe Altos.

Could you make some tests on early chipsets with P75? To compare it with supercharged 486: 1 2

Your Am5x86 180MHz system is more on par with a P90 or P100 on Triton FX. I did some tests with P90 and P100 with Neptune and 256kB, and it was equal with my POD100 (DOS P90<POD100<P100, Windows WS95 P100<POD100). The Am5x86 @160 is faster than POD @100 (except Quake) and you have it @180. So it should be comparable with a Pentium between 90-100MHz on a Triton FX.

The P75 is definitely not the suitable opponent. 😉

The P75 is touching one of my other projects. I will have some numbers soon... But it is mostly like it was said and even the FX won't change the landscape that much.

For the chipset comparison I've chosen P133 since it is the CPU that allows me to compare all Pentium chipsets (Socket 4, 5 and 7) which is very convenient.

Your Am5x86 180MHz system is more on par with a P90 or P100 on Triton FX.

P75 is quite on the mark when it comes to FPU heavy scenarios, i.e. lots of stuff with 3D acceleration. Or probably and more precisely - P82 with 55 Mhz FSB. But that's on 430FX board and only when PB cache not used. I'm curious how "nextgen" SIS chipset will behave in such comparison.

I've added two more chipsets to my survey.

• SiS 5511
• Intel 430TX (with both SDRAM, EDO)

Obviously the 430TX has taken the crown. However, the SiS5511 while not eaten into top-tier showed some good performance.

mpe wrote on 2020-09-09, 21:30:

I've added two more chipsets

Haven't you tested yet your Socket 5 VLSI board?
Re: Bought these (retro) hardware today

Sadly. That board isn't booting. It is using proprietary Dell PSU and I still need to troubleshoot where the problem is.

I finally got a motherboard with VIA Apollo Master Plus chipset - FIC PA-2000 and updated the article.

It looks like the VIA took the crown for the slowest Pentium Socket 4/5/7 chipset I tested. But I am sure there are even slower chipsets, just need to find them 😀

With 25fps in Quake we are slowly getting into 486 territory. And that's with Pentium 133!

mpe wrote on 2021-03-28, 11:06:

I finally got a motherboard with VIA Apollo Master Plus chipset - FIC PA-2000 and updated the article. […]
Show full quote

I finally got a motherboard with VIA Apollo Master Plus chipset - FIC PA-2000 and updated the article.

It looks like the VIA took the crown for the slowest Pentium Socket 4/5/7 chipset I tested. But I am sure there are even slower chipsets, just need to find them 😀

With 25fps in Quake we are slowly getting into 486 territory. And that's with Pentium 133!

Nice chart!
Looks to me that HX with 512kB L2 would take the performance crown.
Well atleast in Quake.

Roman555 wrote on 2020-06-03, 08:52:

But quality of the board is really good. The manufacture used solid-state capacitors that is very unusual.

those aren't solid caps, they're just SMD electrolytics. in fact they are worse than regular thru-hole electrolytics. the only solid caps i've seen on certain 90s boards are those purple OS-CONs, and even those supposedly don't age the best.

regarding the comparison, definitely nice work, just one note (and i think i've made this point in another thread already): comparing stock OEM LX/NX boards with aftermarket hand-tuned HX/TX and listing the results on a chipset level basis doesn't quite sit right with me. IMO it would be very useful to show which results fall into which category using asterisks or something.

i'm still pretty interested in whether anyone ever managed to improve results on the OEM LX boards with BIOS editing software.

On OEM boards and where chipset datasheet is available I am always checking/re-programming chipset registers to make sure the tightest possible memory/cache/PCI settings are applied (just in case the BIOS doesn't enable everything). That's being said there is very little to tune on a 430LX board. Intel Batman or other Intel boards might lack advanced BIOS settings, but turns out to be maxed out everything important elready. I tested other 430LX boards, but Batman turn out to be fastest or just as fast. Other boards might have 512kB of cache, but often also EISA or buffered memory which slow things down.

Futhermore, I tested several boards and where there are differences I picked the fastest one. Such as the UMC8891. My other UMC boards (PINE PT-730A and Shuttle HOT-539) don't show as good results as the Gigabyte GA-586AM. Sadly UM8891 is one of the chipsets where the datasheet is nowhere to be found so hard to say why exactly.