@rg100
I have no idea. I have found package containing these tests here:
http://homepage.virgin.net/roy.longbottom/dostests.zip
Package contains optimized and non-optimized version. @feipoa which version did you use?

Are you refering to the Biostar MB-8433UUD? You now have two of them? Which versions? Are they both version 2.0? If you get a version 3.0 or 3.1 and want to give it good home, let me know. I don't like being without a backup.

What are differences between rev 2.0 and 3.0/3.1?

EDIT:

On the first comparison sheet, there was a mention that those benchmark programs were from Roy Longbottom's dos benchmark pack. You can find them here:
http://homepage.virgin.net/roy.longbottom/oldones.htm

He he he you were faster. [:

Ok, I have reread your document carefully - optimized versions were used. (;

Thanks guys. BTW, udam_u, your exercising rodent is fascinating. 😀

feipoa wrote:

Are you refering to the Biostar MB-8433UUD? You now have two of them? Which versions? Are they both version 2.0? If you get a version 3.0 or 3.1 and want to give it good home, let me know. I don't like being without a backup.

Both version 2.0.

Tetrium wrote:

I went through my pile of hardware I have yet to test and somehow a 8433UUD-A version 2 made it into the pile. No idea how I got it 🤣, except that it must've been pretty recent. No idea if it works though, but it seems to look fine.

That's what happens when you buy bulk lots of old computer parts. You end up with things you never knew you had.

retro games 100 wrote:

Tetrium wrote:

I went through my pile of hardware I have yet to test and somehow a 8433UUD-A version 2 made it into the pile. No idea how I got it 🤣, except that it must've been pretty recent. No idea if it works though, but it seems to look fine.

Hehe, nice find! Very roughly, how many items have you got to test?

Uhm...remember those kiwi boxes I told you about that I use for storage?

Well....a stack of about a meter high or so of motherboards...and a bag of memory modules (and I mean like one of those larger plastic bags 😜), about 12 to 15 graphics cards of all ages (AGP, PCI and ISA) and a LOT of processors, including lots of processors I've had for a longer while now but didn't bother to test...or no time to do the testing.

I did a quick count on the motherboards (so I could look up drivers, manuals etc), the stack itself has 24 boards and I got a few more waiting elsewhere.

Edit:The motherboards are really a mish-mash of generations, from a couple 486's, couple (Super) Socket 7, couple Tualatin boards, couple P4's, some Socket A SDRAM, a few Socket A DDR AGP 4x and that odd VC820 Slot 1 RIMM board.

The graphics cards include a couple ISA Tseng Labs, couple S3 VLB (I think one of them is dead though), one dual head Matrox AGP made in 1999, TNT2, TNT2 Ultra, those 2 Voodoo 3 3500TV's I need those adapters for, 3D Prophet 4500 64MB TV-Out and then some GF2MX's or Vanta's. And somekind of Radeon 9000? and I'm pretty sure I "should" have a couple GF's laying around, so I'm not quite sure if 12 to 15 is correct.

Edit2:But I might wait till my last batches of chips (and 1 board) arrive before I begin full scale testing. I'm currently tidying the attic bit by bit and it's time I start storing more in 3D 😁
Ever since I moved to my new house, I basically put stuff "somewhere" and ended up leaving lots of those "relics" as-is, so high time to finish what I've started 😉

@udam_u
We discussed some of the differences between v. 2 and 3 here,
Is this a good 486 PCI board?

Speaking of hard to find old benchmarks, has anyone had any luck finding the original 3DMark99? The non-Max version.

Here is something intesting I stumbled upon during my benchmark tests -- USB PCI cards in a 486 motherboard may drop your Windows-based benchmark scores by 15% or more. Is continual software search occuring whereby the USB is constantly checking if a USB device has been inserted? I'd have thought the sample rate would have been at least 1Hz and would not have affected the benchmark programs.

Disabling the USB host controller in the device manager seems to fix the dropped scores.

feipoa wrote:

Speaking of hard to find old benchmarks, has anyone had any luck finding the original 3DMark99? The non-Max version.

Here is something intesting I stumbled upon during my benchmark tests -- USB PCI cards in a 486 motherboard may drop your Windows-based benchmark scores by 15% or more. Is continual software search occuring whereby the USB is constantly checking if a USB device has been inserted? I'd have thought the sample rate would have been at least 1Hz and would not have affected the benchmark programs.

Disabling the USB host controller in the device manager seems to fix the dropped scores.

For benchmarking, it's in theory best to make the OS as lean as possible, even if the system becomes unusable for anything but that particular benchmark 😜
It depends on how far you want to go.

@rg100

Thanks guys. BTW, udam_u, your exercising rodent is fascinating. Happy

He he he I think that this is another incarnation of Duke Nukem! ^^
I would like to be as tenacious as he is! [:

@feipoa
Thanks for the link! I have just bought Biostar rev. 2 motherboard.
I wonder if there are other differences than chipset supporting EDO RAM and newer BIOS version...

I have found 3D Mark 99 Lite on one of my old dusty CHIP CD. It also contains Fog City Benchmark and Tirtanium Benchmark. Here you are link to 3DMark 99Lite:
http://www.mediafire.com/?4x7p4i54u7vn208

Regards! (:

EDIT:

Is continual software search occuring whereby the USB is constantly checking if a USB device has been inserted? I'd have thought the sample rate would have been at least 1Hz and would not have affected the benchmark programs.

I think that such modern hardware as USB cards work always in interrupt mode that should consume much less CPU power than software polling. However some operating systems allow to share one interrupt between several peripherals (for example my favourite operating system QNX). In this case it is possible that your graphics card share one interrupt with USB card and when interrupt occurs operating system has to check both cards to determine which caused this event. This can significantly reduce performance if device driver has been made carelessly.

@udam_u

Thanks for the link to 3dmark99. Unfortunately, the "lite" version is not any more reliable than the max version. I wonder if any Pentium instructions are sneaking their way into the code? I can only get 3dmark99max to sometimes complete. I added a fan cooler to the graphics cpu, but that didn't help. I get 14 3dmarks and 5 cpu marks. Unfortunately, this benchmark program is not reliable enough on 486's to be added to the list.

You just bought an MB8433-UUD v2.0? There seem to be a surplus of v2.0's. Other differences between v2.0 and 3.x are below.

3.x fixed issues with NCR SCSI cards and added support for memory parity checking. I have found, however, that if you had parity RAM, and you performed a soft-reboot (i.e. ctrl-alt-del), you'd get a parity error until you hit reset. Maybe this was fixed in BIOS version UUD 960520 found on some v3.1 boards? Ver 3.x also had faster IDE I/O for some reason. Some reports say a 20% increase. I wonder if that's the difference between PIO4 and PIO3? Version 1.x and 2.x of this board have been known to have intermittent use issues with combined PS/2 mouse usage and keyboard strokes.

I hope none of this discourages you testing our your Biostar board as I'd like to see v2.0's be usable. I don't recall exactly what I tossed mine out in the past (7 years ago), but I'd like to hear encouraging results from your tests as this would make me reconsider 2.0 as a backup to my 3.0 board.

As far as the USB PCI card goes, it wasn't sharing any IRQ's. I think its pulling the PCI bus at some regular interval which slows down traffic. I hadn't noticed a speed decrease with leaving the ISA sound and network card in place.

@tetrium
Yes, you are correct. If I could do it all over again, I'd have pulled everything out which was not necessary, but it is too late. Too many hours of testing have been put in. As long as the same conditions exist for all cpus tested and the results are comparative, it shouldn't matter.

Does anyone know what the gametics/realtics in the doom demo mean? I am not seeing a trend that I would expect. The quake demo trend is a lot more believable.

feipoa, I'm sorry that it has taken me this long to get the 486 Biostar & "X5-200" system up and running again. I have got it working now. I'm a bit short on time ATM, but I have run one test, and everything looks stable for me to now run through all of the other tests. I'll get this done within the next 48 hours.

To begin with, I have got the "crucial" Quake DOS test stable. So I have got Quake 1 (1.06 shareware) set to full screen, 200 MHz CPU clock speed, BIOS PCI divider option set to "1:1", BIOS cache timings set to 2-1-1-1, the Read DRAM timing set to "1" (not "0"), 50 MHz mobo FSB, and I get a score of 18.3 FPS. That's the score I remember getting a month or two back.

Please note these are aggressive timings for an OC'd P75 operating at 200 MHz. I'm going to add some more benchies to your "mega test list", and include WinTune 2.0 for Windows 3.x, WinTune 97 for Windows 95, and WinTune 98 for Windows 98. For those last two tests, I will probably have to loosen up the cache timings to something like 3-1-1-1...

I've done all of the DOS tests. BTW, I'm using 1 stick of 32MB EDO, and a "crappy" Virge 325 PCI VGA. To do: I will configure the system to run in "3x 60 FSB = 180 MHz" mode, and rerun all of these tests below, before going on to do the Windows tests.

Quake (shareware 1.06, timedemo demo1, fullscreen) = 18.3 FPS
Doom (shareware 1.9s, timedemo demo3, fullscreen) = 2134/1122
PcpBench (VGA mode, not LFB mode) = 29.0
PcpBench (Mode 100 - 640x400, not LFB mode) = 11.7
Norton Utilities version 8.1 SysInfo = 432.2
PC-Config v9.33 = 126% (% of Pentium 100). Also - Dhrystones, KWhetstones = 96000, 24528. VideoRAM throughput (text) = 9500 KB/sec.
3DBench v1.0c (version 1.0 went off the scale) = 117.4 (With a PCI Diamond Viper, I got 118.1)
Cachechk v4.0 & also v7.0 (I've only got 256Kb of 15ns cache) = L1: 206.2, L2: 93.1, Memory: 46.4, RAM read: 90ns, RAM write: 40ns.
Dhrystone benchmark v1.1 = 195.21
Linpack benchmark = 6.73
Whetstone benchmark = 50.401, and then the N1 to N8 scores are: 15.521, 12.048, 18,893, 19.595, 1.625, 8.110, 16.218, 1.053, total seconds = 99.601.
CPU index v2.3 = 18
PiDOS = total time: 00d 00h 00m 13s
Landmark v2.0 = This computer performs like a 669 MHz AT with a 163 MHz 80287. CPU: 669.30 MHz, FPU: 1636.03 MHz, Video: 20480.00 chr/ms
Bytemark v2, 32-bit DOS = overall integer index: 1.285011, floating point index: 0.741390, 90 mhz dell pentium = 1.00

SpeedSys v4.78 = 75.00, memory bandwidth = 108.75 MB/s, VESA memory = 14756 KB/s, L1 = 176.64 MB/s, L2 = 74.34 MB/s, Memory throughput = 52.04 MB/s.

Justin benchmark for DOS = Memory operations - step 108 Current: 13.65, Average: 13.67, Float points operations - step 171 Current: 16.73, Average 16.51. Average PC index = 12.2 but ignore this final score, because I'm using a slow compact flash drive, and I quit the disk test fairly quickly.

I have started the "3x 60 FSB = 180 MHz" tests. In order to get the Quake test to run without errors, I had to do 2 things. 1) Set the BIOS PCI clock divider option to "1:2/3". This sets the PCI bus to 40 MHz. I tried "1:1", but MS-DOS would not boot. 2) Set the cache timings to 3-1-1-1. Previously, they were set to 2-1-1-1. For Quake, I get 17.0 FPS. In my previous tests, I got 18.3 FPS with the system set to 200 MHz. For PcpBench (mode 100) I get 11.3. In my previous tests, I got 11.7 with the system set to 200 MHz.

Because the scores for the 180 MHz system are slower than the 200 MHz system, I will now abort these "3x 60 FSB = 180 MHz" tests.

I have run WinTune 2.0 for Windows 3.11 (WfWG). The resolution used was 1024x768x8. The system speed has been put back to "4x 50 FSB = 200 MHz", with cache timing set to 2-1-1-1. The BIOS PCI divider option was also reset back to "1:1". The results are attached below, inside a .TXT file.

I'll have to burn through your work in a day or two; I'm pretty busy today. Here is a more specific list of exactly what I am testing for. If you could replicate these conditions, that would be fantastic. This way, I can add your results to the sheet if I cannot get my X5 running at 200 Mhz. I haven't gotten this far yet. It has taken a lot more time that I first thought to work through all these tests on 2 dozen cpus.

***DOS***

1) Symantec (Norton) Sysinfo v8.0 (arb. units)
2) PC-Config v9.33 (% of Pentium 100)
3) Cpu Index v2.3 (arb. units)
4) PiDos [25k digits] (seconds)
5) Landmark v2.0 - ALU, FPU, Video
6) Justin Benchmark Dos v1.0 - Memory and Floating Point

7) Bytemark v2.0, 32-bit DOS - All items, everything quoted in (iterations/second), except for the final Integer/Floating Point index, which should be quoted in (normalized index of P90)

Roy Longbottom
8) DHRY1OD (VAX MIPS Rating)
9) LINPCOD (MFLOPS)
10) WHETCOD - all items quoted in MFLOPS and MOPS

11) Speedsys v4.78 - General score, Video bandwidth, system bandwidth, L1, L2 cache, and ave. RAM speed

12) Cachechk v4.0 - L1, L2, Memory, Ram access time (read and write)
13) 3dBench v1.0 or 1.0c
14) Doom v1.9s timedemo demo3
15) pcpbench v1.04 (VESO Modus 100, 640x480 8bpps LFB)
16) Quake v1.06 timedemo demo1

***Windows 98SE***

17) SuperPi v1.1 (32k digits)
18) Justin Benchmark Win v1.0 - all items, averaged over 3 trials

19) Ziff-Davis Winbench 96 v1.0 - CPUmark32 and Graphics Winmark
(make sure you right click your taskbar, properities, and uncheck "always on top" for the graphics winmark to work)

Ziff-Davis Winbench99 v1.1
20) CPUmark99 -v1.0, this is the standalone addition, not bundled with Winbench99
21) FPU WinMark99 v1.1, this is the edition bundled with Winbench99 v1.1

22) WinTune98 - Integer (MIPS), FLoating Point (MFLOPS), Video2D (Mpixels/s), Direct3D (Mpixels/s), OpenGL (Mpixels/s), Memory (MB/S) - It is easiest to refer to the html the program creates after running the tests.

23) Sandra99 - CPU ALU/FPU, Multi-Media ALU/FPU, Memory ALU/FPU
24) PassMarkv4.0 - 2D graphics Mark, Memory Mark, Math Mark, Integer Add/subtract/multiply/divide, and FPU Add/subtract/multiply/divide, and Max Math MFLOPS.

25) 3DMark99Max or Lite - If you can get it to finish.

FYI, Cyrix 5x86-120 gets 17.2 fps in quake, X5-160 gets 17.3 fps, POD83 gets 19.7, and POD100 gets 23.6.

feipoa wrote:

It has taken a lot more time that I first thought to work through all these tests on 2 dozen cpus.

That's completely understandable. 😀

feipoa wrote:

FYI, Cyrix 5x86-120 gets 17.2 fps in quake, X5-160 gets 17.3 fps, POD83 gets 19.7, and POD100 gets 23.6.

Wow, the overclocked POD is the Quake champ. I've got one here I could try. I remember having difficulty getting the POD to overclock and run Quake successfully, although I'm not entirely sure which mobo I was using. I'll try it again on the Biostar. I'm guessing you had to loosen up the BIOS timings. Are you using any fancy cache chips, eg 12 or 10 nanoseconds?

While waiting for my wife to get ready, I was able to read thru your work. I have a few comments.

For the Windows98SE tests, can you use 1024x768x16bit, DirectX 6.1a, and with all the latest security updates? ;) Is the X5-200 long-term stable in Windows 98SE?

I have used DOS 7.10 included with Win95 and 98's. I doubt this will have any impact on your results using 6.22. What do you think?

Did you confirm that the ADZ does not overclock at all to 200 Mhz and that only the ADW does? What led you to this determination? How many versions of the ADZ did you try? The two I have are of different revisions (as seen by the CPUID) whereby the 4F4 seems to overclock better to 160 Mhz than the 494 one.

Why are you not using LFB mode in pcpbench. My pcpbench defaults to VEST Modus 100, 640x480 8bpp LFB? How do you change the modes?

Also, some of my X5 results from the previous sheet were for L1 WT mode. I don't know how it got set to that, but I've corrected it. The results were lower than they should have been. I've doubled checked all L1/L2 cache modes for the second revision.

Does ctcm, and chkcpu16 show your POD83 as L1 Write-thru even though you put Write-back for the BIOS setting? I thought the POD83 was WB. I wonder what is going on here.

One more thing, did you set PiDOS to 25k digits?

@feipoa
Thanks for detailed explanation! (:
Yes, I bought it recently for 10$ (mobo+shipping). I'm sure that I spend some time with Biostar mobo. I'm especially curious about 1:2/3 PCI divider...

I am grateful to you for the selection of benchmarks listed above. Would have been nice if they have become a standard here.

Did you confirm that the ADZ does not overclock at all to 200 Mhz and that only the ADW does? What led you to this determination? How many versions of the ADZ did you try? The two I have are of different revisions (as seen by the CPUID) whereby the 4F4 seems to overclock better to 160 Mhz than the 494 one.

I have tested five am5x86 ADZ on diferrent motherboards. All worked fine at 160MHz, two were stable at 180MHz and one booted up at 200MHz but was unstable. I think that ADZ version behaves better with standard/lower voltage but ADW version shows the claw when you apply 5V. ^^

@rg100
Your am5x86 is insane! [: I tested today my am5x86@180 on ga-486am/s mobo with fsb:pci divider set to 1:1. It booted to dos correctly from flopy drive. However, I haven't tried with hard drive.
Did you try to boot from floppy drive?

@rg100

I did not have to loosen up any cache or memory timings. All at 0/0 and 2-1-2. It would not be a fair comparison for the other cpus if I had to change some settings on some and not on others. The POD100 was stable enough in Windows to run all the Windows98SE-based tests I mentioned, but I am still not convinced as to the long-term stability. I have 3 POD83's and only 1 of them was stable enough to run all the tests at 100 Mhz. They all have the same stepping/rev though. I think this may be where the dislocation ghost comes into the picture.

I also suspect that the Cyrix 5x86-133 will beat, or be onpar with the AMD X5-200 in Quake -- I haven't gotten to testing it though. I have a sexy matrix ready that I'll post when testing concludes. There's also a few processors that I'm waiting to come in the mail to add to the list, like the Cyrix 5x86-80, AMD DX4-120, AMD DX2-66, and X5 ADW. The DX2/DX4's would be nice to see compared against the Intel DX2/DX4 and Cyrix DX2/DX4. I have already overclocked the Intel DX4-120 (16KB WB) and ran all the tests, the results were quite good, and generally better or on par with the X5-133.

@udam_u

Does wherever you got the Biostar MB-8433UUD v2.0 from have another one? I'd like to rigurously test this against the 3.0 board. I've been thinking more about it, and I may have tossed my few v2.0's because they didn't work well with the Cyrix 5x86's, but memory fades with the years, and even lies to you. It may just be that I was having a zen moment and wanted to reduce burdensome objects in my life.

@feipoa
I bought this mobo from another source than ebay. MB-8433UUD is very rare in Poland - it was the first time I saw this motherboard here. I really doubt seller has one more but I'll ask him. (;

I wonder if is it possible to limit some malfunction listed earlier by updating the bios?

Regards! (: