Would it have been possible to replace the missing RAM clips on the board that was missing them with clips from a junk motherboard?

Thank you for uploading 4.00A BIOS. Unfortunately, it did not fix my L2 and memory speed issues. Perhaps you can upload a Speedsys image of your 4DPS w/AMD X5-133 and of which chipset settings you are using?

You do not need a Cyrix 5x86-133, you can use an IBM 5x86-100HF and run it at 2x66 for 133 MHz.

Using 10 ns cache vs. 12 ns or 15 ns does not seem to improve the benchmark results at all.

I use the same Samsung 60 ns FPM RAM you are refering to. They are 64 MB per module, with parity and gold contacts. It is the best FPM RAM I've ever found. I have 4 sticks of them and would like another 4, but don't see them for sale anymore. Can you interleave with only 2 FPM sticks? I may be wrong, but I thought 2 SIMMS only constituted a single bank, whereby 4 SIMMS would be 2 banks. Please correct me if I remembered this incorrectly. Also, how do you test for proper interleaving functionality?

My multi-meter did not show 3.6 V when shorting JP30 and leaving JP27 open. It showed 4.96 V.

BIOS 1.72f also supports IDE drives greater than 8 GB. Any idea which enhancements come with 4.00A?

I'm having some trouble getting reliable Windows performance when using a Cyrix 5x86 and the PCI400-4. It is possible that this is due to a low core CPU voltage of 3.35 V as opposed to 3.45V+. Only one of several Cyrix 5x86's would even work on this board, which again points to the voltage being too low.

BIOS 4.00A seems to have slower cachechk and speedsys results. I'm using a Winbond 5V EEPROM. I'd be very interested in which settings you are using.

The PCI400-4 was quite stable with the Voodoo3 3000 PCI, however the Matrox Millennium G200 had some issues with certain benchmarks, which was not the case on a UMC chipset motherboard.

I will post the screens later, no worries.

My L2 is 62.2 MB/s and
Memory 42.9 MB/s both according to CacheChk v40

I think this is on par for the chipset but can be upped a bit with a 1.71 or 1.72 version.

2x66 is impossible with the Tomatoboard, even 50MHz is hardwire-disabled (JP18) but the chipset is still capable of 50MHz if you just cut the wire which is soldered on JP18). I think if 50MHz is enabled there is also a build in PCI 1/2 divider which gives you poor 25MHz clock on the PCI Bus but still better than nothing.

Btw. the jumper for Clear CMOS ist between the DIN connector and the coin battery. Default is 1-2 for normal operation opposed to 3-4 for clearing the cmos data. This helped me at lot once because the board refused posting to the DOS after updating the BIOS.

Well I don't know HOW low your memory speed is but it MIGHT have to do with the turbo switch. This board is a bit tricky when it comes to the jumper configuration. Just short J1 2-3 (2-3 is the position to the edge of the board). And bench again. Maybe your scores climb up.

Well that explains a lot! The 4DPS board requires a short for the turbo button, whereas the PCI400-4 board has removed that jumper header. The 4DPS board is now performing as expected. The original poster of this thread should also short their J1 2-3.

Our cachechk scores are now in agreement. By the way, cachechk scores and average speedsys scores are not comparable.

Interestingly, when I flash BIOS 1.72f and 4.00A, I get this message, "The program file's part number does not match with your system!" I will still let me flash though.

EDIT: Using BIOS 4.00A took away PS/2 mouse control in Windows. Mouse still worked with 1.72f, so I have gone back to 1.72f. 4.00A added the CPU Burst Write feature.

I'm happy my hint has helped you! Maybe you gonna start to like the Tomatoboard a bit more right now, hmm ? 😉

The TO has already 2-3 short on J1. His (bad) performance is most likely due to BIOS mismanagement and/or 70ns modules.

One strange thing is still the shown Vesa Speed, because with 4.00A it is limited to 22289KB/s which would be quite low if true. So I would handle the Vesa bandwidth output of SpeedSys with care.

System Specs:

Zida 4DPS Rev. 2.11, Bios 4.00A
2x16MB SS FPM (60ns)
Am5x86 133ADW @33MHz*4
Voodoo3 2000 PCI, 16MB
DOS 6.22, Jemmex

If unreadable: The CPU score is 50.37

#

Your Speedsys L1/L1/Memory are right where mine are for an AMD X5-133. These are the results for a good quality/fast socket 3 board. Cachechk results are also right in line with a fast socket 3 board. With my Zida board and a Voodoo3 3000, I get VESA 22241 KB/s. Same value using a Matrox Millennium G200 16MB. With the Matrox card in my Biostar MB-8433UUD, I get 34400 KB/s. I'm not sure how reliable these are.

Your PCPBench score seems a little low; I get 9.6 with a Voodoo3 3000 and a Matrox Millennium G200.

My particular Zida 2.1 board doesn't have a 3.6 V setting, however the PCI400-4 board has settings for 3.30V, 3.45V, 4.97V as measured by my multi-meter. The PCI400-4 board was not very stable with Cyrix 5x86 CPUs for whatever reason. It may need a BIOS update. It was very stable with AMD X5-133 and the VOODOO3. On the other hand, after fixing the turbo issue with the Zida Tomato 4DPS 2.1 board, I get stable Windows results with the Cyrix 5x86 and AMD X5-133.

After fixing that turbo jumper issue, this Zida Tomato board is turning out to be quite the little champion. With a slow CPU like the AMD X5-133, this Voodoo3 3000 can play Quake 2 at 800x600 w/OpenGL at 8.7 frames per second. It may even be playable with a POD-83/100 or Cyrix 5x86-120/133. Unfortunately, the Matrox Millennium G200's 486-capable drivers didn't get the OpenGL drivers working well enough to play Quake II, so I cannot compare the two graphic cards. With MDK D3D, on the other hand, the Voodoo3 ripped the pants off the Matrox G200, with a score of 239 compared to 117. There were several other Windows graphics benchmarks where the Voodoo3 really outshined the G200. Surprisingly, the Voodoo3 was a bit slower at Quake I in DOS, with 14.3 fps, compared to 14.4 fps for the G200.

I also checked to see if an Nvidia TNT2 would work, but the motherboard wouldn't turn on.

Here's a quick comparison of my favourite UMC board with the 4DPS.

Biostar MB-8433UUD
UMC 8881/8886 chipset
AWARD BIOS 4.51PG - 20 May 1996
512 KB cache (Write-back only)
256 MB of FPM or EDO RAM
4 SIMM slots
Cacheable memory limit is 64 MB write-back
Working PS/2 mouse port
3 PCI slots
4 ISA slots (none shared)
Socketed Dallas RTC module
3.45V, 4.0V, and 5.0V CPU support
20, 25, 33, 40, 50, 60, 66, 83 MHz FSB support
Works with all Cyrix 5x86 special features
Correctly supports POD-83/100
Latest graphics card to work: Matrox Millennium G200 (Voodoo3 will not work)
Cachechk RAM Speed: 39.7 MB/s

Zida Tomato 4DPS 2.1
SiS 496/497 chipset
AWARD BIOS 4.51PG - 1.72f
512 KB cache (Write-back, or Write-through)
128 MB FPM only
2 SIMM slots
Cacheable memory limit is 128 MB write-thru, or 64 MB write-back
Working PS/2 mouse port
3 PCI slots
3 ISA slots (1 shared)
Coin-cell battery for RTC
3.35V, and 5.0V CPU support
25, 33, 40, 50 MHz FSB support [overclocked settings not yet determined]
Works with all useful Cyrix 5x86 special features
Correctly supports POD-83/100 [not yet determined]
Latest graphics card to work: 3dfx Voodoo3 3000 PCI and Matrox Millennium G200
Cachechk RAM Speed: 42.8 MB/s

I'm leaning slightly towards the 4DPS being more desirable than the Biostar, however more testing is needed. If the 4DPS board works well with SCSI bus mastering and proves itself as stable as the Biostar, it may even be a better all around board due to the ability to cache 128 MB of RAM and use of a Voodoo3.

I had set a wrong parameter in the chipset features: LBD# Sample Point should be 2 not 3. That makes ne PCPbench score climb from 8.9 to 9.5. Memory speed is not affected. It just increases the PCI bandwidth. This option is always "2" with an older BIOS (and not shown in the menu IIRC).

Well the SiS chipsets were always great when it comes to the cacheable area. IMO 128MB cached in WT mode is all you need. Nice comparison you did . I would prefer the Zida over the Biostar because it has a very good layout and works with faster VGA cards. But I think both boards are excellent performers and I'm sure in about ten years or so you cannot choose which one you wanna get because they will start to get rare some day. 486 boards with PS/2 headers are very very much desireable.

Could you let me know if you can use your PS/2 mouse port in Windows with BIOS 4.00A?

Works fine in W98SE, never tried W95 though.

Interesting... I used Award Flash Programmer v5.4. Which version did you use? Do you also receive this message when you flash the BIOS?: "The program file's part number does not match with your system!
Are you using a Winbond flash chip?

Did you see any benchmark improvement when using these?:
DRAM RAS to CAS Delay, from 3 to 2 CCLK
DRAM Write CAS Pulse, from 2 to 1 CCLK

I didn't, so I left them at 3 and 2.

Have you thought about adding 512KB cache to your board?

Well I actually never ever flashed a BIOS in the Zida itself. This is because many 486 chipsets / boards require special revisions of the flash programs to work and that makes things unnessesarily difficult.
I love uniflash and flash all the BIOSes in a hotflash-device (it' a K6plus ready SiS based PC-Chips, btw.)

I also never tried RAS to CAS Delay 3 now CAS Pulse 2 because it works so well and stable with 2 and 1. And as I know that higher delays doesn't make a system faster I always kept them running. I remember that I had to higher the RAS to CAS Delay fron 2 to 3 when running the 133@160.

Of course I thought about 512KB cache in at least one of my boards but I don't have the SRAM chips and I don't know where to get them. So I stick with 256K in WB mode and have just 32MB installed for daily operation (2x16). My supergood Samsung FPM memory floats around where needed.. One day in a Gigabyte 486GAM/S another day in a Pentium Pro..

In my opinion

The Voodoo3 is the best video card for a SIS 496/497 motherboard however, the whole point of the Voodoo3(arguably) in a 486 is for games that take advantage of its 3D hardware acceleration and the best cpu for that job is the POD83.

While its a good setup for GLquake and other unknown 3D accelerated titles the better CPU for non accelerated gaming would be a cyrix 5x86 (generally).

Ive recently aquired a UMC based motherboard and threw in a S3 Virge DX card together with the cyrix 5x86-120, the two systems achieve different performance in different appilcations.

My optimised testing with a POD83/SIS 496/497/Voodooo 3

Quake 2 demo1.dm2

Quake 2 640x480 - 12.8fps
Quake 2 800x600 - 12.7fps

Quake timedemo demo1

DOSQuake 320x200 (18.6fps)
GLQuake 640x480 - 27.0fps
GlQuake 800x600 - 26.9fps

PCPBENCH - 8.0fps
3dbench - 62.5fps

Speedsys

Vesa memory - 13,931KB/S
CPU - 61.52

Datacache L1 (16kb) - 163MB/S (Im sure this is an error)
Datacache L2 (256kb) - 45.17MB/S
Memory throughput - 37.83MB/S

Obviously the POD 83 is no speed demon for pcpbench, 3dBench or speedsys, this would indicate a cyrix 5x86 being the better cpu for those type of tasks, as to whether its worth having a UMC based system im not sure, perhaps the Cyrix 5x86 will perform better in some 3d accelerated titles than the POD83? It would make for an interesting comparison.

IMO a V3 in a 486 is more or less just a demonstration of feasibility. Almost all games that take advantage of the V3s' capabilities run so much better on Pentium or faster.

Of course it's cool that it works with the SiS496/497 chipset and of course the great POD has a better FPU for the more demanding games. The bottleneck is really the slow memory/cache subsystem unless you are able to setup a UMC system with 66MHz FSB and a 1/2 PCI divider. But these boards cannot use the V3 because they only support PCI 2.0. cards.

So for me the V3 in a 486 is a nice to have option but not for everyday use because other platforms just perform so much better with it.

I agree that its an exercise in feasibility but some of the games are quite playable such as GLquake or Outlaws (direct3d), not perfect I should add, really a shame the POD is a poor overclocker as running at 40mhz fsb/100mhz would really help in those games.

Are you sure it is the PCI specification which limits the use of the Voodoo3 on the UMC board? According to the Tomato manual, it also uses the PCI 2.0 specification. I think there is more to it than that, but I don't know exactly what the limiting factor is. I've heard reports of some SiS boards being able to run with a 66 MHz FSB. In fact, the clock controller chip on the PCI400-4 is the same one that is on the UMC boards, so it will output 20, 25, 33, 40, 50, 60, 66, 83 MHz if the jumpers are set accordingly. The main question is if the SiS chipset can handle 60/66 MHz FSB's and if there is an auto 1/2 PCI divider. I will eventually get to testing this for myself, however for now, I've found the frequency of diaper changes and feedings to be too much for retro tinkering.

With a POD running at 100 MHz (yes, it is rare to find one which will run stable at that speed, but I have one), I think the Quake 2 scores would jump to about 16 fps, so it is definately playable at 800x600. In my mind, the main advantage of the Voodoo 3, even in slower games, is the fact that you don't need to use up two PCI slots with a Voodoo1/2 and a 2D card.

Another item on my list is to modify the POD's regulator module to increase the voltage going to the CPU. This should really open the door for POD's at 100 MHz. On the other hand, there is something about using a Pentium in a 486 board that just doesn't feel quite right to me. You can always use a Pentium 100 in a socket 5/7, however you cannot use a Cyrix 5x86, Intel DX4, or AMD X5 in a Pentium board.

I am in general agreement with dirkmirk, I think the Voodoo3 opens up a few more games to the 486. Direct3D and OpenGL games seem work surprisingly well with the V3 on a SiS.

EDIT: Has anyone tried a TNT card in this motherboard? I only have a TNT2 and it didn't work. Will a Voodoo 4 or 5 work?

Referring to the SpeedSys readout the board is PCI 2.1 compliant but I doubt it supports the enhanced bandwidth (33MHz *2).

I can tell you that the TNT works without flaws (ELSA Erazor II: http://www.amoretro.de/2012/03/elsa-erazor-ii … t-16mb-pci.html)

The Geforce 2MX (ELSA Gladiac 511 PCI) also works but of course it has the big GeForce BIOS and requires an alternative BIOS and memory configuration if you want to play with it in DOS.

Both cards have excellent Vesa 3.0 extensions.

I don't have PCI versions of VSA100 cards so I cannot comment on that.

Thanks to everyone for your kind comments! I'm glad to see I sparked a discussion about the 4DPS motherboard! 😀

I haven't replied earlier because I've been messing around with some older builds (a 386 and a 286) and the 286 in particular is giving me a hard time (soon you'll see me asking for help around here 😵) so I haven't had the time to play with my 5x86 build, but I have been very pleased with it, everything I've thrown at it just works and it's been pretty much rock solid stable (unlike my previous 486 build).

nforce4max wrote:

Get a tube of arctic silver 5 and apply some to the heatsink with a layer thick enough so that over time it becomes hard enough to glue the heatsink to the cpu. It retains it's thermal properties while being strong enough to keep the heatsink on without needing hot glue or even a bracket. Be sure to clean that psu and the fan inside so that it may continue to run well into the future.

I've been using AS5 for years and I would've never thought of that! Gotta buy some more and try that out. Thanks for the tip! 😉

And yes, just like I thoroughly cleaned the case inside & out, I also opened up the power supply, evicted the dust bunnies and oiled the fan to ensure I don't get problems down the road.

badmofo wrote:

Wow that motherboard is amazingly compact - the ziff socket hangs over the edge! Nice work on getting it going again, very tidy looking system you have yourself there. I like the window-key-less keyboard.

Indeed it is compact! I like the keyboard too, although it's a cheap membrane type, very light and not great to type on, it looks great. The only clicky keyboard I've ever had now lives in my MAME arcade cabinet (ease of hack-ability). I still hope to find another one someday but I haven't had much luck. 😢

GXL750 wrote:

Good job on that repair project. BTW, is that a Dreamcast sitting on top of the three Nintendos?

It is a Dreamcast sitting on top of three NES consoles! Good eye!
It's a friend's Dreamcast I was repairing, my own Dreamcast is in storage at the moment.

FGB wrote:

I can highly recommend you to update your 1.6 BIOS to Version 4.00A, this version should work with your boards hardware revision (printed between the slots). The 4.00A Version supports 40GB HDD's, (real) Y2K compliance, has full PnP Support and a POST device listing and last but not least prevents the board from "flickering" with some VGA cards.
You can use any 128KB flash rom (maybe picked from a Socket 7 board and hotflashed there using uniflash) in the ZIDA, I've tested SST, Winbond, Amtel and Intel with 100% success.

Nice! I wasn't able to find an image for that version so I was going to ask you, but others have beaten me to it! As soon as I get back to this rig I'll update the BIOS with the image you've provided.

FGB wrote:

You can also vastly improve your memory performance up to ~40MB/s throughput ( SpeedSys 4.78 ) by finetuning the advanced BIOS settings.

I'm using the settings I found on this page. I haven't gotten around to compare them with yours but I'll do and I'll also check the RAM.

FGB wrote:

If you have any question about the 4DPS, just send me a message or ask in this thread.

I have a question, is the LPT header pinout non-standard like the COM headers? I've installed a regular parallel port bracket from my parts box but haven't actually tested the port, so I presume if I try to use the port later (with my parallel Zip drive for example) I might be in for a surprise, just like when I tried a serial mouse and found out I had to re-wire my COM brackets.

feipoa wrote:

@133 MHz, could you also test your board with the HIMEM memory test? I found that with my solder work on the HOT-433, Memtest also passed with flying colours, but HIMEM had a fit.

I'll try that out and I'll let you know.

feipoa wrote:

By the way, I really appreciate the matching beige case you have there. I used to have those same speakers with my X5-133 purchased in Jan. 1997.

Thanks, I've been working hard to get good-looking vintage peripherals, sometimes I get lucky and people just give them to me but other times I've had to search everywhere with little success. It saddens me a little how I used to have tons of stuff like these and one day I just threw pretty much all away, thinking it was so abundant and worthless I'll find it again later if I wanted to, which doesn't seem to be the case nowadays. 😢

sliderider wrote:

Would it have been possible to replace the missing RAM clips on the board that was missing them with clips from a junk motherboard?

Not only the clips were missing, but the plastic stands which would keep the clips in place were broken off, so in this case the whole slots had to be replaced for reliability purposes.

FGB wrote:

This board is a pleasure to work with - because it just works!

Totally agree! 😀

@FGB
Speedsys also notes that my UMC boards are PCI 2.10. I've never seen Speedsys say a board was PCI 2.0, so I'm not sure how reliable it is. At any rate, the UMC boards won't turn on with a Voodoo3 in place, so I'm not sure if the issue is entirely PCI revision related.

That is good to know that the TNT works fine in the SiS board. Do you know if it will work in a UMC board? Did you test the TNT2 on your SiS board? What are the strengths and weaknesses of the TNT vs. the Voodoo3 3000 on a SiS 486?

@133 MHz
Would you mind sending me an image of your 1.6 version 4DPS BIOS? I like to keep an archive of all the older BIOS images for the boards in my collection. I have found that some old PCB revisions of certain motherboards will only work well with the older BIOS versions.

You can use the awardflash version I have attached here. It seems to work well with the 4DPS boards. When the program asks you for the filename of what you want to program, just type anything in, which will bring you to the next screen asking what you want to save your existing BIOS as. Safe to a floppy and upload!

Everything looks fine from that website for BIOS settings, with the exception of these:

LD Sample Point End of T3 (T2 causes not so obvious read/write disk errors)
Cache Write Cycle 3 CCLK

If you are using a 33 MHz FSB, set LD to T2 and Cache Write Cycle to 2.

I sometimes think SpeedSys shows crap here and there.. On some machines the CPU MHz is displayed wrong (eg. Cyrix DLC systems), on other machines the memory bandwidth (not the troughput) is displayed wrong, on other machines the VESA Speed is just ridiculous low and/or the memory amount is displayed to small (Diamond SiS 6326 cards), it freezes on many Cx6x86 systems on the MSR testing, ...[?] so I really doubt the the output from time to time.

And from my perspective I think you well may be right that the 4DPS only has the 2.00 compliance just because the 2.1 compliance includes the bandwidth-doubling the 4DPS cannot provide. So it maybe 2.00 on Paper but in pratice it well detects and runs quite a few 2.1 cards.

I'm sorry but I cannot test any TNT2 PCI card - I have none. And I'm very sorry that I don't have the time to test the Voodoo3 against the TNT. But I think it will be very close on this machine because there are limiting factors like CPU and Busspeed.

Would also be very interesting to test a Kyro based card like this litte diamond: http://www.amoretro.de/2011/10/hercules-3d-pr … wervr-kyro.html

Maybe I'll find the time to check out if it works in our little Tomatoboard.

FGB wrote:

it freezes on many Cx6x86 systems on the MSR testing, ...[?] so I really doubt the the output from time to time.

I can tell you how to fix that for good! First, if your BIOS has the option to disable Cyrix CPUID, disable it. If not, use the ibmm1.exe or ibmm2.exe software to disable the CPUID in DOS. Try to run Speedsys again. If Speedsys hangs at MSR still, run Speedsys without PnP detection by typing at the dos prompt, C:\speedsys /sp

I'd be very interested if anyone has a Riva TNT and a UMC board who can test those two for cooperation. If the TNT works on a UMC board, it would make a nice alternative to the Matrox Millennium G200.