Zida Tomato TD60C 286 motherboard, based on (what appears to be a hybrid 286/386) Citygate chipset with support of up to 16Mb RAM.
A QA sticker on the back says year 1993 - late product for 286 hardware.
It came with soldered on-board Harris 25MHz rated CPU.
SIMM and on-board memory can be combined. Tested with 4Mb (SIMM only) and 5Mb (combined).
There are no available jumpers, except 3 - clear BIOS, monitor select and external clock.
BIOS is 386-like with options to adjust timings, etc.

Runs stable at up to 27.5MHz, but the system is apparently chocked by internally imposed wait-states.
Performance is pretty bad. On top of that there are issues with write access to CompactFlash cards, which adds inconvenience.

The usual set of stats and benchmarks:

benchmark results

A good looking piece of retro hardware, but quite disappointing overall.

That's an interesting (but odd!) set of results. The comptest data, especially the RAM throughput and Video speed (enormous throughput but an effective 6 wait states!), suggests that the board could be an absolute screamer... but then the FPU results are terrible and by comparison the Landmark video results are pretty poor (my 16MHz board gets better Landmark video results).

It's a bit of a conundrum. It's an odd, contradictory set of results, that's for sure!

Yes, not all is well with this board.
It has the chops to be a hot-shot, but something is choking it.
Too bad.
So many missed opportunities in the land of 286.

---

And while on the subject of missed opportunities ...

FIC 86-VIP-i02

486 motherboard with character and attitude.
Voltage regulator is on a daughterboard - a bit unusual.
No DIN-4 keyboard connector, instead PS/2 keyboard and mouse ones.
Jumpers are scattered all over the place.
Supports wide range of processors - all the Intels, AMDs, Cyrixes, from its time.
Picky about RAM. Got stuck at couple of trusted sets, but accepted 60ns Micron modules at the end.
The upgraded BIOS 1.175G701 is full of options.
The system works well with all BIOS and jumper settings at their optimal (best performance) values.
Now the bad part.
Cannot run AMD DX5 at 4x40MHz no matter what. Started looking online to see if i am missing something fundamental here and saw another thread here developing at the same time (lucky coincidence) where others confirm that 3x50MHz is the best they could get out of it. Ok. So be it.
There are sockets for up to 1Mb L2 cache, but the board refuses to recognize more than 512Kb. Tried multiple sets of trusted chips - no-go.
In addition - hangs hard with 128k x 8 TAG chip, even it is supposed to be supported.

https://www.petershipkov.com/temp/retro_pc_im … _86-vip-io2.jpg

Speedsys:

Overall, interesting motherboard that gives a bit different feeling than the usual SiS, UMC and Opti based ones.
Performance is also not that bad.
IDE is lagging.
DOS graphics are pretty ok - favors VLB.
Windows graphics are in the ballpark of the competition (especially considering the 10MHz lower CPU speed) - favors PCI.
While the computation results are not that great, it actually spanks the Biostar board running at 160MHz (4x40) in the Windows/LightWave3D test.
If the L2 cache and 4x40MHz issues can be eventually worked-out, this mobo will be pretty decent all-rounder.

[EDIT]
The story with getting the system running at 160MHz (4x40) continues here: Re: 3 (+3 more) retro battle stations

I am a little familiar with your 5x86 problem. That chip changes its CPU id depending on how the multiplier is set. You can verify this in speedsys. When set to 3X mode, it reports itself as an am486, but when in 4X mode it becomes am5x86. My guess is that the BIOS doesn't support am5x86 and chokes.
See if there is a way to switch into 4x mode after the system has booted.

You are correct.
This is what i observed.

Not sure how can switch to 4x after boot.
Not sure it can be done in a jumpered board, can it ?

Hm does the board has any jumper for multiplier? if not you have to manually
ground a pin or use an interposer with that option. check this site here:
http://web.inter.nl.net/hcc/J.Steunebrink/amd5x86.htm

I had the same issue with my jetway J-435 board - it uses same VIA chipset
(and it also its award bios reported it differently like your FIC)

When I added a socket that manually grounded that pin,
the Am5x86 switched to 4x multiplier

Thanks for the info AndrewK2685.

Take a look here:
FIC 486-VIP-IO2 CPU support
and here:
https://stason.org/TULARC/pc/motherboards/F/F … 86-VIP-IO2.html

It should be able to do the 4x40 trick, but cannot.
Will try your suggestion soon. Hopefully i can get this guy running with its full potential.
What adapter your are using ?

--------------------

As of today - 286 motherboard based on GST GOLD chipset.
Looks like the official name and model are Snobol Mini SNB-M008.
Wonder why 286 class hardware was named after programming languages, or was it just a coincidence ?
Anyway.

It is a highly integrated motherboard from year 1990.

According to the manual it is supposed to support up to 5Mb of RAM - combining on board and SIMM memory.
Was not able to accomplish this, no matter how much i tried. It can be one, or the other only.

There is an empty socket in the upper middle section with unclear functionality.
Looked online - there is some inconclusive information about it. Appears to me that it is a remnant from an earlier PCB version.

Maxes out at 20MHz.
This is the first 286 mobo that didn't like ET4000AX VGA, but Cirrus Logic GD05426 worked fine.
CF card write access problems - a red flag.

The usual set of stats:

benchmark results

Need to think a bit about positive finale wording. 😀
Ok, here:
It works well at 16MHz. Can be pushed to 20MHz. Performance is lacking.
Can be a great starting point for the "classic" 286 experience.

A really simple one actually - I just ordered a 168-pin socket, like this:
https://www.ebay.com/itm/2pcs-700-168-17-0SPF … v4AAOSwzmZf8JaC

cut the pin from socket to break connection with motherboard
and soldered a cable to it that connected it on a ground pin in motherboard
(e.g. the ground on power led). That did the trick. - can send some pics later on
I didn't really need the socket for that, but I didn't want to damage my 160MHz Am5x86.

Also, for your graphics bench, if possible include some ATI cards, esp. from Mach series
like Mach64 or Mach32. These were top ones for ISA and VLB.
There are also in PCI, but I'd go to rage series for PCI

Thanks for the link.
I can "extract" one of these from a dead board. Somehow i didn't connect what you are saying in your previous post to the type of socket in question. Heh.
If you can supplement a photo of your setup that will be helpful - will save me time.
Hope this simple mod works here.

Quoting myself from the first post of this thread:

No ATI video cards.
Looking back, i realized that somehow i never owned ATI hardware.
Well, except for a FirePro v7900 couple of years ago, but it was donated by ATI, so it kind of does not count. Wonderful card btw !

I still don't have any ATI video cards, which makes it easy to keep the tradition alive. 😀
But if you are saying that there are VLB ATI ones that do better than ARK1000VL in DOS and S3 Trio64 in Windows, i can reconsider my stance real quick.

Here are the pics I promised. Hope these help a bit
You need to ground pin R17.
I broke that pin on socket to cut connection with
motherboard and soldered a cable to ground it.

You may also findsome 486 interposers sold by Amazon
that do voltage, multiplier and cache selections,
but are expensive (I think from TrinityWorks)
I highly recommend you go through the Am5x86 site
It has a lot of useful information for it.

For ATIs, I understand. I thought you might already had some.
On VLBs I don't think these go above Trio64 or ARK1000, never
had these, but should be close to them - I only had a
Diamond Stealth with Vision968 and they were going head on head.
Unfortunately, I didn't run many tests before my Stealth died 🙁.
I only have a Trio32 VLB, but this test is not fair.
For ISA though I'd say might be top ones.

Their quality with LCDs is also excellent
Another thing is they go up to 4MB (Both for VLB & ISA).
enabling high resolutions with true colors.
Don't know if any Trio64 VLBs exist with 4MB.

Does the Trio (any variant) actually exist in an ISA interface? Wikipedia lists it but I've never seen an example of one. I was sure it was VLB/PCI/AGP only.

Thanks for the pics+info Andrew.
Now i don't even have to go search for R17.

Remember seeing comments from others that Vision968 is faster than Trio64 in Windows GUI at high resolutions.
In my tests Trio64 seems to be equal, or faster at any resolutions. The observed difference is not big and can be attributed to the used different tests.
Still, your comments are making me interested in ATIs. Let's see what's available out there.

Never seen Trio ISA either.

I guess there may be some paper metrics that state the VRAM on the 968 is technically superior to the EDO/DRAM on the Trio - maybe the 968's had better ramdacs as they were positioned higher in the range compared to the 864, 868, but the Trio is such a better all-round, all-in-one chipset that it's really hard to argue that the older, higher-end card is superior in any meaningful way.

Agreed.

Can this be the fastest, long-term stable 286 PC ?

Posted brief info about this subject in another thread Re: 286 dead, long live the 286! Sourcing and building a fast 286 replacement, but here is the full story:

The first post in this thread showed a 286 PC running at 25MHz and 0-wait states.
Its performance was impressive for a 286 class machine - numbers and other statistics are still there.

Later on I improved ventilation for complete long-term stability: Re: 3 (+3 more) retro battle stations

Then tested a CPU upgrade module with Ti486SXL2 running at 60MHz.
Long story short: technically CPU operations are much faster, but the system is compromised in several ways = paper tiger.
Full story here: Re: 3 (+3 more) retro battle stations

Recently was able to achieve complete long-term stability at 27.5MHz and 0-wait states.
This required memory curation on an entirely different level.

In addition, swapped the IIT 2c87 20MHz rated FPU that was used until now:
https://www.petershipkov.com/temp/retro_pc_im … iit_2c87-20.jpg

with the fastest 287 one out there (as far as i know) - Cyrix CX-82s87-np-sv:
https://www.petershipkov.com/temp/retro_pc_im … 82s87-np-sv.jpg

Few more words about the subject.
CX-82s87-np-sv is based on the original FasMath 83D86 core introduced in year 1989.
There is a second variant of the FPU named 287XL+, based on the newer FasMath 83D86 core from year 1991 (also known as 387+)
The later version (black top) overclocks better (can reach 50MHz+), but is clock-to-clock slower than the earlier one (gray top).
Here is set of tests from the first post in this thread showing the perf difference between the two: https://www.petershipkov.com/temp/retro_pc_im … hmarks_lw3d.png
Also, there are other posts on the forums talking about that.
Since the 286 machine cannot go above 32MHz - Cyrix CX-82s87-np-sv is preferrable.
Or to put it in pixels:

Ok, enough theory. Lets look at some stats:

27.5MHz, 0-wait states, IIT FPU:

27.5MHz, 0-wait states, Cyrix FPU:

And some benchmarks:

Also some videos showing test runs with full gear:
https://www.petershipkov.com/temp/retro_pc_im … 6_vlsi_27.5.mp4
https://www.petershipkov.com/temp/retro_pc_im … vlsi_27.5_2.mp4
One of the kids named this PC Timmy.

I am pretty sure that this particular chipset/motherboard can roll with 30MHz at 0-waits, but there is no memory to withstand it.

---

Those who bothered to skim through this thread probably noticed that i examined multiple 286 motherboards, motivated by personal curiosity about who is who in the 286 world.
I think i achieved good sampling rate and distribution in my exploration, by targeting different chipsets and brands.
There are 4 boards that stand out.
3 of them are based on VLSI 200 chipsets, one is based on Headland HT12P.
One of the 3 is the board used in the 286 PC in question here.

I usually stay away from bombastic statements, assuming everything shaking out there, but this can very well be the fastest 286, or at least one of the top pretenders.

EDIT: The above statement is further underscored by evolution to 30MHz and 0 wait states.

One thing I'd like to know is how your disk IO metrics are so good. I'm barely scraping 1.5MB/sec at the moment, that's with a very fast Sandisk Ultra CF card that reads at 90+MB/sec under modern benchmarks. On my non-working HT12 board I could top out just over 3MB/sec.

I do wonder how much of a difference the standard Multi-IO controller makes in this scenario; it's an Acer M5105 based card I'm using right now (can't find my original one - it has disappeared into void... just as odd socks do).

In my experience all 16-bit ISA IDE controllers are equal.
The main performance difference comes from the motherboard/chipset.
A minor, but still noticeable factor are the used CF cards.

Never did a proper "study" on them, but after enough go-rounds with different CF brands/models and hardware scenarios, it became apparent that CF performance rates are not a defining factor. There is no apparent formula to follow. Trial and error FTW. So, long story short:

Over time i converged on these 3 models from everything i tried so far - they cover best the retro "junk" i am interested in.

For example - a Transcend 133x is used in the 286 PC in question.
It delivers top perf and reliability in 286 and 386 computers equipped with ISA IDE adapters.
Compared to it, the other two types lower disk i/o by about 300-400Kb/s.
The SanDisk cards can cause write access problems with some VLB EIDE/SCSI controllers and 386 DLC (especially SXL2) CPUs + overclocking, so Transcend is preferable. Both deliver same performance.
But the table turns when in the context of fast 486 class PCs, especially in combination with PCI IDEs.
There SanDisk ULTRA 50Mb/s >> Transcend 133x.
The Lexar Professional 1066x are similar to the SanDisk ULTRA, but behave better with later than year 2002-2003 hardware.
I use one in a late 486 PC rig ... for bragging rights.

The usual "pick your poison" situation. 😀

I have that exact Sandisk card in my 286. In a modern PC it's an absolute monster:

I was expecting a bit more oomph from it in the 286, but in both the boards I've used it the IO performance has been just 'okay' (up to 1.3MB/sec at stable clock speeds) . I suspect it's because both of them have no ISA bus clock control options.

My original 286 board had an option for overclocking the ISA bus from 8MHz, so I suspect that is where most of the performance came from (the ~3.4MB/sec I mentioned).

pshipkov wrote on 2020-12-12, 20:54:

https://www.petershipkov.com/temp/retro_pc_images/components_lo/motherboard_486_chicony_tk8880f.jpg […]
Show full quote

Forgive me for not answering for a long time, and got into the old message. But I wanted to pay special attention to AMD 5x86 cache L1, it can be either WT or WB.

Judging by the graph from SST4.7, the cache is installed here by WT and it is 1-2% slower than WB. This can be seen in the SST 4.78 "cache" plot where the green line is lower than the yellow one.

As far as I remember, in the WB cache the green line should be higher. Below is a picture of what the WB L1 cache looks like, where green line above yollow linies (small area from 0 to 16kb).

If you can't fix it using the jumper method, then you can try to solder the Pin## in socket 3 (I don't remember which one, I need to google) the wires to GND. The processor will automatically switch to WB.

If you check the link I posted above about the multiplier pin,
it also tells about the cache pins (4 pins) and how to manually enable WB
I haven't tried that as it was more complicated and my motherboard already enabled WB
http://web.inter.nl.net/hcc/J.Steunebrink/amd5x86.htm