I recently got an U5SX @ 40Mhz and installed it on a Chicony CH-498B mainboard (Socket 2 and no VRM) replacing a 486DX/2-S. This board has an UMC chipset and 3 VLB slots. I havetwo VLB cards installed on this system: a Trident TGUI9400Cxi graphic card and a Promise DC-4030VL IDE caching card. Both work really well with a 40Mhz FSB without any WS.

Performance of the system is very good and, certainly, in the range of the DX/2 it has replaced (since it does not have a FPU floating point math is not even near but running integer code is not much slower). Both VLB card show a clear improvement running at 40Mhz and the system is really stable.

Here is a comparison between several 486SX CPUs:

https://www.youtube.com/watch?v=kQEdPQDXdLY

The UMC CPU is allways the faster one. It would be interesting to make a comparison with a SX2 @ 66Mhz.

Looks nice.

Awesome build. I don't think I've ever seen a 486 board with a soldered-on CPU before

They were more common in early 486 era I think, Intel sx25s and sx33s, with about the only real use case for a 487 socket. I did see a "high end" EISA system with a dx40 on once, at time of production it would have been the fastest AutoCAD system you could get. Later in the game there might have been some with AMD 5x86 on.

OK, so the new PSU is in, and it lives. Although there are issues, so case shots will have to wait..

It POSTs, so we are onto something!

Finally we have a BIOS ID, now I can do a lookup on this and see if there are any updates..

And here is the motherboard model, absent on the PCB. The manufacturer seems to be Biostar, and the board is likely an OEM model.

In the meantime, I have some problems and questions 😀

1. The Turbo Switch header on the motherboard is a two pin (Presumably V+/V-) while the case connector is a three pin (V+/G/V- judging by the cable colors); I am considering using dupont cables to connect the V+/V- to the switch headers. Any objections? I shorted the header with a jumper for now.
2. As you can see in the third photo above, the memory is detected as 3712KB (Presumably 384KB is Shadow Memory if memory serves) but the module in the SIMM slot is an 8MB one (2Mx32) so it should have been 7808K - why could this be? The board can only use single sided SIMMs perhaps? Or address only up to 4MB per bank? Now that would be strange!

For the record,this is the motherboard I am using:

Well, I have some answers for my own second question, as I had some time to tinker with the system tonight. The SIMM I originally had in it was 1Mx64 8MB FPM with 16 TMS444000J-70 chips. To test compatibility I popped in two more pieces of RAM. First, I tried an 1Mx64 8MB EDO with 4 VG2618165BJ-6 chips, and it worked fine. I then tried a 4Mx32 16MB EDO with 8 HM5117405S6 chips, and that worked fine as well. I kept the 16MB SIMM in the system. My inconclusive testing tells me that the board does not accept composite SIMMs with more than 8 modules, or that it simply does not work with FPM memory. Regardless, I got it to work and that is fine for me.

For the record, I also managed to get it to run at 40Mhz, apparently stably, by changing the triple jumper block on the upper right side of the board from (top to bottom) SHORT/SHORT/SHORT to SHORT/SHORT/LONG on a hunch. I don't know what the first two jumpers do but I am guessing Vcore or Vmem so I did not dare touch them.

I will do some internal tidying on the weekend and try to get the compact flash card reader to play ball with the IDE controller. If that works, I'll enjoy doing some benches on this Frankenstein ISA 486 PC 😀

So what might be going on, is that the motherboard doesn't have enough refresh line capability to supply a refresh for more than a handful of RAM chips. Now in some BIOSes, there is a cryptically named refresh strength setting in the advanced chipset setup, maybe playing with that might get you something.

BitWrangler wrote:

So what might be going on, is that the motherboard doesn't have enough refresh line capability to supply a refresh for more than a handful of RAM chips. Now in some BIOSes, there is a cryptically named refresh strength setting in the advanced chipset setup, maybe playing with that might get you something.

I will do a deep down of the BIOS, it's the AMI BIOS that looks like Windows 2.0 - I never used one before, so fun times. I also seem to have completely forgotten all those weird memory shadowing and addressing settings in 386/486 motherboards so expect lots of questions in the future 😊

Ok, now that I seem to have mostly resolved the IDE issues, time for some benchmarks.

Do these benches look OK?

The CPU does indeed seem to be on par with a DX2-50, at least when running at 40MHz. As for the 3D performance, I'm not impressed. Doom was running slower than it ran on my DX-33 with a VLB Cirrus Logic card for sure. I need to find a better ISA VGA for this system.. Yet, overall, I'm pretty happy - I finally have a 486, after 20 years!

Perhaps resize the images a bit. I can only see less then half of them 🤣
This makes it hard to Judge the outcome of the benchmarks 😀

Nvm1 wrote:

Perhaps resize the images a bit. I can only see less then half of them 🤣
This makes it hard to Judge the outcome of the benchmarks 😀

Done 😉

appiah4 wrote:

Nvm1 wrote:

Perhaps resize the images a bit. I can only see less then half of them 🤣
This makes it hard to Judge the outcome of the benchmarks 😀

Done 😉

Super! 🙄
CPU speed is indeed pretty good. Perhaps try to find something like a S3 801 isa vga card and try the vga bench then again.

Not bad, they do a bit better in a board that has L2.