if i understand this correctly the ti only has 1kb of cache

In theory, the UMC CPU should be faster. The TI is actually a 386 to 486 upgrade with an added 1KB L1 cache.

Without a doubt, the UMC CPU is faster.
I was fortunate enough to acquire such a CPU from a friend a couple of months ago and I benchmarked it.
I cannot recall the exact results, but it came close to that of a 486DX-50.
The TX486DLC-40 (which is just a rebranded Cyrix 486DLC-40) is basically able to run on par with an Intel 486DX(or SX)-33 (in terms of integer performance).

mrau wrote:

if i understand this correctly the ti only has 1kb of cache

Afaik, the TI variant has 8 kb of cache. The Cyrix one only has 1 kb.

derSammler wrote:

mrau wrote:

if i understand this correctly the ti only has 1kb of cache

Afaik, the TI variant has 8 kb of cache. The Cyrix one only has 1 kb.

I believe they are the exact CPU as at the time Cyrix was using TI fabs, and TI just got to rebrand Cyrix CPUs and sell them in their own brand as part of the deal. Both are 1K L1 cache. I was just wondering if the 40Mhz vs 33Mhz would make up for the 1k vs 8k cache difference. Apparently it won't?

I can probably also overclock the 33MHz UMC chip to 40Mhz comfortably as 33F chips are often also sold as 33F/40 with little difference in actual silicon quality.. Just not sure if (and how) the board lets me do it because I have no jumper settings for it.

derSammler wrote:

mrau wrote:

if i understand this correctly the ti only has 1kb of cache

Afaik, the TI variant has 8 kb of cache. The Cyrix one only has 1 kb.

Only the SXL version has 8kB. DLC has 1kB.

Not only is the UMC way faster than the TI chip, from what I understand it's actually even faster than a real 33MHz intel 486. The 40MHz version is said to be about on par with a DX/2-66.

Anonymous Coward wrote:

Not only is the UMC way faster than the TI chip, from what I understand it's actually even faster than a real 33MHz intel 486. The 40MHz version is said to be about on par with a DX/2-66.

Heres to hoping I can run it at 40MHz then 😀

Attached some comparative benchmarks that I've run on my own systems.

Not all of these were performed on the same motherboard, due to different architectures, but at least one does get an idea of the overall performance (Doom's timedemo 3 a good benchmark).
As you can see from the results, the UMC 33 MHz CPU is somewhere between your Intel 486DX 33 MHz & an Intel 486DX2 66 MHz.
So, the UMC 40 MHz CPU should definitely get close to the performance of an Intel 486DX2 66 MHz.

jesolo wrote:

As you can see from the results, the UMC 33 MHz CPU is somewhere between your Intel 486DX 33 MHz & an Intel 486DX2 66 MHz.

When I tried a socketed UMC U5S I found the same so the CPU itself is potentially faster than a DLC and even an Intel SX.
On the other hand once I found a relatively intact (little to no acid damage) m/b with a 72-pin SIMM socket and a soldered UMC CPU and it wasn't just as tiny as a late 386 board it was also as slow. As a 386SX. It was so disappointing that despite its curiosity it ended up back in the scrap 😵

What I wanted to say is, it strongly depends on the board which is, in this case, soldered to the CPU. Or the other way around.

I have found the UMC 486 chips to be very fast, I have a U5S 40, and it's much quicker than AMD DX40 486 or intel DX33. It also runs cooler than both of those.
My buddy has a soldered 40mhz UMC on a PCCHIPS VLB board, that one is very swift too.

Do any of the motherboards which contain a U5S also have a socket for an FPU?

I suspect the U5S at 33 Mhz is similar to a TI 486SXL2-50, but I have not run benchmarks as of yet.

I tested my U5S against a AMD SX/2 66, they were pretty much equal.
Lot more info about the UMC herehttp://forum.dxzeff.com/forums/thread-view.as … osts=10&start=1

appiah4 wrote:

Which one would be faster?

Are you not curious to test both?

If it was me, I would want to test them both against each other, having said I would agree with other posters the UMC is a foregone conclusion... but it would be interesting to see by how much!?

386_junkie wrote:

appiah4 wrote:

Which one would be faster?

Are you not curious to test both?

If it was me, I would want to test them both against each other, having said I would agree with other posters the UMC is a foregone conclusion... but it would be interesting to see by how much!?

I have the umc but was considering getting the ti with biard abd ram.. I will probably get it as I have no other 386 boards or cpus so why not ; just another project in the backlog..

Just a quick follow up on the UMC U5S 40 MHz performance.
Found some benches I did earlier in the year on a Zida tomato 4DPS motherboard, Diamond S3 Trio 64, pure DOS 6.2 no CPU enhancements loaded.

3D bench 2 PCP Bench Doom

UMC U5S 40 33.7 6.8 16.2
Intel DX 33 22.5 5.6 12.7
Intel DX/2 66 WB 33.8 7.4 19.95
Intel DX/4 75 35 9.7 15.57
AMD DX/4 100 59.3 12.2 21.33
Cyrix 586 100 51 10 22.03

How about running the Intel DX at 40 MHz?

I have a 40Mhz UMC U5SX running on a Chicony CH-498B motherboard (UMC 498F chipset) with two VLB cards:

- Promise DC4030 IDE caching controller
- Trident TVGA 9400CXi

both running stable with 0 WS. I have benchmarked it and performance is a bit over a 486DX 50 and much better than AMD 486DX 40 Mhz parts. Definitively it does not reach 486DX2 66 performance levels but it should perform much better than Cyrix/TI 486DLC parts which are nothing more than very fast 386s. It's another league.

In that case, I wouldn't mind seeing a comparison of the UMC U5S-40 versus these fast 386's: TI 486SXL2 50-80 MHz , DRx2-66, and IBM DLC3 66-100 MHz. It would be interesting to see where the meeting point is.