Nice find. Maybe UMC U5D would be the real "super 486" CPU if it had a FPU.

analog_programmer wrote on 2024-11-20, 08:46:

Nice find. Maybe UMC U5D would be the real "super 486" CPU if it had a FPU.

Unless you were doing some very specific things very little 486-era software used an FPU let alone required it. Intel pretty cynically made use of the bad rep the 386SX had had for its castrated 16b bus to convince people they needed to shell out on a 486DX when the vast majority of them would have had the exact same experience on a cheaper 486SX. The first game that absolutely required an FPU was Quake, and good luck running that acceptably on any 486 even if you had an FPU.

So no, marketing aside I'd not call an FPU a requirement for a 'super 486', particularly not if we're not including all the clock doubled/tripled/quadrupled CPUs (this thing is nice, but CX5x86 and Am5x86 will blow it out the water of course).

dionb wrote on 2024-11-20, 14:44:

Unless you were doing some very specific things very little 486-era software used an FPU let alone required it. Intel pretty cynically made use of the bad rep the 386SX had had for its castrated 16b bus to convince people they needed to shell out on a 486DX when the vast majority of them would have had the exact same experience on a cheaper 486SX. The first game that absolutely required an FPU was Quake, and good luck running that acceptably on any 486 even if you had an FPU.

So no, marketing aside I'd not call an FPU a requirement for a 'super 486', particularly not if we're not including all the clock doubled/tripled/quadrupled CPUs (this thing is nice, but CX5x86 and Am5x86 will blow it out the water of course).

From a purely practical point of view, you are right. Bur it would be interesting if someone could compare UMC U5D running at 50 MHz vs i486/DX2-50 or "the classic" i486/DX2-66 including their FPU performance.

Last time I had one running a couple of decades back, it was stepping on the toes of a Pentium 60 at 50Mhz, though that Pentium system had an early chipset so IPC would be falling somewhere between a POD on a 486 board and a P100 on 430FX

I haven't found it again yet, really want to run it on boards known to do 60, 66.... be cool also if I could remember which board had a 30Mhz setting, then I could stick my Cyrix 5x86 on that in 2x mode for internal clock of 60 and do head to head small integer benchmarks with the UMC at 60... I think the UMC might win actually. IPC of the cyrix is approx 1.3x 486 and 0.66 of Pentium, where the UMC seems to do maybe 1.5 and 0.8

Fun little ripper of a chip anyway, would have been so awesome to see these in DX4 form, they'd probably steal the crown from the AMD DX5 that hit 200

analog_programmer wrote on 2024-11-20, 15:15:

dionb wrote on 2024-11-20, 14:44:

Unless you were doing some very specific things very little 486-era software used an FPU let alone required it. Intel pretty cynically made use of the bad rep the 386SX had had for its castrated 16b bus to convince people they needed to shell out on a 486DX when the vast majority of them would have had the exact same experience on a cheaper 486SX. The first game that absolutely required an FPU was Quake, and good luck running that acceptably on any 486 even if you had an FPU.

So no, marketing aside I'd not call an FPU a requirement for a 'super 486', particularly not if we're not including all the clock doubled/tripled/quadrupled CPUs (this thing is nice, but CX5x86 and Am5x86 will blow it out the water of course).

From a purely practical point of view, you are right. Bur it would be interesting if someone could compare UMC U5D running at 50 MHz vs i486/DX2-50 or "the classic" i486/DX2-66 including their FPU performance.

Here are some results of mine running phil's pack at 50MHz in a M912 (with real L2 cache fitted):
3dbench 1.0c: 47.5
pcp320: 10.8
pcp640: 4.4
doom: 2782
lm60 cpu: 307.22
lm60 vid: 9733.07
topbench: 225
cachchk l1: 10
cachchk l2: 18
cachchk mem: 29

And here is the speedsys results:

Well done, maxtherabbit! So, if you have one of those ultra rare UMC U5D CPUs, I'm sure you can find at least one i486/DX2-50 or i486/DX2-66 to test them in your M919 board against it 😀

maxtherabbit wrote on 2024-11-20, 16:23:

Here are some results of mine running phil's pack at 50MHz in a M919 (with real L2 cache fitted): 3dbench 1.0c: 47.5 pcp320: 10. […]
Show full quote

analog_programmer wrote on 2024-11-20, 15:15:

dionb wrote on 2024-11-20, 14:44:

Unless you were doing some very specific things very little 486-era software used an FPU let alone required it. Intel pretty cynically made use of the bad rep the 386SX had had for its castrated 16b bus to convince people they needed to shell out on a 486DX when the vast majority of them would have had the exact same experience on a cheaper 486SX. The first game that absolutely required an FPU was Quake, and good luck running that acceptably on any 486 even if you had an FPU.

So no, marketing aside I'd not call an FPU a requirement for a 'super 486', particularly not if we're not including all the clock doubled/tripled/quadrupled CPUs (this thing is nice, but CX5x86 and Am5x86 will blow it out the water of course).

From a purely practical point of view, you are right. Bur it would be interesting if someone could compare UMC U5D running at 50 MHz vs i486/DX2-50 or "the classic" i486/DX2-66 including their FPU performance.

Here are some results of mine running phil's pack at 50MHz in a M919 (with real L2 cache fitted):
3dbench 1.0c: 47.5
pcp320: 10.8
pcp640: 4.4
doom: 2782
lm60 cpu: 307.22
lm60 vid: 9733.07
topbench: 225
cachchk l1: 10
cachchk l2: 18
cachchk mem: 29

Here are the results from some quick benchmarking with my lab setup that I annotated.
Sorry for the lack of screenshots. I was in a hurry yesterday; I was just checking that the CPU was stable and functional.

Motherboard: Soyo SY-025P2
Chipset: SiS 85C471/85C407
BIOS Date: 9/26/94
External Cache: 8 x 32KB, 15ns
Main Memory: 2 x 16MB, 72 Pin, 60ns
I/O Card: Acer SAB-560 Maximum AT
Video card: California Graphics Suntracer 2000V - Tseng Labs ET4000W32P 2MB VLB

UMC U5S @ 50MHz
Wolf3D (286 ver.) : 535 Ticks - 93.1 FPS
Doom (screenblocks=10) : 2673 RealTicks - 27.9 FPS
NSSI Dhrystone: 33503
Cachechk: 10/21/22 uS/KB
Speedsys: 27.33

For comparison purposes, same setup:

Intel 486DX-50
Wolf3D (286 ver.) : 630 Ticks - 79.1 FPS
Doom (screenblocks=10) : 3058 RealTicks - 24.4 FPS
NSSI Dhrystone: 26248
Cachechk: 21/29/30 uS/KB
Speedsys: 18.82

AMD 486SX2-66
Wolf3D (286 ver.) : 601 Ticks - 82.9 FPS
Doom (screenblocks=10) : 2528 RealTicks - 29.5 FPS
NSSI Dhrystone: 32011
Cachechk: 16/24/32 uS/KB
Speedsys: 24.40

analog_programmer wrote on 2024-11-20, 17:02:

Well done, maxtherabbit! So, if you have one of those ultra rare UMC U5D CPUs, I'm sure you can find at least one i486/DX2-50 or i486/DX2-66 to test them in your M919 board against it 😀

mine's a U5S not D sadly

analog_programmer wrote on 2024-11-20, 17:02:

Well done, maxtherabbit! So, if you have one of those ultra rare UMC U5D CPUs, I'm sure you can find at least one i486/DX2-50 or i486/DX2-66 to test them in your M919 board against it 😀

There was some info somewhere on here about the D being a DX socket type, set the board for DX but no co-pro, whereas S was for SX type/setting. So basically same core with slight pinout variance.

yes, to my knowledge they are identical FPU notwithstanding

analog_programmer wrote on 2024-11-20, 17:02:

Well done, maxtherabbit! So, if you have one of those ultra rare UMC U5D CPUs, I'm sure you can find at least one i486/DX2-50 or i486/DX2-66 to test them in your M919 board against it 😀

Why compare with DX2-50 and its lacklustre 25MHz bus? The fair comparison would be with the DX-50.

So even the D version would have been without fpu but just with the internal multiplier set to 2?
I'm a great fan of those cpus, some years ago I did some tests too for fast doom.
Clock to clock it was a beast

BitWrangler wrote on 2024-11-20, 18:18:

There was some info somewhere on here about the D being a DX socket type, set the board for DX but no co-pro, whereas S was for SX type/setting. So basically same core with slight pinout variance.

Maybe this is valid for U5SD model, but not for U5D. Who knows...

maxtherabbit, what is your model? U5S?

dionb wrote on 2024-11-20, 18:25:

Why compare with DX2-50 and its lacklustre 25MHz bus? The fair comparison would be with the DX-50.

Because I think i486/DX2-66 can run Doom at playable framerate and it's a widespread CPU 😀 And also because I want some FPU comparison with some other CPU/FPU testing software (of course Doom doesn't use the FPU), if U5D has FPU.

Ok, let's find at least one UMC U5D first 😁

BitWrangler wrote on 2024-11-20, 18:18:

analog_programmer wrote on 2024-11-20, 17:02:

Well done, maxtherabbit! So, if you have one of those ultra rare UMC U5D CPUs, I'm sure you can find at least one i486/DX2-50 or i486/DX2-66 to test them in your M919 board against it 😀

There was some info somewhere on here about the D being a DX socket type, set the board for DX but no co-pro, whereas S was for SX type/setting. So basically same core with slight pinout variance.

To add some gossip to the rumour mill, the U5SD supposedly is a CPU fully compatible with the 486DX, but the built in FPU turned out to be defective, so it was deactivated. It still retained the 486DX pinout. By the time the FPU was fixed and the x2 version was ready to ship (there are even some pics of it floating around...) it seems UMC decided to shut down its x86 business. I guess being mainly a foundry for third parties, UMC management must have thought it was better to be on good terms with everyone in the industry biz.

analog_programmer wrote on 2024-11-20, 18:32:

Maybe this is valid for U5SD model, but not for U5D. Who knows... […]
Show full quote

BitWrangler wrote on 2024-11-20, 18:18:

There was some info somewhere on here about the D being a DX socket type, set the board for DX but no co-pro, whereas S was for SX type/setting. So basically same core with slight pinout variance.

Maybe this is valid for U5SD model, but not for U5D. Who knows...

maxtherabbit, what is your model? U5S?

dionb wrote on 2024-11-20, 18:25:

Why compare with DX2-50 and its lacklustre 25MHz bus? The fair comparison would be with the DX-50.

Because I think i486/DX2-66 can run Doom at playable framerate and it's a widespread CPU 😀 And also because I want some FPU comparison with some other CPU/FPU testing software (of course Doom doesn't use the FPU), if U5D has FPU.

Ok, let's find at least one UMC U5D first 😁

U5S-40

UMC U5S Super40 vs. U5SX-40 + FPU?

Re: Identifying an old intel CPU
"40MHz UMC was available in Poland in December 1994, cost ~$84, half the price of AMD DX/40
Month prior also listed 33MHz version at ~same price."
Poland prices in December 1994:
Intel 486 sx/25 $71
UMC SX/40 $84
Intel 486 sx2/50 $140
Intel DX/33 $160
AMD DX/40 $160
Pentium 60 $500
Pentium 90 $740
1MB of ram ~$40

UMC was the clear performance/$ winner.

MSxyz wrote on 2024-11-20, 18:47:

it seems UMC decided to shut down its x86 business. I guess being mainly a foundry for third parties, UMC management must have thought it was better to be on good terms with everyone in the industry biz.

Intel sued https://www.proquest.com/docview/209696670?so … rade%20Journals
and forces UMC into settlement in January 1996 https://www.sfgate.com/business/article/off-t … ker-3160833.php

How super was it? https://x86.fr/the-uca-486-adapter-now-supports-umc-486s/
> While Intel 486s requires 40 cycles to perform a INT divide, UMC 486s only need 7 cycles.

Too bad companies rarely did use-case analysis when optimizing CPUs at that time. Nowadays Intel/AMD gather CPU instruction and instruction sequence stats from popular OSes/programs/games and use that data to concentrate on the hottest paths for most impact. Here it looks like UMC did some impressive optimizations of things that dont matter in real life (div).

dionb wrote on 2024-11-20, 14:44:

analog_programmer wrote on 2024-11-20, 08:46:

Nice find. Maybe UMC U5D would be the real "super 486" CPU if it had a FPU.

Unless you were doing some very specific things very little 486-era software used an FPU let alone required it. Intel pretty cynically made use of the bad rep the 386SX had had for its castrated 16b bus to convince people they needed to shell out on a 486DX when the vast majority of them would have had the exact same experience on a cheaper 486SX. The first game that absolutely required an FPU was Quake, and good luck running that acceptably on any 486 even if you had an FPU.

So no, marketing aside I'd not call an FPU a requirement for a 'super 486', particularly not if we're not including all the clock doubled/tripled/quadrupled CPUs (this thing is nice, but CX5x86 and Am5x86 will blow it out the water of course).

I disagree here. Quake infamously requires an FPU (and ran terribly on a 486, but could eek out acceptable results with a DX50, and vlb video), as does the period-correct 68k mac emulator Fusion.

wierd_w wrote on 2024-11-21, 00:44:

I disagree here. Quake infamously requires an FPU (and ran terribly on a 486, but could eek out acceptable results with a DX50, and vlb video)

No it couldnt, 5 fps is not acceptable 😀 But Duke3D does actually run acceptable on dx2-66, same for UMC U5S 486-40 except for parts of the map with slopes as those actually do use FPU https://youtu.be/-dNaFbm_hL0?t=1115 You get 20-30 fps down to choking when ramps are on screen because game uses FPU emulation.