According to this legitimately looking document, the VM386SX+ is its own design that is 10-40 % faster than baseline Intel 386SX.
Suddenly the 386SX field is more interesting.

The same document contains very interesting information about Cyrix and Texas Instruments.
There is a diagram showing how their offerings started diverging around year 1993, because of IP rights dispute.
This led TI to develop the SXL chips, which are the better place compared to SLC, DLC, and DRx2.
And so on.

pshipkov wrote on 2023-08-30, 06:09:

According to this legitimately looking document, the VM386SX+ is its own design that is 10-40 % faster than baseline Intel 386SX […]
Show full quote

According to this legitimately looking document, the VM386SX+ is its own design that is 10-40 % faster than baseline Intel 386SX.
Suddenly the 386SX field is more interesting.

The same document contains very interesting information about Cyrix and Texas Instruments.
There is a diagram showing how their offerings started diverging around year 1993, because of IP rights dispute.
This led TI to develop the SXL chips, which are the better place compared to SLC, DLC, and DRx2.
And so on.

The article:

Performance on system-level, PC-oriented benchmarks is 5–15% better than a standard 386SX at the same clock rate; on some benchmarks, such as Dhrystone, it is as much as 35–40% faster

I assume it's mostly still bottlenecked by its 16 bit bus?

That was a pretty interesting read.
In the report, there is no mention of a 386DX pinout version of the SXL2.
It also states that the 168-pin version of the SXL2 is 486SX pin compatible, which is obviously not the case.

I wonder if there are other reports from this magazine on TIs later 486 developments.
I am pretty interested in the TI486DX2 and DX4. Based on what was printed in this article, it sounded like TI couldn't expect any further technology sharing from Cyrix, so that implies that their DX2 and DX4 were based on the SLC/DLC but developed independently. Did TI have access to the Fasmath FPU? Has anyone ever compared FPU performance of the TI and Cyrix DX2/DX4?

This article is from year 1993, but the PGA132 DLC/SXL2 models came to market in 1994.
Apparently both Cyrix and Texas Instruments saw market opportunity for 386DX upgrades and went for it.

Early 486 components are in my blind spot, so cannot comment on your questions about TI/Cyrix 486 designs.
But their 486 silicon was clock-to-clock slower than corresponding Intel stuff.

pshipkov wrote on 2023-08-29, 22:37:

Very interesting. Few questions […]
Show full quote

Very interesting.
Few questions

Never heard of VM Technology VM386SX+ : )
Did you solder it in the Alaris Leopard board ?
I don't think there are Leopard boards with sockets for the 386SX CPUs.

Can you specify which Matra motherboard you tested with ?
This one ?
Notice that you can easily get it to 45/90 MHz, but it is not fully stable there.

Yes i soldered it to the Alaris. The performance is the same as the ALi M1386SX CPU.
Yes that Matra was it, mine has Efar chipset.
The IBM 386SLC is 20 or 25MHz and it seems it doesnt like 30+MHz. Maybe i should use better cooling.
My second try to the fastest 286 config: 33MHz (i just found that you made it faster (but less MHz at least), mine has some bottlenecks: still real HDD (WD Caviar 420MB cant catch up the speed, Seagate 520MB almost working), my fastest Tseng W32i Video card also didnt works, just the slower ET4000AX one, the IDE controller seems fine, maybe the 60ns memory also slow for this config.

https://www.youtube.com/watch?v=S_19KosuGY0

If ALi M1386SX has the same perf as VM386SX+, this is a strong indication they are the same on the inside.
Online searches doesn't return anything meaningful enough to support, or disprove it.

Noted about 386SLC.

About 286 - very nice. Good video (bombastic name ! 😉 ).
What motherboard ? If you can edit your previous post with this info, i would like to link it in the dictionary at the beginning of the thread.

The best i got out of 286 CPU was 35 (or was it 36 ?) MHz, but at the expense of increased DRAM wait states.
It was with this motherboard.
Still, the system was hopelessly slow.

https://websrv.cecs.uci.edu/~papers/mpr/MPR/A … CLES/071702.pdf
https://websrv.cecs.uci.edu/~papers/mpr/MPR/A … CLES/080101.pdf
https://websrv.cecs.uci.edu/~papers/mpr/MPR/A … CLES/080201.pdf
https://websrv.cecs.uci.edu/~papers/mpr/MPR/A … CLES/080602.pdf
https://websrv.cecs.uci.edu/~papers/mpr/MPR/A … CLES/090101.pdf

Here's a few more Microprocessor Report articles I've cherry picked concerning Cyrix/TI.
I still don't have a definite answer about the TI DX2/DX2, but according to the last article I found they were likely gifted to TI by Cyrix as part of the lawsuit settlement.
The answer is likely somewhere in a 1995 article.

pshipkov wrote on 2023-08-30, 18:28:

If ALi M1386SX has the same perf as VM386SX+, this is a strong indication they are the same on the inside. Online searches doesn […]
Show full quote

If ALi M1386SX has the same perf as VM386SX+, this is a strong indication they are the same on the inside.
Online searches doesn't return anything meaningful enough to support, or disprove it.

Noted about 386SLC.

About 286 - very nice. Good video (bombastic name ! 😉 ).
What motherboard ? If you can edit your previous post with this info, i would like to link it in the dictionary at the beginning of the thread.

The best i got out of 286 CPU was 35 (or was it 36 ?) MHz, but at the expense of increased DRAM wait states.
It was with this motherboard.
Still, the system was hopelessly slow.

It should correct however my experience is that ALi M1386SX works on 3.3V and seems to get hot on 5V, the VM386SX+ didnt run on 3.3V and not so hot on 5V.

My motherboard is the same as yours 😀. I also can reach 35MHz but it was with floppy drive only (i didnt try Seagate then just the WD, maybe i will switch to CF next time), and with a slow VGA card. Yes i saw that the memory handling is not that fast. Maybe I should try more motherboards...

The different thermal "profile" of the two 386 CPUS at 3.3 and 5 volts is interesting indeed.
So, maybe they are different after all and simply deliver similar performance.

Hah, using the same board. Okay. That makes sense.
Switching to CF cards makes it much easier. Also, they tend to be more resistant to overclocking activities than old HDDs.
About testing more motherboards - that will be welcome, but try to pick models that have not been examined already - will be more interesting in general.
It has been established that boards based on the VLSI200 series chipset are top performers, but there is always a chance for new findings.

Thanks for the links @Anonymous Coward.
Very good information there.

Micro Express Inc. Forex 386 Cache v3.61 based on Forex FRX46C411 / FRX46C402 chipset hinted at some interesting capabilities with 386DX processors but was not great with TI486SXL2 and didn't support IBM BL3.
It was not obvious if that particular assembly was problematic or is the chipset itself, so kept looking for other Forex mobo, especially later 46C421/46C422, or 46C451 models.
Not far ago i managed to obtain such motherboard from the same manufacturer.
There was some damage in the upper left area, but price was low-enough so decided to buy it..

Micro Express Inc. Forex 386/486 rev. B2/B3 based on Forex FRX46C421A, FRX46C422 chipset.

It has soldered on board AMD 386DX-40 processor which is not ideal, but for some reason most/all boards based on FRX46C421 and FRX46C451 are that way.
Still, it will be sufficient to see where things are with classic 386 CPUs and try to extrapolate the assembly/chipset qualities from there.

First had to fix quite a bit of damage caused by leaked battery.
That turned well and the board came to life.
The on-board soldered clock generator goes up to 40MHz only. Luckily there is wiring for crystal oscillator.
Socketed the clock generator and added 2 more female pins for the oscillator.

Overly complicated setup of the level 2 cache.

System is unpretentious - takes 60/70 ns RAM with ease, but tripped a bit on level 2 cache chips. Trusted WinBond and ISSI sets didn't go well. UMC 15ns ones worked fine.
BIOS offers 4-5 wait state parameters - ISA bus divider, SRAM, DRAM delays.

Assembly does not overclock well. Maxes out at 45MHz.
Fully stable in DOS, but fails complex tests and Windows.
This matches what the other Micro Express board (FRX46C401/2) was able to achieve.
At the same time performance is much lower. Didn't expect that. Disappointing.
40MHz is completely uninteresting, so i mostly lost interest.
Finished testing 45MHz despite the compromised stability.
This way could at least compare the two boards/chipsets more directly.

All BIOS settings on max, except:
ISA bus divider = CLK2/4 (best is /3), results in 11.25MHz
CACHE READ WAIT STATE = 1 (best is 0)
DRAM READ WAIT STATE = 2 (best is 0)

Used STB Nitro 2Mb video card (Cirrus Logic GD-5434) and standard IDE controller, 16Mb 60ns RAM, 256Kb level 2 cache.

Updated the combined benchmark results accordingly.
Look for Micro Exp 386/486 B3.

Tried different BIOS versions - results didn't change.

---

In the past i had 1-2 wide-format 386 boards based on Forex chipsets - none of them were good. Ebay took them before this thread started, so they are not presented here.
This board's performance is not much different. In fact its performance is even lower than the previously examined one, which is based on earlier Forex chipset and was not great to start with.
Maybe it is time to make bets about FRX46C451 - will it go further that path, or correct the course ...

pan069 wrote on 2023-08-30, 06:48:

Performance on system-level, PC-oriented benchmarks is 5–15% better than a standard 386SX at the same clock rate; on some benchmarks, such as Dhrystone, it is as much as 35–40% faster

I assume it's mostly still bottlenecked by its 16 bit bus?

The article talks about SX derivatives, so 16-bit bus is a limiting factor for all of them.
But the later DX derivatives are different. Namely the 132 pinned 486DLC, 486DXr2, 486SXL2 processors.

Unknown brand TDVIP-2 based on IMS IM8848, IM8849.

Lots of PCI/VLB/ISA slots.

There is no information online about this motherboard and its chipset. At least i was not able to find anything meaningful.
Clock generator frequency configurations are printed on the silk screen.
Level 2 cache jumpers are easily identifiable. The config for 1024Kb buffer can be derived after some experimentation.
White jumper JP41 is for level 2 cache in WB mode.
Jumpers for 3/5 volts to CPU.
Jumpers for VLB wait states nearby the slots - easily identifiable.

486 class processors with level 1 WB cache don't work on this board.
Didn’t find a working jumpers configuration. The few BIOSes available online for this chipset were not helpful either.
There is a chance that i missed something, but the probability for that is low after the long game of musical chairs with the jumpers. Disappointing.

Intel Pentium Overdrive P24T processors work, but only with up to 33MHz FSB. Attempts for 40MHz result in no lights.
Pins 2-3 of JP25 were shorted by default. This was a problem with P24T - POST hangs with corrupted display. Setting the jumper to 1-2 fixes it, but no idea why.

Board does not support EDO RAM.

Has anyone tried running the Asus PVI-486SP3 at 200 MHz? I was going to try running this board at 180 MHz, only to be surprised that the PLL has options for up to 50 MHz. Unfortunately, I don't have an Am5x86 which can run at 200 MHz.

If you've run yours at 200 MHz, what was the outcome? There's only two voltage selections, 3.45 V and 3.6 V. The manual lists 5V CPUs as supported, so it probably reads the VOLDET signal then does some automatic switching. Best thing would probably be to modify the 3.6 V voltage set resistor and turn it into 4 V or variable.

JP20 JP18 open provides 5V to CPU.
System is unstable in this configuration.
At least i was unable to make it work.

Thanks.

When I look at the jumper configuration for some 5 V CPUs on the PVI-486SP3, I don't see JP20 all open. For example, P24T has JP20 set to 5-6; or U5S has JP 20 set to 4-5; or i486SX has JP20 set to 1-2 and 5-6.

Thus, my guess is that you are saying to remove JP20 1-2 for just the Am5x86 settings to enable 5V? Did you measure the voltage coming into the CPU?

I have removed all pins on JP20 and measured the voltage directly at the CPU, for the case that a) JP7 is on 1-2, and b) JP7 is on 2-3. The results were that JP20 does nothing to the voltage at the CPU. Are you sure JP20 controls 5V? Maybe another jumper on JP20 is needed? There are header pins 1-6.

There exists JP6, which is near the VRM and without a header. I wonder if I should solder on a header here to get the 5V.

Deleted

JP16 on my rev 1.8 board is one of the 3 cache jumpers. Removing JP16 didn't alter the voltage.

I see you are using version 1.22. Looking at your board photo, JP16 is still the middle cache jumper.

Could you confirm the jumper?

If using the onboard TIP127, the highest no-load voltage it can output to the CPU is 4.15 V. With load, it is closer to 4.08 V. I beleive changing the 14K or 15K resistor at R3 or R4 with a 10K will allow for this 4.08 V output. However, I was hoping to experiment with 5 V.

EDIT: Based on your response about JP16, I think you are referring to the VLI board. I've been talking about the PVI board. You talk about it on the first page of this thread.

Some initial thoughts on this board:
I've run some tests on the PVI-486SP3 using a Voodoo3, Am5x86, 256K, and 64M FPM. DOOM benchmarks reveal that this board is faster at 150 MHz (1319 realtics) compared to 160 MHz (1412 realtics), likely because the L2 and RAM throughput's are substantially faster. A major problem I've run into is that I cannot have the Cache Write Cycle set to 2 and use the Voodoo3 card in Windows 95. With this setting at 2, Windows cannot find or cannot assign drivers to the Voodoo3 card. I get a message saying that there is a problem with my display adaptor, which cannot be resolved. If I set the Cache Write Cycle to 3, there are no issues in Windows with the Voodoo3.

Unfortunately, setting the Cache Write Cycle to 3 kills the DRAM performance, e.g. DRAM Read speeds go from 59.7 MB/s to 45.5 MB/s, and DRAM Write speeds go from 104 MB/s to 68.2 MB/s. Whether I am running the system at 160 MHz or 150 MHz, I must set Cache Write Cycle to 3 for the graphics card to work in Windows 95. I don't understand why. I'm not sure if this issue is specific to the Voodoo3, or if other graphics cards are afflicted in the same manner.

feipoa wrote on 2023-09-24, 08:42:

JP16 on my rev 1.8 board is one of the 3 cache jumpers. Removing JP16 didn't alter the voltage. […]
Show full quote

JP16 on my rev 1.8 board is one of the 3 cache jumpers. Removing JP16 didn't alter the voltage.

I see you are using version 1.22. Looking at your board photo, JP16 is still the middle cache jumper.

Could you confirm the jumper?

If using the onboard TIP127, the highest no-load voltage it can output to the CPU is 4.15 V. With load, it is closer to 4.08 V. I beleive changing the 14K or 15K resistor at R3 or R4 with a 10K will allow for this 4.08 V output. However, I was hoping to experiment with 5 V.

EDIT: Based on your response about JP16, I think you are referring to the VLI board. I've been talking about the PVI board. You talk about it on the first page of this thread.

Some initial thoughts on this board:
I've run some tests on the PVI-486SP3 using a Voodoo3, Am5x86, 256K, and 64M FPM. DOOM benchmarks reveal that this board is faster at 150 MHz (1319 realtics) compared to 160 MHz (1412 realtics), likely because the L2 and RAM throughput's are substantially faster. A major problem I've run into is that I cannot have the Cache Write Cycle set to 2 and use the Voodoo3 card in Windows 95. With this setting at 2, Windows cannot find or cannot assign drivers to the Voodoo3 card. I get a message saying that there is a problem with my display adaptor, which cannot be resolved. If I set the Cache Write Cycle to 3, there are no issues in Windows with the Voodoo3.

Unfortunately, setting the Cache Write Cycle to 3 kills the DRAM performance, e.g. DRAM Read speeds go from 59.7 MB/s to 45.5 MB/s, and DRAM Write speeds go from 104 MB/s to 68.2 MB/s. Whether I am running the system at 160 MHz or 150 MHz, I must set Cache Write Cycle to 3 for the graphics card to work in Windows 95. I don't understand why. I'm not sure if this issue is specific to the Voodoo3, or if other graphics cards are afflicted in the same manner.

It is the jumper "on top", for both PVI and VLI
So on the PVI remove JP 18 1-2, then you will get 5 volts.
Also on the VLI the position is the same, but the number is JP 16 on the VLI.
EDIT: For the VLI it is jumper 16 for Rev.2.x, I don't know the location on the older revisions at the moment.