MikeSG wrote on 2025-04-18, 12:04:

Yes should be able to run a resistor there.

Thank you, I tried a 27 Ω which gave 3.58 volts, booted into NT @ 100 MHz and started SIV. Next I switched the bus to 40 MHz, SIV updated the speed to 120 MHz and a couple of seconds later I got a BSOD.

I guess I need to go for a higher voltage, but am reluctant to do this. I think I will get a Model 9 DX4-100 (SX209, SK896,SK096), if I can find one, and see if that does any better. The SX896 is listed as '89, '94, maybe it will do better, but I wonder if '94 is an error and thus far have been unable to find a picture of one to check.

Out of interest what does the AMD2X jumper do?

red-ray wrote on 2025-04-18, 12:57:

Out of interest what does the AMD2X jumper do?

One of that jumpers should do the clock multiplier, the other may interfer with cache mode (WT/WB).
But this is depending on the processor. Am486DX4-SV8B and Am486DX4-NV8T behave different with the clock multiplier jumper!

If AMD2X is on R17 it's CLKMUL for i486DX4, and AMD486DX2(W) and DX4(W) write-back models. If it's on B13 it's either CLKMUL for AMD non write-back, or WB-WT selection for AMD/intel write-back CPUs.

3.75v would be safe for the CPU... I have the datasheet for the low profile embedded intel DX4-100 and it says 4.6v is the absolute max before damage occurs. But I wouldn't go past 4.0v.

AMD DX4 CPUs may overclock better, even the DX4-100 ones.

MikeSG wrote on 2025-04-18, 16:24:

AMD DX4 CPUs may overclock better, even the DX4-100 ones.

Thank you for the AMD2X information, my motherboard does not support write-back so I may need to use it for Model 9 CPU, then again my SX955 supports write-back and reports a CPUID of 0436 rather than 0470 meaning it's in write-through mode.

My AMD Am5x86 happily runs @ 160 MHz. I recall there is an AMD Am4 with 16KB of L1 cache that's really an Am5x86 and overclocks well.

I suspect I will be brave over the weekend and try a higher voltage.

MikeSG wrote on 2025-04-18, 16:24:

3.75v would be safe for the CPU...

I tried 3.70 and it was worse than 3.60 !, as soon as a switched to 40 MHz NT gave a BSOD. I also tried starting at 40 MHz, NT Loader started and got stuck before it got very far.

I switched back to 3.60 and when I switch to 40 MHz SIV updated the speed to 120 MHz and it ran for about 20 seconds before a BSOD.

MikeSG wrote on 2025-04-18, 16:24:

If AMD2X is on R17 ...

If it's on B13 it's either CLKMUL for AMD non write-back, or WB-WT selection for AMD/intel write-back CPUs.

I know that interposer. AMD2X is on B13. That's why they call it "AMD2X", as it is only ever a "2X" selector for AMD CPUs. If you set it, an AMD non-write-back DX4 ("NV8T") will run at 2X, if you leave it open, that type of AMD CPU will operate at 3X. MikeSG is correct that this pin is used for WB/WT selection on the later unified 486 pinout.

J1 (top right corner in the first photo of that interposer in this thread) is on R17.

mkarcher wrote on 2025-04-18, 21:57:

it is only ever a "2X" selector for AMD CPUs.

Thank you, what should I set for my AMD Am5x86 please?

I also noticed that when I change the multiplier from x4/x2 to x3 the CPU changes to an AMD Am486

red-ray wrote on 2025-04-18, 22:08:

mkarcher wrote on 2025-04-18, 21:57:

it is only ever a "2X" selector for AMD CPUs.

Thank you, what should I set for my AMD Am5x86 please?

I also noticed that when I change the multiplier from x4/x2 to x3 the CPU changes to an AMD Am486

Use the Intel jumper and set it to x2. Then you have x4.
If you set it to x3 the CPU identifier is changed as you already have noticed.
Please ensure how to enable write back cache on it too. This perhaps has to be done on the motherboard, if there is support at all.

red-ray wrote on 2025-04-18, 22:08:

mkarcher wrote on 2025-04-18, 21:57:

it is only ever a "2X" selector for AMD CPUs.

Thank you, what should I set for my AMD Am5x86 please?

For the Am5x86, that jumper selects L1WB or L1WT mode. The BIOS might display "Enhanced Am486" in WB mode and just "Am486" in WT mode. To not get issues in WB mode, the mainboard has to support that mode, and BIOS needs to configure the chipset correctly. If the POST screen displays 5x86 correctly, it is likely that WB mode is supported. OTOH, there were no 5V processors with the "standard" pinout used by the Intel enhanced DX4, the Am5x86 and (most?) 3.3V Cyrix 486/5x86 processor, so a board without 3.3V supply possibly does not support these CPUs in write-back mode.

red-ray wrote on 2025-04-18, 22:08:

I also noticed that when I change the multiplier from x4/x2 to x3 the CPU changes to an AMD Am486

That's normal. An 5x86 at 3x identifies the same way as an Am486DX4. Originally, the DX4 had just 8KB cache, while the 5x86 had 16KB cache. Later, AMD also sold the 5x86 dies with 16KB cache as 486DX4 processors. The only difference between these very late 486DX4 and the 5x86 is how you jumper them (4x vs. 3x) and what's printed on the case. The die and the bonding is identical.

mkarcher wrote on 2025-04-18, 22:27:

it is likely that WB mode is supported

Thank you, my DEC Venturis 4 V1.0 does not support write-back, AFAIK the DEC Venturis 4 V2.0 does, but I have just noticed it does not seem to have a 40 MHz FSB option. It does, I got the link fixed.

What is the brand and model of the power supply you are using? How old is it?

cyclone3d wrote on 2025-04-18, 23:11:

What is the brand and model of the power supply you are using? How old is it?

At the moment it's a Thermaltake Smart DPS G 700W, it reports itself as a SPG-0700D-BZ and it's maybe a couple of years old, its good for 18 amps @ 5 volts. I use it for all the motherboards I have that aren't in cases. I just plugged it into this W10 system and it's supplying < 20 watts to the i486.

I just installed the AMD Am5x86, connected the AMDx2 jumper and it's not in write-back mode. The BIOS reports it as a 486DX4

mkarcher wrote on 2025-04-18, 22:27:

OTOH, there were no 5V processors with the "standard" pinout used by the Intel enhanced DX4, the Am5x86 and (most?) 3.3V Cyrix 486/5x86 processor, so a board without 3.3V supply possibly does not support these CPUs in write-back mode.

Isn't the P24D aka 486DX2-66WB such a processor?

Ok, PSU is def not problem then. Just wanted to make sure it wasn't some old or poorly regulated PSU with dirty output.

jakethompson1 wrote on 2025-04-18, 23:47:

Isn't the P24D aka 486DX2-66WB such a processor?

I assume this is SX954 + SX955 and yes, it's a 5volt CPU that supports WB cache. I guess SIV should be reporting P24D rather than P24, should it?

red-ray wrote on 2025-04-19, 00:06:

I assume this is SX954 + SX955 and yes, it's a 5volt CPU that supports WB cache. I guess SIV would be reporting P24D rather then […]
Show full quote

jakethompson1 wrote on 2025-04-18, 23:47:

Isn't the P24D aka 486DX2-66WB such a processor?

I assume this is SX954 + SX955 and yes, it's a 5volt CPU that supports WB cache. I guess SIV would be reporting P24D rather then P24, should it?

https://archive.org/details/bitsavers_intel80 … e/n465/mode/1up

Because the cpuid changes based on how it's jumpered, I don't think you could distinguish a P24D in write through mode from a plain 486DX2.

jakethompson1 wrote on 2025-04-19, 00:19:

Because the cpuid changes based on how it's jumpered, I don't think you could distinguish a P24D in write through mode from a plain 486DX2.

Yes you can, the stepping is 6 rather than <= 5 and it also supports PSE .

Do you happen to know which other i486 CPUs support PSE (Page Size Extension) please?

Makes sense, I meant just by family/model

jakethompson1 wrote on 2025-04-18, 23:47:

mkarcher wrote on 2025-04-18, 22:27:

OTOH, there were no 5V processors with the "standard" pinout used by the Intel enhanced DX4, the Am5x86 and (most?) 3.3V Cyrix 486/5x86 processor, so a board without 3.3V supply possibly does not support these CPUs in write-back mode.

Isn't the P24D aka 486DX2-66WB such a processor?

This is likely correct, and red-ray also confirmed it, but I'd like to try the excuse "the P24D is likely not older than the &EW DX4 processor, and I'd be surprised if a mainboard at the time the P24D was available at the market chose to omit 3.3V support" next. I guess, someone will come around the corner soon and Link a board in TRW that does exactly this, so yeah, there is a 5V WB processor, and there might be a revision of a mainboard that just supported the PODP at 5V with WB in older revision and just got some extra traces to connect the "&EW" pins to the corresponding P24T pins, without adding a 3.3V regulator, as the latter would likely require major layout changes.

red-ray wrote on 2025-04-18, 22:38:

AFAIK the DEC Venturis 4 V2.0 does, but I have just noticed it does not seem to have a 40 MHz FSB option.

The silkscreen on the photo by Breakpoint does indicate a 40MHz setting, so I guess as long as a Venturis 4 V2.0 is equipped with J14 and J15, J14=ON, J15=OFF will yield 40MHz, even if that setting has been redacted from newer documentation.