mkarcher wrote on 2023-04-03, 21:29:

So possibly 7 is only used by &EW processors jumpered into WB mode.

Yes, I the model changes when it's in WB mode. I would very much like to know which Intel DX2 stepping's have CPUID and which don't my 0436 + a 0435 have is all I currently know.

Thank you for the link , I can see the Am486DX5-133 has CPUID and also changes model depending on if WB is active. I suspect the WB capable Am486DX2-66 + Am486DX4-100 have CPUID as well, but it's not 100% clear as to if they do.

I just checked and it looks like Windows 9X does not report Level and Revision when we really need it 🙁

Level is no big deal as for 386 + 486 it's the same as Type.

Windows 95 always returns Revision zero, 98 + Me return the same as CPUID, but it looks like if there is no CPUID they also return zero 🙁

I felt I needed an Intel 486 that did not have CPUID support and noticed my local computer recycler had an Intel i486SX-33 so I popped round on my bike and got it. It's CPUID 0423 and had CPUID support, I wonder do all Intel i486SX have CPUID support?

mkarcher wrote on 2023-04-03, 16:22:

Writing to CR0 to enable WB is a Cyrix thing, accompanied by locking the NWT "no write-through" bit to keep the cache in WB. For Intel/AMD processors, the processor enables the WB protocol if the WB/WT pin is pulled high on RESET. If the WB protocol is enabled, the processor samples the WB/WT pin during every cache-line fill, and if the pin is high at that point in time, write-back is enabled for that line. Typically, that pin is jumpered to either "high" or "low" all the time, so in WB mode, all cacheable memory is write-back capable.

Yesterday I noticed there was a P24D jumper, so I selected it, installed my DX2 CPU and checked what happened. Below you can see WB does not seem to be enabled, do you have any thoughts as to why? The Venturis 4 manual is hazy about WB on the V1 board I have and I wonder if it only works for Intel CPUs on the V2 board.

red-ray wrote on 2023-04-04, 23:33:

mkarcher wrote on 2023-04-03, 16:22:

Writing to CR0 to enable WB is a Cyrix thing, accompanied by locking the NWT "no write-through" bit to keep the cache in WB. For Intel/AMD processors, the processor enables the WB protocol if the WB/WT pin is pulled high on RESET. If the WB protocol is enabled, the processor samples the WB/WT pin during every cache-line fill, and if the pin is high at that point in time, write-back is enabled for that line. Typically, that pin is jumpered to either "high" or "low" all the time, so in WB mode, all cacheable memory is write-back capable.

Yesterday I noticed there was a P24D jumper, so I selected it, installed my DX2 CPU and checked what happened. Below you can see WB does not seem to be enabled, do you have any thoughts as to why? The Venturis 4 manual is hazy about WB on the V1 board I have and I wonder if it only works for Intel CPUs on the V2 board.

Take a look at this chart: https://x86.fr/wp-content/uploads/2021/06/P5O … out-768x758.png
and see whether your board has connectivity between C14 and pin T1. Note that these are Pentium OverDrive pin numbers, not 486 pin numbers. If not, you have one of the boards mkarcher was alluding to (I have one of them also). You can also check whether either side of the jumper has connectivity with either of those two pins.

jakethompson1 wrote on 2023-04-05, 03:03:

see whether your board has connectivity between C14 and pin T1.

If not, you have one of the boards mkarcher was alluding to (I have one of them also). You can also check whether either side of the jumper has connectivity with either of those two pins.

Thank you, C14 is connected to T1, but I can't find connectivity to any jumper pin, which pin should it be connected to please?

The board I have the the DEC Ventutis 4 V1, Service Maintenance Manual attached.

How rare/expensive are PODP5V83? I half fancy one, but there don't seem to be any for sale in Europe, let alone England.

red-ray wrote on 2023-04-05, 07:41:

jakethompson1 wrote on 2023-04-05, 03:03:

see whether your board has connectivity between C14 and pin T1.

If not, you have one of the boards mkarcher was alluding to (I have one of them also). You can also check whether either side of the jumper has connectivity with either of those two pins.

Thank you, C14 is connected to T1, but I can't find connectivity to any jumper pin, which pin should it be connected to please?

This likely means that enabling WB mode is not implemented on the V1 board. Unless there is a short circuit between C14/T1 and ground, try connecting that pin to Vcc (to be safe, you can use a 4.7K resistor or something like that). If you are lucky, all other provisions for the WB mode are already present. If you are unlucky, your system will crash as soon as you try to access the floppy, or it will read/write garbage data. In the latter case, the board is likely not designed for L1WB on P24D-type processors (Intel "&EW") at all.

mkarcher wrote on 2023-04-05, 08:47:

try connecting that pin to Vcc (to be safe, you can use a 4.7K resistor or something like that)

Thank you, I gave it a try, with 4K7 to VCC I saw about 4.6 Volts on T1, the system posted, said (WB) and corrupted the CMOS. I also noted the 463 meaning a speed of 63 rather than the usual 66 MHz.

Is what I did what you had in mind? What do you think would happen if with the 4K7 installed I used a 486SX-33 or 486DX-33?

After I removed the 4K7 and restored the CMOS all is back to as it was before.

red-ray wrote on 2023-04-05, 11:31:

mkarcher wrote on 2023-04-05, 08:47:

try connecting that pin to Vcc (to be safe, you can use a 4.7K resistor or something like that)

Thank you, I gave it a try, with 4K7 to VCC I saw about 4.6 Volts on T1, the system posted, said (WB) and corrupted the CMOS. I also noted the 463 meaning a speed of 63 rather than the usual 66 MHz.

Is what I did what you had in mind?

Yes, it is. It seems to show that the board is physically not compatible with the Intel &EW WB protocol, even though the chipset and the BIOS would support it if the required traces and jumpers were present. You basically have to add some bodge wires between the chipset and the processor socket to make the WB protocol work. Unless you specifically need WB in that board in that machine, or you enjoy physically tinkering in with your hardware, getting a different board for 486 WB support is likely the smarter approach.

red-ray wrote on 2023-04-05, 11:31:

What do you think would happen if with the 4K7 installed I used a 486SX-33 or 486DX-33?

I'm unsure whether the latest 486SX-33 include WB support (it would say "&EW" on the case), but if not, the 486DX or 486SX just will ignore this resistor.

mkarcher wrote on 2023-04-05, 14:40:

I'm unsure whether the latest 486SX-33 include WB support (it would say "&EW" on the case), but if not, the 486DX or 486SX just will ignore this resistor.

Thank you for confirming things and for now I won't be making any hardware mods. With my Intel DX4 100 installed the system is fast enough for all the testing I wish to do.

My DX-33 just arrived and with that installed the the 4K7 in place no (WB) is reported and the system runs OK. I also noted the voltage was a about 4.9 rather than 4.6 volts. I decided not to try the DX4 as the interposer get's "in the way", there is on WB on my i486 SX so I did try it.

It's also startling to look like the Revision returned by Windows NT is next to useless for x32 CPUs 🙁 I wonder if it depends on the motherboard/BIOS and guess time may tell.

mkarcher wrote on 2023-04-05, 14:40:

Yes, it is. It seems to show that the board is physically not compatible with the Intel &EW WB protocol, even though the chipset […]
Show full quote

red-ray wrote on 2023-04-05, 11:31:

mkarcher wrote on 2023-04-05, 08:47:

try connecting that pin to Vcc (to be safe, you can use a 4.7K resistor or something like that)

Thank you, I gave it a try, with 4K7 to VCC I saw about 4.6 Volts on T1, the system posted, said (WB) and corrupted the CMOS. I also noted the 463 meaning a speed of 63 rather than the usual 66 MHz.

Is what I did what you had in mind?

Yes, it is. It seems to show that the board is physically not compatible with the Intel &EW WB protocol, even though the chipset and the BIOS would support it if the required traces and jumpers were present. You basically have to add some bodge wires between the chipset and the processor socket to make the WB protocol work. Unless you specifically need WB in that board in that machine, or you enjoy physically tinkering in with your hardware, getting a different board for 486 WB support is likely the smarter approach.

red-ray wrote on 2023-04-05, 11:31:

What do you think would happen if with the 4K7 installed I used a 486SX-33 or 486DX-33?

I'm unsure whether the latest 486SX-33 include WB support (it would say "&EW" on the case), but if not, the 486DX or 486SX just will ignore this resistor.

Taking a second look at the jumper, it's odd that P24T (Pentium OverDrive) & P24D (486DX2WB) are lumped together with AM486DXL, which is a WT-only CPU. Perhaps it's actually for SL-enhanced CPUs and has nothing to do with WT/WB?

I didn't think any of the non-clock doubled 486 CPUs had support for WB internal cache (well, perhaps except Cyrix). Isn't the payoff it offers greater as the clock multiplier increases?

jakethompson1 wrote on 2023-04-05, 22:30:

I didn't think any of the non-clock doubled 486 CPUs had support for WB internal cache

AFAIK it's just the later Intel DX2 + DX4 + PODP5V that have support for WB L1 cache. Do you mean "didn't", if so which other CPUs do you think have?

red-ray wrote on 2023-03-31, 21:44:

I wonder why my SX807 has CPUID 0436 which is different to the 0435 specified in https://www.cpu-world.com/sspec/SX/SX807.html

I wanted to see if my Intel DX2 CPUID 0436 really has WB support given cpu-world says SX807 does not have WB support.

red-ray wrote on 2023-04-05, 22:50:

jakethompson1 wrote on 2023-04-05, 22:30:

I didn't think any of the non-clock doubled 486 CPUs had support for WB internal cache

AFAIK it's just the later Intel DX2 + DX4 + PODP5V that have support for WB L1 cache. Do you mean "didn't", if so which other CPUs do you think have?

The ones that have WB L1 would be:

• The "P24D" version of the Intel 486DX2
• The "&EW" version of the IntelDX4 (other than the voltage and multiplier, the jumpers for P24D and write-back IntelDX4 should be exactly the same.
• Enhanced Am486 & Am5x86
• All Cyrix Cx486 and 5x86. As Cyrix introduced WB L1 before Intel did, they came up with their own pinout, before later switching to the "Standard Pinout" once Intel released P24D. This is why 486 boards end up with so many jumpers to switch between them.
• The Pentium OverDrive ("P24T") of course.

jakethompson1 wrote on 2023-04-05, 22:56:

The ones that have WB L1 would be:

Thank you for the summary, in many ways it's harder to know what these old CPUs support than Ryzen/Epyc as the datasheets are no longer on the AMD site, not even the NDA one.

red-ray wrote on 2023-04-05, 23:07:

jakethompson1 wrote on 2023-04-05, 22:56:

The ones that have WB L1 would be:

Thank you for the summary, in many ways it's harder to know what these old CPUs support than Ryzen/Epyc as the datasheets are no longer on the AMD site, not even the NDA one.

Yeah. There are also docs like this: http://datasheets.chipdb.org/SGS%20Thomson/x86/486/4193.PDF
Which were intended for motherboard manufacturers, and show which pins of the socket vary between CPUs, and which need a jumper (and why).

jakethompson1 wrote on 2023-04-05, 23:09:

There are also docs like this: http://datasheets.chipdb.org/SGS%20Thomson/x86/486/4193.PDF

Thank you, very handy. It says the Intel DX4 supports x2.5, so I just tried it and it does not work, my CPU ended up @ 66 MHz rather than 83 MHz, do you know if x2.5 should work please ?

It's connected using the interposer shown in Re: Digital Venturis PSU type, Video DRAM chips and CPU Upgrade Options

red-ray wrote on 2023-04-05, 23:32:

Thank you, very handy. It says the Intel DX4 supports x2.5, so I just tried it and it does not work, my CPU ended up @ 66 MHz rather than 83 MHz, do you know if x2.5 should work please ?

The Vogons Wiki (and I have read it elsewhere) says that was supposed to be for a DX3 (thus the DX4 being x3) but it was never released.

mkarcher wrote on 2023-04-05, 14:40:

Yes, it is. It seems to show that the board is physically not compatible with the Intel &EW WB protocol, even though the chipset and the BIOS would support it if the required traces and jumpers were present..

I just spotted a cheapish AMD Am5X86 CPU on eBay which made me wonder if this would work in WB mode on my Venturis 4, do you think it would and could I switch it between WB and WT mode please?

As long as it should work in WT mode I am inclined to get it and try.

red-ray wrote on 2023-04-06, 08:08:

mkarcher wrote on 2023-04-05, 14:40:

Yes, it is. It seems to show that the board is physically not compatible with the Intel &EW WB protocol, even though the chipset and the BIOS would support it if the required traces and jumpers were present..

I just spotted a cheapish AMD Am5X86 CPU on eBay which made me wonder if this would work in WB mode on my Venturis 4, do you think it would and could I switch it between WB and WT mode please?

As long as it should work in WT mode I am inclined to get it and try.

It should work fine in WT mode. It will fail in WB mode in a similar way as the DX2 CPU, because the Am5x86 uses the same &EW protocol as the WB-enabled Intel DX2/DX4 CPUs. Possibly the failure mode is different, because the BIOS might not recognize the 5x86 in x4/WB mode as WB-capable processor, so it might not print "WB" and might not set the chipset to expect the &EW protocol, but DMA stuff is still going to fail if the processor operates in WB mode and the chipset expects a WT processor.

To enable x4 mode, put the x2/x2.5/x3 jumper into the x2 position. Am5x86 CPUs do not have an x2 mode, and interpret that jumper setting as x4.

Oh, this dealer also has some Am486 DX4 NV8T - that are the ones without CPUID.
Too bad he has no other interesting non-CPUID 486 processors.

When you have the CPU above, the Am486 DX2 80 isn't interesting because the DX4 can be jumpered to behave as DX2.

And the Am486 DX4 V8T seem to be very rare.