mkarcher wrote on 2023-04-06, 08:14:

To enable x4 mode, put the x2/x2.5/x3 jumper into the x2 position.

Thank you for confirming what will happen, I recall reading x2 means x4 somewhere, but can't recall there. I just looked in the datasheet and can't find anything about which multiplies are supported.

I assume an ADZ is preferable to an ADW as it allows 85° C rather than 55° C, is there any downside to getting an ADZ and wonder, why does the ADW exist at all ? I guess chips what failed in testing.

red-ray wrote on 2023-04-06, 10:15:

mkarcher wrote on 2023-04-06, 08:14:

To enable x4 mode, put the x2/x2.5/x3 jumper into the x2 position.

Thank you for confirming what will happen, I recall reading x2 means x4 somewhere, but can't recall there. I just looked in the datasheet and can't find anything about which multiplies are supported.

I am looking at AMD publication #19751, Rev C, issued March 1996. Looking at page 14, it says:

AMD 5x86 data sheet wrote:

CLKMUL

Clock Multiplier (Input)

The microprocessor samples the CLKMUL input signal at RESET to determine the design operating frequency. An internal pull-up resistor connects to VCC , which selects Clock-tripled mode if the input is High or left floating. For Clock-quadrupled mode, the input must be pulled Low. For 133-MHz processors, this input must always be connected to VSS to ensure correct operation.

red-ray wrote on 2023-04-06, 10:15:

I assume an ADZ is preferable to an ADW as it allows 85° C rather than 55° C, is there any downside to getting an ADZ and wonder, why does the ADW exist at all ? I guess chips what failed in testing.

The ADZ is indeed preferrable, and you are likely correct that ADW ones are "B grade" processors that fail to work reliably at elevated temperatures.

I've read that ADW ones may be easier to overclock (to 200 MHz) than the ADZ, but I have no clue why some people here in the old threads of this forum say that.

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

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

The last one to sell on CPU-World with fan was $13. I have one sat on the desk in front of me. Seen them as high as $60 as well.

debs3759 wrote on 2023-04-06, 11:00:

The last one to sell on CPU-World with fan was $13. I have one sat on the desk in front of me. Seen them as high as $60 as well.

Thank you, I look @ CPU-World one in a while and feel most of the time it's hard to find is a given CPU is for sale, maybe I am doing something wrong, when I searched for PODP5V63 they all look to have been sold, I could not find a way to search for unsold ones, is there one please?

I searched in the For Sale forum, and checked the last few.

It might be worth posting in the Want to Buy (Collectible Items) sub forum (https://www.cpu-world.com/forum/viewforum.php?f=4)

I am surprised that the model number changed from 14 (0E) to 8 (08) depending on that the multiplier is.

I am wondering if it's possible to tell is a CPU is an AMD Am5x86 in Am486 mode from the stepping or some other thing, if you have a real AMD Am486 please will you post what SIV reports for it?

SIV looks to have a bug on the cache reporting and looking at [Cache-0 Latency] SIV should have got it right so I will fix this.

red-ray wrote on 2023-04-12, 12:42:

SIV looks to have a bug on the cache reporting and looking at [Cache-0 Latency] SIV should have got it right so I will fix this.

Yes, a silly bug at the level of a typo.

Real AMD Am486 CPUs have an 8KB L1 Unified cache so should SIV use this to detect an AM5x86 in AM486 mode? I feel it should, but what should it report?

I can't see any issue with what the latest SIV 5.70 Beta-05 development reports, the real test be what it reports on other systems.

red-ray

Did you read this section

Am486DX4-100SV8B (remarked 5×86)

on this website ? https://x86.fr/the-odd-story-of-factory-downgraded-486s/

weedeewee wrote on 2023-04-12, 18:55:

Did you read this section
Am486DX4-100SV8B (remarked 5×86)
on this website ? https://x86.fr/the-odd-story-of-factory-downgraded-486s/

weedeewee, what's about processors like AMD Am486DX4-100V16BGC or AMD Am486DX4-100V16BGI ?

red-ray wrote on 2023-04-12, 12:42:

I am surprised that the model number changed from 14 (0E) to 8 (08) depending on that the multiplier is.

It's according to the data sheet. See https://www.amd.com/system/files/TechDocs/20736.pdf , page 55. The model number is dependent on the multiplier. Also note that the WB/WT affects the CPUID as well.

red-ray wrote on 2023-04-12, 12:42:

I am wondering if it's possible to tell is a CPU is an AMD Am5x86 in Am486 mode from the stepping or some other thing, if you have a real AMD Am486 please will you post what SIV reports for it?

That depends a lot on the variant of the Am486 you are looking at. The (preliminary) data sheet I linked above is for the "AMD Enhanced 486 familiy", and contains specifications for 16KB L1 processors at x2, x3 and x4. They are still named "DX2", "DX4" and "DX5", the name "5x86" is not yet mentioned in the data sheet, but it's exactly what this data sheet calls "DX5". There are a lot of different AMD 486 processors, even just looking at the DX4 family: The original one containing the Intel ICE microcode ("V8T"), the one after the lawsuit without that microcode ("NV8T"), the one with SMM and CPUID support ("SV8T"), the one with SMM support and L1WB support ("SV8B"), and the late one with 16KB L1 cache ("V16BGC", sometimes marked as V8BGC, see some amd486dx4-sv8b do have 16kb cache as if remarked from v16bgc). There is a comprehensive overview on the AMD 486 dies and packages at https://x86.fr/the-ultimate-amd-486-die-packaging-guide/ .

You can possibly identify the die variant and the stepping using CPUID on the processors with SMM (and CPUID) support. You won't be able to tell a Am486DX4-100V16BGC jumpered for x3 from an AM5x86-133V16BGC jumpered for x3, though. They both use the same 350nm die with 16KB L2 cache in the 25544 package, just with a different printing on the case! You also won't be able to tell an Am486DX2-66V16BGC jumpered for x2 from an Am486DX4-100V16BGC jumpered for x2.

So, the whole 5x86 name is just marketing. It's the latest generation Am486 with clock quadrupling support, and it identifies as such. Keep in mind that the CPUID instruction was not meant to tell the computer what is printed on the case to display it to the user, but to tell the computer how the processor behaves. This is why the multiplier and the caching mode influence the CPUID. And in this regard, a late 16KB L1 cache Am486DX4 behaves exactly the same as an Am5x86.

mkarcher wrote on 2023-04-12, 19:30:

red-ray wrote on 2023-04-12, 12:42:

I am surprised that the model number changed from 14 (0E) to 8 (08) depending on that the multiplier is.

It's according to the data sheet. See https://www.amd.com/system/files/TechDocs/20736.pdf , page 55. The model number is dependent on the multiplier. Also note that the WB/WT affects the CPUID as well.

Would this be for BIOS compatibility for BIOSes that either reject DX values not on a known list, or that have a catch-all that programs the chipset with super-conservative settings if it can't identify the bus speed?

mkarcher wrote on 2023-04-12, 19:30:

So, the whole 5x86 name is just marketing. It's the latest generation Am486 with clock quadrupling support, and it identifies as such. Keep in mind that the CPUID instruction was not meant to tell the computer what is printed on the case to display it to the user, but to tell the computer how the processor behaves. This is why the multiplier and the caching mode influence the CPUID. And in this regard, a late 16KB L1 cache Am486DX4 behaves exactly the same as an Am5x86.

The Cyrix 5x86 came first so was this about AMD having a "586-class" product for PCI 486 budget PCs, with the K5 either late or not competitive?

jakethompson1 wrote on 2023-04-13, 02:16:

mkarcher wrote on 2023-04-12, 19:30:

red-ray wrote on 2023-04-12, 12:42:

I am surprised that the model number changed from 14 (0E) to 8 (08) depending on that the multiplier is.

It's according to the data sheet. See https://www.amd.com/system/files/TechDocs/20736.pdf , page 55. The model number is dependent on the multiplier. Also note that the WB/WT affects the CPUID as well.

Would this be for BIOS compatibility for BIOSes that either reject DX values not on a known list, or that have a catch-all that programs the chipset with super-conservative settings if it can't identify the bus speed?

Measuring the CPU clock speed is straightforward, but measuring the FSB clock is difficult, unless you know the specific number of wait states used. The stanadard approach to identify the bus speed is to measure the CPU clock and take the multiplier from DX-after-reset. Then you just divide the CPU clock by the multiplier and obtain the bus speed and use that for auto-configuration.

jakethompson1 wrote on 2023-04-13, 02:16:

mkarcher wrote on 2023-04-12, 19:30:

So, the whole 5x86 name is just marketing.

The Cyrix 5x86 came first so was this about AMD having a "586-class" product for PCI 486 budget PCs, with the K5 either late or not competitive?

That's very likely. Selling an "enhanced DX4" is difficult one competitior sells an 5x86 for the same kind of boards, and the other competitor tells the customers that 486 is dead and Pentium is the way to got. And in my experience, the Cyrix 5x86-100 with jump prediction enabled is about as fast as the Am5x86 at 133 MHz ("P75"), so just from a performance point of view, the 5x86 moniker isn't even that misleading.

mkarcher wrote on 2023-04-13, 05:38:

That's very likely. Selling an "enhanced DX4" is difficult one competitior sells an 5x86 for the same kind of boards, and the other competitor tells the customers that 486 is dead and Pentium is the way to got. And in my experience, the Cyrix 5x86-100 with jump prediction enabled is about as fast as the Am5x86 at 133 MHz ("P75"), so just from a performance point of view, the 5x86 moniker isn't even that misleading.

Imagine, a Cyrix 5x86 with all CPU features active at the same time. But in real it is more like a downsized 6x86... in beta phase.

The thing with the detection of the enhanced Am486DX4/Am5x86 is that the relevant jumpers are interpreted right after reset and remain until next reset.

red-ray wrote on 2023-04-03, 18:42:

Ah, that makes sense as for it being a 486, that's easy , SYSTEM_INFO->wProcessorLevel is 3 for a 386, 4 for 486, 5 for Pentium […]
Show full quote

jakethompson1 wrote on 2023-04-03, 17:16:

FFD0 means: Model 3, Stepping 0. This does not even indicate the presence of a 486 processor at all.
Wouldn't model 3 conform with DX=0430h at reset for a 486DX2?

Ah, that makes sense as for it being a 486, that's easy , SYSTEM_INFO->wProcessorLevel is 3 for a 386, 4 for 486, 5 for Pentium and 6 for PentiumPro and later, 31 for Itanium 2.

There is also SYSTEM_INFO->dwProcessorType which can be such as PROCESSOR_INTEL_386, PROCESSOR_INTEL_486 , PROCESSOR_INTEL_PENTIUM, PROCESSOR_INTEL_IA64 , PROCESSOR_AMD_X8664, PROCESSOR_ALPHA_21064, ...

Things change when the CPU has CPUID, my DX2 has Type 486 Level 4 Revision 0306, my DX4 Type 486 Level 4 Revision 0800, my i9-9980XE Type 8664 Level 6 Revision 5504, my Itanium 2 Type 2200 Level 31 Revision 0105, my Alpha 21164A Type 21064 Level 21164 Revision 0002 all of which match what the native CPUID returns.

I just ordered a 486 DX-33 so will see what that reports later in the week. It would be good if we knew what was returned for a new other 486/386 CPUs that do not have CPUID.

So I am happy we can get the model, but suspect AMD vs. Intel will be tricky to impossible.

As we were talking about NT CPU identification a while back, this talks about how NT distinguishes pre-CPUID 486 CPUs: https://www.geoffchappell.com/studies/windows … pu/precpuid.htm

jakethompson1 wrote on 2023-06-02, 02:10:

red-ray wrote on 2023-04-03, 18:42:

Ah, that makes sense as for it being a 486, that's easy , SYSTEM_INFO->wProcessorLevel is 3 for a 386, 4 for 486, 5 for Pentium […]
Show full quote

jakethompson1 wrote on 2023-04-03, 17:16:

FFD0 means: Model 3, Stepping 0. This does not even indicate the presence of a 486 processor at all.
Wouldn't model 3 conform with DX=0430h at reset for a 486DX2?

Ah, that makes sense as for it being a 486, that's easy , SYSTEM_INFO->wProcessorLevel is 3 for a 386, 4 for 486, 5 for Pentium and 6 for PentiumPro and later, 31 for Itanium 2.

There is also SYSTEM_INFO->dwProcessorType which can be such as PROCESSOR_INTEL_386, PROCESSOR_INTEL_486 , PROCESSOR_INTEL_PENTIUM, PROCESSOR_INTEL_IA64 , PROCESSOR_AMD_X8664, PROCESSOR_ALPHA_21064, ...

Things change when the CPU has CPUID, my DX2 has Type 486 Level 4 Revision 0306, my DX4 Type 486 Level 4 Revision 0800, my i9-9980XE Type 8664 Level 6 Revision 5504, my Itanium 2 Type 2200 Level 31 Revision 0105, my Alpha 21164A Type 21064 Level 21164 Revision 0002 all of which match what the native CPUID returns.

I just ordered a 486 DX-33 so will see what that reports later in the week. It would be good if we knew what was returned for a new other 486/386 CPUs that do not have CPUID.

So I am happy we can get the model, but suspect AMD vs. Intel will be tricky to impossible.

As we were talking about NT CPU identification a while back, this talks about how NT distinguishes pre-CPUID 486 CPUs: https://www.geoffchappell.com/studies/windows … pu/precpuid.htm

That's a useful link, thank you. I'll refer to that when I get round to working on my code again. I'll probably never be adding code to identify most modern CPUs, as that was never my intention, but I do intend to add code to detect different 808x versions, and possibly different 286 vendors (such as Harris, for example). Being able to determine what 386 or early 486 is present without a reset will be handy. My code was originally intended as the basis for being able to optimise for whatever CPU is present. I still need to look up how to detect what FPU is present, never got round to that 😀