Created 3 reports hope it's helpful. The V30 10Mhz and 486DX2 66Mhz look okay.

For the 386DX 25Mhz it didn't detect the correct FPU. It's an IIT 4C87DLC-40 running at 25Mhz. Not a 287 😜. Also the "Running DOS version" is a bit strange. It's running on DR-DOS 6.0 btw.

Another V20 report, detected correctly! (10MHz)

I'll try to run it on some of my other machines this weekend (386, 486, etc.).

maxtherabbit wrote on 2022-12-04, 20:00:

Seems to be working just fine for me, though I'm not sure why it wrote the same information to the report file 4 times. I only ran report.bat once.

Same system, new version

And here it is on the 5162, dead on accurate

Finally on my 386sx, the speed is bang on but the co-processor is actually a 387sx

Thanks again for all these new reports! 😀
It showed me that a lot of the new logic I put in this updated CHKCPU works as expected. Only the FPU detection on 386 systems is clearly bugged.

When Intel introduced the 80387 FPU in 1986, the 80386 CPU was already a year on the market. During that time, there were motherboards that allowed to install an 80287 FPU next to the 80386, while also providing a 80387 socket for a future upgrade FPU. So in those days a 386 could be paired with a 287 or 387 FPU. I never saw a board like that however, did any of you?

Anyway, as I read about a way to tell the 287/387 apart, I put this logic in CHKCPU.
Thanks to your testing I know now that this logic doesn’t work!

In hindsight, it is unlikely that we ever encounter such a rare 386+287 system after all these years. Most of them will have vanished or have been upgraded with a 387 FPU.
Therefore I decided to delete the 287/387 detection on 386 systems and consider the 386+287 indication obsolete. The next CHKCPU version will then always show a 387 when an FPU is detected on a 386 system...

As I haven’t found out yet how to identify other FPUs, like IIT 3C87, 4C87 and Cyrix 83D87, those will also be shown as 387 for the time being. 😉

I will report back when the next CHKCPU beta is ready.
Jan

Suggest a on screen notification that actual fpu may vary

Chkcpu wrote on 2022-12-10, 17:59:

Thanks again for all these new reports! :) It showed me that a lot of the new logic I put in this updated CHKCPU works as expect […]
Show full quote

Thanks again for all these new reports! 😀
It showed me that a lot of the new logic I put in this updated CHKCPU works as expected. Only the FPU detection on 386 systems is clearly bugged.

When Intel introduced the 80387 FPU in 1986, the 80386 CPU was already a year on the market. During that time, there were motherboards that allowed to install an 80287 FPU next to the 80386, while also providing a 80387 socket for a future upgrade FPU. So in those days a 386 could be paired with a 287 or 387 FPU. I never saw a board like that however, did any of you?

Anyway, as I read about a way to tell the 287/387 apart, I put this logic in CHKCPU.
Thanks to your testing I know now that this logic doesn’t work!

In hindsight, it is unlikely that we ever encounter such a rare 386+287 system after all these years. Most of them will have vanished or have been upgraded with a 387 FPU.
Therefore I decided to delete the 287/387 detection on 386 systems and consider the 386+287 indication obsolete. The next CHKCPU version will then always show a 387 when an FPU is detected on a 386 system...

As I haven’t found out yet how to identify other FPUs, like IIT 3C87, 4C87 and Cyrix 83D87, those will also be shown as 387 for the time being. 😉

I will report back when the next CHKCPU beta is ready.
Jan

There definitely is a way to tell them apart. Norton System Information, NSSI, and PC Config are all able to properly distinguish between the 287 and 287xl(387) on my various 286 systems.

The author of PC Config claims to have source available for his tool, but the link seems broken ):
https://holin.com/indexe.html

EDIT: found it on archive.org

Chkcpu wrote on 2022-12-10, 17:59:

Thanks again for all these new reports! :) It showed me that a lot of the new logic I put in this updated CHKCPU works as expect […]
Show full quote

Thanks again for all these new reports! 😀
It showed me that a lot of the new logic I put in this updated CHKCPU works as expected. Only the FPU detection on 386 systems is clearly bugged.

When Intel introduced the 80387 FPU in 1986, the 80386 CPU was already a year on the market. During that time, there were motherboards that allowed to install an 80287 FPU next to the 80386, while also providing a 80387 socket for a future upgrade FPU. So in those days a 386 could be paired with a 287 or 387 FPU. I never saw a board like that however, did any of you?

Anyway, as I read about a way to tell the 287/387 apart, I put this logic in CHKCPU.
Thanks to your testing I know now that this logic doesn’t work!

In hindsight, it is unlikely that we ever encounter such a rare 386+287 system after all these years. Most of them will have vanished or have been upgraded with a 387 FPU.
Therefore I decided to delete the 287/387 detection on 386 systems and consider the 386+287 indication obsolete. The next CHKCPU version will then always show a 387 when an FPU is detected on a 386 system...

As I haven’t found out yet how to identify other FPUs, like IIT 3C87, 4C87 and Cyrix 83D87, those will also be shown as 387 for the time being. 😉

I will report back when the next CHKCPU beta is ready.
Jan

I have an old ATi 386 card, unfortunately it's quite complicated to fix, both because I don't have a service manual, and because there are some corroded tracks, apart from that it had on an i386DX 20 (or 25?) and an i387, but in the card is the socket for the 287.
I put the 387 on an AMD 386DX 40MHz, and I think it runs at 40MHz (even if at most it should go to 33), to try your program on that PC I have to wait a while (same for the DLC), but in the next few days I could try it on another AMD 386DX40, and on an AMD 386SX 33, I have these at hand and are still under construction, unfortunately they don't have a coprocessor, but if I wanted I could start an emulator and see if the program detect something.

286 looks good now.

mwdmeyer wrote on 2022-12-12, 10:06:

286 looks good now.

Great! Thanks for reporting back with latest CHKCPU beta. 😀

Jan

PC@LIVE wrote on 2022-12-11, 14:08:

Chkcpu wrote on 2022-12-10, 17:59:

Thanks again for all these new reports! :) It showed me that a lot of the new logic I put in this updated CHKCPU works as expect […]
Show full quote

Thanks again for all these new reports! 😀
It showed me that a lot of the new logic I put in this updated CHKCPU works as expected. Only the FPU detection on 386 systems is clearly bugged.

When Intel introduced the 80387 FPU in 1986, the 80386 CPU was already a year on the market. During that time, there were motherboards that allowed to install an 80287 FPU next to the 80386, while also providing a 80387 socket for a future upgrade FPU. So in those days a 386 could be paired with a 287 or 387 FPU. I never saw a board like that however, did any of you?

Anyway, as I read about a way to tell the 287/387 apart, I put this logic in CHKCPU.
Thanks to your testing I know now that this logic doesn’t work!

In hindsight, it is unlikely that we ever encounter such a rare 386+287 system after all these years. Most of them will have vanished or have been upgraded with a 387 FPU.
Therefore I decided to delete the 287/387 detection on 386 systems and consider the 386+287 indication obsolete. The next CHKCPU version will then always show a 387 when an FPU is detected on a 386 system...

As I haven’t found out yet how to identify other FPUs, like IIT 3C87, 4C87 and Cyrix 83D87, those will also be shown as 387 for the time being. 😉

I will report back when the next CHKCPU beta is ready.
Jan

I have an old ATi 386 card, unfortunately it's quite complicated to fix, both because I don't have a service manual, and because there are some corroded tracks, apart from that it had on an i386DX 20 (or 25?) and an i387, but in the card is the socket for the 287.
I put the 387 on an AMD 386DX 40MHz, and I think it runs at 40MHz (even if at most it should go to 33), to try your program on that PC I have to wait a while (same for the DLC), but in the next few days I could try it on another AMD 386DX40, and on an AMD 386SX 33, I have these at hand and are still under construction, unfortunately they don't have a coprocessor, but if I wanted I could start an emulator and see if the program detect something.

PC@LIVE, Thanks for the time you put in to test CHKCPU on your 386 systems. 😀
Amazing you have that rare 386 board with both 287 and 387 sockets! I hope you can get it fixed.

All CHKCPU beta versions have already been tested on various emulated machines in 86Box before I put it up here for testing. So thanks for offering emulator help, but that isn't necessary. 😉

Jan

maxtherabbit wrote on 2022-12-10, 18:39:

There definitely is a way to tell them apart. Norton System Information, NSSI, and PC Config are all able to properly distinguis […]
Show full quote

Chkcpu wrote on 2022-12-10, 17:59:

Thanks again for all these new reports! :) It showed me that a lot of the new logic I put in this updated CHKCPU works as expect […]
Show full quote

Thanks again for all these new reports! 😀
It showed me that a lot of the new logic I put in this updated CHKCPU works as expected. Only the FPU detection on 386 systems is clearly bugged.

When Intel introduced the 80387 FPU in 1986, the 80386 CPU was already a year on the market. During that time, there were motherboards that allowed to install an 80287 FPU next to the 80386, while also providing a 80387 socket for a future upgrade FPU. So in those days a 386 could be paired with a 287 or 387 FPU. I never saw a board like that however, did any of you?

Anyway, as I read about a way to tell the 287/387 apart, I put this logic in CHKCPU.
Thanks to your testing I know now that this logic doesn’t work!

In hindsight, it is unlikely that we ever encounter such a rare 386+287 system after all these years. Most of them will have vanished or have been upgraded with a 387 FPU.
Therefore I decided to delete the 287/387 detection on 386 systems and consider the 386+287 indication obsolete. The next CHKCPU version will then always show a 387 when an FPU is detected on a 386 system...

As I haven’t found out yet how to identify other FPUs, like IIT 3C87, 4C87 and Cyrix 83D87, those will also be shown as 387 for the time being. 😉

I will report back when the next CHKCPU beta is ready.
Jan

There definitely is a way to tell them apart. Norton System Information, NSSI, and PC Config are all able to properly distinguish between the 287 and 287xl(387) on my various 286 systems.

The author of PC Config claims to have source available for his tool, but the link seems broken ):
https://holin.com/indexe.html

EDIT: found it on archive.org

@maxtherabbit, Thanks for the PC Config program and source!
Together with the COMPTEST tool and source that kixs sent earlier, I have now some solid information on how this FPU detection can be done.

It bothered me that my code always detected a 387 as a 287 and seeing that PC@LIVE has that rare 386 board with 287 and 387 sockets, it stimulated me to find a fix.
So last night, I started a thorough debugging session (again) to discover why my CHKCPU logic won’t detect a 387. To make a long story short, I found the bug! 😀
It was a logic error, easily overlooked because the your brain is telling you this logic should work, and the assembler doesn’t generate an error when building the program.

The first tests in emulated 386 machines in 86Box were successful and the fixed logic will be part of the next CHKCPU v1.26.19 build. I expect to have this version ready tomorrow, after I’ve added FPU detection for 386 upgrade CPUs like Cx486SLC/DLC and TI486SXLC/SXL.
This was still missing in v1.26.18 beta. 🙁

Jan

Here's an Intel 8088 at 10MHz, measured as 10.8

Chkcpu wrote on 2022-12-03, 17:12:

Hello, fellow retro computer enthousiasts,

I’m Jan and you may know me as the guy who patched all those socket 7 BIOSes for K6-2+/III+ support.

Mad respect and hats off to you! I run your patched BIOS on my TX97-XE and it's fantastic.

Chkcpu wrote on 2022-12-03, 17:12:

...I’m also the author of CHKCPU, a small DOS tool for CPU Identification. I’m planning to release the 25th anniversary update of CHKCPU within a few weeks.
Now that the update is nearing completion, I’m inviting you all to try it out.

Sadly my oldest is a 486, so I can't help testing with the 16-bitters.

Still, I wanted to say thanks and in doing so push the thread up so the largest amount of potential testers will see it.

@leonardo, Thanks for the nice words. This Retro computing is a great hobby but takes a lot of my spare time and it’s nice to hear that the work is appreciated. 😀

@konc, you were the first to report how CHKCPU worked on both the V20 and 8088, so a big thanks for your assistance!
I will correct the slightly high speed reading on the 8088.

The next CHKCPU v1.26.19 beta is ready now and can be downloaded from the first message of this thread. Please use this new version for further testing.
I’ve addressed the following 4 issues:
- Fixed the incorrect 287 FPU detection on a 386 + 387 system
- Fixed the slightly high speed reading on the 8086/8088
- Fixed the slightly low speed reading on Cyrix design 486 CPUs
- Added an FPU indication for Cx/TI 486SLC/DLC/SXLC/SXL systems with a coprocessor.

Although correct from a software standpoint, CHKCPU’s indication of a separate FPU is a generic 8087, 80187, 80286, or 387 at the moment.
I’m planning a more precise FPU detection in the future, but this will take me a while to figure-out.

Cheers, Jan

Jan,
do you check for speed differences between DOS clock and RTC clock (if present)?

Disruptor wrote on 2022-12-12, 23:48:

Jan,
do you check for speed differences between DOS clock and RTC clock (if present)?

I'm pretty sure DOS uses the BIOS / RTC for its time.

debs3759 wrote on 2022-12-12, 23:57:

Disruptor wrote on 2022-12-12, 23:48:

Jan,
do you check for speed differences between DOS clock and RTC clock (if present)?

I'm pretty sure DOS uses the BIOS / RTC for its time.

No, that's wrong. DOS just initializes its timer by BIOS / RTC. But then its clock is based on the timer interrupts. It is a relict from the past since the first IBM PCs did not habe a RTC at all.
I refer to this topic, where a user got his 25 MHz 386 measured as 8 MHz. The reason was that there was a failure on his mainboard so that the timer ran with a triple harmonic frequency of a about 42 MHz instead of 14,318 MHz. https://www.dosforum.de/viewtopic.php?f=1&t=13687 (in German)
The old diagnostics software CheckIt did compare both clocks.

So the bugfix you alluded to in the previous post regarding the FPU type has not yet been implemented?