Generally it is better to report no data than wrong data. If you can't read it back then it is just PS/2 SIMMs with FSB speed.

As for the COMCTL32 error, maybe in the release version just disable tooltips if the DLL does not support it.

As for the Alpha software, I have to see if I can obtain some.

Actually I checked also my NexGen system and my BIOS has support for CPUID:

This indicates some other bug as well.

Though for SIV it would be nice to have generic support for NexGen, not only for systems with BIOSses that have CPUID support.

elianda wrote:

As for the COMCTL32 error, maybe in the release version just disable tooltips if the DLL does not support it.

I did some checking and what you need is IE 5.00.2919.800 which I think comes as standard with W2K RC1, RC2 has 5.00.2919.3800 which is also OK.

DRAM-21 should now say DRAM everywhere and I have adjusted the tooptip, have all the EDOs gone? Please may I have new save files.

At the moment I am getting SIV to report ISA GPUs, thus far it's working on W2K + NT 4.00, do you have a system you could easily try it on?

elianda wrote:

Actually I checked also my NexGen system and my BIOS has support for CPUID.

The issue is mainly that the NexGen does not say it has CPUID so SIV does not use CPUID. The standard code to detect is as below so why does this fail on NexGen?

At the moment all my CPUs have CPUID and it would be much easier if I had one what did not. I am half inclined to get one, but feel they are overpriced, what should a NexGen system with PCI slots cost and where would I get one?

Talking of systems I would like reminds me I would like a Samsung UP2000, see https://mattst88.com/blog/2008/12/23/UP1500%2 … _CPUs_For_Sale/

1ULONG cpu_has_cpuid( ULONG mode )               // decide if CPUID should be used
2{        ULONG   val;                           // work value
3
4  __asm  pushfd
5  __asm  pop     ecx
6  __asm  mov     eax,ecx
7  __asm  xor     eax,0x00200000
8  __asm  push    eax
9  __asm  popfd
10  __asm  pushfd
11  __asm  pop     eax
12  __asm  xor     eax,ecx
13  __asm  mov     val,eax
14
15  return val;                                               // return if CPUID should be used
16}

The code you quoted for detecting CPUID support is valid only for 486+ CPUs. Though I can have a look later. I expect that it runs on the NexGen as on any other 386.

Should I run chkcpu also?

red-ray wrote:

I am half inclined to get one, but feel they are overpriced, what should a NexGen system with PCI slots cost and where would I get one?

Luck, patience, and timing have more to do with it than a price. A year or two ago, I think I saw a PCI NexGen sell for around $500 on CPU-World. They don't pop up very often, but if one does, don't hesitate or *pooof*

red-ray wrote:

Talking of systems I would like reminds me I would like a Samsung UP2000, see https://mattst88.com/blog/2008/12/23/UP1500%2 … _CPUs_For_Sale/

Intrigued, but I didn't see a price (or a photo)...

feipoa wrote:

Should I run chkcpu also?

No, please try Beta-21. I removed -NEXGEN and made some other adjustments and with luck SIV32O 5.41 Beta-21 will just work. Please post new save files from NT 3.51 NexGen system.

elianda wrote:

I expect that it runs on the NexGen as on any other 386.

I decided to __try again, please try SIV32L 5.41 Beta-21 on your W95B NexGen and SIV32A on your Alpha and post new save files

Have a look @ http://unimatrix.gomtuu.net/ADP/api/up2000-doc.html for some manuals and https://www.mikerohard.de/up2000spec.pdf for an image of a UP2000

Beta21 on NexGen:

CPU is detected!

notes:
- the Logo is wrong. Original Logo is https://de.wikipedia.org/wiki/NexGen#/media/D … NexGen_Logo.svg
- L2 cache of 256 kB is not detected, also not by the Cache-0 Latency tool.

elianda wrote:

L2 cache of 256 kB is not detected, also not by the Cache-0 Latency tool.

Thank you and at last! I will change the logo, but it's not as pretty. Is the speed correct? It would be interesting to see what Menu->Help->Conversions->CPU MHz reports.

Does CPUID report Model 15 Stepping 15 for all NexGen CPUs? What should SIV report as the name?

Cache-0 Latency without RDTSC is tricky. The trigger for a size change was a tripling of the time, I will change this to doubling and see what happens.

How do I read the L2 cache size?

The screen may be better if you install the Tahoma font, it's attached to the first post.

First of all I tried on a 386DX-40 with Win95 and could not get it running.
With 16 MB memory behavior is random:
- It freezes the system after a while loading
- It crashes
- It sometimes seems to quit the process without freeing the memory.

Crash example:

It seems to me that due to the limited memory it gets unstable. Have you tested SIV32L with low memory configurations?
(I deleted PCIDEVS.TXT and USBDEVS.TXT as these make no sense on a 386)

Same system with 32 MB freezes the system after a while loading.
See log.

The NexGen:

red-ray wrote:

elianda wrote:

L2 cache of 256 kB is not detected, also not by the Cache-0 Latency tool.

Thank you and at last! I will change the logo, but it's not as pretty. Is the speed correct? It would be interesting to see what Menu->Help->Conversions->CPU MHz reports.

The CPU clock is 84 MHz, so I would say with 83 MHz detected it is correct.
I will report that later.

Does CPUID report Model 15 Stepping 15 for all NexGen CPUs?

If the BIOS of the Nx586 implements the CPUID instruction then most likely yes. The hypercode for the CPUID instruction is in
https://www.memotech.franken.de/cgi-bin/AsmCo … urce/DECODE.ASM at line 2600. It is the same for all CPUID supporting BIOSes.

I can not tell what CPUID on a NexGen Nx686 returns.

What should SIV report as the name?

Nx586-P<rating> where rating depends on the frequency.
The 84 MHz CPU is named Nx586-P90
There exists a P80, P90, P100, P110, P120 and P133, I don't have the full corresponding actual frequency table.
If the Nx586 has an internal FPU then it is a Nx586-PF<rating>. I think starting from P100 there were also PF100.
Supposedly there were also external Nx587 FPUs.

I think feipoa uses a Nx586-PF110, so he knows the actual frequency of the 110 rating.

Cache-0 Latency without RDTSC it's tricky. The trigger for a size change was a tripling of the time, I will change this to doubling and see what happens.

How do I read the L2 cache size?

For the Nx586 L2 cache size can only be 256 kB or 1 MB. I do not know how to read it back, probably it is just a string in the BIOS. I have not seen a board with 1 MB cache. What you can read from the system controller is documented here:
http://bitsavers.org/components/nexgen/NexGen … ary_Mar1994.pdf
Page 101 ff.

At the moment I am getting SIV to report ISA GPUs, thus far it's working on W2K + NT 4.00, do you have a system you could easily try it on?

I could try on the 386, it is currently freezing when starting SIV though. Does that feature requires that the driver for the ISA graphics card is installed?

ALPHA:

elianda wrote:

Does GPU reporting require that the driver for the ISA graphics card is installed?

Thank you and I think the Alpha is now good enough, do you?

For SIV to report any graphics card it needs to be listed on Menu->System->Device IDs. AFAIK this is only possible with W98 and on W95 you get Failed to Load SetupDiGetClassDevsA() from setupapi.dll.

elianda wrote:

First of all I tried on a 386DX-40 with Win95 and could not get it running. With 16 MB memory behavior is random: - It freezes t […]
Show full quote

First of all I tried on a 386DX-40 with Win95 and could not get it running.
With 16 MB memory behavior is random:
- It freezes the system after a while loading
- It crashes
- It sometimes seems to quit the process without freeing the memory.

It seems to me that due to the limited memory it gets unstable. Have you tested SIV32L with low memory configurations?
(I deleted PCIDEVS.TXT and USBDEVS.TXT as these make no sense on a 386)

I think that the freeze happens while W9x is loading SIV so it's impossible to add code to SIV to warn about this, if you do -DBGINI and no file get's created it means SIV never got started.

For Beta-21 I changed SIV so it only uses/loads PCIDEVS.txt if the system has >= 32 MB and the same for SIVRES.dll. It does not cache any of the other files so you don't need to delete them.

SIV runs OK, but rather slowly, only having SIV running. With all 192 MB available then 39.68 MB was used.

How many other programs did you have running? How much memory was free before SIV was started?

First we need to make sure SIV will run on an i386 as I have never tested this. Can you setup MSVC 6 remote debugging to the i386 and install > 32 MB of memory? If so then I can make the .PDB available so we can see exactly what it's doing when it crashes. Is there a DrWatson report of the crash?

elianda wrote:

The CPU clock is 84 MHz, so I would say with 83 MHz detected it is correct. Nx586-P<rating> where rating depends on the frequenc […]
Show full quote

The CPU clock is 84 MHz, so I would say with 83 MHz detected it is correct. Nx586-P<rating> where rating depends on the frequency. The 84 MHz CPU is named Nx586-P90 There exists a P80, P90, P100, P110, P120 and P133, I don't have the full corresponding actual frequency table. If the Nx586 has an internal FPU then it is a Nx586-PF<rating>. I think starting from P100 there were also PF100. Supposedly there were also external Nx587 FPUs. I think feipoa uses a Nx586-PF110, so he knows the actual frequency of the 110 rating.

For the Nx586 L2 cache size can only be 256 kB or 1 MB. I do not know how to read it back, probably it is just a string in the BIOS. I have not seen a board with 1 MB cache. What you can read from the system controller is documented here:
http://bitsavers.org/components/nexgen/NexGen … ary_Mar1994.pdf
Page 101 ff.

Thank you and I have just uploaded Beta-22, does it get your PR90 and how does [Cache-0 Latency] do? It should also report PF if the FPU is present.

I had already checked that datasheet and still can't find anything about the cache size. I am also pondering how to know the NxVL exists on the system before I start reading from it.

I think Alpha is ok.

NexGen:

Cache-0 does not work, tried several times.
One thing I noticed is that it shows for memory DIMM, but should be PS/2 SIMM.

I think you can only be sure to have NxVL chipset if CPU is a NexGen and a VLB bus is present.

The 386:

For 32 MB it generated the linked DBGOUT.LOG and the last thing there was CPUID detection before it froze.
I think the freeze with 16 MB RAM is happening at the same point.

Dr.Watson is AFAIK only running as default in NT, not Win9x.

First we need to make sure SIV will run on an i386 as I have never tested this. Can you setup MSVC 6 remote debugging to the i386 and install > 32 MB of memory? If so then I can make the .PDB available so we can see exactly what it's doing when it crashes. Is there a DrWatson report of the crash?

>32 MB is not an option. I mean 32 MB on a 386 is already a lot, usually they had 4 or 8 MB. This particular system uses already 4x 4MB SIMMs and I am pretty sure the mainboard does not support higher capacity SIMMs.

I might be able to install MSVC6 on the 386. Though do you think that makes sense with 32 MB RAM?
It would be quite an effort though. The system has 2x 504 MB HDDs where the second uses Stacker 4.0 compression, which are also filled.

ISA Graphics cards:
Ok so you say this needs at least Win98?
I could setup that on my testbench (P3 800MHz), but I see your system is also a P2 300MHz. Would not be so much of a difference then.

How many other programs did you have running? How much memory was free before SIV was started?

I did not had any other programs running.

I checked that with Quarterdeck Manifest:
after bootup:
Total Global Heap 25,464K
Free Global Heap 18,090K
Free Physical 7,060K

Manifest itself takes 496 kB.

elianda wrote:

[…]
Show full quote

1. Cache-0 does not work, tried several times.
2. One thing I noticed is that it shows for memory DIMM, but should be PS/2 SIMM.
3. I think you can only be sure to have NxVL chipset if CPU is a NexGen and a VLB bus is present.
4. For 32 MB it generated the linked DBGOUT.LOG and the last thing there was CPUID detection before it froze. I think the freeze with 16 MB RAM is happening at the same point.
5. I might be able to install MSVC6 on the 386. Though do you think that makes sense with 32 MB RAM? It would be quite an effort though. The system has 2x 504 MB HDDs where the second uses Stacker 4.0 compression, which are also filled.
6. ISA Graphics cards: Ok so you say this needs at least Win98? I could setup that on my testbench (P3 800MHz), but I see your system is also a P2 300MHz. Would not be so much of a difference then.

1. Thank you for your testing and my thoughts are:
2. It came up with 192 rather than 256 KB which makes sense given the times, what happens if you select [_]Repeat and let it run (it may be tricky to deselect!).
3. I fixed it an hour ago, it will be in Beta-23
4. But how do I know it's VLB? I can easily do no PCI + NexGen CPU.
5. OK, so it's not the memory size. There are about 481,637 lines of code in SIV so without a debugger it will take ages to find.
6. No, Install MSVC 6 on a recent system, I have it running on my i9 9980XE which runs Windows 10 x64 Enterprise V10.00 Build 18362 19H1, see https://www.codeproject.com/Articles/1191047/ … udio-on-Windows. You just need the attached remote server on the target system, see if it runs, it may need the .DLLs from NT4.
7. Yes, >= W98 or >= WNT 3.51. I suspected you will have different ISA GPUs to me, I only have a Kelvin 64 ISA.

[/list]

Did one more run with the 486DLC-40.
Not exactly right...

GigAHerZ wrote:

Did one more run with the 486DLC-40. Not exactly right...

Thank you, so the Cyrix detect code must need adjusting. Please post the Menu->Hardware->CPUID->CPU-0 screen or ideally the save file from doing Menu->File->Save Local so I can use it to run SIV in test mode.

red-ray wrote:

GigAHerZ wrote:

Did one more run with the 486DLC-40. Not exactly right...

Thank you, so the Cyrix detect code must need adjusting. Please post the Menu->Hardware->CPUID->CPU-0 screen or the save file from doing Menu->File->Save Local.

There you go!