Splendid idea! 😀

use MS-DOS 3.3 which is smaller
use himem.sys
use a ramdisk driver and put all benches as self extracting exe on the disk
copy the exe to the ramdisk on bootup and depack it there.
The Win98 bootdisk is a good example how to find the drive letter of the installed ramdisk.

Use choice to
1) Run the benchmarks automatically + generate report files
2) run them manually

Oh and Speedsys does not run on MCA Systems I think.

Sounds good!

What other benchmarks (that scale will) could we add?

Is there a SuperPi for DOS?

edited and deleted...

Try these benches: http://homepage.virgin.net/roy.longbottom/oldones.htm
Especially Dhrystone, Whetstone and LinPack.

Good idea!

I have one question though, is it really necessary to use 2 different versions of the program?
What happens if we use just one for 486 and post-486?

I've been looking for a DOS version of SuperPi forever, but I just think it doesn't exist 🙁

Offering non freeware programs for download is not the ideal way to do things. At least Symantec is still around and might or might not care at some point...

SuperPI for DOS does not much sense.
Do you want to use this as FPU/memory bench?

SuperPI runs in Win 3.x though but it takes ages on an older CPU (days to weeks).
Also for a missing FPU it uses the Win 3.x fallback x87 emulation dll that calculates the commands in integer and is much much slower than a real FPU.
This fallback would not be available in DOS though if you don't install a x87 emulation TSR.

And this is with FPU already. (memory requirements do not allow >256K calc)

Edit: same 256K calc on a 386DX-40 without FPU needs about 30.6h.

elianda wrote:

SuperPI for DOS does not much sense. Do you want to use this as FPU/memory bench? […]
Show full quote

SuperPI for DOS does not much sense.
Do you want to use this as FPU/memory bench?

SuperPI runs in Win 3.x though but it takes ages on an older CPU (days to weeks).
Also for a missing FPU it uses the Win 3.x fallback x87 emulation dll that calculates the commands in integer and is much much slower than a real FPU.
This fallback would not be available in DOS though if you don't install a x87 emulation TSR.

And this is with FPU already. (memory requirements do not allow >256K calc)

Edit: same 256K calc on a 386DX-40 without FPU needs about 30.6h.

The thing is, I'd want a SuperPi that can run from a bootdisk...any bootdisk 😉

I'm not sure if it's possible to create a Win 3.1 + SuperPi that can boot and work from a floppy.
And yes, the basic idea was to have a cpu-only benchmark (and including the fpu if it's actually available).

I agree however that SuperPi isn't ideal for this, but it is the kind of calculation I was looking for.

My basic idea was this:
Suppose you are somekind of hacker, or a scientist, or something, and you need to crack a password or do somekind of calculation or something. How fast would any cpu (+fpu if available) be to complete the task?

Basically I'd want a benchmark/program that can emulate this (or something, sorry for sounding a little vague here), a cpu being used for somekind of general (maybe even non-optimized) calculation.

The reason why I'd want something bootable is that you can very quickly set up a motherboard+CPU+RAM+Video and be on your way 😁
Also I'd want a single program that is able to run on all these different CPU's

Edit:and indeed, on old hardware, SuperPi can take a VEERY long time to complete. But at least it's result is somewhat realistic and I think also very scalable.

I believe Superpi was made for windows only. I recall having around here a PI calculation for DOS. Maybe we could use that for bench. I'll look for it.

EDIT: Found it. It takes 83KB on floppy for 4 files (including batch file to run it).

Dominus wrote:

Offering non freeware programs for download is not the ideal way to do things. At least Symantec is still around and might or might not care at some point...

Good point...

Tetrium wrote:

I have one question though, is it really necessary to use 2 different versions of the program?
What happens if we use just one for 486 and post-486?

Yes, version 1.0 stops at 99 frames. And version 1.0c is unreliable on slow machines (you get a much higher score).

A 486 won't crack 99 fps, but a Pentium does. We applied the same for the 3dbench database and it worked quite well.

elianda wrote:

Try these benches: http://homepage.virgin.net/roy.longbottom/oldones.htm
Especially Dhrystone, Whetstone and LinPack.

Ok let's pick one test:

Which one? Should we go Non-optimised?

Dhry1ND.exe - Dhrystone 1 Benchmark Non-Optimised Dhry1OD.exe - Dhrystone 1 Benchmark Optimised Dhry2ND.exe - Dhrystone […]
Show full quote

Dhry1ND.exe - Dhrystone 1 Benchmark Non-Optimised
Dhry1OD.exe - Dhrystone 1 Benchmark Optimised
Dhry2ND.exe - Dhrystone 2 Benchmark Non-Optimised
Dhry2OD.exe - Dhrystone 2 Benchmark Optimised
LinpCND.exe - Linpack Benchmark Non-Optimised
LinpCOD.exe - Linpack Benchmark Optimised
LiveCND.exe - Livermore Loops Benchmark Non-Optimised
LiveCOD.exe - Livermore Loops Benchmark Optimised
SystemD.exe - Input file for description of system
WhetCND.exe - Whetstone Benchmark Non-Optimised Single Precision
WhetCOD.exe - Whetstone Benchmark Optimised Single Precision
WhetDCD.exe - Whetstone Benchmark Optimised Double Precision

Use the optimized. Software is usually also compiled with -O2. (optimization level switch for gcc example)
So Dhry1OD, Dhry2OD, LinpCOD, WhetDCD.
I think this would be a good estimation of the CPU speed. Especially think of 386 with different FPUs.

Mau1wurf1977 wrote:

Yes, version 1.0 stops at 99 frames. And version 1.0c is unreliable on slow machines (you get a much higher score).

But wouldn't that mean that 486's are simply faster then what the old version benches to? Or are the results like "486 33Mhz benches as good as Pentium 1 100Mhz" much higher?

I mean, if a 486 can bench higher using the new version, doesn't that still mean the 486 simply can run this fast??

sprcorreia wrote:

I believe Superpi was made for windows only. I recall having around here a PI calculation for DOS. Maybe we could use that for bench. I'll look for it.

EDIT: Found it. It takes 83KB on floppy for 4 files (including batch file to run it).

Could you upload it please? Or at least tell us how to find it?

Tetrium wrote:

Could you upload it please? Or at least tell us how to find it?

Here it is. Both original files, cwsdpmi for DPMI server, and pi.bat set to calculate 100000 digits.

There is only one bad thing... The internal timer keeps running while displaying the result. A different graphic card may change the total value...

How about Landmark Speed Test? I like it, it scales well.. There's also BYTECPU - scales really well, works on brand new 64-bit Windows and old 16-bit DOS. Thing is it's 32-bit, so you need the HX extender to run it on DOS, and it won't work on pre-32-bit boxen. Other than that caviat, it's great.

Also, it's worth noting that you can boot Win3x from a single floppy with some space left over - not a full copy, but enough to let SuperPi run easily I'm sure.

Tetrium wrote:

But wouldn't that mean that 486's are simply faster then what the old version benches to? Or are the results like "486 33Mhz benches as good as Pentium 1 100Mhz" much higher?

I mean, if a 486 can bench higher using the new version, doesn't that still mean the 486 simply can run this fast??

486 shouldn't be much of a difference, but if you run 1.0c on a 386, you get a score that's way to high.

And if you run 1.0 on a Pentium you get a score of 0 or something else that tells you something isn't right 🤣

Now I remember something I had done back in 1994. My first pc was a 286/16 with no hard disk, which my parents bought me in late 1991. After a couple of years, I upgraded it to a 386dx/40 with 4MB ram, but still had no hard disk in the beginning. Then a friend of mine gave me a copy of windows 3.1 in 10 disks. Not setup disks, but a copy of all the files that were installed on his system. And a night that I had nothing to do, I tried this. I started from copying the file win.com on to a disk, and I would run it. Of course it wouldn't start, but every time it mentioned a different file that was missing. I kept copying every file needed, one by one, and in the end, I had all the necessary files that would run win 3.1 from a disk. Of course it was very slow. It took ages to load, so I thought to try something else. Using ramdrive, I made a ramdisk of 1,5MB. I copied all the files of the disk on it, and the result was astonishing. Because of the ram being very much faster than a physical disk, I typed win, I hit enter, and it took about only a second to load. I brought my friend that gave me the copy of windows here to watch it, without having told him what I'd done, and he was amazed when he saw that.
So I guess, it is possible to make a disk that can run windows...

Yes, but:
SuperPI requires Win32s and a RAMdisk would take RAM that would be useful for SuperPI.

Yushatak wrote:

How about Landmark Speed Test? I like it, it scales well..

Sounds good!

Could you upload it here and a few of us try it out?