For benchmarking, does anyone have results from the following tools?:

• Norton SysInfo
• Landmark Speed Test

If you have any benchmark results from these tools, please share them.

Modern CPUs are just more ineffective in 16bit real mode, but not slower than a Tulatin or a Athlon XP or even Core2, that I concider the last "classic" cpu. After that they started to remove legacy functionalities.

That's what I think, too, but it would be great to see benchmarks confirming this.

Also, how do you know what the fastest MS-DOS 16bit real mode CPU in the world is? Is it simply the fastest single-threaded modern x86 or are there differences between manufacturers? How to the fastest optimized 16 bit real mode CPUs compare etc…

Under virtualization or bare metal? I don’t know if there is a difference, just asking. My 8th gen Intel chip is plenty fast under virtualization, and I have never tormented it, nor do I plan to, with real real mode.

Wasn't the Netburst (Pentium 4) architecture less efficient than Pentium III without special compiler optimizations? Wasn't that the reason why they actually went back to the Pentium III architecture (or a variant of it) in the form of Pentium M (Banias/Dothan) after the P4 couldn't scale in performance, particularly in an energy-efficient way?

I guess it would be interesting to compare real mode (and 32-bit protected mode) DOS benchmarks between those architectures, as well as later ones, to see how legacy mode peformance developed.

But measurements should be operation per megabyte also. Not just raw power.

When I compare CPUs for DOS these days I run PI by @Rayer and just check the time elapsed.
Would be curious to see what you folks get in terms of time of computation on those fabled Athlons of old.
One other benchmark I could think of is using Doom (e.g. MBF, I think, has sufficient capabilities for that), but personally I don't even go there.

Well in terms of a chipset that can natively run Win2k and perhaps NT,98, etc., and may be able to boot a dos disk (can't swear to it though), consider a socket 771 motherboard. At least 1 has a floppy controller on board, Supermicro.

This might lead someone to ask whether socket 775 are equally capable and as fast. 771 and 775 mobos both can have dual cpu's. Not that it would benefit real mode performance (don't think anyway).

Appreciate not exactly what the OP has asked for, but here are couple systems I run PI against today.
Don't have the OP mentioned benchmarking tools though...

This PI calculator runs in 32-bit protected mode

Apologies, have overlooked this; and from description and trying it in DOSBox, it's compiled for later CPUs, so not quite a universal comparison tool, so my bad.

For testing I can recommend earlier AIDA16 versions that still have the 'Benhcmark' menu option with MIPS16 and MIPS32 sections. AIDA16 also gives a 'Pentium rating' at each banchmark section so you can compare microarchitectures in case of both 16-bit and 32-bit code.
From the Pentium Pro/PII the trend is that all x86/x64 microarchitectures are much faster in 32-bit code than in 16-bit code (compared to a baseline P5 100 MHz CPU).
AFAIR the K6 2/3 line from AMD was the last x86 microarchitecture where 16-bit code was nearly as fast as 32-bit code. But of course this does not mean that a K6-2 at 400 MHz would be faster in case of 16-bit code than a Core I7 at 3Ghz.
This just means that e.g. an Ivy Bridge Core I7 at 1 GHz performs as a hypothetical P5 500 MHz in case of 16-bit code but performs as a hypothetical P5 3 Ghz in case of 32-bit code (while a K6-2 400 MHz performs as a hypothetical P5 500 MHz in case of 16-bit code and as a P5 550 MHz in case of 32-bit code).

ludicrous_peridot wrote on 2024-07-27, 12:47:

When I compare CPUs for DOS these days I run PI by @Rayer and just check the time elapsed.
Would be curious to see what you folks get in terms of time of computation on those fabled Athlons of old.
One other benchmark I could think of is using Doom (e.g. MBF, I think, has sufficient capabilities for that), but personally I don't even go there.

I can see no program called PI on RayeR's homepage. May you please provide a link or attach that program here?

An excellent DOS program which calculates Pi to large numbers of digits is Carey Bloodworth's PiAGM version 2.4

http://rayer.g6.cz/programm/programe.htm, "PI.EXE"

edescourtis wrote on 2024-07-24, 13:14:

For benchmarking, does anyone have results from the following tools?: […]
Show full quote

For benchmarking, does anyone have results from the following tools?:

• Norton SysInfo
• Landmark Speed Test

If you have any benchmark results from these tools, please share them.

I compared two systems because they are so close; Core 2 Duo T7600 2.33 GHz Merom vs Core i5 520m 2.4 GHz Arrandale.

Aida16
Core2 1096 - Core i5 1108

RayeR's PI
Core2 24.82 - Core i5 20.31

I was interested in testing this because the Core i5 was going to replace the Core2 in my ultimate 98 system but it scores 200 realtics lower in Shareware DOOM. Is DOOM shareware a 16bit application?

Dothan Burger wrote on 2024-07-29, 22:39:

...Is DOOM shareware a 16bit application?

No, DOOM is a 32-bit protected mode application (the shareware version is not different from this aspect).

Falcosoft wrote on 2024-07-29, 22:53:

Dothan Burger wrote on 2024-07-29, 22:39:

...Is DOOM shareware a 16bit application?

No, DOOM is a 32-bit protected mode application (the shareware version is not different from this aspect).

Thank you! I wasn't sure how to find that out.

Is realmode achieved by restarting into ms-dos mode and running Aida16?

Ringding wrote on 2024-07-29, 05:35:

http://rayer.g6.cz/programm/programe.htm, "PI.EXE"

Thanks!

For comparison, Pi-AGM 2.4 above is able to calculate 2,097,152 digits in about 12 sec on my i5 8250 laptop @ 3.4 Ghz in bare metal DOS of course, with CWSDPMI.

I'm gonna go with the interpreted Basic and compiled basic versions of the whetstone on Roy Longbottoms pages for clues, the efficiency table here, first two columns only of results are 16 bit I think, http://www.roylongbottom.org.uk/whetstone.htm#anchorPC3

Which makes it look like Phenom II might be the one to beat, needing near 6 Ghz of intel core to best it at 3Ghz, and the Phenoms will go to 4Ghz on air given a bit of luck and 6 or 7 ghz with cryogenic liquids. So unless you can get over 6 ghz on air on recent core i-doesn't-matter-it's-single-core, or 12 Ghz or so with LN2 I think it will be the phenom II.