James-F wrote:

Each run is randomly different?

I wondered the same thing, but I think what he meant is that his results are varying from other systems on the list with similar CPUs.

bluejeans-can you clarify? Are your results varying from run to run, or just different from other systems with similar CPU to yours?

clueless1 wrote:

James-F wrote:

Each run is randomly different?

I wondered the same thing, but I think what he meant is that his results are varying from other systems on the list with similar CPUs.

bluejeans-can you clarify? Are your results varying from run to run, or just different from other systems with similar CPU to yours?

Each run gives the same realtics. Just different to cpu's on the list that should be similar to mine, with cache disabled. Especially according to topbench.

On a pentium 166 with 16mb ram, internal cache off, this is the result with landmark:

I get 6836 realtiks.

I'm confused by the results on landmark, do I ignore that it's telling me it performs like a xxmhz AT, and just go by the 81mhz cpu clock? Or the 96mhz figure?

Landmark is meant for much older systems. I would use the benchmarks that are on the spreadsheet of this thread: 3DBENCH2, PCPBench, Speedsys, and Doom. Then you can compare to other systems in the spreadsheet with the same benchmarks and their results. 😀

clueless1 wrote:

Landmark is meant for much older systems. I would use the benchmarks that are on the spreadsheet of this thread: 3DBENCH2, PCPBench, Speedsys, and Doom. Then you can compare to other systems in the spreadsheet with the same benchmarks and their results. 😀

Speedsys claims 26.4. Since I only really want to experience doom slowed down with wads it was never really meant to run, I might just test how many realtiks I get for moslo etc set at 10% intervals.

bluejeans wrote:

clueless1 wrote:

Landmark is meant for much older systems. I would use the benchmarks that are on the spreadsheet of this thread: 3DBENCH2, PCPBench, Speedsys, and Doom. Then you can compare to other systems in the spreadsheet with the same benchmarks and their results. 😀

Speedsys claims 26.4. Since I only really want to experience doom slowed down with wads it was never really meant to run, I might just test how many realtiks I get for moslo etc set at 10% intervals.

26.4 is 486DX2-66 territory. But to get the full picture of how your system is performing with L2 disabled, the other benchmark results will also be telling.
Please share your results. I'm curious how moslo works for you.

clueless1 wrote:

bluejeans wrote:

clueless1 wrote:

Landmark is meant for much older systems. I would use the benchmarks that are on the spreadsheet of this thread: 3DBENCH2, PCPBench, Speedsys, and Doom. Then you can compare to other systems in the spreadsheet with the same benchmarks and their results. 😀

Speedsys claims 26.4. Since I only really want to experience doom slowed down with wads it was never really meant to run, I might just test how many realtiks I get for moslo etc set at 10% intervals.

26.4 is 486DX2-66 territory. But to get the full picture of how your system is performing with L2 disabled, the other benchmark results will also be telling.
Please share your results. I'm curious how moslo works for you.

MTRF: failed
3dbench:250
pcp: 7.9
Quake: failed
Chkcpu: 164.5

You've got other issues that should be addressed first if benchmarks are failing. Can you be more specific about how Quake is failing? What's MTRF?

clueless1 wrote:

You've got other issues that should be addressed first if benchmarks are failing. Can you be more specific about how Quake is failing? What's MTRF?

Oh, MTRR.

I actually don't have a working pen in the house, so here are the screens.

MTRRRLFBE should work on your slot 1 cpu. If you're still underclocking, set it back to stock speeds and try again. Same thing for Quake, see if it works without any changes to your system. BTW, are you running these in pure DOS, MS-DOS mode, or from a Win98 command prompt? You should be running in either pure DOS or MS-DOS mode (if using Win9x).

clueless1 wrote:

MTRRRLFBE should work on your slot 1 cpu. If you're still underclocking, set it back to stock speeds and try again. Same thing for Quake, see if it works without any changes to your system. BTW, are you running these in pure DOS, MS-DOS mode, or from a Win98 command prompt? You should be running in either pure DOS or MS-DOS mode (if using Win9x).

These results are on a pentium 166. Running that unofficial 7.1 version of dos, though apparently it's in protected mode, not real mode.

MTRFFLBE (or FASTVID) does not work on a Pentium 1. I don't have experience with 7.1, so can't help there. I use 6.22 for DOS benchmarking.

clueless1 wrote:

MTRFFLBE (or FASTVID) does not work on a Pentium 1. I don't have experience with 7.1, so can't help there. I use 6.22 for DOS benchmarking.

Can I simply downgrade to 6.22 by overwriting all system files? I don't want to reinstall everything in config.sys and autoexec.bat.

bluejeans wrote:

clueless1 wrote:

MTRFFLBE (or FASTVID) does not work on a Pentium 1. I don't have experience with 7.1, so can't help there. I use 6.22 for DOS benchmarking.

Can I simply downgrade to 6.22 by overwriting all system files? I don't want to reinstall everything in config.sys and autoexec.bat.

I'm pretty sure that is not a good thing to try. 😀 Do you have a spare hard drive? Just install 6.22 from scratch on it, then copy over your autoexec, config and programs (you can temporarily connect both to a Windows machine for the file transfers).

clueless1 wrote:

bluejeans wrote:

clueless1 wrote:

MTRFFLBE (or FASTVID) does not work on a Pentium 1. I don't have experience with 7.1, so can't help there. I use 6.22 for DOS benchmarking.

Can I simply downgrade to 6.22 by overwriting all system files? I don't want to reinstall everything in config.sys and autoexec.bat.

I'm pretty sure that is not a good thing to try. 😀 Do you have a spare hard drive? Just install 6.22 from scratch on it, then copy over your autoexec, config and programs (you can temporarily connect both to a Windows machine for the file transfers).

How about running 6.22 from a boot cd and running the benchmark from the hard drive?

I've never tried that, not sure it will work. But MS-DOS 6 is hardware-agnostic. I have a spare hdd with DOS and games, apps, benchmarks installed that I move from machine to machine to run benchmarks. Once you set it up on a device that you can boot from (compact flash, hdd, etc) then its easy to use it on different systems. You need to figure out how that will work for you.

clueless1 wrote:

I've never tried that, not sure it will work. But MS-DOS 6 is hardware-agnostic. I have a spare hdd with DOS and games, apps, benchmarks installed that I move from machine to machine to run benchmarks. Once you set it up on a device that you can boot from (compact flash, hdd, etc) then its easy to use it on different systems. You need to figure out how that will work for you.

Dos 6.22 is showing almost identical results, only I still couldn't get Quake to work. This time it didn't show an error at all, I just got kicked back to the menu.

I would consider starting a new thread on the Quake issue. Not everyone is going to read this because of the thread title.

A couple of things to check:
what does your autoexec/config look like? Are you running EMM386 or Extended memory? What is your free conventional? What other components are installed in your system. In general, when trying to troubleshoot stuff like this, strip your system down to the bare essentials. Use barebones autoexec/config, remove sound cards, set BIOS defaults, disable unused serial/parallel ports, use a single stick of RAM, etc. Eliminate variables. If the system starts working as expected, add back one variable at a time and retest til you find what was causing the problem.

To run quake just upgrade cwsdpmi.exe to a new version, download csdpmi5b.zip.

I'm starting to love the "generic" W6BXA Slot 1 440BX motherboard I have around here, and it got me thinking about how flexible a Pentium III 450 might be for older games on this board. Its really sad that you can't even underclock these things, so that's a major bummer, but I do have options for disabling internal and external cache. From what I understand, internal is L1 and external is L2.

I just did some very quick and dirty tests to see what effect it would have and it seems like disabling L1 cache on a P3 obliterates its performance more than it does on a K6-2... is that normal? Going by the spreadsheet, a K6-2 500 drops from ~118 to ~11 when disabling L1 cache. By disabling L1 (whether I disable external or not) on my PIII 450 however, I go from ~129 in PCP to 1.7!? The CPU is a wonderful overclocker (and I've added cooling, just because...) so I bumped it up to 124Mhz FSB just for kicks, giving me a 558Mhz P3... and it only got me up to 2.0 in PCP with L1 disabled. Enabling fastvid (if I did it right) gave me 2.1... which is within the margin of error, so it didn't seem to do anything.

I was also using a Velocity 4400 TNT 16MB AGP.

Is this normal, or have I possibly missed something? I have a really nice FIC-PA2013 v2.1 + K6-2 500 system here that is quite flexible when it comes to adjusting clock speed and caches, I just wondered if there was any hope for this 440BX, since it has a lot going for it. It has AGP, 4 PCI and 3 ISA, SB Pro header and options for using DIP switches or the BIOS for clock adjustment (the BIOS limits me to 100,112,124 and 133 though sadly... would have liked to see 66Mhz too). On top of all that, a fairly flexible ATA33 controller (the one on the PA2013 is useless for any newer drives) and seemingly decent USB performance. And... well, it is a Slot 1 system, and I have a soft spot for them.

Hey there!

Your scores are normal. Check out the Pentium II 333, Celeron 466, and Celeron 333 on the spreadsheet. I've never tested a Pentium III myself, but from what I've heard Phil say, it performs like yours does. It seems like a common behavior in all Pentium II and Pentium III class cpus (including Celerons). My P2-333 is unlocked, so I was able to run it at 333, 300, 266, 233, and 133 Mhz. It's PCPBench scores with L1 disabled ranges from 1.2 to 0.7. So I'd say your scores are consistent. I'd guess you'd need somewhere around 800-933Mhz in order to get L1 disabled scores in the fast 386/slow 486 range.

Regarding your video card, it doesn't matter. When you disable cpu cache, the cpu becomes the bottleneck. You could have have a slow Trident in your system and it would give the same results as your TNT with cpu cache disabled.

You are welcome to fully bench your cpu and add the results to the spreadsheet. Might help someone else down the road with the same questions.

Cheers!