running DOS games in a 9x VDM gives them access to much improved I/O performance via 32-bit disk/file access and DMA, if it is enabled, so it's quite normal to see improvements in certain games. build games like duke3d are a prime example and can really stutter heavily under DOS when smartdrive is not enabled, even on much newer harddrives and fast systems. it's curious that smartdrive didn't improve things.

you say that you already tried swapping hard drives already and that both machines are identical, but is that true for the BIOS as well? maybe the one on the "well-performing" machine enables DMA while the other doesn't. i would try to run speedsys on both systems to compare hard drive performance.

Jorpho wrote on 2021-03-24, 22:25:

I don't have much experience with it, but my understanding is that Fastvid is specifically intended to compensate for shortcomings in certain revisions of Pentium II processors, and that these problems would otherwise be mitigated by Windows. If the processors in your two devices are subtly different, then that might explain the problem.

AFAIK it's not so much windows itself that controls the MTRRs as the video drivers, and since they just need to be set once, rebooting into DOS will retain the performance boost. and as they were introduced with the pentium pro, really all pentium iis should have them.

I am running speed sys now on both systems, and getting a .txt file report as well. I will update this post when it is finished.

EDIT: I have ran the program twice on my problematic machine (C1XS) and one full test on my functional machine (C1XF). I have attached the reports in this edit, and it seems to me that it seems to be a processor issue with my C1XS. In the first picture is an initial test run on my C1Xs, and look at the MMX writing and moving lines in the bottom right graph. It seemed that each time I did the test that the processor MMX's speeds would change, sometimes being normal and sometimes showing drops as my pictures show.

Also seems odd that the memory bandwidth changes from ~450MB/s to ~210MB/s on the C1XS between the initial and second run. It is worth trying swapping the RAM over, or have I missed something?

i think OP mentioned taking out some memory he put in himself (?), but yeah, making a memtest boot floppy/usb (whatever that thing can boot) and testing the memory on that "bad" machine is a good idea. actually i'm reading the datasheet for the C1XS now and it states max 128 MB of RAM but there seems to be 192 in there?

worth noting that the hard drives actually aren't identical but if the bigger one performs badly too in that C1XS i guess that can be ruled out.

Yeah I can go ahead and do memtest when I get back from work. The machine officially only had a 64mb chip sold but unofficially, I-O data sold a 128mb chip to go in it. Hard to find nowadays.

EDIT: I am still running memtest, but it has done 1 pass already with no errors being found, which indicates it's most likely not a ram issue.

Brute389 wrote on 2021-03-25, 19:20:

Yeah I can go ahead and do memtest when I get back from work. The machine officially only had a 64mb chip sold but unofficially, I-O data sold a 128mb chip to go in it. Hard to find nowadays.

EDIT: I am still running memtest, but it has done 1 pass already with no errors being found, which indicates it's most likely not a ram issue.

EDIT 2: I have let it run for 3 hours and still no errors, which leads me back to potentially a processor or hard drive issue. Will swap drives and perform benchmarks to make sure.

I went ahead and performed speedsys on the problem machine with the swapped hard drive and I noticed an interesting finding with the linear read speed going up and down constantly. I went ahead and performed a second test to make sure it was a constant effect. I ended up canceling the second test partially through when the linear read speed was doing the same thing.

Lastly, I went ahead and did one more speedsys with thr original hard drive as I never did a full one on the problematic machine. On the picture, notice the memory bandwidth at the top, seems weird to me how that number went from 400~ mb in previous tests to in the 50's, but I only noticed it that low on this one test, so it seems to be giving volatile results. Also, it seems me that linear read speed is partially an issue on the original drive as well.

did you ever check the capacitors on that unit? not that i would necessarily think sony would cheap out here, but there were certain types of problematic SMD caps in the 90s.

otherwise, maybe the processor is overheating? are there any thermal readouts in the BIOS? i don't think thermal throttling was quite a thing yet but that's almost what this reminds me of...

maybe try a couple of runs of pcpbench to see whether the scores fluctuate there as well.

I think there is a thermal indicator LED on the front of the Picturebook (according to a YouTube teardown).

The slow downs seem to effect everything. The SpeedSys VESA memory rate drops from ~20MB/s to ~3.5MB/s on that last run. And on some the L1 and L2 cache rates drop as well. I guess that if a slowdown hits when SpeedSys is measuring then it shows in the result. I wondered about thermal stuff as well, but then the question is what's Win98 doing to stop that being a problem?

There's always the possibility that this does come down to the CPU. The cpu-world link that Jorpho gave does list the QB59 as an engineering/qualification sample. Maybe some interrupt triggering that Windows handles but DOS doesn't? Does it matter which version of DOS?

Do you have a USB floppy drive or CD drive? I think the laptop can boot from USB so it might be interesting to try with no hard drive or the extra RAM (although on a second read through I did see you still had the pauses after removing the extra RAM).

Also, I know you've tried booting to a safe-mode command prompt, with the same problem. What happens if you boot to safe-mode Windows and then try running games from there? It's been too long since I played with Win98, so I can't remember what options there are for choosing which drivers to load. Maybe there's a way to figure out what Win98 does that stops the lag.

Oh, linked to Auron's point, what BIOS options are there?

Can you bench them in Windows? Perhaps Windows throttles the system giving your a smoother experience but limiting the max. performance where DOS goes all out and running into trouble.

Perhaps related to real/protected modes? Although Doom runs in protected mode...

looking back, OP already clarified that he turned off throttling in the BIOS. i assume he's been using the device with power adapter connected to make sure it's not throttling for power saving.

unfortunately at least the hard drive results can't be meaningfully compared between windows and DOS for speedsys, as the tool complains about being run under windows, but pcpbench should be good for comparison.

Thanks everyone, I appreciate all the comments. I have a couple things to report.

To first address the bios question, I took a picture of the power section of it and included it here. As you can see, it is very barebones and simple, much like the rest of the bios is too. I have the throttle set to off.

Second to address, I went into windows 98 safe mode, and booted Doom from there, and I did get stuttering. Not as much as in pure DOS but it certainly would happen when I was moving to new areas.

Thirdly, I ran pcpbench. In both pure DOS and in windows 98, and I got stuttering occuring in both instances. My playerbench number in windows 98 the first time was 81, second time it was only 54.7, third time was 56, and fourth time was 59.

The fact I can replicate this in windows 98, and get varying results in the pcpbench, leads me to think this is a throttling issue. Any other ideas?

Interesting that it explicitly notes that the settings are not effective under ACPI mode. Maybe you can enable ACPI in DOS with FDAPM?
https://www.ibiblio.org/pub/micro/pc-stuff/fr … html/fdapm.html

If the stuttering appears when booting Windows 98 normally but with ACPI disabled, then that would also explain the problem. Unfortunately I don't think there's a way to disable ACPI without re-installing Windows.

Variability in Win98 testing could just be what passed for multitasking getting in the way? We know that Doom runs ok in normal Win98, but not when Win98 is in safe mode. So there must be some driver or something that gets loaded in normal mode that somehow avoids the slow downs.

Unless, of course the stuttering in Safe Mode is caused by something else entirely. Don't suppose there are any RAM or cache settings in the other BIOS pages?

did you do a few runs of pcpbench on the C1XF under DOS to make sure it's really stable there? also i'd check autoexec.bat/config.sys/dosstart.bat for whether there's any differences to the ones on the C1XS.

The bios does not have any ram or cache settings in it. It pretty much has what I showed you as well as changing the system time, plug n play bios, NTSC/pal settings, boot logo animation, ide settings, and security settings.

I went ahead and ran fdapm in resident with the command fdados, and I have to say that the machine gave back similar pcpbench results as previously. First one in the 80s, next two in the high 60s.

I will go ahead and run pcpbench on the C1XF and edit this post as an update. I will also say that I have booted from a clean windows 98 boot disk with the same issue to rule out the autoexec/config issues.

EDIT: I ran pcpbench on the C1XF and all three tests came back with one consistent score: 96.0.

are the BIOS versions on the two units identical?

i have found a teardown video here: https://www.youtube.com/watch?v=8QIaMFxrtWQ it looks like there are two electrolytic SMD capacitors on the mainboard. i don't know how plausible it is that those could be the culprit and to be honest haven't seen reduced performance as a result of bad capacitors myself, but if you were out of options and wanted to try to fix the unit at all cost, swapping those would be a possibility. wouldn't be easy with that connector right next to them, though.

or maybe the CPU really does need new thermal paste - that perhaps would be easier. you could try running prime95 on there and see if you get any instability, and while doing so, you can also monitor clocks with aida64 or a similar tool.

'fraid I think I'm getting stumped here. One last thing that might give some useful information, try running the doom timedemo on both machines, booting to a command prompt, in win98 safe mode and in win98 normal. We can use the C1XF as a known good reference. If all three figures on the C1XS come back lower then we know there's a problem even in win98 normal mode. If Win98 normal mode is about the same for both then we know that win98 is fixing the problem somehow.

OP already said that pcpbench gives him erratic scores and stutter even under win98.