That bigger heatsink is definitely helping if the CPU is being overvolted my guess is that it was running hotter than that smaller heatsink could dissipate effectively. Reading your thread this sounds like it's going to be a great glide gaming system. I'm looking forward to seeing how the benchmarks go and how it plays your games.

I think the main difference between the two CPUs is that the K6-2 requires 2.2V while the K6-3+ needs 2.0V. Your board provides 2.5V at the lowest setting which is too much for the K6-3+ but may be within tolerance range for the K6-2. As noted above, overvolted CPUs tend to run hotter and generally behave more erratic, so it makes sense that a larger heatsink would help.

BTW, if you have the audio bracket for that motherboard, the CMI8330 integrated sound card is well worth using. It offers SBPro, SB16 and WSS compatibility which is a very rare combination. It also has excellent sounding FM synth.

chrismeyer6 wrote on 2021-07-17, 02:04:

That bigger heatsink is definitely helping if the CPU is being overvolted my guess is that it was running hotter than that smaller heatsink could dissipate effectively. Reading your thread this sounds like it's going to be a great glide gaming system. I'm looking forward to seeing how the benchmarks go and how it plays your games.

Thank you for your interest.

I've ran a few benchmarks from Phil's DOS benchmark pack.

So far so good everything seems to work great. I'm still in the process of getting all drivers/USB etc working right. Hopefully I can get some time to post up some numbers this weekend.

I also would like to mess around with getting 83mhz FSB working.....but don't want to press my luck too much either.

Allright here are a few quick benchmark numbers from Phil's Benchmark Pack as well as a few Windows ones as well.
DOS benches ran from pure DOS mode under Windows 98SE

Superscape 1.0 Low Res: Comes back 0.0 but runs super fast?
Superscape 1.0 High Res: 530.9
Chris' 3D Bench Low Res: 566.8 (340.0FPS)
Chris' 3D Bench High Res: 104.7 (62.8FPS)
PC Player Bench Low Res: 146.5FPS
PC Player Bench High Res: 34.7FPS
DOOM Minimum: 2134 in 154 realtics
DOOM Maximum: 2134 in 636 realtics
Quake Low Res: 79.9
Quake Medium Res: 34.3
Quake High Res: 19.5FPS
Topbench: 438
Speedsys CPU: 509.61
Speedsys Memory Throughput - 110.48 MB/s

Windows benchmarks:

3DMark 99 Max Default Run 800x600 16bit:
2551 3D Marks
5930 CPU 3D Marks

3DMark 2000: Default Run 1024x768 16bit:
1417 3DMarks ( I might do a 800x600 run later who knows)

Unreal (Gold Edition UT99 engine I believe?)
800x600 16bit Glide Mode
3 pass average built in benchmark: AVG 28.3 FPS (Highest - 60.0, lowest 10.7)

These results are with the FSB at 75mhz. I did test 83mhz and while the PC will POST and boot into DOS fine, and also run some benches it isn't stable. It also will not boot into Windows.
I might could play with PCI/ISA speeds and dividers but honestly i doubt the results would be that dramatic and it is 100% stable at 75mhz/6X
I did tighten all the RAM timings as much as possible and disable on board sound/serial/parallel ports.
Motherboard cache is enabled. I doubt it would make much difference to turn it off.

Unreal results a bit lower than what I was hoping for as I know this CPU at 550MHZ with a 100MHZ FSB and the same Voodoo will do better.
But this is also the most demanding game this PC will likely ever run, and if I want to run it faster I can use my P3/Voodoo rig or just run it on my modern PC.
It might also bench a bit higher with a PCI sound card and i might try that in the future.
I have several including a Vortex1/Vortex2/SB Live!/Turtle Beach Santa Cruz etc.

I doubt that this thread gets many views but to those of you that have helped and offered advice as well as your own experiences thank you so much. I've been messing with PC hardware with 20 years but it is easy to either forget something you one knew or just to be ignorant of something you never encountered.

So again thank you all.

If anyone has any requests of benches or games you would like me to run just let me know and I'll try to make it happen!

Looking at your results honestly that system has got quite a good level of performance. It may not be a super duper power house but it's got solid performance that you can load your games on and just have a great time with it.

Joseph_Joestar wrote on 2021-07-17, 02:49:

I think the main difference between the two CPUs is that the K6-2 requires 2.2V while the K6-3+ needs 2.0V. Your board provides 2.5V at the lowest setting which is too much for the K6-3+ but may be within tolerance range for the K6-2. As noted above, overvolted CPUs tend to run hotter and generally behave more erratic, so it makes sense that a larger heatsink would help.

BTW, if you have the audio bracket for that motherboard, the CMI8330 integrated sound card is well worth using. It offers SBPro, SB16 and WSS compatibility which is a very rare combination. It also has excellent sounding FM synth.

This all good information. I wish I could confirm 100% the voltage being provided but I agree it is very likely being overvolted.

I don't have a audio bracket off hand. If I can locate one I will give the integrated audio a try.

I have several PCI audio cards but only a handful of ISA ones. The ESS and my SB16 that is currently in my 486 build.

chrismeyer6 wrote on 2021-07-18, 01:23:

Looking at your results honestly that system has got quite a good level of performance. It may not be a super duper power house but it's got solid performance that you can load your games on and just have a great time with it.

Yeah it should be very fine for what I HOPE to find the time to use it for.

I have plenty of other much faster rigs I could use for 3D games.....Athlon XP, A64 as well as two complete Pentium III builds.

But I hope to maybe use this as a DOS/Windows gaming catch all up to about 2000. Much like Phils 4n1/5n1 builds.

There are a ton of games I played through back in day that i'd like to experience again on somewhat similar hardware that I originally used it on.

Now I just have to find the time

Sadly finding the time is the hardest part no doubt about that.

Jasin Natael wrote on 2021-07-18, 02:03:

I don't have a audio bracket off hand. If I can locate one I will give the integrated audio a try.

The on-board audio connector is proprietary, so finding the original bracket would be the best.

That said, I was able to hook up a generic audio bracket to the connector using jumper cables and this pinout but that was somewhat tricky. It needs to be done very carefully as connecting the bracket to the wrong pins could potentially short out the board.

chrismeyer6 wrote on 2021-07-18, 02:19:

Sadly finding the time is the hardest part no doubt about that.

Absolutely

Joseph_Joestar wrote on 2021-07-18, 05:24:

Jasin Natael wrote on 2021-07-18, 02:03:

I don't have a audio bracket off hand. If I can locate one I will give the integrated audio a try.

The on-board audio connector is proprietary, so finding the original bracket would be the best.

That said, I was able to hook up a generic audio bracket to the connector using jumper cables and this pinout but that was somewhat tricky. It needs to be done very carefully as connecting the bracket to the wrong pins could potentially short out the board.

Good to know. I didn't get a bracket with this board.
But in light of having several decent discreet sound options I'll probably never explore it.
It is good to know that it is a decent option to have though so thanks.

Some rambling, I hope some is useful...

I'm surprised nobody mentioned it (or they did and I missed it) many PATA (and I imagine SATA) simply don't handle the IDE controller (southbridge) running at 41Mhz. I remember having to try a few different drives to find one(s) that don't lose their marbles at above like 37.5 (75/112) - heck some hated even that.

If your drives end up working when you set bus speed down to 66 (33) then... that's it. Another thing to try is... shove a random CD-ROM drive on there while it's at 83/41 - see if it works (though w/o a boot drive, dunno maybe try an old linux live CD?). Optical drives seemed way less picky about it. It's also hilarious on "TX Pro" chipset board cause it seriously thrashes that drive for all it's worth! (I've never seen an old drive work that hard 🤣).

(side ponder - This also makes me wonder in this modern age, all the people using the IDE > CFcards if they run into this too? Man now I wanna see someone actively test for this - sadly I no longer have a board that is able to drive PCI that high w/o something else taking a dump first)

You can also try lowering DMA/PIO modes ..maybe? or maybe just giving up on DMA - but yuck that's of course not ideal.

I remember having the best luck with Samsung and IBM drives tolerating high IDE/southbridge/PCI speeds. So if you have any old Spinpoint F1s or Deskstars around worth a shot? But note this is like the era from 2GB to 80GB drives tops.
It IS totally worth it to find a drive that does tolerate it though.

Last thing I remember running into, some PATA PCI add-on cards also choke if PCI too high. While others using same exact same controller work brilliantly. I had a Promise ATA100 branded card that absolutely choked if PCI speed was over like 35 (lame!) but an Iwill branded card (using exact same controller!) worked brilliantly like past ridiculousness (I was trying to force an RMA of a drive I didn't trust, and it dropped off somewhere insane like 91 - and no the drive didn't throw any codes either).

(amusing giggle - ATA133 you ask? well considering the Maxtor drives I tried wouldn't even work at ATA66 when PCI was even at a tame 37 sadly it doesn't look promising. That's not even considering if there is an ATA133 controller that would tolerate this. That said, how cool would that be 🤣)

random trivia - I actually had a pair of the later infamous IBM "deathstars" as my RAID-0. 2 45GB 75GXP's. Yes they both died 7 months later. Which really made me sad because they were awesome performers and super tolerant of overspec ATA speeds.

xrror wrote on 2021-07-21, 02:31:

Some rambling, I hope some is useful... […]
Show full quote

Some rambling, I hope some is useful...

I'm surprised nobody mentioned it (or they did and I missed it) many PATA (and I imagine SATA) simply don't handle the IDE controller (southbridge) running at 41Mhz. I remember having to try a few different drives to find one(s) that don't lose their marbles at above like 37.5 (75/112) - heck some hated even that.

If your drives end up working when you set bus speed down to 66 (33) then... that's it. Another thing to try is... shove a random CD-ROM drive on there while it's at 83/41 - see if it works (though w/o a boot drive, dunno maybe try an old linux live CD?). Optical drives seemed way less picky about it. It's also hilarious on "TX Pro" chipset board cause it seriously thrashes that drive for all it's worth! (I've never seen an old drive work that hard 🤣).

(side ponder - This also makes me wonder in this modern age, all the people using the IDE > CFcards if they run into this too? Man now I wanna see someone actively test for this - sadly I no longer have a board that is able to drive PCI that high w/o something else taking a dump first)

You can also try lowering DMA/PIO modes ..maybe? or maybe just giving up on DMA - but yuck that's of course not ideal.

I remember having the best luck with Samsung and IBM drives tolerating high IDE/southbridge/PCI speeds. So if you have any old Spinpoint F1s or Deskstars around worth a shot? But note this is like the era from 2GB to 80GB drives tops.
It IS totally worth it to find a drive that does tolerate it though.

Last thing I remember running into, some PATA PCI add-on cards also choke if PCI too high. While others using same exact same controller work brilliantly. I had a Promise ATA100 branded card that absolutely choked if PCI speed was over like 35 (lame!) but an Iwill branded card (using exact same controller!) worked brilliantly like past ridiculousness (I was trying to force an RMA of a drive I didn't trust, and it dropped off somewhere insane like 91 - and no the drive didn't throw any codes either).

(amusing giggle - ATA133 you ask? well considering the Maxtor drives I tried wouldn't even work at ATA66 when PCI was even at a tame 37 sadly it doesn't look promising. That's not even considering if there is an ATA133 controller that would tolerate this. That said, how cool would that be 🤣)

random trivia - I actually had a pair of the later infamous IBM "deathstars" as my RAID-0. 2 45GB 75GXP's. Yes they both died 7 months later. Which really made me sad because they were awesome performers and super tolerant of overspec ATA speeds.

Good information to have.

If I am correct however the M571 does have a PCI clock divider option by jumpers. So it is possible to run the bus speed at 75/83mhz without overclocking the PCI bus, at least not to much.

I used these resources (linked below) to do so. But looking at it again it appears that I might actually be underclocking my PCI bus from 33mhz to 30mhz by running the board at 75mhz FSB.

I need to check and see what I have JP5 set to...

At any rate I actually did try just optical drives on the IDE channels as well as a known good Samsung Spinpoint HDD, which I never got the board to detect.

http://m571.com/m571/franczabkar/freq.htm

http://m571.com/m571/jumpers.htm

These were linked from this page obviously

http://m571.com/m571/

I've used numerous Asus TXP4s and P55T2P4s @ 83MHz with all sorts of hard drives, video cards, sound cards, and PCI IDE and PCI SATA cards with no issues. I'm not saying that all hardware will work with a 41 MHZ PCI bus, but it's never been an issue for me.

I need to do some more checking.

I really don't think this is what was causing my stability issues....all I can say as it didn't work and then it did. Nothing at all really changed.

But I also think that my memory bandwidth is way worse than it should be, and I think that it is affecting my performance.

Most of the DOS benchmarks look fine to me, but I really think that the Unreal results in particular are far worse than they should be. Also the 3dMark 99 results look a bit low as well.

Curious what others with similar builds think.....

Well I got the FSB to run more or less stable at 83mhz.

Thanks xrror for kickstarting my brain. I reset the divider and seems to be working now.

A few benchmarks are @ 83x6=500mhz

Superscape 1.0 Low Res: Comes back 0.0 but runs super fast?
Superscape 1.0 High Res: 587.5
Chris' 3D Bench Low Res: 627.6 (376.5FPS)
Chris' 3D Bench High Res: 116 (69.6FPS)
PC Player Bench Low Res: 162.3FPS
PC Player Bench High Res: 41.2FPS
DOOM Minimum: 2134 in 140 realtics
DOOM Maximum: 2134 in 510 realtics
Quake Low Res: 88.5
Quake Medium Res: 38.0
Quake High Res: 22.2FPS
Topbench: 519
Speedsys CPU: 564.46
Speedsys Memory Bandwidth - 186.69 MB/s
Speedsys Throughput - 122.57MB/s

Windows benchmarks:

3DMark 99 Max Default Run 800x600 16bit:
2750 3D Marks
6501 CPU 3D Marks

3DMark 2000: Default Run 1024x768 16bit:
1509 3DMarks ( I might do a 800x600 run later who knows)

Unreal (Gold Edition UT99 engine I believe?)
800x600 16bit Glide Mode
3 pass average built in benchmark: AVG 29.91FPS (Highest - 61.05, lowest 12.25)

Improvements are notable in the DOS benchmarks, not so much I think in the Windows ones.
I did get a crash the first time trying to run 3dmark 2000....could be coincidence I'm not sure.
But the more I use the board the better it seems to be.

It is known that caps with lots of life in them, left unused, even if new never used, gradually lose capacitance, but some short periods of use will "reform" them and they gain back their original spec. So that would be a possible explanation for it getting better as you use it.

BitWrangler wrote on 2021-07-23, 01:50:

It is known that caps with lots of life in them, left unused, even if new never used, gradually lose capacitance, but some short periods of use will "reform" them and they gain back their original spec. So that would be a possible explanation for it getting better as you use it.

That is interesting. The board was used, but appeared to be basically new so there might be something to that.

Probably my last update....but I installed OG 1998 Unreal to compare again Unreal Gold. And wow...what a difference. For one it is super unstable and I could only get it running with the 226 patch. I also had to remove the ESS Audio Drive and install my Aureal Vortex 1 PCI card.

But man I went from a average framerate of about 29FPS at 800*600 to a average of 57fps at the same settings.

This is in Glide. I knew that the Gold version used the UT engine but I didn't expect nearly twice the performance. I guess there are a lot more differences in the game engines than I knew.

Anyway still buggy and crashes all the time. .....good ol classic Unreal.

For fun, maybe try installing win2000 - even better, co-install it with win9x on FAT32 and make them share the Program Files directory 😉
In the win2000 installer just set the base folder to WIN2K for max jank ;p

Sorry to others where my IDE/southbridge "speeds" experience doesn't line up. The two systems I distinctly remember banking into this is that aforementioned "Super TX" socket 5/7 board I really wish i still had (If only to actually identify it!), and an ABIT KR7A-RAID (socket A)... that I also really wish i still had.

This is something I'd love to see more testing/info on - it very well may be that my experience was heavily tainted by fluke/quirks of these specific boards. I could also very well be doing things wrong - I never had luck with bus speeds that pushed PCI / southbridge over 75/112 - but then I was mostly on the AMD side and after nForce2 made everything "strapless" I never really dug back into it.