The ATA133 card with the CF card only benchmarks as fast as 20 MB/s, so this test is inconclusive. The CF card is 133X, which is limited to 20 MB/s.

I may have figured out the solution. It seems that only one of my MB8433 motherboards will work stably with the Adaptec SCSI card set at above 40 MB/s. It is a version 3.0 board, and the southbridge is UM8886BF with datecode 9634-FXO. The northbridge is UM8881F with datecode 9633-EYT. This board also was only overclocked to 66 MHz for less than 1 hour. I'm not sure if this minimal overclocking played a role in its greatness.

That is the only southbridge with FXO I've seen. Most UMC 1996-era southbridges are FXS or FXA. 1995-era southbridges are often DYS. Most 1996-era northbridges are EYT or EYA. Most 1995-era southbridges are ETA or ETO.

I thought that only the first 2 letters had any possible impact on the chip's revisions, not the 3rd letter.

rg100, what are your north/southbridge datecodes?

This recent finding really seems to solidify my plans to order all new 1997-era chipsets. I'll probably still persue my plans for a clock amplifier and divider since this will allow for running the FSB at 66 MHz and the use of the CPU's 4x multiplier. I may also continue looking into a PCI clock amplifier, but these curiosities are now less urgent.

I now have 3 slot fans, a front case fan, and a CPU cooler on my backup 486 rig, running a Cyrix 5x86-120 at 133 MHz (2x66MHz). I also put in a much more powerful power supply fan. I'm hoping this rig now can be run with the case on and branch prediction. I think I'm going to leave my main rig alone (the one I'm typing on). I probably got a little too greedy trying to run it at 66 MHz too. I'm going to put my Cyrix 5x86-133GP back in there; I should probably weld the case shut to prevent itchy hands.

feipoa wrote:

rg100, what are your north/southbridge datecodes?

Luckily, the photo I took of my mobo shows the datecodes. 😀 It's here. BTW, have you tried a different SCSI HDD yet? I have read your posts, and you have done a lot of testing, but I don't recall seeing a mention about swapping out a SCSI HDD for a replacement. I'm sorry if you have mentioned this.

OK, so you have an FXA and an EYT with 1996, 40th week date codes. My v3.1 board also has FXA. The board I'm using now is the only FXO board I've ever seen.

For completeness, I should mention that in my working FXO system, my osciliscope showed the PCI clock signal with:

Vmin = 0.78 V (just barely enough to qualify for a low signal)
Vmax = 2.50 V
Freq = 33.33 MHz

My main belief is still that the PCI minimum voltage isn't falling well enough below 0.8 V, and this is why the system is generally not very stable. I placed a resistor in there to drop the voltage down a bit, but it also drops the high level voltage, hence an amplifier is needed. The addition of the resistor also messes up the matched impedence of the lines, causing reflections. Again, an amplifier will fix this by having a high imput impedance and a low output impedance, whereby the output impedance.

The mentioning of the changing the HD's may have gotten mixed up in here, or in my head, but I have tried different SCSI hard drives. It is all probably a bit confusing because I've been working on two identical Biostar systems simultaneously, so in effect, 2 Biostar systems showed the same symptoms, therefore 2 different hard drives have been tested.

For others who run into these problems, there are a few no-brainer solutions:

1) In the SCSI BIOS, set the maximum speed to 20 MB/s and leave your motherboard BIOS set to 1:1 FSB-to-PCI divider.

2) In the motherboard BIOS, set a 1:2/3 FSB-to-PCI divider and leave your SCSI BIOS set to max speed (40 MB/s - 160 MB/s). This is probably only an option if you plan on running 40 MHz x 4 (AMD X5), 40 MHz x 3 (Cyrix 5x86), 50 MHz x 3 (AMD X5), or in the rare case, 50 MHz x 4 (AMD X5).

3) For those who want to run an overclocked FSB system, set the FSB to 60 or 66 MHz, and set a 1:1/2 FSB-to-PCI divider in the motherboard BIOS while leaving your SCSI BIOS speed set to max. 60/66 MHz FSB works well with Cyrix 5x86 CPU's on 2X mode. Some AMD X5-133's will work fine on 3X mode and 60 MHz FSB.

4) Stay tuned for news of a possible custom solution for a CPU clock amplifier/multiplier and/or a PCI clock amplifier.

EDIT: Well, that didn't last long. The BSOD's have returned and this time the 20 MB/s SCSI setting didn't help. This is awefully random. I"m going back to option 3, Cyrix 66 MHz x 2.

EDIT: I'm trying a wild card now: I had prevously move the PC speaker to right under the harddrive to make room for the front case fan. I've removed the front case fan and put the PC speaker back to a position far from the harddrive. I figure, the speaker is magnetic, the HD is magnetic... interference? Also, the fan contains a magnetic field, which is also pretty close to the hard drive. Couldn't hurt to remove them both. I came to this idea when I noticed that having the case on greatly increased the frequency of BSODs. Wild cards sometimes work. The PC speaker and the fan are the only other changes I had made to the system.

EDIT: 7 consecutive reboots OK. Not sure if it was the PC speaker, the fan, or the repositioning of the SCSI cable, but things seem OK now. Time will tell. I'm still suspicious about the PCI clock voltages.

feipoa wrote:

OK, so you have an FXA and an EYT with 1996, 40th week date codes. My v3.1 board also has FXA. The board I'm using now is the only FXO board I've ever seen..

I just pulled out one of mine and it's a ver.2 board with an FXO chip on it.

feipoa wrote:

EDIT: 7 consecutive reboots OK. Not sure if it was the PC speaker, the fan, or the repositioning of the SCSI cable, but things seem OK now. Time will tell. I'm still suspicious about the PCI clock voltages.

Has this repositioning of magnetic components solved the entire problem with these mobos? In other words, will your mobo now function correctly at a bus speed of 33 MHz, if you use PCI SCSI based equipment set to operate at a fast speed?

Sliderider, interesting... another FXO. Well, fire it up at 1:1 with an Ultra SCSI adapter and let me know how it goes.

rg100, I am not sure about the other MBs; I didn't test them. This FXO has always been my good performer. It is the only one of the lot, for example, which will run long-term stable at 40 MHz FSB and 2-1-1-1 cache settings. 2-1-1-1 at 33 MHz is okay on all boards though. For some reason, the other motherboards need 3-2-2-2 at 40 MHz to be stable. What is striking is that all boards seem ok at 3-2-2-2 and 66 MHz. I think some new chipsets would fix this randomness right up.

I suspect that this repositining fix is only for this particular MB/case. I did 9 reboots with no crash and the case off, then I put the case on, and it crashed on the first reboot. So the problem had to be related to interference coming from inside the farady cage (the case). This is what led me to move the PC speaker, fans, and SCSI cable.

I remember in my other system I had 1:1 issues as well at 33 MHz. Perhaps it is due the close proximity of the front case fan? I have no ide and I still find this solo arguement hard to swallow. I remember testing other SCSI drives in this case which were laying on the floor and also had the 1:1 BSOD's.

For now, I am satisfied. I got my Cyrix 5x86-133/4x back and I don't want to touch it as it is my everyday rig. For my other duplicate system, I have my Cyrix 5x86-120 running at 133 MHz just fine with 1:1/2. I ran 3 hours of the Windows game Subspace on it without a hickup and with the case on, so I'm pleased. It has 3 slot fans, a front case fan, and a monster PSU fan.

For reasons of curiosity, I'll probably get around this duplicate rig at 33 Mhz and 1:1 with 10 reboots to see if there are any pending issues -- I suspect there still are. The PC speaker in that case is far from the HD and it had issues. I think there is a lot of sensitivity and EM noise inside the case which is just sufficient to occasionally throw off the PCI clock voltages.

Some alternate testers would be good here. Unfortunately, I can't touch this stuff again until after the holidays or I'll be looking at either 1) a new wife, or 2) a pile of smashed computers.

4 days, 12 hours consecutive uptime with no BSOD to report under regular use.

I don't want to say conclusively that the cause of the BSOD's were due to the PC speaker being directly under the HD (about 1 cm from the bottom of the HD) or the front case fan being right next to the HD, but things are sure pointing in that direction.

Sometimes the cause of random BSOD's are the unexpected. I still haven't gotten around to testing out the other MB8433 boards to determine if they are all well. I may look into this again later when I begin work on the Ultimate 586 Benchmark Comparison. I plan on including the Cyrix 5x86-100, 120, and 133 for this comparison as well since they are really 486/686 hybrids. I am planning on including results for them with a higher FSB to offer a fairer Pentium comparison. So, 2x50 (100 MHz), 2x60 (120 MHz), and 2x66 (133 MHz). I am very curious to see how they will compare up to a same-frequency Cyrix 6x86.

I've been waiting months for Amazon to restock on ATX-to-AT power cable converters. I need this for my socket 7 rig to get started. They offer free shipping. The only other option is to make on myself, or to pay some eBay dude $22 for shipping to Canada (that will never happen).

For my socket 7 rig, I'm going to see if I can get Win98SE, WinNT4.0, W2K, and WinXP all on one HD. The trick here will be to have an 8 GB FAT32 partition for Win98SE and have NT4 still bootable. I think I can do this with Partition Magic and using the W2K boot loader. I also want NT4 to have a 7.X GB partition. Both of these are tricky to achieve, but it is doable. Instead of SCSI, I am planning on using the TX2plus ATA133 PCI controller card with a spare 200 GB IDE drive.

My plan is to setup four 2 GB FAT partitions. Install Win98SE on the first partition, and NT4.0 on a 5th 4 GB NTFS partition. Once NT 4.0 is all installed, I'll create a 6th NTFS partition and use partition magic to merge the two 4's into one 8 GB NTFS partition. Then I'll then use Partition Magic to convert the Win98SE partitions from FAT to FAT32, then merge the first 4 partitions to make one 8 GB FAT32 partition. After converting Win98SE to FAT32, Windows NT 4.0 will not be bootable, so I'll then install Windows 2000, which will install the FAT32 readable W2K boot loader. All 3 OS's will now be bootable. Then I can slap on XP if I wish.

I wonder if there is an easier way to do this? I know NT 4.0 can only have up to a 4 GB partition when installing. Or perhaps it can work with an 8 GB partition as long as it is preformated w/NTFS? I am pretty sure I cannot install NT 4.0 after W2K.

Uptime 10+ days. Case closed, for now anyway.

feipoa wrote:

For my socket 7 rig, I'm going to see if I can get Win98SE, WinNT4.0, W2K, and WinXP all on one HD. The trick here will be to have an 8 GB FAT32 partition for Win98SE and have NT4 still bootable. I think I can do this with Partition Magic and using the W2K boot loader. I also want NT4 to have a 7.X GB partition. Both of these are tricky to achieve, but it is doable. Instead of SCSI, I am planning on using the TX2plus ATA133 PCI controller card with a spare 200 GB IDE drive.

My plan is to setup four 2 GB FAT partitions. Install Win98SE on the first partition, and NT4.0 on a 5th 4 GB NTFS partition. Once NT 4.0 is all installed, I'll create a 6th NTFS partition and use partition magic to merge the two 4's into one 8 GB NTFS partition. Then I'll then use Partition Magic to convert the Win98SE partitions from FAT to FAT32, then merge the first 4 partitions to make one 8 GB FAT32 partition. After converting Win98SE to FAT32, Windows NT 4.0 will not be bootable, so I'll then install Windows 2000, which will install the FAT32 readable W2K boot loader. All 3 OS's will now be bootable. Then I can slap on XP if I wish.

This scheme was working out just fine until I converted the Win98SE partition (C: drive) to FAT32 using the drive converter included with Windows 98. After conversion from FAT to FAT32, not even the Windows 2000 boot loader came up. The machine booted straight into WIn98SE only. The driver converter warnings said it should still work with NT5. OH well, I guess I'll just leave Win98SE as an FAT partition.

You need to write the Win2000 boot sector again.

How does one re-write the harddrive's boot sector with the NT5 boat loader without reinstalling NT5?

The repair mode of the Win2000 installation CD can do that.

I did system restore for the W2K CD, but it didn't work.

Windows 2000 could not start because of a computer disk hardware configuration problem. Could not read from the selected boot disk. Check boot path and disk hardware.

Please check the Windows 2000(TM) documentation about hardware disk configuration and your hardware reference manuals for additional information.

I made sure the boot.ini was unchanged after conversion to FAT32 for the Win98SE partition.

C:\Win98SE
D:\WINNT 4.0
E:\W2K

I wonder if the issue could be related to the WinNT 4.0 partition being an extended partition from C? Everything works fine when C is DOS FAT (2 GB). I should probably mention I'm doing these tests with a 16 GB CF card (200x) and the built-in IDE/PIO-4 port on the Biostar motherboard.

feipoa wrote:

I don't want to say conclusively that the cause of the BSOD's were due to the PC speaker being directly under the HD (about 1 cm from the bottom of the HD) or the front case fan being right next to the HD, but things are sure pointing in that direction.

I may have finally resolved this issue for good and wanted to share it incase anyone else has this issue. The removal of the front case fan and speaker speaker under the HDD was not a permanent fix. The symptoms of BSOD's and HDD errors eventually returned. When experiencing the BSOD symptoms, I would open the case, inspect components, then remove the HDD connectors, reinsert them, and the system would usually work fine for a few days, then BSODs with HDD-related driver errors would return.

The biggest clue I received was when the HDD would sometimes shutdown on reboot then power back up. This sounded to me like the HDD wasn't getting enough current from the power supply, or that the power connector was loose. So I thought to myself, if the connector was loose, wouldn't it not work right from the beginning? Perhaps not. It seems that when I would press the connector back on, it would often work fine, but over time, case vibrations would loosen an already weak connector. The source of the vibration is from: 1 single-wide slot fat, 1 double-wide slot fan, a front case fan, the CPU fan, the PSU fan, and the HDD vibrations.

I then tried to pry the connector a bit so that the mating pieces would make a tighter fit. This lasted another few days, but I eventually got another BSOD. As a final effort, I decided to solder on a whole new power connector and cable directly onto the PSU's PCB. This connector came off an old PSU and fits very tight. I have had an uptime of over 15 days now without rebooting and continual use. This explains why using a different HDD and cable didn't help the issues noted earlier.

I wonder how much of these PSU's we determined are junk because we didn't swap out the connectors? For my PSU, I tried every other connector, but all of them didn't make a good enough connection to avoid the HDD BSODs. With us retro testing folk, we tend to connect and disconnect PSU connectors quite a bit, causing the connection to loosen.

I am not sure if this solution will also resolve the issue I noted earlier about not being able to use a 33 MHz PCI bus after the MB has been overclocked to 66 MHz. This result is for my Cyrix 5x86-133/4x box and I haven't revisited the 33 MHz PCI issue on my boards which had been overclocked to 66 MHz, so the previously noted caviat is still in effect until further notice.