Every PC I own, including 5x86-133@150, works fine when it's +30 degrees outside. Don't you think memory timings are a little too tight for 70ns at 40Mhz bus?

The problems you are describing sounds like memory problems, not CPU issues.

I don't think so guys, because it pretty much works in the beginning, it takes time to spot the first CRC errors, nevertheless I am conducting a test that could prove us right or wrong:

Am playing SC1 now. It's been like an hour, this should be sufficient to give me like 6-7 errors, am going to try it. Right after this, I will setu the timings to the lowest possible option, just to short the testing process. The problem is that it has finally started to rain and the temperatures keep decreasing. Will report back soon.

OK, so it seems, for now, that it is the cache - no errors at the possible lowest settings.

But as I said, I'm gonna wake up tomorrow, run the installation and it will give me no errors at the possible highest settings (because of the cooldown). But this test is ruined by the good weather, so it is not objective to classify both of the possible causes now. The weather forecast seems to be bad again, somewhere around 34 degrees, so I'll report back tomorrow with the findings.

I never had an issue with using old computers in that type of weather. Our original 486 back in the 90s was subjected to it as well without any problems for years. Being that most of the hardware is reaching 20 years old, problems are bound to crop up though. 90-95 degree (F) days are typical every summer where I live. My solution would be air conditioning if possible, it keeps the computer AND the user comfortable! That could be a problem though as apparently Europeans aren't familiar with window mount A/Cs that are dime a dozen in North America. They aren't cheap to run, but its better than sweating in the humidity, particularly at night.

Read this topic, Help with SiS 496/497 Tomato 4DPS 486 motherboard

I have troubleshooted this issue on SiS 486 motherboards using an AMD X5-160. The Zida Tomato 4DPS and Asus PVI-486SP3 behave similar in regard to the issues you are experiencing with a 40 MHz bus. Ensure that the ISA bus is at PCI/4, and more importantly, that "cache write cycle" is set to 3. This seems to be the key if using 512K of cache. Increasing the CPU voltage to say, 3.8 V does improve stability somewhat with 512K and cache write cycle at 2, however it was not enough for long-term stability. If using double-banked 256K of cache, 5u3 noted that he could get away with the cache write cycle set to 2 and it was stable.

How much cache do you have in the system? 1 bank or 2 banks? Double-banked 256K seems to be able to handle the fastest timings for all PCI 486 boards I've tested. Single-banked 512K usually needs slower timings when using 40 MHz and above. Double-banked cache in general (512K-double or 1024K) can usually handle faster timings. The problem is amplified further if you are using a Cyrix 5x86 with cache greater than 256K and FSB's greater than 33 MHz. I suspect this has to do with the Cyrix benchmarking faster per-clock at L1/L2/RAM compared to the AMD X5.

It is unfortunate that most of the later good PCI 486 motherboards could only accept 256K-double banked cache, or 512K-single banked cache. I'm working on a mod now to get my Biostar MB8433-UUD working with 1024K double-banked cache. Some initial tests were successful, but I have more work to do on this still. It is not so straight-forward as I once thought.

NJRodafan - Did you also put a 15k SCSI right under under an SCSI Magneto-Optical drive in those days also? No irony in here, but can you imagine the heat generated by this drive in such case? These drives run pretty hot, I guess we all know that, and as you can see on the pictures, the heat dissipated from the drive can be easily absorbed by the CPU fan - not good.

feipoa - thanks for your ever exhausting much info into this, much appreciated. I'm not gonna babble around here, showing my memory setting in BIOS by this pictures:

These are the settings I've been using all the time and have never encountered an error of any sort. With or without the SCSI drive BEFORE these thays (that's like Christmas and months after):

And these are the ones that never produce any error (with the new cooler, not the original one):

------------------------------------------

The thing is that, as I said in the morning, I woke up today and immediately ran the installation of the game - no errors. Tried this 3 times consecutively, rebooted, did again those 3 times - no errors. Been playing the SC1 for a couple of hours and now it gives me 3 errors. So guys, do we still think the problem may reside in the memory subsystem? By lowering these settings, thus having no errors, IMHO the CPU is ''working'' easier, thus not generating that much heat that could cause these 3 errors and NO errors in games after hours of playing.

The weather is good for now, alas the test is not objective now. I can, of course, use the original heatsink, but I superglued the CPU with the P1 pretty much, so I gotta find my spare ADZ for this purpose.

So I turn on the PC, ran that DOS game installation, went through. Then (again) played SC1 for an hour (with my bro also playing his 3D intensive games), followed by few files copy and some win95 games installation and their testing - NO error, NO crash, NO inappropriateness.

Then I came back to the DOS installation again - 18 CRC errors. So I immediately reset the PC and put up the lowest memory settings. NO errors. Then, immediately again, I rebooted it and set up the settings back to the original state - 7 errors.

Well this (not that clearly - I still believe in your advice) shows us that there has to be some sort of temperature issue. If not of the CPU, then of RAM modules? Is it possible these chips can generate such heat they could not stand and behave incorrectly?

I've had instances in a 486 w/SCSI system whereby after some time of being turned on (as it heats up), the power connectors would make poor contact and cause erratic behavior. It was most puzzling.

What cache configuration are you using? 512K-single-bank, 256K-double-bank, or 512K-single-banked?

Shodan486 wrote:

So I turn on the PC, ran that DOS game installation, went through. Then (again) played SC1 for an hour (with my bro also playing his 3D intensive games), followed by few files copy and some win95 games installation and their testing - NO error, NO crash, NO inappropriateness.

Then I came back to the DOS installation again - 18 CRC errors. So I immediately reset the PC and put up the lowest memory settings. NO errors. Then, immediately again, I rebooted it and set up the settings back to the original state - 7 errors.

Well this (not that clearly - I still believe in your advice) shows us that there has to be some sort of temperature issue. If not of the CPU, then of RAM modules? Is it possible these chips can generate such heat they could not stand and behave incorrectly?

I still think that your 70ns memory modules can't handle the 2-1-1-1 timings, regardless of the heat. Those are fast timings, and 70ns isn't that quick. I'd say get 60ns or better yet, 50ns memory modules if you can.

EDIT:
Did you have the SCSI controller when the system was working two months ago? I don't know what kind of PCI bus speed you get with 40MHz FBS, but I imagine it's possible that it's the SCSI controller causing problems, running at a bus speed it doesn't like. Just throwing out some ideas. I'm close to 100% certain it's not CPU overheating.

LunarG wrote:

I still think that your 70ns memory modules can't handle the 2-1-1-1 timings, regardless of the heat. Those are fast timings, and 70ns isn't that quick. I'd say get 60ns or better yet, 50ns memory modules if you can.

I agree with this. 2-1-1-1 is very tight for 70ns modules which I can't get running on my motherboards. I can only get that with my 60ns RAM.

OK, I'd rather verify whether I got 70ns or less since ya'll point at the memory timings. Just gimme a sec.

Feipoa - I've read about your experience thoroughly months back, just in the time when I was inspired by using a SCSI. Yes, the drive is running pretty hot, thus I think the heat is being engulfed by the CPU fan, which causes my problems.

Other than that, how do I figure out whether I have double or single banked? ... I currently use half of the sockets, thus probably single banked and it's 512 I think, I'm gonna make some pictures for you.

PVI-SP3 has no PCI divider, thus PCI runs at 40MHz - but thats the thing of the SCSI - to my knowledge (and a little of experience), SCSI should be quite immune to the host bus frequency, because it itself is a bus with it's own logic rules and electrical attributes running the only way the controller is able to.

Well... I was wrong with the memories - they're 60ns. As for the cache, please clarify. It's 256KB (4) 64K x 8.

I will take a better look on the power plug on the HDD, but if the HDD was problematic, I would face some corrupted data - after 2 weeks of fuzzing with this rig, seeing those errors and stuff, I end up with no lost data - everything is in its place as it should be and OK.

Is the system stable if you drop the bus speed to 33 MHz?

5u3 wrote:

Is the system stable if you drop the bus speed to 33 MHz?

It has been all the time until this very summer hell on 40MHz. Played the whole Diablo & Hellfire on it for a few weeks, which is a pretty demanding game in the last stages of it - no errors.

It, of course, came up to mind just test out the 33MHz, but I received this mobo with an ADW cpu (which cannot be clocked) working at 133MHz, thus the 33MHz bus - no errors back then.

I will test it out since it's a different time of the year. Cannot tell if today....naah, for the heck of it I shall do it today. I plan to build 2 more 486 ONLY DOS based machines, wanna have this one work 100% ASAP.

I recommend using double-banked 256K cache instead of single-banked 256K cache, that is, four pieces of 32Kx8. Double-banked cache seems to handle faster timings better.

^^^ Agreed.
Another possible problem might be the voltage regulator (this is weaker on early revision boards, check if it gets very hot).

So it's pretty cool outside nowadays, thus bearable for all living beings and plus extremely positive for the system - so far no errors after doing whatsoever.

Will try to locate to possible source of the issue, but since I kind of neglected this inconvenience, the chances to trace it are diminished for this year, I guess things will change in the next summer. For now, we're fine.

Cache, regulators, RAM, chipsets - will have a close lookup.