Shodan486 wrote:

I'm trying to concentrate on the highlighted ones, because I often get Memory Test Fail during POST error using soft resets, the cold one always erases this problem.

Did you try removing your RAM modules one at a time?

Many things could cause your problem...

But the first thing I noticed was that you have quite a few expansion cards, a high powered SCSI drive...and a shitty power supply. I'm sure the age of the components in that thing doesn't help either. I would get a power supply from a reputable company...and if you still have instability problems I would look into recapping some of your components.

But yeah, check the easy stuff first...memory and expansion cards.

Oh also, is there some kind of incompatibility between The 2940UW and the 486SP3? I seem to remember reading something along those lines many years back. For the record I did run a 2940U2W in my 486SP3 without any problems...but that's a newer card.

If such an incompatibility exists, you can always get something from Compaq or Bus Logic. Adaptec is overrated anyway.

This doesn't relate to your immediate problem, but is it safe to run a 15K Cheetah without a fan? I've never owned one, but older 10K Cheetahs definitely got hot.

A Voodoo in a 486, keeps the dust out of the slot I guess! 😜

a voodoo 1 would be enough to saturate a 486. even a 133mhz. the super fast scsi drive is just a noise and heat generator. Use one of these instead: http://www.ebay.com/itm/360590436017

Its going to be just as fast, is going to make NO noise and NO heat by comparison. having a 15k scsi drive without a direct fan is putting a blanket over a space heater.

love the ark2000 video card. A better power supply will certainly help as well.

Would Banshee work? I always thought it would be the least CPU demanding chip from 3dfx, and all in one.

What memory and cache timings are you using in the BIOS? Your problem is similar to when I ran a SiS 486 with a 40 MHz FSB and the fastest possible BIOS settings. Oddly, I had to ramp up the CPU voltage (at 160 MHz) to get into Windows, but it would still eventually crash. I had to lower the timings to get the system stable. Refer to this thread, Help with SiS 496/497 Tomato 4DPS 486 motherboard

What voltage are you sending to the AMD X5-133?

Are you running the motherboard at 40 or 33 MHz? Have you tried different RAM modules with a different layout density? I have found PCI 486 motherboards to be rather fussy with memory, even when running at 33 MHz. They are even fussier with 512 KB single-banked cache. What cache configuration are you running? It is always good to have a dozen different types of RAM in your test kit. By far the best RAM modules for any 486 I have tested are the 64 MB FPM Samsung pieces with gold contacts. Refer to the World's Fastest 486 link for a photo of this module. It brought back to life motherboards I thought were useless at the fastest memory timings.

Memory errors on a soft resent sounds like some expansion card is not letting go of some resource, or that your BIOS is not handling this let-go correctly. For troubleshooting, I would remove all other expansion cards and use a single simple graphics card like a Virge. I also have only used the 2940U2W on my 486's so cannot comment on the 2940UW.

Have you run MemTest 4.0 (before and after the soft reset)? Stick with MemTest 4.0 instead of going with the latest v4.2. I've noticed that v4.2 doesn't work well with 486's.

How would you rate the noise level of your 15K SCSI drive? I've noticed that to get a quiet Ultra320 drive, the trick is to get a later generation unit. I have a 15K Ultra320 drive on a PCI-X Ultra320 adapter on a dual Tualatin 1.4 GHz machine and don't find it loud at all. In terms of heat, I just have a front and rear case fan in that computer. It has been running great for years. For a mini 486 tower that is packed full of expansion cards, perhaps you will want to angle a recirculation fan onto the HDD. Before going to that trouble, you might want to monitor the HDD temperature under typical load conditions with the case on using a cheap multimeter-coupled thermocouple.

I have used a 10K Ultra160 SCSI drive in a densely packed mini-486 tower for more than a decade as an HTTP server and did not have any issues with it failing, or with any component failing for that matter, not even the unbranded PSU.

Are you using an Ultra320-LVD SCSI terminated cable? On a side note, I try to only use a 2-connector SCSI cable for Ultra 160/320 LVD when I only have 1 HDD connected. I don't know how much, if any, impact this will have, but it was pointed out to me by Adaptec tech support some time ago when I was questioning why I wasn't getting the rated benchmark speed on their host controller, using their cable, and their benchmark program. I think this is more critical if you are aiming for max throughput on your HDD/host controller, which will never happen on a 40 MB/s host controller, especially on an AMD 486.

One last IMPORTANT note concerning SCSI - I did have some other issues with Ultra160 on my 486 recently, the same 486 which worked fine for over a decade. I started to get random BSOD's upon booting into WinNT. This would be similar to your Win95 protection error message. I had no idea what it was, I tried everything over the course of 2 weeks. At one point, I thought I figured it out, but the symptoms eventually returned. The only clue that finally shed any light was that the HDD, after weeks of random BSOD's, started to shutdown and power up again. I beleive that the vibration from the fans and HDD were loostening the HDD's power connector over time.

I determined this because when I would tigthen the PSU's HDD cable by bending the female retaining holes smaller, the system would not show any symptoms for a few days, then the problems came back all of a sudden. It turned out that all the connectors on my old 486's PSU were too loose due to years of fooling around in the case. The looseness is probably more apparent with the HDD because of the larger continuous current draw than from a floppy or DVD drive. How did I fix it? I had an old broken ATX PSU that I cut one power connector off of. This connector was extremely tight as it came from a machine whereby the owner did not have itchy hands. The connector ends were also gold plated. So I opened up my PSU and soldered the cut cable directly to the PSU's PCB. There was even 1 empty unused solder through-hole for the 5, 12, and GND volt sections. After this fix, I never got another BSOD in NT upon boot, and it has been working for over a year in a 24/7 environment after this fix. Side note: a few months ago I finally took this 24/7 Cyrix 5x86-133 system offline.

Putas wrote:

Would Banshee work? I always thought it would be the least CPU demanding chip from 3dfx, and all in one.

I beleive I tested the Banshee on SiS 486's. It at least powered up correctly. I did not test for Windows drivers on the Banshee. I tested more extensively the Voodoo3, which worked great. It is a nice way to free up a PCI slot if you insist on having a Voodoo in your 486. Freeing up 1 PCI slot lets you use either USB or a PCI ethernet card. The PCI ethernet card greatly improved network file transfer speeds (600-800 KB/s vs. 2.5 MB/s - these are measured values, not theoretical ones).

Thanks all ! Feipoa again questions extensively. I should have clarified a few things firstly and read the previous thread more thoroughly, as I have forgotten to mention few facts:

First is that everything from the HW point of view is fine in the rig. Before placing the SCSI adapter in, I was using the onboard VLB IDE controller with the very same configuration as I've stated above, but there was an USB card wich didn't work out too well with the drivers provided, so I ignored the card, didn't even bother to install the drivers. Everything worked, played Tomb Raider 1 for 2 horus, nothing, finished BS 1&2 on it...

I bought this mobo with RAM & CPU altogether from ebay, guy claiming absolute stability throughout his usage of these parts. So I wanna highlight the fact that before using the SCSI subsystem, the PC ran absolutely flawless on the onboard ''supposedly'' VLB IDE controller except the soft / cold reset issue. I guess it would come in handy to discover the reason of this.

The bus clock is set to 33MHz, downclocked from my 40MHz just to exclude the possible OC issue - same behavior on both settings. The CPU voltage is 3.3v. Memory timings have always been 2-1-1-1, works OK both on 33MHz & 40MHz.

Let's get back to the PSU & environment issues:

Forgot to mention that the PSU is also powering a 20'' HP LCD , since it has the output for displays, saves me one outlet on my surge protector. BUT still, the PSU with all those cards having their respective drivers installed, thus being enabled and powered up, manages to drive the PC & the LCD with the IDE HDD. MAYBE this is more electrically intense since I use my 15K HDD, so I'll remove the LCD from the PSU power to lower the pressure.

The temperature should not be a problem - As you can see, I've added the front 5.25'' double fan assembly, which actually may cool the HDD right underneath, but I've never felt this drive as hot nor noisy. I also added the slot fan right above the V2 card, because most of the V2 PCBs tend to get very hot and become the overall source of heat, so I wanted the fan to be close. When running Q1 @ glide, the fan exhausts quite a warm air, so I improved the environment I guess.

One little thing - I have the box laid on a side, so the HDD is actually vertically positioned - do you guys have any experience with SCSI drives (new ones) that require being laid properly - horizontally. At such speeds I can imagine the HDD having a problem sometimes...

I guess this may be a simple incompatibility of the mobo and the controller 🙁

Shodan486 wrote:

Forgot to mention that the PSU is also powering a 20'' HP LCD , since it has the output for displays, saves me one outlet on my surge protector.

I thought that 'second' power thingy many AT PSU's have was merely a diversion of 220V/110V power, same as from the outlet.
Does having a monitor attached to the PSU actually increase the burdon on the PSU? Because I thought it didn't.

• Remove all other expansion cards except for simple graphics and the SCSI card.

• Set the cache timings to 3-2-2-2.

• Set the CPU voltage to 3.45 V.

• Do not discount the issue with a poor power connector; these high-end SCSI drives are very sensitive to this.

• You can also add a $20 Kilowatt meter to your AC recepticle to see how much power is being consumed. You want it less than 80% of the PSU's max rating.

• What was the result with MemTest with and without soft-reset?

• What happens if you only use a single stick of RAM?

• Maybe your board has poor PCI implementation or a conflict with that 2940UW. Try a 2940U2W as this is what I've confirmed works well on 486 boards. I seriously doubt the 2940U2W would work fine and not the 2940UW though.

• What happens when you use a PCI ATA adapter with an ATA HDD?

Tetrium wrote:

Shodan486 wrote:

Forgot to mention that the PSU is also powering a 20'' HP LCD , since it has the output for displays, saves me one outlet on my surge protector.

I thought that 'second' power thingy many AT PSU's have was merely a diversion of 220V/110V power, same as from the outlet.
Does having a monitor attached to the PSU actually increase the burdon on the PSU? Because I thought it didn't.

You're right, if you open up a PSU, you can see that the outlet line bypasses the PSU.

Referring to Feipoa

-Will do this, it´s definitely better way to insert, install and test it immediately afterwards, to scope the real issue.

- Irrelevant

- Irrelevant

- Thought this also could be irrelevant in my case, because I have got two of two of those HDDs which display the same behavioral error. Anyway I will take a look on these plugs.

- I borrowed one once from my ex-colleague, it displayed 70W while letting TR1 play demo.

- Irrelevant, but debatable (will run the test as I got some other stuff to do)

- Why would I do that?

- Yes, this is where I definitely agree with you, although I am not that doubtful as you are.

- Ive got two, but these seem to be those SATA\PATA combo controllers, for which I have no win95 drivers and win98 is unnecessary for me, thus not even bothering myself to test.

Feipoa - I clarified that the mobo, CPU and the RAM come from one guy altogether and when I received it at first it worked like hell. So verifying the RAM and Cache chips \ settings is a waste of time according to my experience with this mobo - in the previous thread I was trying to get the SATA PATA USB combo card to get working, at least the USB subsystem - failure. Then I tried a pure PCI USB card from Axago based on a NEC controller - That was almost successful - I initialized the USB system of the card, but only Flash drives were recognizable, not saying only intermittently. I closed this case as many discussions referred to a poor PCI implementation with multiple possibilities of resource handling problems with newer PCI cards, regardless of their purpose.

Thanks for the info about bypassing the PSU´s electronics when using the outlet of it, never occured to me this way.

Two last things - so the HDD at its vertical position is a go? And as for the soft reset problem - it stops at POST saying Memory Test Fail - this time I tried to continue ignoring this, but the win hangs during the boot process - this being done with my IDE drives. Performing a cold reset vanishes this problem.

Okay guys, I'm really at loss here (again)...

So I tried everything that I posted above. I even swapped the memory modules, same behavior, and I'm 100% sure the single module would make no difference. Set the memory timings to auto - no change. Disabled both caches, nothing. Using only the SCSI and VGA, even the onboard IDE is disabled (have the installation files + drivers on the HDD), FDD is used for booting. The box is now standing properly, thus providing comfort for the HDD - don't mean shit.

But I've discovered I am unable to run any memtest familiar utilities. Used images or bare executables on the FDD - it just hangs right in the beginning, doesn't even display itself much.

I'm gonna fuss with it for a while, if no change, then I'm declaring this mobo incompatible with AHA-2940UW PCI SCSI Controller. Been enough time and I've seen this mobo to have problems with many other cards regardless of their purpose. In my experience, it's the TNT riva, then I tried other cards that didn't even beep...stuff like that, stuff I like not.

I think I got it guys...actually I've got two, not one, problems, but seem to be resolved (workarounded).

Verifying...

So it was due to image managers - namely Daemon tools, Alcohol 120% just proved my theory - hangs up during boot-up process. But I'm kinda left with nothing in regards to image managers, do you got any ideas?

The second problem is being verified still...

cd emulators were ALWAYS buggy on win98 for me. look for an older version than what you are using.

Finding daemon tools lower than v3.47 is quite tricky, anyways I used som windows 95 supporting version of alcohol and that gave me the results mentioned above, but nevertheless I am going to look for more drivers...

Could be either simply not enough CPU power from that 486 or, what's more likely with the crashes, missing Pentium instructions. I'm running DAEMON Tools 3.47 on 98, works fine.