V-Box Problem

[Newtun]

I have a Compaq DeskPro EN Pentium 550 PC from work, running Win95B.

No, that's not the problem in and of itself - there's more.

It has been having problems when I VPN in to work: random blue-screens, freezes, or screwy error-boxes popping up. Since it's a work PC, I can't do anything (detectable) to the OS or software. I had upgraded the cruddy 6.4 MB hard drive to a D740X, and upped the memory from 64 to 256 MB, though.

MemTest86 runs okay for days, even the high-end tests. But I recently burned Prime95 onto a CD so I could test without installing it. Prime95 kills the PC within a few hours - it blue-screens, or has some other failure, like Explorer failing - need to re-install Windows.

Does this seem like a hardware problem, or more like a problem in the Win95 install?

Are there any diagnostic programs on floppy and/or CD that could help to narrow down the problem to hardware vs. bad OS install?

Is it possible that the memory passes MemTest86, but Windows stresses it harder?

Thanks.

 

[blakerwry]

i think it's probably not a memory problem...

maybe it's a problem wiht the NIC or the drivers.. or just win95 networking in general... you say you changed the HDD but cant make any software changes?

is it possible you can just keep the old HDD with the old software on it and reformat to winNT 4 or 5 on the d740x?... so when the PC goes back to work it has the same ol software on it...

 

[Mercutio]

You might want to throw a real fan in that thing. Those old ENs have passive-cooled CPUs.

 

[Handruin]

As long as the deskpro EN isn't stuffed in a corner, I doubt it will over-heat. At work we have a two (used to have 4) of the Deskpro EN's running W2K Pro and they run very reliable during their 24x7 operation. Those machines have been powered on for at least the past year. I've opened them up at one point or another and I believe they are passively cooled as you mentioned Mercutio. If I'm not mistaken it has a single internal 80mm fan that channels air across the CPU. That deskpro EN case is a PITA to open, BTW. It was a bit deperssing to install a Emulex LP8000 fibre channel adapter into this machine since the HBA was worth way more than the entire Deskpro EN. (somewhere around $1300 for the HBA)

It could be the memory even if it passes your tests. I remember an incident many years back with my parent machine where the guy we bought our computer from gave us parity ram instead of EDO. (At the time I didn't know how to tell the difference, nor did I have a internet connection)

Once installed, the machine had one stick of EDO and the other was parity. (This was a cyrix 686 166 MHz) The win 95 box would randomly blue screen constantly, sometimes it was hours, sometimes minutes. It wasn't until we took the RAM to a computer show to trade it in that we found one was EDO and the other was parity. Both sticks worked fine on the tester and even tested fine in the machine, yet it BSOD like crazy.

I'm not saying the same applies to you, but even though the memory looks fine in the machine, it could be a combination of things...

 

[Mercutio]

Keep in mind that we're talking about a passively cooled box with a 740X in it. In my experience the BB version of that drive isn't hot in a SCSI/early Barracuda ATA sort of way, but it does get warm.

Most of the ENs I've seen had WD AA-series or Seagate U drives in them. Slugs both, but they also operate at essentially room temperature. The 740X could be just enough to push things out of whack in that Qmachine.

Incidently, the Deskpro EN is probably Compaq's best recent effort. Maybe not "good" hardware, but certainly not in the same territory as Presarios, Pavilions, or Emachines.

This message brought to you by the Society for Prevention of Vomit-Box Phrase Dilution.

 

[blakerwry]

As long as the deskpro EN isn't stuffed in a corner, I doubt it will over-heat. At work we have a two (used to have 4) of the Deskpro EN's running W2K Pro and they run very reliable during their 24x7 operation. Those machines have been powered on for at least the past year. I've opened them up at one point or another and I believe they are passively cooled as you mentioned Mercutio. If I'm not mistaken it has a single internal 80mm fan that channels air across the CPU. That deskpro EN case is a PITA to open, BTW. It was a bit deperssing to install a Emulex LP8000 fibre channel adapter into this machine since the HBA was worth way more than the entire Deskpro EN. (somewhere around $1300 for the HBA)

It could be the memory even if it passes your tests. I remember an incident many years back with my parent machine where the guy we bought our computer from gave us parity ram instead of EDO. (At the time I didn't know how to tell the difference, nor did I have a internet connection)

Once installed, the machine had one stick of EDO and the other was parity. (This was a cyrix 686 166 MHz) The win 95 box would randomly blue screen constantly, sometimes it was hours, sometimes minutes. It wasn't until we took the RAM to a computer show to trade it in that we found one was EDO and the other was parity. Both sticks worked fine on the tester and even tested fine in the machine, yet it BSOD like crazy.

I'm not saying the same applies to you, but even though the memory looks fine in the machine, it could be a combination of things...

lol, a Cyrix 6x86 with mixed types of RAM and win95... I think I already found 3 problems...

 

[Handruin]

The cyrix ran great once we fixed the ram issue. It almost never BSOD'd after that. There were a few other game compatibility issues, but for the most part it was a good little machine.

 

[blakerwry]

I just have to give cyrix a hard time... their 586/686 CPU's were known for heat issues... stability issues.. and compatibility issues...

I'm sure if you ran them on a 66mHz bus (some were speced to run 75 and maybe even 83 mHz bus) and used an adequate heatsink then all you had left were a few incompatibilites. :mrgrn:

 

[Mercutio]

My experience with Cyrix is much like the one I've had with Nividia. Tannin and I have a brief arguement about it several times a year, but in my experience they were... not good. Opposite of good, in fact. I think that's called "bad" or maybe "evil". Either way.

 

[time]

Sorry, but I would have to vehemently disagree. I found the Cyrix solutions we supplied to be very stable. Only the early 6x86 versions had software incompatibility issues worth noting, and not enough to stop them being effective for most situations.

In fact, we used one as a co-hosted web server for a year or so, running NT3.51 (SP5). Zero problems. Of course, good motherboards made a big difference in those days.

The heat issue was largely BS. Yes, the 6x86 version did put out more heat than the smaller process Pentium, but they also remained stable at far higher temperatures - check the specs. I've had much bigger problems with a Pentium 120 - the onboard sensor typically read >50C with the ridiculous little Intel cooler, and the machine would inevitably become unstable over 56C. I solved the problem by fitting a Cyrix CPU.

Of course, the third party 50mm fans had a nasty habit of dying, thereby making the PC unstable. It used to be the first thing I'd check. But the only model with any appreciable heat issues was the original 6x86/200 (166MHz), which was of course the fastest desktop CPU available at the time. Once the L and MX versions became available, heat was less important. And from about the M2 onwards, the Cyrix chips were putting out less heat than their Intel equivalents.

It's also worth noting that with a good HX mainboard (and 512kB cache), the early Cyrix CPUs performed well in excess of their PR rating (unlike the M2, and unless you had a floating-point-intensive app).

 

[Tannin]

What Time said.

Only I'd substitute the 83MHz bus ones. We had a great run with the 200s, even the early ones so long as you didn't buy some cheap crap motherboard, but some trouble in the really early days with the 120s and 133s before there were any properly compatible motherboards, and near the end some trouble with the IBM 83MHz ones ... er ... PR266 they were, and not actually Cyrix, all IBM. (This was towards the end of the Cyrix/IBM marriage, when the two companies had started releasing seperate products.)

After them came the 300s, and although they didn't even come close to their PR ratings (i.e., no faster than a PR266) they were dirt cheap and as solid as a rock. Last of the good Cyrix CPUs. The 333s were a last-gasp of the old architecture sort of thing, and anyway the K6-2 was hitting its straps by that time, so it was bye-bye Cyrix.

Pity about the National Semiconductor idiots: they ruined a great little company.

The secret to the 686 was always using the right motherboard. FIC had some excellent ones about that time. By the way, Time, I have some interesting benchmark figures around here somewhere, involving a 3 x 3 matrix of CPUs and chipsets: Intel, Cyrix and AMD 166s, with HX, SiS and VIA chipsets. (I used a few different boards too, but the results with different boards with the same chipset were much the same as eaxh other.) Essentially, the HX chipset hated Cyrix chips and ran very slow. The VIA hated Intel chips and ran very slow. The AMD CPU didn't really care which chipset it had, and the SiS 5571 didn't care which CPU you gave it. The 686MX-166 was the fastest of the lot by quite a margin (faster than a P-200 MMX & HX combination!) but only with the FIC VA502 (VIA) or Gigabyte 586S (SiS 5571) boards. The slowest of the lot (if my memory is to be trusted) was also the 686MX-166, but on the HX board. And yet, with an Intel or AMD CPU, the HX really hummed along.

Weird stuff. Anyway, I guess that's where I picked up the habit which I still have: if I'm going Intel, I go all the way: CPU & chipset both. If I'm not going Intel, I try not to use any Intel components: all Intel or none is my rule.

Mind you, I break it now and then, mostly for crappy Celerons where I need a cheap board to move them out. Then I buy anything, and don't care as long as it runs.

As for trouble with a 5x86-100? Huh? Those things ran like trains! Same as all the 486ish chips, actually, bar the ones that tried to push the bus speed places it was never meant to go: Intel DX-50 (the worst of them all), AMD DX-40 and DX/2-80 (but the 120 was OK) and Cyrix 120. Anything over 33MHz on a 486 board nearly always spelled trouble.

 

[Buck]

I had good success with the PR200s too.

 

[blakerwry]

I ahd good success with my 486 dx2 66mHz Cyrix, but I have heard so many problems about the 5x86 and 6x86 that I enver even want to try... I have heard the m2 and above remedied the problem, although perform sub par compared to an intel or AMD solution.

 

[James]

Doubt it's a heat issue too - I have a Deskpro EN SFF even which doesn't overheat with 2x WD 1200BBs and an aircooled P2 350 all running 24/7 in it.

 

[Newtun]

Okay, I've been off in a corner drooling with rapture about the raptor.

But I did come up with a possible solution to my diagnostic problem, namely, using Linux on a CD to run Prime95. If that works for an extended period, it probably narrows it down to the OS install.

I found such a beast, Knoppix, downloaded the ISO and burned it on my non-work-PC. Then I booted the work PC into Linux/KDE and played around with it for a while, then started Prime95. We'll see how it goes - it doesn't use a lot of memory (.2%), but is working the CPU over 99%.

Pretty interesting - you can get a taste of Linux without disturbing your hard drive. I'll have to fool around a little more with it. I'm somewhat used to Unix (HP and Sun), but it's still a little foreign - bash instead of the ksh I'm used to.

 

[Newtun]

Okay, Prime95 ran for 7 hours on Knoppix, seemingly without a problem. I'll try a longer test, but I think it's pretty conclusive.

 

[Tea]

I have heard so many problems about the 5x86 and 6x86 that I enver even want to try... I have heard the m2 and above remedied the problem, although perform sub par compared to an intel or AMD solution.

You heard wrong, Blake. Big-time wrong.

The 5x86-100 was as zolid a performer as you could ask for. The very rare 5x86-120, like all 40MHz 486-zocket products, required care.

The 6x86 Classic was a handful in the very early days - not as bad as the Pentium 60 but close - for about two, maybe three months until three things happened (a) Motherboard manufacturers got their act together (new chips are always hard for motherboard makers), (b) techies learned the ins and outs of them - getting the voltage right was Real Hard Stuff in those days, I mean, you actually had to move a jumper on the motherboard and to make it worse they would actually run on the wrong voltage so you didn't have that handy blank screen that was the normal way of telling f you had the jumpers wrong - and (c) Cyrix/IBM did a core shrink which made them much more forgiving.

The 6x86-200 Classic was different. It was the very first chip ever to run a 75MHz bus speed, and there were only two or three motherboards that were factory-certified to go that fast. The FIC PA-2005 was the pick of them: a superb board. A lot of people tried to overclock their older-model boards using uncertified chipsets (especially the Intel VX) and came to grief because of it.

The 6x86MX was late to market, but when it finally arrived it blew the Pentium MMX out of the water - the 200 was visibly faster than Intel's 233 - and comfortably outperformed the K6 Classic as well.

The M-II was exactly the same chip under a new name. I mentioned IBM's ridiculously optimistic 6x86MX-266 with the 83MHz bus - running that big a board overclock just wasn't on as a production product - and trust me, I used to know a bit about boards in those days, I was a hands-on man. Not like now where I sit in an office and count the money.

(Tannin! Get the hell out of my post! Here I am enjoying myself on Ztorage Forum and I look away for a minute and there you are pretending to be me! Stop it!)

(Sorry Tea. I just got a bit bored waiting for you to get your paw out of your mouth and type something.)

Anyway, as Tannin was zaying, he even uzed the exact same boards that IBM certified the 83MHz 266 on. He got them all going reliably in the end, but they were alwayz a struggle.

The 300 was as solid a chip as you could hope to lay your paws on, though uniquely among the PR-rated Cyrix chipz, it didn't match up to its PR number. The K6-2/300 came out about that time, and that was that for Cyrix. Once the K6-2 hit, nothing else was in the same street for performance. We forget just how good the K6-2/300 was. (And NRG can ztick his zlimy old Zluggerons up his arze - K6 ruLeZ!!!!)

(Tea! Settle down you fool!)

(Zorry. I got a bit carried away there.)

(Are you going to apologise to NRG?)

(Nope.)

(Tea, if you don't start behaving like a normal, sensible ape, I'm going to call Mercutio and have him come over and help you with your browser problems.)

(Eeeek! NO! 17 BILLION POP-UPZ! Not that again!)

I'm zorry, NRG.

(That's better.)

(Anyway, I didn't mean "arze". NRG's not from Australia, he wouldn't know what one of those waz. I meant to zay "azz")

(Right. That's it! Bedtime for you. And no banana!)

(Awwww.)

 

[NRG = mc²]

(And NRG can ztick his zlimy old Zluggerons up his arze - K6 ruLeZ!!!!)

No thanks. (though with the Slot 1 version one might manage, but the PPGA versions would be painful I think )

Whatever you say, massively overclocked Celerons rocked :mrgrn:

 

[The Grammar Police]

Of course, one would wish to remove all that mounting hardware from the Slot One product first, my friend. On the ... er ... whole ... it might be best to chose a more modest CPU, such as one of those nice little socketed 386 parts.

 

[Mercutio]

Personally I think a P4 with a tricked-out heat sink would be the worst. Or maybe one of those Zalman sharkfin jobs that has the 120mm fan on the end. Ouch!

 

[Tea]

I waz zort of expecting you to be in this thread to announce that you were ztepping in to moderate it, Mercutio. In a way, I'm zort of zorry you didn't.

 

[Mercutio]

Well, as long as we don't find a need for inline images for all this.

 

[Jan Kivar]

The most badass wink HSF would be one in the newer Shuttle SFF's with the heat pipes...

 

[blakerwry]

I agree... those things are great.

I suppose someone could make one with bendable arms between the CPU and the radiator part so that you could bend it to fit in almost any case... that *would* be badass.

 

[NRG = mc²]

What I never understood is why they don't make heatpipes like that which will utilise the entire case for dissipation. It would not be easy on a normal case for several reasons, the key reason being that each side panel etc is removable and separate from the rest, but certainly a suitable case can be made....

 

[Jan Kivar]

I agree... those things are great.

I suppose someone could make one with bendable arms between the CPU and the radiator part so that you could bend it to fit in almost any case... that *would* be badass.

Blake, that HSF suggestion was a reply to Merc's & NRG = mc²'s posts...