TheAbandonwareGuy wrote on 2021-01-02, 12:37:

Could I get an idiot-check please? […]
Show full quote

Could I get an idiot-check please?

I was going through a haul of computers I got and among them was an Barton Athlon XP 2800+ powered Compaq. I pulled the CPU, burned black over the back of the processor, burned brown around the front of the core (tan colored CPU for ref). Figured it was the CPU so I threw my only other Barton 2800 on it and turned it on with the heatsink off and it immediately blew up the 2nd CPU. Smoke column within literally 3 seconds of the power button coming from the CPU core. It actually appears to have blown open what looks like traces on the top of the die. I usually do this and pull the power as soon as I see a BIOs screen or after 5 seconds.

The board and PSU (since I'm not sure whos fault this is) are both going in the garbage, but do Bartons burn up that quick without a heatsink? This is from 100 percent cold, and the CPU was pretty much cool to the touch by the time I had it pulled 30 seconds later. or is this motherboard just a murderer? I wanted to double check what the safe time is on AXPs without a heatsink for future trouble shooting purposes. I know I've heard that these are prone to thermal issues and easy to burn out but this seems excessive even for that.

Also is it even worth trying a CPU that has smoked out on another board? I'm thinking that if it failed violently enough to smoke within 3 seconds that its surely dead, and might even damage my other 462 boards (which I dont really have many of)

Edit: https://www.youtube.com/watch?v=vQG1Lr1DmZk 200f/93c degrees within 5 seconds, but no smoke for him at least. Hit 500f within 15 seconds though.

Imho + memory_from_the_times, the basic answer is Yes, running those "naked die" (<-- that perhaps being one of the key contributors in thermal terms, I here speculate) Durons/Athlons/AthlonXP's sans-heatsink was a recipe for quick, expensive disaster.

Whether it takes 3 seconds or 15 seconds is probably just a variable of the components.

Working on my Packard Bell 486 tower, which has to be run "topless" due to missing parts...first it was giving me screens of random junk characters when booted and now I just get a completely white screen. I have no clue what's wrong with this thing.

EvieSigma wrote on 2021-01-03, 04:07:

Working on my Packard Bell 486 tower, which has to be run "topless" due to missing parts...first it was giving me screens of ran […]
Show full quote

Working on my Packard Bell 486 tower, which has to be run "topless" due to missing parts...first it was giving me screens of random junk characters when booted and now I just get a completely white screen. I have no clue what's wrong with this thing.

Most likely the built in video has gone bad. Which P. Bell model is it ?
If it has everything on board and uses a riser card for the ISA slots like my old P.Bell then fixing it is not worth the effort in the long run.. just my opinion
Please do post picture of the motherboard

Shreddoc wrote on 2021-01-03, 03:34:

TheAbandonwareGuy wrote on 2021-01-02, 12:37:

Could I get an idiot-check please? […]
Show full quote

Could I get an idiot-check please?

I was going through a haul of computers I got and among them was an Barton Athlon XP 2800+ powered Compaq. I pulled the CPU, burned black over the back of the processor, burned brown around the front of the core (tan colored CPU for ref). Figured it was the CPU so I threw my only other Barton 2800 on it and turned it on with the heatsink off and it immediately blew up the 2nd CPU. Smoke column within literally 3 seconds of the power button coming from the CPU core. It actually appears to have blown open what looks like traces on the top of the die. I usually do this and pull the power as soon as I see a BIOs screen or after 5 seconds.

The board and PSU (since I'm not sure whos fault this is) are both going in the garbage, but do Bartons burn up that quick without a heatsink? This is from 100 percent cold, and the CPU was pretty much cool to the touch by the time I had it pulled 30 seconds later. or is this motherboard just a murderer? I wanted to double check what the safe time is on AXPs without a heatsink for future trouble shooting purposes. I know I've heard that these are prone to thermal issues and easy to burn out but this seems excessive even for that.

Also is it even worth trying a CPU that has smoked out on another board? I'm thinking that if it failed violently enough to smoke within 3 seconds that its surely dead, and might even damage my other 462 boards (which I dont really have many of)

Edit: https://www.youtube.com/watch?v=vQG1Lr1DmZk 200f/93c degrees within 5 seconds, but no smoke for him at least. Hit 500f within 15 seconds though.

Imho + memory_from_the_times, the basic answer is Yes, running those "naked die" (<-- that perhaps being one of the key contributors in thermal terms, I here speculate) Durons/Athlons/AthlonXP's sans-heatsink was a recipe for quick, expensive disaster.

Whether it takes 3 seconds or 15 seconds is probably just a variable of the components.

Agreed. I've boiled water on an Athlon (or maybe Duron?) die before. Happened almost instantly. For the record, this was done in a PC repair class in high school back in 2002 or 2003 as an experiment to show why we use heatsinks. A drop of water on the die + power on = instant steam. I believe the board and chip were both bad already, but I'm not 100% sure. After that the CPU was toast, for sure.

I would never power on anything newer than a Pentium 1 without a heatsink. And even a P1 would be a quick test... Though I have never risked even doing this personally.

I am fine powering on a 486DX 33 without a heatsink, but a DX2 or DX4 or a Pentium, i usually use a heatsink.

When you think about it (/No More Jockeys), the fact a heatsink makes the difference between Go and Blow in the space of <5 seconds is quite an impressive testament to their efficacy in those scenarios.

That is, they really take the heat fast.

Shreddoc wrote on 2021-01-03, 13:55:

When you think about it (/No More Jockeys), the fact a heatsink makes the difference between Go and Blow in the space of <5 seconds is quite an impressive testament to their efficacy in those scenarios.

That is, they really take the heat fast.

Athlon = MUST HAVE H/S ON IF NOT DO NOT TURN ON, put "True" at "value" string.

Yeah, that's true. Lesson learned. I'll have to buy some Athlon XP 2400+ and mod it to SMP capable, then put into this motherboard with proper H/S.

I've soldered in an ISA slot to the Asus CUBX motherboard. There were pads on the board but my version doesn't had the slot.

Removing the solder from pads with a solder wick was the most time consuming part.

I need to compare it with photos from the versions with ISA factory mounted, to check if I need to solder some other elements. Then I'll test the solution in a few next days.

Horun wrote on 2021-01-03, 06:06:

Most likely the built in video has gone bad. Which P. Bell model is it ? If it has everything on board and uses a riser card fo […]
Show full quote

EvieSigma wrote on 2021-01-03, 04:07:

Working on my Packard Bell 486 tower, which has to be run "topless" due to missing parts...first it was giving me screens of ran […]
Show full quote

Working on my Packard Bell 486 tower, which has to be run "topless" due to missing parts...first it was giving me screens of random junk characters when booted and now I just get a completely white screen. I have no clue what's wrong with this thing.

Most likely the built in video has gone bad. Which P. Bell model is it ?
If it has everything on board and uses a riser card for the ISA slots like my old P.Bell then fixing it is not worth the effort in the long run.. just my opinion
Please do post picture of the motherboard

I'm actually missing the riser card for this system, along with a few other parts, which means I can't even try using a dedicated video card.

EvieSigma wrote on 2021-01-03, 17:22:

Horun wrote on 2021-01-03, 06:06:

Most likely the built in video has gone bad. Which P. Bell model is it ?
If it has everything on board and uses a riser card for the ISA slots like my old P.Bell then fixing it is not worth the effort in the long run.. just my opinion
Please do post picture of the motherboard

I'm actually missing the riser card for this system, along with a few other parts, which means I can't even try using a dedicated video card.

Ahh. What model is it ?
My P.Bell is a Pack Mate 733 (5v only 486DX) that uses a straight wired ISA only riser and can fit a small ISA VGA into the riser slot if I remove the back tang off the ISA VGA card.
If yours uses a PCI+ISA riser then you cannot do that with out risk of ruining the board more than it is. Mine has lots of corroded traces around the onboard memory and IDE connectors (gives mem errors and no HD detected) plus has no L2 cache so is not really worth trying to fix but
yours might just be a video ram issue fixable by a dedicated vga card if you can figure out if yours uses a straight wired riser.....

Horun wrote on 2021-01-03, 17:36:

Ahh. What model is it ? My P.Bell is a Pack Mate 733 (5v only 486DX) that uses a straight wired ISA only riser and can fit a sma […]
Show full quote

EvieSigma wrote on 2021-01-03, 17:22:

Horun wrote on 2021-01-03, 06:06:

Most likely the built in video has gone bad. Which P. Bell model is it ?
If it has everything on board and uses a riser card for the ISA slots like my old P.Bell then fixing it is not worth the effort in the long run.. just my opinion
Please do post picture of the motherboard

I'm actually missing the riser card for this system, along with a few other parts, which means I can't even try using a dedicated video card.

Ahh. What model is it ?
My P.Bell is a Pack Mate 733 (5v only 486DX) that uses a straight wired ISA only riser and can fit a small ISA VGA into the riser slot if I remove the back tang off the ISA VGA card.
If yours uses a PCI+ISA riser then you cannot do that with out risk of ruining the board more than it is. Mine has lots of corroded traces around the onboard memory and IDE connectors (gives mem errors and no HD detected) plus has no L2 cache so is not really worth trying to fix but
yours might just be a video ram issue fixable by a dedicated vga card if you can figure out if yours uses a straight wired riser.....

It's a Force 2133. All ISA, 486DX. The factory riser card is a 5(?) slot all ISA riser but so far the only one I've found for sale online was a 4 slot from a different Packard Bell.

I eventually managed to get the 2600Jr to work, replaced its on/off switch and did the composite mod. That's two dead machines rescued! 😜 Since the output quality was much cleaner than RF, I returned to the six-switch one to do the corresponding AV mod with the 2N3904 transistor. Now to service the joysticks and clean those paddle controllers!

Today I did some slight modifications to the BASIC code of a game for the Commodore VIC-20. Nothing spectacular, just simple tweaks to the old 1982 quiz game BBC Mastermind - fixing a glitch when you get a question correct and the "Correct" appears jumbled over the answer. I simply entered the down cursor function to that particular line in the code, to move it onto the next line instead.

A big problem the game has is repetition of questions, as the rounds are only 2 minutes long, with the program ending after the 2nd round and of course it tells you the answers to the ones you get wrong so you can memorise them. So I did experiment with altering the code so that it wouldn't tell you the answer unless you deliberately passed on the question - passes of course count against you in the case of a points tie. Whilst the modification was simple enough & succesful, I decided not to keep it as the game wouldn't really be faithful to the TV show "Mastermind" anymore, where the host always tells you the correct answer.

I guess to avoid repetition, I'll just have to create multiple .d64 versions of the game with different question packs. 🤣

I've been messing around with my newly restored Supermicro P54VL-PCI board:

The board is using the Opti "Premium" aka Opti Cobra chipset. Seems like it's an earlier Opti Pentium/VLB chipset with Opti Python being released a year later. I'm yet to see any info on which one of them is faster, but it does seem like Python allows setting different VLB speeds while Cobra limits VLB to 33 MHz.

This board supports up to 512KB L2 cache and takes Pentium 90 and Pentium 100. My current setup is:

Pentium 100
64MB EDO RAM
512KB cache
Tekram DC-680C cached IDE controller with 4MB cache and a 2.5GB Fujitsu HDD
Ark Logic ARK1000VL VLB video card with 2MB RAM
Voodoo 1 from Diamond
Turtle Beach Tropez Plus

Windows 95 RTM

The only thorough review of a Pentium VLB system I've found so far seems to be from WaybackTech (link).

A few differences:

- his board uses Python chipset, but he does not mention any settings for VLB
- his board allows tweaking cache/RAM timings, while my board wouldn't boot into Windows if I did that or tried to enable CPU Address Pipeline Mode. Essentially any tweaks I do in the BIOS at the moment make my system unstable
- his board does not POST with cache switched to write-back mode. Mine does, and actually offers two "adaptive write back" modes as well, not sure what they are
- I get almost two times the memory bandwidth (79.71 MB/s vs 45 MB/s)

Currently it runs Quake at 20.7 FPS (timedemo demo1) and at 22.1 FPS with Voodoo (640x480). Weirdly enough it feels very playable, but it does stutter a bit whenever there are explosions on the screen or other effects that generate a lot of particles.

Descent 2 framerate really jumps from 40 something to 27FPS but still feels very playable (the Voodoo version is almost unplayable). Duke Nukem 3D doesn't really go below 30FPS, the actual framerate of course depends on the scene (outdoors/indoors, etc). Doom timedemo gives 1682 realtics (44FPS).

So far, the system gives a pretty nice experience in most games of the era, but it does feel like Pentium 100 performs a lot worse than it should.

Compared to mpe's benchmarks on 430FX I can see that P100 performs on the Supermicro somewhere between P75 and P90! And it only gives a slight edge over AMD 5x86 and only in some cases. No wonder these board weren't too popular! Right now it feels a bit like a 486 board with a Pentium Overdrive (which it kind of is).

I'd like to upgrade the BIOS to the latest version 3.2, but for some reason the amibios upgrade procedure (hold Ctrl+Home before you turn the machine on until it starts to read the ROM file from the floppy) doesn't work. I've decided to finally buy a TL866, so I'll try to flash a new BIOS then. Hope it will allow me to tweak cache/RAM timings.

I will also try to get my hands on a WinChip IDT C6 chip. WaybackTech scored 17.7FPS with it in Quake and 21.4FPS wih P-MMX@166 which kind of makes me hope that my motherboard is less of a bottleneck than his is (I do have almost twice the memory bandwidth after all).

Oh, and I hope to restore my other S5/VLB motherboard using an Opti Python chipset, but I'm yet to figure out all the missing resistors.

jheronimus wrote on 2021-01-04, 04:29:

I've been messing around with my newly restored Supermicro P54VL-PCI board: […]
Show full quote

I've been messing around with my newly restored Supermicro P54VL-PCI board:

The board is using the Opti "Premium" aka Opti Cobra chipset. Seems like it's an earlier Opti Pentium/VLB chipset with Opti Python being released a year later. I'm yet to see any info on which one of them is faster, but it does seem like Python allows setting different VLB speeds while Cobra limits VLB to 33 MHz.

This board supports up to 512KB L2 cache and takes Pentium 90 and Pentium 100. My current setup is:

Pentium 100
64MB EDO RAM
512KB cache
Tekram DC-680C cached IDE controller with 4MB cache and a 2.5GB Fujitsu HDD
Ark Logic ARK1000VL VLB video card with 2MB RAM
Voodoo 1 from Diamond
Turtle Beach Tropez Plus

Windows 95 RTM

The only thorough review of a Pentium VLB system I've found so far seems to be from WaybackTech (link).

A few differences:

- his board uses Python chipset, but he does not mention any settings for VLB
- his board allows tweaking cache/RAM timings, while my board wouldn't boot into Windows if I did that or tried to enable CPU Address Pipeline Mode. Essentially any tweaks I do in the BIOS at the moment make my system unstable
- his board does not POST with cache switched to write-back mode. Mine does, and actually offers two "adaptive write back" modes as well, not sure what they are
- I get almost two times the memory bandwidth (79.71 MB/s vs 45 MB/s)

Currently it runs Quake at 20.7 FPS (timedemo demo1) and at 22.1 FPS with Voodoo (640x480). Weirdly enough it feels almost playable, but it does stutter a bit whenever there are explosions on the screen or other effects that generate a lot of particles.

Descent 2 framerate really jumps from 40 something to 27FPS but still feels very playable (weird enough, the Voodoo version is almost unplayable). Duke Nukem 3D doesn't really go below 30FPS, the actual framerate of course depends on the scene (outdoors/indoors, etc). Doom timedemo gives 1682 realtics (44FPS).

So far, the system gives a pretty nice experience in most games of the era, but it does feel like Pentium 100 performs a lot worse than it should.

Compared to mpe's benchmarks on 430FX I can see that P100 performs on the Supermicro somewhere between P75 and P90! And it only gives a slight edge over AMD 5x86 and only in some cases. No wonder these board weren't too popular! Right now it feels a bit like a 486 board with a Pentium Overdrive (which it kind of is).

I'd like to upgrade the BIOS to the latest version 3.2, but for some reason the amibios upgrade procedure (hold Ctrl+Home before you turn the machine on until it starts to read the ROM file from the floppy) doesn't work. I've decided to finally buy a TL866, so I'll try to flash a new BIOS then. Hope it will allow me to tweak cache/RAM timings.

I will also try to get my hands on a WinChip IDT C6 chip. WaybackTech scored 17.7FPS with it in Quake and 21.4FPS wih P-MMX@166 which kind of makes me hope that my motherboard is less of a bottleneck than his is (I do have almost twice the memory bandwidth after all).

Oh, and I hope to restore my other S5/VLB motherboard using an Opti Python chipset, but I'm yet to figure out all the missing resistors.

Awesome tests and benches. I've benched alot of Socket 5 and 7 motherboards with a Pentium 100 here: Socket 5 & 7 Motherboard VGA Benchmark comparison (so you can compare).

I also have a Socket 5 VLB, planning on building a whole system around it with my Creative 3D Blaster VLB in the center. I have the TMC PAT54PV board, which is a bit special since it is a Socket 5 board, but without PCI!

jheronimus wrote on 2021-01-04, 04:29:

Compared to mpe's benchmarks on 430FX I can see that P100 performs on the Supermicro somewhere between P75 and P90! And it only gives a slight edge over AMD 5x86 and only in some cases. No wonder these board weren't too popular! Right now it feels a bit like a 486 board with a Pentium Overdrive (which it kind of is).

Not really surprising as this board was originally designed for Socket 4 (Pentium 60/66).

Shreddoc wrote on 2021-01-02, 11:51:

Not intentionally!, just part of the universe's random serendipity I guess. […]
Show full quote

Ozzuneoj wrote on 2021-01-02, 05:09:

Is that where you got your username? Shredding documents? Honest question.

Not intentionally!, just part of the universe's random serendipity I guess.

I certainly don't define my life by that long-ago deletion of material, and even if I did, naming myself for a personal mistake would be like Bill Clinton giving himself the nickname Cigar Lewinski. 😉

Hey y'all! I'm Bill, but y'all can jist go ahead and call me Cigar!

🤣

Replaced the mobile rack in Pentium MMX with a period correct one.
Ended up having to replace the infamous JST hard drive . The damn thing is just too long. It only fits in the rack if I remove the front. The new one is a 3GB Quantum Fireball EX I was using in my AMD K5.
Later I'll probably just replace it with a compact flash. I ordered a couple with IDE adaptors and they should arrive this week or the next. Though I'd really like a period correct drive in this one.

vetz wrote on 2021-01-04, 04:34:

Awesome tests and benches. I've benched alot of Socket 5 and 7 motherboards with a Pentium 100 here: Socket 5 & 7 Motherboard VG […]
Show full quote

Awesome tests and benches. I've benched alot of Socket 5 and 7 motherboards with a Pentium 100 here: Socket 5 & 7 Motherboard VGA Benchmark comparison (so you can compare).

I also have a Socket 5 VLB, planning on building a whole system around it with my Creative 3D Blaster VLB in the center. I have the TMC PAT54PV board, which is a bit special since it is a Socket 5 board, but without PCI!

Thanks! Finally picked up a TL866II+ programmer today. Awesome buy — I've tried a similar device before and had issues, but this time everything worked on first attempt using minipro CLI tool under OS X.

Upgraded the BIOS to the latest 3.2 version. There are now a lot more options in the BIOS, especially when it comes to cache and memory timings. Also VLB waitstates are now selectable (ows/1ws). The CPU Address Pipeline Mode option is now gone (the one that didn't work for me before).

So far I've only enabled hidden memory refresh and shadowed everything in memory. Already got around 0.8FPS in software-rendered Quake under DOS (for a total of 21.5), but glQuake is now slower (20.9 instead of 22.1fps). Memory bandwidth went from 75MB/s to 80MB/s. Tried just cranking up all the timings down but couldn't boot (that actually worked for me on an Opti895-based 486 board), so I'll have to tweak timings gradually.

There is one section in the BIOS that confuses me:

DRAM Region A Control Mode: Disabled | Cached | NonCached | Hole
DRAM Region A Size: 512KB | 1MB | 2MB | 4MB
DRAM Region A Base Address
DRAM Region B Control Mode: Disabled | Cached | NonCached | Hole
DRAM Region B Size: 512KB | 1MB | 2MB | 4MB
DRAM Region B Base Address

Never seen anything like this. I imagine I should try caching everything (I do have 64 megs available), but I don't understand which base address should I specify.