Possible that we do have it mixed up, there is no pin numbering on the silkscreen and it's an undocumented setting anyway.
I'm fairly sure that the position closest to the card slot brackets is 1/2 and the position closer to the front of the case is 1:1.

What about your video card? Many of them do bus master, the Matrox Millenium that pshipkov and myself use notably does not.

chrismeyer6 wrote on 2021-08-02, 17:48:

Is it possible to get a water block to mount to the CPUs instead of using Peltier cooler and worrying about the ice build up?

Water cooling system cannot go below ambient (room) temperature, assuming no other active cooling involved.
Good 486 AMD X5 CPU die can be stable at 180MHz around the 15C - give or take few. This implies active cooling of some sort.

pshipkov,

32 MB RAM (19.2 fps) yields higher Quake results compared to 64 MB (19.6 fps).

Replace "higher" with "lower" in that sentence. I get 19.6 fps with a 64 MB module and 19.2 fps with a 32 MB module.

I've seen an AMD X5 run at 50x4 with just air cooling in DOS and Windows. Apparently passed all the 486 benchmarks in the Ultimate 486 Benchmark Comparison without crashing once. It was at 5 V though, so not ideal. Perhaps 180 and 4V would suffice on that chip.

I am trying to determine if I want to invest the effort in replacing my SRAM. Sounds like I might be able to go from 3-1-3 to 3-1-2.

Attached is my variant of the Rev.D BIOS with PS/2, CPU Burst Write unhidden, and a few defaults altered.

Max, thanks for the file. Appreciate it. I am sure it will come handy.

---

I confirm that AMD X5 boots regardless the bus divider state on these LS boards.
Don't have Cyrix 5x86, so no idea about how things behave with it.

maxtherabbit wrote on 2021-08-02, 21:51:

Possible that we do have it mixed up, there is no pin numbering on the silkscreen and it's an undocumented setting anyway.
I'm fairly sure that the position closest to the card slot brackets is 1/2 and the position closer to the front of the case is 1:1.

What about your video card? Many of them do bus master, the Matrox Millenium that pshipkov and myself use notably does not.

The PCI clock multiplier is not an undocumented setting. It is in the official manual, Chapter 2. Installation, top of the page, above the REMARKS section. For whatever reason, they did not place this on the page with all the other jumper settings. The official layout of the board, in the manual, shows pin 1 being closest to the CPU. Jumper on pin 1-2 is for the 1/2 multiplier; pin 2-3 is for no multiplier.

I have attached the official manual for your reference. Go to PDF page 10.

I have used a few graphics cards during these tests, but when the Promise Ultra100 was installed, I was using a noname S3 Trio64V2/DX because this card is small enough to fit in the first PCI slot. Does this graphics card bus master? There is a jumper installed on the card, JP1 on pins 1-2. The other options is 2-3. I have no idea what this jumper is for. With ISA cards, such jumpers were usually used to set IRQ to 9. ALso to set 0 WS / no 0 WS.

feipoa wrote on 2021-08-02, 21:53:

Replace "higher" with "lower" in that sentence. I get 19.6 fps with a 64 MB module and 19.2 fps with a 32 MB module.

Got it.

feipoa wrote on 2021-08-02, 21:53:

I've seen an AMD X5 run at 50x4 with just air cooling in DOS and Windows. Apparently passed all the 486 benchmarks in the Ultimate 486 Benchmark Comparison without crashing once. It was at 5 V though, so not ideal. Perhaps 180 and 4V would suffice on that chip.

Really lucky die apparently !
I didn't explore the boundaries of the few 180Mhz capable CPUs here on air cooling. Both are ok with 3.3V only.
Seem to be fine with most DOS and Windows things, but doubt they will pass for real. Need to be verified tho.
Going Peltier-less will be a good thing.

feipoa wrote on 2021-08-02, 21:53:

I am trying to determine if I want to invest the effort in replacing my SRAM. Sounds like I might be able to go from 3-1-3 to 3-1-2.

Attached is my variant of the Rev.D BIOS with PS/2, CPU Burst Write unhidden, and a few defaults altered.

LuckyStar_LS486E_rev.D_feipoa_ps2.zip

I am going to do it anyway, so will see what's what.

And thanks for the BIOS.

I saw that in the manual, but pretty sure i had to solder a jumper on the C2 board - just noticed that the photo i provided of the motherboard is before that change.

How do you plan on removing the SRAM chips without destroying them? I have Cypress and Alliance SRAMs of 32kx8 at 10 ns on hand. I also have an 8 ns Elite module on a dead super7 board I can salvage, but not sure how to best remove the chips without being destructive. Ph4nt0m is using double irons, but I only have one proper iron; the other is a desodler iron with a triggered vacuum.

I think using QuikChip would require too much product and a lot of cleanup to salvage the 9 existing SRAM chips. I have one of these hair dryer like heat guns from China that I'd like to use. Any idea how much heat (temp ?) and for how long (time) to bake the IC for? There are two SRAMs with SIMM sockets adjacent, so these sockets may melt, but I have plenty spare SIMMs from dead boards I could reuse. Or I could even desolder the SIMMs to begin with.

The most sure route for me would be to just cut the existing SRAM off with a sharp blade, then vacuum suck the leads off, but this of course destroys the existing SRAM.

In my experience heat gun works really well for this stuff. Safe and clean.

By way of comparison, on the Biostar MB-8433UUD w/Quake 1.06 (IBM 5x86c-133/2x)

256K WT = 19.1 fps
512K WT = 19.5 fps
1024K WT = 19.8 fps

the LS-486E Rev.D gets 19.6 fps with 256K WT. Is the extra 0.5 fps due to the faster timings allowed on the LS486E board due to the EDO? If I set the cache burst read cycle from 1 to 2, the score on the LS486E drops to 19.3 fps.

I am getting 20.1 fps in Quake 1 with 256K WT + 64Mb EDO + MGA + DOS 6.22 or Win95.
Why you are seeing a whole frame less ?!
With 1024K L2 you should be further ahead of my 20.1.
On page 18 i shared results and notes with Biostar UUD. Can you spot a BIOS, or other difference between that and your settings ?

feipoa wrote on 2021-08-02, 21:53:

I've seen an AMD X5 run at 50x4 with just air cooling in DOS and Windows. Apparently passed all the 486 benchmarks in the Ultimate 486 Benchmark Comparison without crashing once. It was at 5 V though, so not ideal. Perhaps 180 and 4V would suffice on that chip.

At the risk of repeating something that may have been mythology in the first place... I remember back in the day it was claimed that stress testing and burnin at higher voltages, near limit of stability, actually caused beneficial electromigration, such that a CPU would then run as fast at a lower voltage, or at least be stable at a clock it almost made before. Then came the "Sudden Northwood Death Syndrome" thing and during that it was said, not that previous experience was wrong, but that the newer CPU processes used opposite doping, or went P-channel to N channel or something and thus big volts now did the opposite. I think that theory mostly had currency during the Celeron A overclocking era. So I guess I'm suggesting there's a chance with older CPU that limited high voltage running, coupled with "impractical" overkill cooling lashups like breathing right from an AC unit or something, or really loud forced air, might show a small benefit when you return to sane volts and cooling.

pshipkov wrote on 2021-08-02, 23:19:

I am getting 20.1 fps in Quake 1 with 256K WT + 64Mb EDO + MGA + DOS 6.22 or Win95. Why you are seeing a whole frame less ?! Wit […]
Show full quote

I am getting 20.1 fps in Quake 1 with 256K WT + 64Mb EDO + MGA + DOS 6.22 or Win95.
Why you are seeing a whole frame less ?!
With 1024K L2 you should be further ahead of my 20.1.
On page 18 i shared results and notes with Biostar UUD. Can you spot a BIOS, or other difference between that and your settings ?

I am quoting numbers for the MB-8433UUD with an IBM 5x86c-133/2x and am comparing it with the LS-486E rev.D with an IBM 5x86c-133/2x.

@bitwrangler
Yeah, this sounds a lot like urban fairy tales. 😀

@feipoa
Ugh, somehow didn't understood your post.
From what I see UUD is clock-to-clock faster than LS D.
Not sure the opposite is happening to you.
Perhaps the boards handle Cyrix 586 differently than they do AMD 486DX5 ?

I have found that when pushing the timings to the limits on these socket 3 boards that the Cyrix 5x86 is even fussier than the AMD X5. I guess I'll try to replace the SRAM modules next. I've pulled 3 dead socket 7 boards to test on. I had wanted to pull their 8 ns SRAMs to use as the TAG on the LS-486E, but unfortunately one of them is 12 ns at 5V; the other is 8 ns, but 3.3 V; and the last is 8 ns, but is of 64Kx8 with 4 extra pins. So looks like I'll be using 10 ns for the TAG as well. I'll experiment with desoldering tomorrow. I also feel that the socket 3 is a bit too close to the SRAM for using the hair dryer hot air gun.

EDIT: I am considering using this technique with one soldering iron, https://www.youtube.com/watch?v=xMXJTLWyBCo , just jump straight to 5:00 as the example there is most fitting for removing the LS-486E's SRAM chips. I didn't see any written details under the video, but looks like he is using solid core copper wire, perhaps in the 20-16 gauge range. The gooey stuff he applies to the IC before heating looks like flux paste, kind of like the flux out of your flux pen, but more viscious. I think I ordered some of this flux paste of similar consistency 2 digikey orders ago. After the IC's are heated with the two parallel copper bard, which T-off his single soldering iron, it looks like he is applying regular leaded solder to the chips to help the leads not making contact with the copper wire to make contact. My main concern is the amount of time the iron is on the leads. As there are cuts in the videos, it may appear shorter than it actually is. I don't think I'd want this kind of heat on there for more than 10 seconds. Unfortunately, the guy doesn't specify how much time was required and if the IC's still work after removal. He also didn't specify his iron's temperature setting. I had the feeling that phantom used this copper wire trick, but used two irons instead of one. I may be able to make a copper wrap for my desolder iron and use two as well, but wouldn't this whole job be easier to just use regular leaded solder, which already has a pretty low melting point and quickly fill both sides of the SRAM with solder, going back and forth, then pull the chip off? I did this with a TSOP before.

I ran some tests with the X5 at 2x66 and 66 MHz to the graphics card only ups the Quake result by 0.1 fps, that is, from 15.7 fps to 15.8 fps.

I tried 50 ns EDO, but it didn't allow any timings faster than my 60 ns EDO when using a 66 MHz FSB.

I had a 35W weller iron that I made a "desoldering spade" for. Unfortunately destroyed the iron some time back when it fell off the bench and shattered the ceramic insulator in the heater part, so it's not around for pics. Anyway, I had a burned up tip (screw in type) so I cut about half off that, then the end with the screw I sawed a slot in, then made a short triangle with a ~1 inch base out of copper sheet, put that in the slot, drilled through both, and clamped it together with a tiny nut and bolt. Had that in a jar of screws so not sure what it was, probably M2 or smaller. Just about worked, could have used another 10W maybe, but did whole rows of pins at a time.

I think my technique is much simpler and less invasive.
Prep the motherboard as shown in the picture below. For the actual job i add one more layer around the plastic parts and tuck the aluminum foil under them where possible.
Put a narrow end on the heat gun, wait for it reach 800C or higher temperature and flash solder/unsolder the chips one by one.
Space the chip-by-chip process to not accumulate too much heat in the PCB.
Also, point the head gun at an angle so it does not face the walls of the surrounding plastic components.

In fact i am much less cautious than what i wrote above and so far it was always fine.

feipoa wrote on 2021-08-03, 11:01:

I ran some tests with the X5 at 2x66 and 66 MHz to the graphics card only ups the Quake result by 0.1 fps, that is, from 15.7 fps to 15.8 fps.

I tried 50 ns EDO, but it didn't allow any timings faster than my 60 ns EDO when using a 66 MHz FSB.

Yeah, Q1's performance at 320x200 is almost entirely CPU bound, assuming a decent VGA used.

Sounds familiar - at 50MHz and higher, most of the issues come from L2 cache, assuming decent CPU and RAM in the system.

pshipkov wrote on 2021-08-03, 15:24:

Put a narrow end on the heat gun, wait for it reach 800C or higher temperature and flash solder/unsolder the chips one by one.

Those "hair dryer" style heat guns don't come with a very narrow end. The smallest circular end cap they have contains a diameter of about 1 cm. Does your hair dryer have a smaller end? If so, could you provide a photo of your nozzle and hair dryer? I don't have a proper hot air rework station.

What I have looks something like this: https://www.amazon.ca/Resistant-Temperature-S … 8026755&sr=8-37 With that smallest tip here being around 1 cm in diameter.

feipoa wrote on 2021-08-03, 11:01:

I have found that when pushing the timings to the limits on these socket 3 boards that the Cyrix 5x86 is even fussier than the A […]
Show full quote

I have found that when pushing the timings to the limits on these socket 3 boards that the Cyrix 5x86 is even fussier than the AMD X5. I guess I'll try to replace the SRAM modules next. I've pulled 3 dead socket 7 boards to test on. I had wanted to pull their 8 ns SRAMs to use as the TAG on the LS-486E, but unfortunately one of them is 12 ns at 5V; the other is 8 ns, but 3.3 V; and the last is 8 ns, but is of 64Kx8 with 4 extra pins. So looks like I'll be using 10 ns for the TAG as well. I'll experiment with desoldering tomorrow. I also feel that the socket 3 is a bit too close to the SRAM for using the hair dryer hot air gun.

EDIT: I am considering using this technique with one soldering iron, https://www.youtube.com/watch?v=xMXJTLWyBCo , just jump straight to 5:00 as the example there is most fitting for removing the LS-486E's SRAM chips. I didn't see any written details under the video, but looks like he is using solid core copper wire, perhaps in the 20-16 gauge range. The gooey stuff he applies to the IC before heating looks like flux paste, kind of like the flux out of your flux pen, but more viscious. I think I ordered some of this flux paste of similar consistency 2 digikey orders ago. After the IC's are heated with the two parallel copper bard, which T-off his single soldering iron, it looks like he is applying regular leaded solder to the chips to help the leads not making contact with the copper wire to make contact. My main concern is the amount of time the iron is on the leads. As there are cuts in the videos, it may appear shorter than it actually is. I don't think I'd want this kind of heat on there for more than 10 seconds. Unfortunately, the guy doesn't specify how much time was required and if the IC's still work after removal. He also didn't specify his iron's temperature setting. I had the feeling that phantom used this copper wire trick, but used two irons instead of one. I may be able to make a copper wrap for my desolder iron and use two as well, but wouldn't this whole job be easier to just use regular leaded solder, which already has a pretty low melting point and quickly fill both sides of the SRAM with solder, going back and forth, then pull the chip off? I did this with a TSOP before.

I ran some tests with the X5 at 2x66 and 66 MHz to the graphics card only ups the Quake result by 0.1 fps, that is, from 15.7 fps to 15.8 fps.

I tried 50 ns EDO, but it didn't allow any timings faster than my 60 ns EDO when using a 66 MHz FSB.

that technique looks absolutely ridiculous

this is how to do it - https://youtu.be/yfFM5lTEOMs?t=251