Socket 4 is 5V Pentium P5 though, definitely not 486.

pshipkov wrote on 2021-12-28, 01:35:

@chadti99
I also noticed that Voodoo3 gets us +0.1 fps at 200mhz compared to MGA, but still a bit slower in Doom for example.
What ram you are using to be a problem for 200mhz.

I’ll need to test an MGA card for sure. My 66x3 isn’t as stable as 50x4. If I had this module installed 66x3 would boot but I couldn’t complete a Quake timedemo. It works great at 60x3 and I believe it’s one you had mentioned works well in your setups. I’ve swapped it for a 32MB EDO 60ns module for now.

Any recommendations on fighting condensation? If the humidity would drop here, like it should in winter, would be a huge help 🤣.

Either coat everything in vaseline so the condensation doesn't get to the conductors or do total air control, where you run it in a box fed by an air-con unit, dehumidifier, or pass the intake air through a bed of calcium chloride (The expensive concrete safe ice melter salt) which is sitting on a grid over a drip pan to catch the water. (Which is now a calcium chloride solution if you wanna boil off the water and recover the CaCl salt later)

12v Peltier will need EXTENSIVE air/humidity control.
We can go there eventually but let's leave it for 2022.
As a first step we need to determine the highest temperature past which the cpu gets unstable.
From what I see this is around 5C, give or take a few based on the quality of the die.
If we find magical processor that can push the temp as high as possible, this will lower the air/humidity control requirements and make it more achievable.
On top of that you will need at least 12x12cm heatsink with 3 (ideally 4) rods and fan. Not an easy thing to fasten to socket3 components that don't have the appropriate connection points.
I tried few times on open bench with big fans blowing at the cpu/peltier/heatsink with practically no improvement in reducing water condensation. Vaseline or other vaxes/silicone are not enough.
Don't know what dehumidifier solution may look like and how it can be implemented in a PC case.

Do you have references for that calcium chloride thing ?
I can check online, but right now don't have an idea what really to search for.

5v Peltier with appropriate heatsink/fan are enough for stable 180mhz while avoiding condensation.
This is the best I succeeded with so far. It was very hard to transform the loose test bench contraption into a real PC.

---

Chadti99, I think roughly 50% of these modules don't cut it for proper 180/200 mhz.
So far only 2 of them can do 200mhz here.
Still, they seem to be one of the better models for the purpose.

Hmm there's not a lot of especially clear explanations or construction details, see this for basics... https://commons.wikimedia.org/wiki/File:Dryer … scent_Dryer.jpg try search terms of "dryer tube" "deliquescent air dryer"

pshipkov wrote on 2021-12-28, 04:22:

12v Peltier will need EXTENSIVE air/humidity control. We can go there eventually but let's leave it for 2022. As a first step we […]
Show full quote

12v Peltier will need EXTENSIVE air/humidity control.
We can go there eventually but let's leave it for 2022.
As a first step we need to determine the highest temperature past which the cpu gets unstable.
From what I see this is around 5C, give or take a few based on the quality of the die.
If we find magical processor that can push the temp as high as possible, this will lower the air/humidity control requirements and make it more achievable.
On top of that you will need at least 12x12cm heatsink with 3 (ideally 4) rods and fan. Not an easy thing to fasten to socket3 components that don't have the appropriate connection points.
I tried few times on open bench with big fans blowing at the cpu/peltier/heatsink with practically no improvement in reducing water condensation. Vaseline or other vaxes/silicone are not enough.
Don't know what dehumidifier solution may look like and how it can be implemented in a PC case.

Do you have references for that calcium chloride thing ?
I can check online, but right now don't have an idea what really to search for.

5v Peltier with appropriate heatsink/fan are enough for stable 180mhz while avoiding condensation.
This is the best I succeeded with so far. It was very hard to transform the loose test bench contraption into a real PC.

---

Chadti99, I think roughly 50% of these modules don't cut it for proper 180/200 mhz.
So far only 2 of them can do 200mhz here.
Still, they seem to be one of the better models for the purpose.

A peltier cold side is proportional to the hot side.

If you use a passive heatsink and allow it to heat up a bit and control it by how much airflow it gets you might be able to keep it above dewpoint on the cpu side.

Thanks @BitWrangler.

@Sphere478
Looked at this stuff quit a bit. Also posted bunch of findings in another thread.
Based on my observations I am inclined to think that water condensation happens because of the close proximity of surfaces with significant temperature difference.
If the CPU needs 0-5C for operational stability, but the heatsink surface 1-2 millimeters away from it is let's say 70C, even with gel on top - water or ice will form.
This needs to be looked at from multiple angles.
Increase CPU surface temp as much as possible.
Use the biggest possible heatsink+fan. Number of rods has big impact. 4 or even more will be best.
Use a thermistor to precisely dial the Peltier voltage to make sure the minimum possible amount is provided (for stable CPU).
Proper case ventilation.
Consider some stuff as what BitWrangler said, or possibly other solutions that we are not aware of yet.
Make sure wide enough ambient temperature range is covered - from Summer day noon to night during Winter.
I am not eager to get on this 😀

pshipkov wrote on 2021-12-28, 09:39:

Thanks @BitWrangler. […]
Show full quote

Thanks @BitWrangler.

@Sphere478
Looked at this stuff quit a bit. Also posted bunch of findings in another thread.
Based on my observations I am inclined to think that water condensation happens because of the close proximity of surfaces with significant temperature difference.
If the CPU needs 0-5C for operational stability, but the heatsink surface 1-2 millimeters away from it is let's say 70C, even with gel on top - water or ice will form.
This needs to be looked at from multiple angles.
Increase CPU surface temp as much as possible.
Use the biggest possible heatsink+fan. Number of rods has big impact. 4 or even more will be best.
Use a thermistor to precisely dial the Peltier voltage to make sure the minimum possible amount is provided (for stable CPU).
Proper case ventilation.
Consider some stuff as what BitWrangler said, or possibly other solutions that we are not aware of yet.
Make sure wide enough ambient temperature range is covered - from Summer day noon to night during Winter.
I am not eager to get on this 😀

There is this logic in construction and they try to separate hot and cold with insulation and that works but your logic is side tracked, ultimately it is dew point that draws moisture. And that varies by the weather. In the desert you can cool a cpu more without condensation more than you can in the rainforest so we circle back to insulation and construction technique of separating the cold from the hot. Not because the proximity, but to keep the humid air from reaching something cold enough to make it dew and to slow down the moment of heat to that cold thing.

But..

If you wanted to you could put the board in a fish tank or something and replace the air with dry air and go wild with whatever temp you wanted. Nitrogen is a cheap dry gas that can be bought easily at welding supply stores.

Just seal it up really good and keep it from getting out

Another dry environment , and you may find this quite handy is the bottom of a sealed chest freezer. Very very little humidity in there.

Speaking of this, I do still have a fun experiment that I made a while back, a chest freezer that goes down to -73f by using propylene r1270 (mapp II torch gas) as refrigerant, a upgraded condenser, and 1/2 hp compressor. Lol

Mmmm negative Celsius 🥶

I don’t see why I couldn’t put a computer in there.🤷‍♂️ It wouldn’t reach -73 with a active heat source but I would imagine -40 would be achievable.

BitWrangler wrote on 2021-12-28, 03:34:

Either coat everything in vaseline so the condensation doesn't get to the conductors or do total air control, where you run it in a box fed by an air-con unit, dehumidifier, or pass the intake air through a bed of calcium chloride (The expensive concrete safe ice melter salt) which is sitting on a grid over a drip pan to catch the water. (Which is now a calcium chloride solution if you wanna boil off the water and recover the CaCl salt later)

A slow flow of compressed nitrogen into the case before and during operation will stop any condensation (it’s dry / zero humidity)
A bottle full from welding supplies isn’t super expensive but the security deposit can be steep.

Not recommended in an enclosed area need lots of fresh air coming in.

I have one of those old CPU coolers that uses an A/C compressor and the evaporator is literally the cold block for the CPU. It needs to be recharged, but when I got it , at the price I got it for I couldn't resist. I had wanted one as soon as I saw the reviews when they first came out.

As far as condensation issues, that could mostly be solved by having the computer in a box with a good dehumidifier drying out the fresh air coming into the box.

Now y'all know why Linus of Tech Tips fame gets excited about sticking things in chilled mineral oil 🤣

There's also flourinert if you wanna stick the whole board in the refrigerant.

Boring old compressed air comes out of the top half of tank fairly dry, it's dropping all the moisture in the bottom if it had any. So you could have large tank compressors running, one you're using, one you're draining water out of the tank of, and one that's done draining and is filling up the tank again, but unless you have an industrial electrical circuit you're probably only gonna get 5 cfm or so.

Sonic booms can cause water to condense out of the air, and wings at high angles of attack making larger pressure differences, so maybe just need to connect a 12V fan to a few hundred volts to get 50,000 RPM out of it 🤣

@Sphere478
We are talking about the same thing.
Close proximity of these hot/cold surfaces causes rapid heat transfer through the surrounding air and we get to the dew point quickly.
The Pacific ocean is next door, so air humidity is high here.

We either fill the gaps and surrounding areas with some low thermal conductivity material or have to dehumidify the air inside the PC case.
I am inclined to try the first option after taking all other measures i mentioned above - good CPU, minimum voltage to Peltier, high-quality big-butt heatsink+fan, etc.
Option 2 may end up looking like this ...

... which is not cool for personal computer.

@cyclone3d
Can you share more info/pics please ?

BitWrangler
Didn't think back in Uni that one day will get interested in the "thermal phenomena" subject because of some retro computing hobby.

Here’s what the AMD 5x86 can do at 200MHz (66x3) with faster cache timings on the LS486e Rev. D. Cache timings set to 3-1-2 with 8ns mod. Phils Dosbench Quake timedemo option C. Voodoo 3 2000 PCI. 32MB EDO 60ns. 3.75v to CPU. TEC-12715 Peltier running at 7v. Not Windows stable but maybe it could be with differing cpu/peltier voltages, different ram.

Very nice! Congratulations, looks like the fastest Socket3 486 Quake1 machine 😀
What is "8ns mod"?
Can You run hi-detail Doom bench from Phils Dosbench?

lukas12p wrote on 2022-01-02, 20:41:

Very nice! Congratulations, looks like the fastest Socket3 486 Quake1 machine 😀
What is "8ns mod"?
Can You run hi-detail Doom bench from Phils Dosbench?

Haven’t improved on CPUGalaxy’s Doom score but here’s what I got on all the tests. I’ll try swapping in a Voodoo Banshee later to match what he was using.

The 8ns mod is replacing all the 15ns SMD cache chips and tag with 8ns parts.

14.3 in PCP 480 is pretty nice.

Belarusian attempt to extremely overclock AM5x86: https://overclockers.ru/lab/show/113190/retro … stryj-486-j-cpu

The Serpent Rider wrote on 2022-01-15, 16:05:

Belarusian attempt to extremely overclock AM5x86: https://overclockers.ru/lab/show/113190/retro … stryj-486-j-cpu

Good read, if understand correctly he hit the silicon lottery and found an AM5x86 that runs at 200Mhz under air cooling? It would be interesting to see his cache and memory timings. I think he could improve his scores if he dropped back to 256k cache.

Would have been really cool to see something at 240.

I think he will try to find right CPU for the job, so probably check his new articles every so often. His current results are slightly lower from what was achieved here on vogons. Probably not aware of stable working PCI/FSB 1:1 trick though.

@chadti99
I was able to find a memory module that allows DRAM SPEED = FASTEST for the LuckyStar D motherboard.
Updated the related post with bunch of screenthots and summary in the middle.
Some of the improvements:
+ 0.4 fps in Wolf3D (148.7 fps)
+ 0.2 fps in PC Player Benchmark (29.8 fps)
+ 0.1 fps in Quake 1 (20.5 fps -> so far the highest score at 180MHz)
+ ~1350 in WintTune2 accelerated Windows GUI (17791 kpixels/sec)

With this the system is kind of maxed-out at 180MHz using the built-in tools - jumpers/bios.
From here on Tweak or other hacks will be needed to eventually climb higher.
Another option is to find CL-5480 video card that works at 180MHz - that will bump-up the Quake 1 score at least.

At 200MHz (3x66) this memory stick can potentially allow either 3-1-2 and DRAM SPEED = FASTES or if lucky 2-1-2 with FASTER.
The system booted into DOS prompt at 3-1-2 FASTEST early today but my focus was 180MHz.
Will know more tonight or tomorrow.

Btw, i am seeing Voodoo3 falling behind a tiny bit but consistently in Doom test compared to MGA.
Do you observe the same ?

---

About the overclockers.ru link:
Poor chose of motherboard in my opinion. Wonder why author picked that one ?
Biostar UUD is better assembly than Shuttle HOT-433 and climbs higher. Then UM8881 does not scale as well as SiS496/7 for overclocking.
And as Chadti99 said - getting 1024Kb L2 cache running stably at 3x66 with tight BIOS timings will hurt you more than help. Possible in theory of course.
I think his intention was to max-out the system with RAM and L2 Cache than break some records.
Which is a great angle.