MikeSG:
I was unable to locate 85 MHz and 90 MHz DIP-14 style crystal oscillator cans on Aliexpres. Could you relay some links?

Sphere:
I also have my own variant of the AnyCLK, which is smaller than the original Migron unit, but I prefer the ability to type in exactly the clock frequency I want rather than being stuck to some quantised table, e.g. 81.92 MHz, 85.33 MHz, etc. When I have time, I was planning to make a thread on my replica AnyClk unit so people can make their own. I heard Migron actually sanded away the PLL's part number on his AnyClk. lol? I'm using ICS512MILF for the PLL, which I had bought several of 15 years ago doing a similar project. I have tested both the SI5351 and PLL type extensively in the 80-90 MHz range. Still working on the SI5351 based unit. Rasz is pushing for a 1 MHz increment/decrement button. Time... where has it all gone?

Hi, just to let you know if someone interested in another kind of interposer for TQFP208 Am486DX5 CPU, I just made one...
Re: Data Expert EXP4044 ver. 1.0 (486 VLB) MB repair and VRM add

feipoa wrote on 2024-10-22, 03:29:

Rasz is pushing for a 1 MHz increment/decrement button. Time... where has it all gone?

You can change it to 100KHz or even 10KHz by deleting couple zeroes 😀 could really use few more lines to add some form of auto repeat acceleration so that lets say after 2? 3? seconds of keeping key pressed it switches to 10x the step, otherwise switching frequency might be slower than reflashing 😀

feipoa wrote on 2024-10-22, 03:29:

MikeSG: I was unable to locate 85 MHz and 90 MHz DIP-14 style crystal oscillator cans on Aliexpres. Could you relay some links? […]
Show full quote

MikeSG:
I was unable to locate 85 MHz and 90 MHz DIP-14 style crystal oscillator cans on Aliexpres. Could you relay some links?

Sphere:
I also have my own variant of the AnyCLK, which is smaller than the original Migron unit, but I prefer the ability to type in exactly the clock frequency I want rather than being stuck to some quantised table, e.g. 81.92 MHz, 85.33 MHz, etc. When I have time, I was planning to make a thread on my replica AnyClk unit so people can make their own. I heard Migron actually sanded away the PLL's part number on his AnyClk. 🤣? I'm using ICS512MILF for the PLL, which I had bought several of 15 years ago doing a similar project. I have tested both the SI5351 and PLL type extensively in the 80-90 MHz range. Still working on the SI5351 based unit. Rasz is pushing for a 1 MHz increment/decrement button. Time... where has it all gone?

Where can I see pics/details of your device?

Sphere478 wrote on 2024-10-22, 05:03:

feipoa wrote on 2024-10-22, 03:29:

MikeSG: I was unable to locate 85 MHz and 90 MHz DIP-14 style crystal oscillator cans on Aliexpres. Could you relay some links? […]
Show full quote

MikeSG:
I was unable to locate 85 MHz and 90 MHz DIP-14 style crystal oscillator cans on Aliexpres. Could you relay some links?

Sphere:
I also have my own variant of the AnyCLK, which is smaller than the original Migron unit, but I prefer the ability to type in exactly the clock frequency I want rather than being stuck to some quantised table, e.g. 81.92 MHz, 85.33 MHz, etc. When I have time, I was planning to make a thread on my replica AnyClk unit so people can make their own. I heard Migron actually sanded away the PLL's part number on his AnyClk. lol? I'm using ICS512MILF for the PLL, which I had bought several of 15 years ago doing a similar project. I have tested both the SI5351 and PLL type extensively in the 80-90 MHz range. Still working on the SI5351 based unit. Rasz is pushing for a 1 MHz increment/decrement button. Time... where has it all gone?

Where can I see pics/details of your device?

Which device? I referenced two different devices. For my variant of the AnyClk, I haven't posted about it yet. If referring to the SI5351 based clock chip, I already linked it in my previous post, see:

feipoa wrote on 2024-10-21, 01:34:

Small update:

I have been using this plug-in clock generator for 386 systems, shown here: Re: Project: Full Can Clock Oscillator Replacement It contains a programmable SI5351 PLL chip....

RayeR wrote on 2024-10-22, 04:10:

Hi, just to let you know if someone interested in another kind of interposer for TQFP208 Am486DX5 CPU, I just made one...
Re: Data Expert EXP4044 ver. 1.0 (486 VLB) MB repair and VRM add

I hadn't seen this. Replied in the other thread.

Dip-14, 85MHz:
https://vi.aliexpress.com/item/1005006334846523.html
https://vi.aliexpress.com/item/1005007327230535.html
https://vi.aliexpress.com/item/1005007016705763.html
https://vi.aliexpress.com/item/1005006626640862.html - TCXO (expensive)

Dip-14, 90MHz:
https://vi.aliexpress.com/item/32881663237.html - TCXO (expensive)
https://vi.aliexpress.com/item/33038241631.html - TCXO (expensive)
https://vi.aliexpress.com/item/1005002536685827.html

Just searched for "85mhz crystal", and "90mhz crystal"... but maybe there's some auto-sorting per person that stops them from showing up.

Thank you. I had searched several things, including "DIP-14 85 MHz crystal" and "85 MHz crystal", but I did not end up with the items you found. I'll order the 85 MHz cyrstal 2-pack for $4.

I finally managed to find another one of these Evergreen TI486SXL2-66 QFP interposers. I found it in an old beat up 386-20 system, but the interposer looks almost new, aside from the dust. It has an affixed heatsink and I was wondering if this was an original Evergreen supplied heatsink. Has anyone else found one of these Evergreen SXL2-66 QFP interposers with a pre-attached heatsink? If the heatsink is original, I'd be reluctant to remove it.

The VRM is set for 3.60 V, so overclocking will be limited. It was able to do 75 MHz at 3.6 V, but not 80 MHz. VRM adjustments will be needed for more speed.

That heatsink does not look original for Evergreen but never know. Tried wayback machine but it still having issues with * searches. Once they get that part fixed this may help with some other pics
http://web.archive.org/web/19961225204636/htt … t.exe?group=OVR

I’ve found a torch is a good way to remove heatsinks like that.

Put small amount of pressure on heatsink and heat fins with flame until glue gives way.

This is interesting

Sphere478 wrote on 2024-11-17, 05:39:

I’ve found a torch is a good way to remove heatsinks like that.

Put small amount of pressure on heatsink and heat fins with flame until glue gives way.

Personally I just use isopropyl alcohol, drip a bit of it onto the edge of the heatsink then stick a blade in there which should go into the thermal tape fairly easily if the IPA works on it. Once the blade is worked in a few mm then drip some more IPA down the blade and the heatsink should remove quite easily after that once IPA gets into more places.

Sphere478 wrote on 2024-11-22, 14:22:

This is interesting

PK-A486CX80-2 for NEC PC-98 PC-9801DA/RA & PC-98RL. Can often get them on yahoo.jp auctions (currently one listed at $150).
http://www.amy.hi-ho.ne.jp/nakajima-jr/kikaku … ecial2/cpu4.htm
https://www-amy-hi--ho-ne-jp.translate.goog/n … en&_x_tr_pto=sc
Its interesting because https://necretro.org/PC-9801_RA is from 1988 and definitely didnt have 486 compatible chipset 😀
Looking at other 386-486 upgrades listed at that page the first one ABM 486GT-33 is ridiculously complicated https://xcancel.com/wildcat_300104/status/172 … 491155247501461 includes huge daughterboard mounted in 5 inch drive bay with 128K cache and Intel A82485-33 cache controller.

Second listed cpu board on that site is also interesting, Buffalo HDX-16/HDX-32 apparently lets you arbitrarily add additional CPU FSB multiplier to 486

Maybe works like those cheap 14MHz Amiga accelerators that boost clock when CPU is not touching external bus, but downclock back anytime CPU tries to fetch/write data?

Fun topic https://forum.vcfed.org/index.php?threads/run … oard-386.47554/
IBM 5150 with 486 up to 15 fps in DOOM ?!?!
https://xcancel.com/foone/status/1424418405579325442

^wow…. That’s clever.

Thermalwrong wrote on 2024-11-22, 16:35:

Sphere478 wrote on 2024-11-17, 05:39:

I’ve found a torch is a good way to remove heatsinks like that.

Put small amount of pressure on heatsink and heat fins with flame until glue gives way.

Personally I just use isopropyl alcohol, drip a bit of it onto the edge of the heatsink then stick a blade in there which should go into the thermal tape fairly easily if the IPA works on it. Once the blade is worked in a few mm then drip some more IPA down the blade and the heatsink should remove quite easily after that once IPA gets into more places.

Many ways to do stuff. I like that heat keeps you from having to use physical force. Always makes me nervous to go prying on chips

My best guess is that the heatsink was installed by Evergreen, thus I won't be removing it. Since PGA SXL2-66 units came with a factory heatsink from Texas Instruments, and they are also designed around 3.6 V, I'd guess that the QFP chips also needed a heatsink.

I left the heatsink as is and added a 25 mm fan. Then installed it in the case. Not the most glamorous having a fan wider than the heatsink, but there aren't any standard 20 mm fans as far as I'm aware.

Back to this topic. Finally.

DTK PEM-0030Y (didn't want to disturb the DTK PEM-4036Y in the assembled 386 PC)
STB Nitro 2Mb (CL GD-5434)
All BIOS settings on max except DRAM WS = 2
cyrix -cd -i1 -i2
Buck regulator with 12V input from the PSU

90MHz crystal oscillator
2 x 45MHz to CPU

Wolf3D: 70.2 fps <--- !
Superscape: 31.2 fps
PC Player Benchmark: 7.9 fps
Doom: 23.82 fps
Quake 1: 4.4 fps <--- !
WinTune2: later

Great W3D and Q1 scores.
The Q1 result is on par with the best i have seen so far, produced with IBM BL3 at 110MHz (2x55).

Still not 100% if it is 100% stable but so far so good.

For some reason the IDE stability is very vulnerable with this CPU. Took a moment to find IDE controller that survives the 22.5MHz base frequency.
Similarly for FPUs.
40MHz rated ULSI - flaky.
40MHz rated IIT - flaky.
33MHz rated FasMath (gray top) - flaky.
40MHz rated FasMath (black top) - does well.

Interestingly enough i found an SXL2-66 CPU (well, one of the 4 chips i have) that works just fine at 90MHz in the 3.9 - 4.5 voltage range.
Going below or above that results in instabilities.
Not sure why it was unstable when suppling the voltage through the interposer's trimmer.
I have another interposer, so will compare stability between the two supply mechanisms.

2 x 50MHz feels possible.
Able to get into DOS but with lowered ISA bus frequency which results in lowered performance compared to sharp 2 x 45.
Will know more soon.

What have you done differently this time around? In the past, you couldn't get to 80 MHz stable, now you are feeling 90 MHz is stable. Maybe you should push it to 92 MHz with your Migron?

pshipkov wrote on 2024-12-16, 02:53:

90MHz crystal oscillator
2 x 45MHz to CPU

Doom: 23.82 fps

This is on par with my results at 90 MHz.

pshipkov wrote on 2024-12-16, 02:53:

For some reason the IDE stability is very vulnerable with this CPU. Took a moment to find IDE controller that survives the 22.5MHz base frequency.

Did you try at more conventional ISA frequencies in the past, yet still had instability over 80MHz?

pshipkov wrote on 2024-12-16, 02:53:

40MHz rated FasMath (black top) - does well.

That has been my experience with the SXL2-66.

pshipkov wrote on 2024-12-16, 02:53:

Interestingly enough i found an SXL2-66 CPU (well, one of the 4 chips i have) that works just fine at 90MHz in the 3.9 - 4.5 voltage range.
Going below or above that results in instabilities.

That's a rather large voltage range. Is your SXL2 stable at 90 MHz with 3.9 V? If yes, time to shoot further north! What cooling apparatus are you using?

pshipkov wrote on 2024-12-16, 02:53:

Not sure why it was unstable when suppling the voltage through the interposer's trimmer.

Further testing revealed that the LP regulator is much preferred to the MIC regulator on this interposer. I think I mentioned this previously, but if I forgot, here it is! The LP showed stability at lower voltages compared to the MIC regulator when using the same CPU. I recommend only using the LP in subsequent productions, if overclocking is of interest to you (e.g. LP38501TS-ADJ, LP38503TS-ADJ, or its derivatives).

pshipkov wrote on 2024-12-16, 02:53:

2 x 50MHz feels possible.
Able to get into DOS but with lowered ISA bus frequency which results in lowered performance compared to sharp 2 x 45.
Will know more soon.

Shoot for the intermediate range, like 92 or 95 MHz.

How many SXL2's are you able to get working well at 90 MHz? Was it one of the units you bought without factory heatsink?

Have you tried booting into Windows 3.11 and loading an IE5 google page?

Initially i was using the voltage controls on the interposer. That didn't work well.
Then switched t0 ~5V directly from PSU, bypassing all the wiring on the interposer.
That didn't go well either because the only CPU here that handles 80+ MHz likes lower voltage, as mentioned above.
Switched to buck regulators and things are better now.

Before the CPU was the main obstacle regardless of anything else.
Now is the ISA bus frequency, meaning the CPU seems stable.

100MHz was fussy early on.
First will spend some time within the 80-90 MHz range to see if things are stable-stable before making next moves.

I have only 4 SXL2-66 chips from John.
One of them can do 80-90.
The rest are not having it.

Web browsing in GUI ?
Still far from there. : )

Guessing binning wasn't that much different to today. An sx was binned based on functionality (fpu/whatever other differences), capability (mhz), and finally, market demand (Ie a perfectly good dx might have been crippled because the sx was selling better).

So it's possible you got a late model "dx4" (if it worked properly) that had a dud fpu and couldn't quite hit 100.
Thus, it became the sx66.

Which begs the question, have you tried undervolting at all? Voltage is a delicate balance between "enough to switch /and retain digital integrity" and "not so much the extra heat limits it".

You might find, if it's a later chip that lowering the voltage is beneficial. Ie, maybe as low as 3v.

pshipkov:
In summary, the one SXL2 CPU you have that handles 80-90 MHz:

1) doesn't like 5 V from PSU directly, because of too much voltage
2) likes 3.9-4.5 V range, but not from the MIC linear regulator
3) only likes 3.9-4.5 V from a switching regulator?

If 3) is true, then I'd be curious how the LP regulator works for you in that voltage range. I'm puzzled why the switching reg. would work but not the linear at the same voltages.

myne:
Are you suggesting that if 80 MHz doesn't work at 3.6 V, then to try it at 3.0 V or 3.3 V? Generally, I've seen increasing voltages helped with higher operating frequencies. However, I have had one SXL2 which could not handle more than 3.65 V at any frequency. It could do 75 MHz at a voltage lower than all other SXL2's (3.65 V). Perhaps it would be a good candidate to test 80 Mhz at lower voltages? Or are you suggesting running it at the rated speed of 66 MHz, but with 3.0 - 3.5 V ranges?