Sphere478 wrote on 2022-03-19, 04:48:

Repo Man11 wrote on 2022-03-19, 02:42:

I put the shield back on with Arctic Silver 5 and four dabs of super glue. The result was counterintuitive; all else was the same, but the temp was noticeably cooler. I then decided to swap to the best Socket 7 heat sink I have. Unfortunately, the super glue didn't hold (the shield stuck to the heat sink), so I'll eventually have to try something else. Since right now I only have super glue and JB Weld on hand (JB Weld is forever) I used super glue again for now. The larger heat sink I have cannot be used if the IHS is off since it has ridges that prevent it from being able to make contact with the CPU's core with the shield removed. No surprise that it ran cooler still with a larger heat sink and fan.

J&b weld would probably work great for the corners. I can try it.

I'm sure it would, but it's a bit more permanent than what I'm looking for.

You all are way overthinking what to use to reattach the IHS.

Silicone glue will work just fine. If you want something that is way closer to the look and feel of the original glue, go buy a tube of HondaBond from your local Honda dealer.

cyclone3d wrote on 2022-03-19, 05:45:

You all are way overthinking what to use to reattach the IHS.

Silicone glue will work just fine. If you want something that is way closer to the look and feel of the original glue, go buy a tube of HondaBond from your local Honda dealer.

As a collector, I feel it’s important to use the best stuff, even to maybe return it as close as possible to original after the mod

We can over think if we want.

Looks like hondabond is a silicone of some sort. Have you compared the two? Does it seem to be the same stuff? By my reckoning I was figuring the original stuff to be some sort of hard polyurethane.

One thing I noted about the epoxy is it bonds strong but if it ever broke loose it could leave a gap. I have some new epoxy to try. And can even give j&b a try. (On my test chip.

Others have kinda been swaying me to try AS5 on the core.

Recently I had an idea,

What if I put as5 on the core with a sealed ring of sealant around it?

You’re right we are over thinking. I’m fine with my madness. 😀 it’s fun.

I know Hondabond will not absorb petroleum products and swell up and break down like silicone will.

It is way higher quality than any of the other automotive silicone sealants I have used in the past 25 years.

Not sure that trying to have a sealed ring around the core is a great idea. For one, there won't be anywhere for the offgassing of whatever chemicals are in the glue to go in the sealed off area. Might not cause any issues but who knows what it would do to the TIM over time.

As for JB Weld... that stuff is trash. I've used it before and don't trust it at all. It is super brittle and also breaks down when exposed to solvents.

For metal repair, QuickSteel is infinitely better.

I've been using JB Weld (the original, not the quick set) to repair things for forty years now. The 400 block I used in my Camaro had a dent right where the oil filter seal was - I filled in the divot with JB Weld, and it never leaked again, even after driving it for over twenty years that way. It isn't the right tool for all jobs, but there are definitely applications where I've yet to find anything better.

cyclone3d wrote on 2022-03-19, 06:24:

I know Hondabond will not absorb petroleum products and swell up and break down like silicone will. […]
Show full quote

I know Hondabond will not absorb petroleum products and swell up and break down like silicone will.

It is way higher quality than any of the other automotive silicone sealants I have used in the past 25 years.

Not sure that trying to have a sealed ring around the core is a great idea. For one, there won't be anywhere for the offgassing of whatever chemicals are in the glue to go in the sealed off area. Might not cause any issues but who knows what it would do to the TIM over time.

As for JB Weld... that stuff is trash. I've used it before and don't trust it at all. It is super brittle and also breaks down when exposed to solvents.

For metal repair, QuickSteel is infinitely better.

The early k6 chips totally sealed off the interior with a sealing ring around the entire perimeter

I was thinking if the interior was preserved in a sealed environment the TIM (AS5, for example) could last the decades possibly.

I know many p4 chips did something similar

It would protect the TIM when cleaning the chip.

I may try this on my next mod.

Perhaps as5 on the core and polyurethane around the entire perimeter.

A modification of course to original design, which some won’t like, but perhaps a upgraded method for the overclocking enthusiasts.

Repo Man11 wrote on 2022-03-19, 07:18:

I've been using JB Weld (the original, not the quick set) to repair things for forty years now. The 400 block I used in my Camaro had a dent right where the oil filter seal was - I filled in the divot with JB Weld, and it never leaked again, even after driving it for over twenty years that way. It isn't the right tool for all jobs, but there are definitely applications where I've yet to find anything better.

Oh yeah, I’ve seen j&b used many a time on engine blocks. I have a retired 400sbc block in my living room holding wine bottles and a glass table top. (The side was cracked and repaired with j&b) but never had the desire to try it as I had good blocks, so I retired it to a coffee/wine table. I am fully EV these days though, so no more engine blocks these days 🤣.

As for its fitness for this application, I am unsure. I can maybe test. 😀

Not questioning the durability of JB Weld. Plugged holes in transmissions with it myself. It's kinda "What God has joined together with JB Weld, let no man put asunder" and that is the problem.

I started thinking about the oldschool single part grey and flexible Bondo, came in a tube, set up quick, meant to be a polyester resin. Not sure if current spot and glaze filler is the same stuff, but it's the closest looking.

edit: btw cyclone3d's experience sounds like what happens if you don't get the hardener/resin mixture right.. sets up "crumby" in that it crumbles.

I realized that the counterintuitive drop in temperature with the IHS on must be caused by increased spring pressure. With the shield removed, the stack height is lower so you have less spring pressure holding the heat sink against the CPU's core. This is good because the core didn't get damaged, but less pressure holding the heat sink to the core means lower thermal conductivity. An effort to increase spring pressure would make it more likely that the core would be damaged when installing the heat sink, but putting the IHS back on means you can substantially increase the spring tension with no fear of damaging the CPU's core.

BitWrangler wrote on 2022-03-19, 15:44:

Not questioning the durability of JB Weld. Plugged holes in transmissions with it myself. It's kinda "What God has joined together with JB Weld, let no man put asunder" and that is the problem.

I started thinking about the oldschool single part grey and flexible Bondo, came in a tube, set up quick, meant to be a polyester resin. Not sure if current spot and glaze filler is the same stuff, but it's the closest looking.

edit: btw cyclone3d's experience sounds like what happens if you don't get the hardener/resin mixture right.. sets up "crumby" in that it crumbles.

It doesn't do good in situations where vibration is a factor.

It's not like I only tried it once either.

QuickSteel just works way better in my experience, but it probably is not going to be as good for sealing cracks because it is a putty.

JB Weld also doesn't hold up to gasoline very well.

BitWrangler wrote on 2022-03-19, 15:44:

Not questioning the durability of JB Weld. Plugged holes in transmissions with it myself. It's kinda "What God has joined together with JB Weld, let no man put asunder" and that is the problem.

I started thinking about the oldschool single part grey and flexible Bondo, came in a tube, set up quick, meant to be a polyester resin. Not sure if current spot and glaze filler is the same stuff, but it's the closest looking.

edit: btw cyclone3d's experience sounds like what happens if you don't get the hardener/resin mixture right.. sets up "crumby" in that it crumbles.

Quite agree. Would definitely not use anything resulting in a rigid joint when gluing parts with wildly different thermal expansion coefficients together, that's going to be subjected to large temperature changes. The mechanical stress will be considerable. I'm of course just guessing, but wouldn't be the least bit surprised if this is the main reason the manufacturers always seem to use something flexible.

Also, a flexible joint will make reversal a lot easier.

Would probably use some kind of silane based adhesive myself.

Yes that too, fixing it too well at the edges and have the HS pop up off the core when it's hot is not something you wanna risk.

Its been a few days, decided to test the polyurethane.

https://youtu.be/vf2PQ5X_bbY

Didn’t go so well

The center of the corners wasn’t set up yet

And the thin film on the core let loose very easily (wasn’t cured)

But the parts that were cured were flexible yet strong. More flexible than the stuff AMD used though.

Hi,
Thanks Sphere478 for that mod idea 😀
I have bought 9x K6-2+ 570 from ebay and selected the most stable one to make mod.
I have put Thermal grizzly Kyronaut on die and glued cover with silicone - automotive gasket to 350 C 😀
Now it is more stable than before mod 😀 Can run DOS benchmarks at 5,5*120=660 (no L2 cache on board)
I use 2,2 V for those CPUs, more doesn't make sens.
I hope that it will be stable in Windows98 at 600 MHz at least 😀

lukas12p wrote on 2022-03-20, 22:12:

Hi, Thanks Sphere478 for that mod idea :) I have bought 9x K6-2+ 570 from ebay and selected the most stable one to make mod. I h […]
Show full quote

Hi,
Thanks Sphere478 for that mod idea 😀
I have bought 9x K6-2+ 570 from ebay and selected the most stable one to make mod.
I have put Thermal grizzly Kyronaut on die and glued cover with silicone - automotive gasket to 350 C 😀
Now it is more stable than before mod 😀 Can run DOS benchmarks at 5,5*120=660 (no L2 cache on board)
I use 2,2 V for those CPUs, more doesn't make sens.
I hope that it will be stable in Windows98 at 600 MHz at least 😀

Wow! Stable at 660?!? Amazing.

The next mod I do I think I will try your method.

I assume the higher clocks are from the TIM? Do you agree?

What was your max prior?

660 is actually a exceptional result. Beyond what is usually gotten. Very good job.

Interested to hear your reply, I hadn’t figured we would see this much of a increase because of how little heat the chip makes. But I’m glad to be wrong if so, it will make it worth while to put as5 or whatever under there.

The question now is, do we do a full perimeter seal now in order to keep the TIM fresh? Or keep it open?

Sphere478 wrote on 2022-03-21, 00:51:

lukas12p wrote on 2022-03-20, 22:12:

Hi, Thanks Sphere478 for that mod idea :) I have bought 9x K6-2+ 570 from ebay and selected the most stable one to make mod. I h […]
Show full quote

Hi,
Thanks Sphere478 for that mod idea 😀
I have bought 9x K6-2+ 570 from ebay and selected the most stable one to make mod.
I have put Thermal grizzly Kyronaut on die and glued cover with silicone - automotive gasket to 350 C 😀
Now it is more stable than before mod 😀 Can run DOS benchmarks at 5,5*120=660 (no L2 cache on board)
I use 2,2 V for those CPUs, more doesn't make sens.
I hope that it will be stable in Windows98 at 600 MHz at least 😀

I assume the higher clocks are from the TIM? Do you agree?
What was your max prior?

I haven't tested it for long time, just about 20 minutes in DOS games/tests, so I don't want to take this results so serious.
I will test this CPU in my K6 built (MVP3, 1MB cache), where I have K6-3+ 400@560 with Voodoo3 and will let You know how it goes.
What is "TIM"?
Before mod I was able to get it boot to DOS at 660 MHz, nothing more.

lukas12p wrote on 2022-03-21, 01:04:

I haven't tested it for long time, just about 20 minutes in DOS games/tests, so I don't want to take this results so serious. I […]
Show full quote

Sphere478 wrote on 2022-03-21, 00:51:

lukas12p wrote on 2022-03-20, 22:12:

Hi, Thanks Sphere478 for that mod idea :) I have bought 9x K6-2+ 570 from ebay and selected the most stable one to make mod. I h […]
Show full quote

Hi,
Thanks Sphere478 for that mod idea 😀
I have bought 9x K6-2+ 570 from ebay and selected the most stable one to make mod.
I have put Thermal grizzly Kyronaut on die and glued cover with silicone - automotive gasket to 350 C 😀
Now it is more stable than before mod 😀 Can run DOS benchmarks at 5,5*120=660 (no L2 cache on board)
I use 2,2 V for those CPUs, more doesn't make sens.
I hope that it will be stable in Windows98 at 600 MHz at least 😀

I assume the higher clocks are from the TIM? Do you agree?
What was your max prior?

I haven't tested it for long time, just about 20 minutes in DOS games/tests, so I don't want to take this results so serious.
I will test this CPU in my K6 built (MVP3, 1MB cache), where I have K6-3+ 400@560 with Voodoo3 and will let You know how it goes.
What is "TIM"?
Before mod I was able to get it boot to DOS at 660 MHz, nothing more.

TIM thermal interface material (paste)

Well, I don’t think it’s likely that enabling cache made it more stable so it must have been the TIM

So if you were able to boot dos at 660 and then run benchmarks by changing TIM that is a minor though significant and noticeable difference. Suggests that what AMD used is worth upgrading from.

Le' sigh...

Of course better TIM is going to help transfer the heat away from the CPU better.

Got that higher conductance adhesive. Actually the second time I ordered it. First time seller only sent one tube (it’s an epoxy) second time I got both tubes but the stuff is dry and doesn’t react or stick to anything. Seems to be bad.

The polyurethane I tried earlier I didn’t clean up and let cure more it is super flexible and very very well attached

I’ll summarize my experiences so far with these materials

Higher wattage epoxy: two strikes, useless.

First Epoxy dries hard, dries fast, works good is removable. Was kinda hoping that it wouldn’t be though. Not the greatest at thermal conductivity but it has its perks

Urethane from really early in thread nice and runny/stringy dries flexible, seems to work okay takes a while to dry though. Sticks well.

Polyurethane mentioned earlier isn’t runny, it wants to keep its shape this is an issue. Sticks very well takes a long time to dry. Isn’t super strong, wish it was less flexible.

Marine 5200
Black, messy, takes forever to dry (literally forever) don’t have an update yet.

None of these are really like the stuff amd used.

I may get some windshield adhesive next

Maybe there's a way to know from the source.
Does anyone know someone who work at AMD at that time ?

crazytiti wrote on 2022-03-24, 07:51:

Maybe there's a way to know from the source.
Does anyone know someone who work at AMD at that time ?

Any chance the patent could tell us? How do we look that up?