bloodem wrote on 2022-04-12, 06:13:

Too bad, but I guess it's normal (some might have actually had issues with that portion of the cache). […]
Show full quote

Sphere478 wrote on 2022-04-12, 05:53:

I’ve stubbled across a 570 that isn’t stable unlocked at 550. Or 500 it seems.

Might have to pull it apart and re lock it.

Too bad, but I guess it's normal (some might have actually had issues with that portion of the cache).

Why didn't you test it before reassembling? If you're afraid of damaging the die, don't be, you only need to put a bit of thermal paste and a cooler on top (without using the retention clips).
Just make sure that the cooler doesn't slide. 😀 I've already done this four times and it worked great for 2+ hours during the Prime 95 + 3DMark2000 stability testing.

because I've done a bunch of them, none have had issues till now. I don't feel comfortable burning them in without a IHS anyway.

so does a 570 2+ that can't unlock have any value? I'm gonna try and put the resistor back and reattach IHS
Edit: Re: SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod

Sphere478 wrote on 2022-04-12, 05:53:

I’ve stubbled across a 570 that isn’t stable unlocked at 550. Or 500 it seems.

Might have to pull it apart and re lock it.

edit: not even stable at 400mhz 2.1v (better though)

That's a shame, though wouldn't bug me too too much if a couple of mine are like that as I have some boards I'd like to K6-3-ify that won't go much below 2.5V and are only stable on 75Mhz max.

Sphere478 wrote on 2022-04-12, 21:32:

bloodem wrote on 2022-04-12, 06:13:

Too bad, but I guess it's normal (some might have actually had issues with that portion of the cache). […]
Show full quote

Sphere478 wrote on 2022-04-12, 05:53:

I’ve stubbled across a 570 that isn’t stable unlocked at 550. Or 500 it seems.

Might have to pull it apart and re lock it.

Too bad, but I guess it's normal (some might have actually had issues with that portion of the cache).

Why didn't you test it before reassembling? If you're afraid of damaging the die, don't be, you only need to put a bit of thermal paste and a cooler on top (without using the retention clips).
Just make sure that the cooler doesn't slide. 😀 I've already done this four times and it worked great for 2+ hours during the Prime 95 + 3DMark2000 stability testing.

because I've done a bunch of them, none have had issues till now. I don't feel comfortable burning them in without a IHS anyway.

so does a 570 2+ that can't unlock have any value? I'm gonna try and put the resistor back and reattach IHS

I'm sure it does to anyone who just wants the K6-2+ for whatever reason. Glad I got mine before the price went up (seller doesn't seem to be accepting the same offer now as when first listed)

just got another batch of three.

plopped them in and saw which of them would post at 1.7v 600mhz

only one posted.

modded it

not stable at 600mhz but goes 45min on super pi before error though

these issues started happening soon as I started using this new thermal paste.

artic mx-4

I heard that sometimes paste takes a while to start working? is that true?
is this paste crap? I chose it for 8 year lifespan as advertised.

I have some grizley kryonaut on order

Edit: Re: SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod

Sphere478 wrote on 2022-04-13, 04:43:

I heard that sometimes paste takes a while to start working? is that true?

For all intents and purposes... no. There are cases like that of the old Arctic Silver 5, where the manufacturer states that it will take up to 200 hours and a few thermal cycles to achieve max performance, but... really, we are talking about a few degrees, max.
With modern thermal paste you already have the max performance immediately after applying it.

I used Arctic MX-2 on all four K6-2+ 570 CPUs that I modded, and they're all stable at 633 MHz 😁. Maybe I've just been very lucky. 😀
I also used the same thermal paste on much more power hungry CPUs like Slot A Athlons running at 1 GHz, and temperatures dropped significantly compared to the factory thermal compound.

Well, I have two cpus with issues now, and two that are not yet modded.

Do I take them apart and try again with the kryonaut? (When it comes in)

Edit: Re: SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod

Sphere478 wrote on 2022-04-13, 06:03:

Well, I have two cpus with issues now, and two that are not yet modded.

Do I take them apart and try again with the kryonaut? (When it comes in)

It can't hurt to wait and try the Kryonaut just to rule out any weirdness. I know MX-4 is a good compound but maybe you got a bad batch or something.

Sphere478 wrote on 2022-04-13, 06:03:

Well, I have two cpus with issues now, and two that are not yet modded.

Do I take them apart and try again with the kryonaut? (When it comes in)

Are you sure that there is good contact between the heatsink and the die?
You can easily test this by powering on the PC without a CPU cooler and keeping your fingers on the heatsink (heatspreader).
At 2.0V, in a matter of ~ 20 seconds (maybe even less), the heatsink should become very hot to the touch.

If it is, then thermal transfer between the die and heatspreader is good, so the only plausible explanation is that you got unlucky and bought a bad batch of CPUs (that actually have a partially damaged L2 cache). 😀

These CPUs would probably work even without any thermal compound between the heatspreader and the die if otherwise the contact is good. It would not be a good idea, but they would most likely work.
In fact, at one point, I forgot to connect the fan after modding one of these CPUs, and the damn thing was working fine even after 30 minutes of a 3DMark2000 loop (at which point I realized that there was a strange burning smell in my room 🤣 ). Sure enough, the cooler heatsink was at 80+ degrees C.

I had a bad 570 in my batch. It is stock and completely unstable.

I recorded this almost a month ago, finally managed to upload the video today: https://www.youtube.com/watch?v=5Aw2U_BZnSk
It's a guide with the method I'm using for modding the K6-2+ to a K6-3+, maybe it will help others.

bloodem wrote on 2022-04-13, 12:12:

Are you sure that there is good contact between the heatsink and the die? You can easily test this by powering on the PC without […]
Show full quote

Sphere478 wrote on 2022-04-13, 06:03:

Well, I have two cpus with issues now, and two that are not yet modded.

Do I take them apart and try again with the kryonaut? (When it comes in)

Are you sure that there is good contact between the heatsink and the die?
You can easily test this by powering on the PC without a CPU cooler and keeping your fingers on the heatsink (heatspreader).
At 2.0V, in a matter of ~ 20 seconds (maybe even less), the heatsink should become very hot to the touch.

If it is, then thermal transfer between the die and heatspreader is good, so the only plausible explanation is that you got unlucky and bought a bad batch of CPUs (that actually have a partially damaged L2 cache). 😀

These CPUs would probably work even without any thermal compound between the heatspreader and the die if otherwise the contact is good. It would not be a good idea, but they would most likely work.
In fact, at one point, I forgot to connect the fan after modding one of these CPUs, and the damn thing was working fine even after 30 minutes of a 3DMark2000 loop (at which point I realized that there was a strange burning smell in my room 🤣 ). Sure enough, the cooler heatsink was at 80+ degrees C.

Die without compound is a bad idea.

Yeah they get hot, I’m wondering if there may be a bubble? Idk

Sphere478 wrote on 2022-04-13, 19:45:

Die without compound is a bad idea.

Of course it is, but I still bet it would work. 😀

Sphere478 wrote on 2022-04-13, 19:45:

Yeah they get hot, I’m wondering if there may be a bubble? Idk

If they get hot as fast as they used to before the mod (you can compare with an original K6-2+), then there's no reason to blame the thermal transfer between the die and heatspreader, so something else must be going on.

Only thing that I changed was TIM

So I’m thinking that must be it.

I’ll prob mod one more, this time go back to the epoxy or try the kryonaut

I’m torn on wether or not to pop apart the two that are giving issues. i think I’ll have to though

Edit: Re: SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod

Sphere478 wrote on 2022-04-13, 04:43:

these issues started happening soon as I started using this new thermal paste. […]
Show full quote

these issues started happening soon as I started using this new thermal paste.

artic mx-4

I heard that sometimes paste takes a while to start working? is that true?
is this paste crap? I chose it for 8 year lifespan as advertised.

I have some grizley kryonaut on order

Thermal paste is rarely at it's best the first run after application, a few full thermal cycles help even it out.

edit: Though I should say it's a couple of degrees drop from burn-in/bed-in not "night and day" .... so if it's way out or more than 5C hotter than previous, there's an application problem or extremely bad batch of paste.

Sphere478 wrote on 2022-04-13, 20:47:

Only thing that I changed was TIM […]
Show full quote

Only thing that I changed was TIM

So I’m thinking that must be it.

I’ll prob mod one more, this time go back to the epoxy or try the kryonaut

I’m torn on wether or not to pop apart the two that are giving issues. i think I’ll have to though

I agree that it's better to take one apart and retry with the method that worked for you in the past, just to make sure.

What type of instability are you seeing, though? Because, if rapid overheating of the die is the issue, the only thing that would make sense would be a complete system freeze.
And if die - heatspreader contact is not good, I would actually expect the CPU to die a pretty quick and horrible death, since these CPUs (obviously) lack any thermal throttling protection.

IMO, if you are just seeing BSODs and other types of failures after the mod, then the only plausible explanation is that the additional cache on these two CPUs was faulty (so I'm betting on the 'big coincidence' scenario 😁 ).
I just don't see how the Arctic MX-4 could ever cause this (I mean, this modern thermal paste is probably at least an order of magnitude better than the original silicone glue that AMD used).

Anyway, I recommend always testing the CPU before reattaching the heatspreader. This would eliminate these kinds of doubts.

Ran sims fine for like 12 hours, but fails shortly into a super pi burn in. Program gives error no crash.

The other cpu was crashing the laptop even in bios, these new tests are ok my freeway mobo. Neither one pass super pi at 600

The laptop one won’t pass even at 400.

The laptop issues seemed to be heat related. They got better with cool down period

Edit: Re: SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod

Well, guess what... Just tried to mod another K6-2+ 570 MHz (the fifth!), and this one failed miserably.
After the mod, it barely POSTs (no matter the frequency) and fails to even start to boot Windows (it just freezes).
As with the others, I also tested this one before reattaching the heatsink, so I will just move the resistor back to the original position without too much fuss.

So, after four consecutive successful mods, my luck finally ran out. 😁

I still have four left. I will try to mod one more (the last one), and will leave the last 3 in their original state.

bloodem wrote on 2022-04-14, 13:00:

Well, guess what... Just tried to mod another K6-2+ 570 MHz (the fifth!), and this one failed miserably. After the mod, it barel […]
Show full quote

Well, guess what... Just tried to mod another K6-2+ 570 MHz (the fifth!), and this one failed miserably.
After the mod, it barely POSTs (no matter the frequency) and fails to even start to boot Windows (it just freezes).
As with the others, I also tested this one before reattaching the heatsink, so I will just move the resistor back to the original position without too much fuss.

So, after four consecutive successful mods, my luck finally ran out. 😁

I still have four left. I will try to mod one more (the last one), and will leave the last 3 in their original state.

Update: Moved the resistor back to its original position, all good now.
Too bad, this CPU also works fine at 633 MHz, but the second cache cluster is definitely dead...
So it seems that, with these top of the line higher clocked SKUs, there is a pretty good chance of having a fully working cache (4 out of 5 is not bad).
My guess is that the percentages go down significantly with lower clocked CPUs.

I've got a K6-2+ 533 that I'm considering applying this mod to, I think I've got pretty good chances to have all of the L2 cache working! I've got a couple of questions though:

a) Does the resistor have to be soldered to the pad next to the one it's pulled from? I've seen some reports in this thread that the resistor only has to be removed for 256K cache.

b) I got this to glue the heatspreader back on. Would this work?
Also, semi-related, would the same solution work for gluing a socket 939 A64x2 heatspreader back on?

Garrett W wrote on 2022-04-15, 09:54:

a) Does the resistor have to be soldered to the pad next to the one it's pulled from? I've seen some reports in this thread that the resistor only has to be removed for 256K cache.

Indeed, apparently it does work by simply removing it, however, just to avoid any unforeseen issues, I think it's still better to move it to the pad next to it (which is how it's done on normal K6-3+ CPUs).
I mean, I think it's the easiest part of the whole process, so might as well do it right. 😀

Garrett W wrote on 2022-04-15, 09:54:

b) I got this to glue the heatspreader back on. Would this work?
Also, semi-related, would the same solution work for gluing a socket 939 A64x2 heatspreader back on?

That's regular silicone which remains pretty elastic when it cures, I prefer something rigid (like the Bison Epoxy I'm using), to make sure that there's no wiggle (which might eventually lead to a cracked CPU die).