VOGONS


Reply 480 of 574, by jbenam

User metadata
Rank Newbie
Rank
Newbie
Sphere478 wrote on 2022-05-02, 02:11:

The 550 is certainly a collertor’s item. I’d advise investing in some 570s maybe even sell the 550 to pay for the three 570s

Sadly it seems the auction was for just one CPU (the seller is going to amend the description soon).

I can see plenty of 550ACZ sold for 50EUR on the 'bay, not sure if it would make sense to sell it for 50 EUR (if I can actually manage to do so, since everyone else is probably going to purchase from the other seller) to purchase a very-slightly-faster 570 for 50EUR more.

Sphere478 wrote on 2022-05-02, 02:11:

Grab some 3m 5200 and some kryonaut. That’s the best way I’ve come up with. It’s more work, but these antique cpus are worth it.

It seems I can't get the 3M 5200 for less than 32EUR, which is more than what I did pay for the CPU in the first place! Any cheaper alternatives?

As for the TIM, it seems the MX-5 is pretty much 95% similar to the Kryonaut in terms of thermal transfer and I have already got some 😉

Reply 481 of 574, by Nemo1985

User metadata
Rank Oldbie
Rank
Oldbie
jbenam wrote on 2022-05-04, 20:39:

As for the TIM, it seems the MX-5 is pretty much 95% similar to the Kryonaut in terms of thermal transfer and I have already got some 😉

Well if we consider the thermal conductivity I wouldn't say so.
Artic Silver 5: 8.9 W/mK
Artic Mx-2: 5.6 W/mK
Artic Mx-4: 8.5 W/mK
Thermal Grizzly Hydronaut: 11.8W/mk
Artic Mx-5 (there is no official informations about it) but it appears to be around 5.2W/mK
The half of hydroanut.
I'd read this article: https://www.hwcooling.net/en/arctic-kept-his- … know-the-specs/

Sure probably the difference with such low power cpu will be not noticeable but...

Reply 482 of 574, by DrAnthony

User metadata
Rank Newbie
Rank
Newbie

Ohhh my Ph.D is coming in handy here! Viscosity is going to matter a lot here. The thinnest possible layer of TIM is generally going to be preferable to a thicker, but more conductive one. Then there's mounting pressure and the flatness of the surfaces and everything else going on.

Reply 483 of 574, by Nemo1985

User metadata
Rank Oldbie
Rank
Oldbie
DrAnthony wrote on 2022-05-04, 22:24:

Ohhh my Ph.D is coming in handy here! Viscosity is going to matter a lot here. The thinnest possible layer of TIM is generally going to be preferable to a thicker, but more conductive one. Then there's mounting pressure and the flatness of the surfaces and everything else going on.

That's an important clarification, so an higher value of viscosity is generally better? Do you think that the lower W/mk for the Mx-5 can be compesated from an higher viscosity?

Reply 484 of 574, by Sphere478

User metadata
Rank l33t++
Rank
l33t++
jbenam wrote on 2022-05-04, 20:39:
Sadly it seems the auction was for just one CPU (the seller is going to amend the description soon). […]
Show full quote
Sphere478 wrote on 2022-05-02, 02:11:

The 550 is certainly a collertor’s item. I’d advise investing in some 570s maybe even sell the 550 to pay for the three 570s

Sadly it seems the auction was for just one CPU (the seller is going to amend the description soon).

I can see plenty of 550ACZ sold for 50EUR on the 'bay, not sure if it would make sense to sell it for 50 EUR (if I can actually manage to do so, since everyone else is probably going to purchase from the other seller) to purchase a very-slightly-faster 570 for 50EUR more.

Sphere478 wrote on 2022-05-02, 02:11:

Grab some 3m 5200 and some kryonaut. That’s the best way I’ve come up with. It’s more work, but these antique cpus are worth it.

It seems I can't get the 3M 5200 for less than 32EUR, which is more than what I did pay for the CPU in the first place! Any cheaper alternatives?

As for the TIM, it seems the MX-5 is pretty much 95% similar to the Kryonaut in terms of thermal transfer and I have already got some 😉

There are small tubes of 5200

DrAnthony wrote on 2022-05-04, 22:24:

Ohhh my Ph.D is coming in handy here! Viscosity is going to matter a lot here. The thinnest possible layer of TIM is generally going to be preferable to a thicker, but more conductive one. Then there's mounting pressure and the flatness of the surfaces and everything else going on.

the more and more I play with these fancy pastes, the more I like that cheap white TIM

mx4 and kryonaut you put it on and a day later it's dry as a bone but that cheap white stuff that comes with heatsinks is still gooey

in any case, I'm still not convinced that just using the adhesive isn't a viable option 🤣 AMD did it, I did it on the first one with urethane. seemes to work fine, the kryonaut seems though it got some better OC results on one user's report. so I started using it.

ultimately anything inflexible or dry runs the chance of separating, and a separation is a thermal barrier. i don't know how bad it would be though if not separated very far. but it's undoubtedly worse than not separated.

what did they use on pentium 4s?

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 485 of 574, by jbenam

User metadata
Rank Newbie
Rank
Newbie
Nemo1985 wrote on 2022-05-04, 21:58:
Well if we consider the thermal conductivity I wouldn't say so. Artic Silver 5: 8.9 W/mK Artic Mx-2: 5.6 W/mK Artic Mx-4: 8.5 W/ […]
Show full quote

Well if we consider the thermal conductivity I wouldn't say so.
Artic Silver 5: 8.9 W/mK
Artic Mx-2: 5.6 W/mK
Artic Mx-4: 8.5 W/mK
Thermal Grizzly Hydronaut: 11.8W/mk
Artic Mx-5 (there is no official informations about it) but it appears to be around 5.2W/mK
The half of hydroanut.
I'd read this article: https://www.hwcooling.net/en/arctic-kept-his- … know-the-specs/

Sure probably the difference with such low power cpu will be not noticeable but...

IMHO theoretical numbers aren't that useful, if they don't have a correspondence in real world testing.

Here is a much better article, as it comes with a huge comparison table:
https://www.tomshardware.com/best-picks/best-thermal-paste

And it seems the Kryonaut fares way worse than the MX-5.

Reply 486 of 574, by Nemo1985

User metadata
Rank Oldbie
Rank
Oldbie
jbenam wrote on 2022-05-05, 09:33:
IMHO theoretical numbers aren't that useful, if they don't have a correspondence in real world testing. […]
Show full quote
Nemo1985 wrote on 2022-05-04, 21:58:
Well if we consider the thermal conductivity I wouldn't say so. Artic Silver 5: 8.9 W/mK Artic Mx-2: 5.6 W/mK Artic Mx-4: 8.5 W/ […]
Show full quote

Well if we consider the thermal conductivity I wouldn't say so.
Artic Silver 5: 8.9 W/mK
Artic Mx-2: 5.6 W/mK
Artic Mx-4: 8.5 W/mK
Thermal Grizzly Hydronaut: 11.8W/mk
Artic Mx-5 (there is no official informations about it) but it appears to be around 5.2W/mK
The half of hydroanut.
I'd read this article: https://www.hwcooling.net/en/arctic-kept-his- … know-the-specs/

Sure probably the difference with such low power cpu will be not noticeable but...

IMHO theoretical numbers aren't that useful, if they don't have a correspondence in real world testing.

Here is a much better article, as it comes with a huge comparison table:
https://www.tomshardware.com/best-picks/best-thermal-paste

And it seems the Kryonaut fares way worse than the MX-5.

That's right real world stuff is the most important.
According to the chart the MX-5 seems the best except for the liquid metal (which I do not think it can be used in k6-2+).

Reply 487 of 574, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

file.php?id=136764&mode=view

Re: AMD K6 3DMARK, aiming for stars.

In case you missed this^

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 488 of 574, by BitWrangler

User metadata
Rank l33t++
Rank
l33t++
Sphere478 wrote on 2022-05-05, 07:06:
DrAnthony wrote on 2022-05-04, 22:24:

Ohhh my Ph.D is coming in handy here! Viscosity is going to matter a lot here. The thinnest possible layer of TIM is generally going to be preferable to a thicker, but more conductive one. Then there's mounting pressure and the flatness of the surfaces and everything else going on.

the more and more I play with these fancy pastes, the more I like that cheap white TIM

Back in the day it seemed to me that the reasonable quality zinc oxide pastes (Not super cheap unknowns, but like Radio Shack, or RS Components ) outperformed all the fancy crap on high pressure direct die installations (small area) and the fancy crap only beat them on large area stuff, heatsink to heat spreader or package.

Unicorn herding operations are proceeding, but all the totes of hens teeth and barrels of rocking horse poop give them plenty of hiding spots.

Reply 489 of 574, by bofh.fromhell

User metadata
Rank Oldbie
Rank
Oldbie
BitWrangler wrote on 2022-05-08, 00:08:
Sphere478 wrote on 2022-05-05, 07:06:
DrAnthony wrote on 2022-05-04, 22:24:

Ohhh my Ph.D is coming in handy here! Viscosity is going to matter a lot here. The thinnest possible layer of TIM is generally going to be preferable to a thicker, but more conductive one. Then there's mounting pressure and the flatness of the surfaces and everything else going on.

the more and more I play with these fancy pastes, the more I like that cheap white TIM

Back in the day it seemed to me that the reasonable quality zinc oxide pastes (Not super cheap unknowns, but like Radio Shack, or RS Components ) outperformed all the fancy crap on high pressure direct die installations (small area) and the fancy crap only beat them on large area stuff, heatsink to heat spreader or package.

I think its more a case of what the surfaces look like.
A lot of modern gear tend to have very nice finish, almost mirror like in some cases.
So a low viscosity paste with very fine particles makes sense, after all you don't have any large pits to fill.
If you have rougher surfaces on the other hand thicker viscosity and larger particles may be what you want to use.

Reply 491 of 574, by ChrisK

User metadata
Rank Member
Rank
Member
Sphere478 wrote on 2022-05-06, 21:49:
“ https://www.vogons.org/download/file.php?id=136764&mode=view” […]
Show full quote

file.php?id=136764&mode=view

Re: AMD K6 3DMARK, aiming for stars.

In case you missed this^

First thought: WOW! World's first 64bit 4GHz socket 7 CPU. 🤣

Reply 492 of 574, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

Idk why super pi stopped working for me, but yeah those chips I was talking about earlier, are all successfully modded 3+ chips now.

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 493 of 574, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

Latest method with 3m 5200 marine sealant

Attachments

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 494 of 574, by BitWrangler

User metadata
Rank l33t++
Rank
l33t++

Cool, is it proofed against algae and marine borers now? 🤣

Unicorn herding operations are proceeding, but all the totes of hens teeth and barrels of rocking horse poop give them plenty of hiding spots.

Reply 495 of 574, by Sphere478

User metadata
Rank l33t++
Rank
l33t++
BitWrangler wrote on 2022-06-01, 15:10:

Cool, is it proofed against algae and marine borers now? 🤣

sadly no, I have to scrub the moss off them monthly still.

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 496 of 574, by kitten.may.cry

User metadata
Rank Member
Rank
Member
bloodem wrote on 2022-03-16, 11:14:
I have to say, @Sphere478, when you first mentioned the possibility of modding a K6-2+ to a K6-3+, I kinda laughed it off, think […]
Show full quote

I have to say, @Sphere478, when you first mentioned the possibility of modding a K6-2+ to a K6-3+, I kinda laughed it off, thinking that something like this would've been discovered a long time ago.
Of course, in hindsight, it does make sense that such an option was never explored in the past.
All in all, kudos for pursuing this with so much zeal! 😀

Anyway, I also modded one of my K6-2+ 570 CPUs to a K6-3+ 570.

My steps:
1. As you suggested, I used an old traditional thin razor to remove the heatsink (pretty straightforward, took me < 5 minutes).
2. Then I installed the CPU on the first Socket 7 MB that I could find, to make the cleanup process easier (I don't have a vise in my house).
3. The wife enjoyed the whole process as well, and helped me with cleaning up the old epoxy/silicone (I actually filmed this very "special" procedure 🤣), which was surprisingly easy for her... nails.
4. Switching the resistor to the other pad was the easiest part, it took me 30 seconds with the hot air station (I do have quite a bit of soldering experience, though).
5. After that, I applied some Arctic MX-2 paste on the die, mixed a bit of standard transparent "Bison" epoxy that I had lying around, and applied 4 small drops on all four corners. I used the MX-2 paste, because I don't want this to be a permanent thing (+ I don't have any thermal epoxy, anyway).
6. Immediately after that, very carefully, I installed the CPU on a motherboard (Gigabyte GA-5AA rev 3.2), and mounted a beefy cooler on top (which puts a nice even pressure on the heatsink, until the epoxy cures).

And... succes! The CPU works flawlessly at the exact same frequency that I had it running before on this MB (110 x 5.5 = 605 MHz). It passed a couple of hours of Prime 95 (small FFTs) without a hiccup.
Also, I've gotta say, the performance difference after modding it to a K6-3+ really surprised me (especially the 3DMark99 scores):

BENCHMARK RESULTS BEFORE MOD:
3DMark99 default: 4836 3DMarks / 8418 CPU Marks
3DMark2000 default: 4328 3DMarks / 267 CPU Marks

BENCHMARK RESULTS AFTER MOD:
3DMark99 default: 5412 3DMarks / 9282 CPU Marks
3DMark2000 default: 4872 3DMarks / 307 CPU Marks

Some pics:

You gained a new sub!

Reply 497 of 574, by feipoa

User metadata
Rank l33t++
Rank
l33t++

So you are using soft paste on these lids? I haven't done a lot of de-lidding, but I have delidded my Opteron 185 which was overheating. After the operation, the Opteron 185 worked cool for some time, but in less than a year, it started to overheat again. I suspect the hardening of soft thermal paste may be to blame as well as continued clamping on/off with the socket retainer. From this experience, I was thinking that perhaps thermal epoxy might be a better approach compared to soft paste. The Opteron is an everyday use 24/7 system, so maybe you won't come across the same issues with your occasional use K6-3 chips. Have you?

Plan your life wisely, you'll be dead before you know it.

Reply 498 of 574, by Sphere478

User metadata
Rank l33t++
Rank
l33t++
feipoa wrote on 2022-06-01, 22:43:

So you are using soft paste on these lids? I haven't done a lot of de-lidding, but I have delidded my Opteron 185 which was overheating. After the operation, the Opteron 185 worked cool for some time, but in less than a year, it started to overheat again. I suspect the hardening of soft thermal paste may be to blame as well as continued clamping on/off with the socket retainer. From this experience, I was thinking that perhaps thermal epoxy might be a better approach compared to soft paste. The Opteron is an everyday use 24/7 system, so maybe you won't come across the same issues with your occasional use K6-3 chips. Have you?

Currently I am using thermal grizzly kryonaut for the die and 3m 5200 for the corners.

I am contemplating switching to as5 for the core as it doesn’t dry out as easily?

The kryonaut dries out within days. But doesn’t seem to be an issue.

I did the thermal epoxy, and stopped after I had a incident where it popped off again with heatsink suction from paste.

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 499 of 574, by debs3759

User metadata
Rank Oldbie
Rank
Oldbie
feipoa wrote on 2022-06-01, 22:43:

So you are using soft paste on these lids? I haven't done a lot of de-lidding, but I have delidded my Opteron 185 which was overheating. After the operation, the Opteron 185 worked cool for some time, but in less than a year, it started to overheat again. I suspect the hardening of soft thermal paste may be to blame as well as continued clamping on/off with the socket retainer. From this experience, I was thinking that perhaps thermal epoxy might be a better approach compared to soft paste. The Opteron is an everyday use 24/7 system, so maybe you won't come across the same issues with your occasional use K6-3 chips. Have you?

I've noticed a few people talking about epoxy. Is that not a rather permanent solution?

See my graphics card database at www.gpuzoo.com
Constantly being worked on. Feel free to message me with any corrections or details of cards you would like me to research and add.