With shims you have to keep in mind that not all S462 CPUs are the same height, so you can either end up not making good contact with the CPU die or the shim not taking any of the pressure. 3D printing them also has problems, considering that every printer has its own margin of error and making thickness consistent is practically impossible.

Absolutely. Forgot about that in earlier post. Shims I talked about were used with thoroughbred 2400+ and 2700+

TrashPanda wrote on 2023-02-27, 05:29:

The socket A CPUs don't have a IHS to protect the die so its a direct to die heatsink and AMD in their brilliance didn't really […]
Show full quote

The socket A CPUs don't have a IHS to protect the die so its a direct to die heatsink and AMD in their brilliance didn't really provide a great way to keep the cooler level or a way to prevent too much pressure when installing so one minor slip and it chips the die a bit too much pressure and it cracks it. now they did put soft felt pads on the CPU to help but they are woefully inadequate for the task. IIRC there were shims produced that keep the cooler level and from applying too much pressure but getting them nowadays is pretty much impossible.

a 3d printed bracket would be awesome, not sure how easy it would be to make one.

You can see on the picture below what I mean by pads and direct die.

Athlon-M 3000.png

Not just socket A motherboards but Socket 370 and 423 boards as well, there was a whole range of them afflicted by the cap plague but Socket A being smack in the middle caught the worst of it.

I wasn't aware it was that bad. It sounds like it's more trouble and fragile then delidding a more modern CPU. I wonder if AMD assumed everyone would buy a complete system and never touch it.

I should find a board and find out.

acl wrote on 2023-02-27, 10:40:

Yes but even without the foam pads, using a shim is advised on these CPUs. I still have one or two. And i'm using one on my XP32 […]
Show full quote

TrashPanda wrote on 2023-02-27, 05:29:

The socket A CPUs don't have a IHS to protect the die so its a direct to die heatsink and AMD in their brilliance didn't really […]
Show full quote

Asininity wrote on 2023-02-27, 05:14:

What's the common cause of CPU death? I wonder if some 3D printer brackets could help with some of the cooler frustration.

I recently purchased an electric pump desoldering gun. It sounds like Socket A boards would be a good candidate for taking it through the paces.

The socket A CPUs don't have a IHS to protect the die so its a direct to die heatsink and AMD in their brilliance didn't really provide a great way to keep the cooler level or a way to prevent too much pressure when installing so one minor slip and it chips the die a bit too much pressure and it cracks it. now they did put soft felt pads on the CPU to help but they are woefully inadequate for the task. IIRC there were shims produced that keep the cooler level and from applying too much pressure but getting them nowadays is pretty much impossible.

a 3d printed bracket would be awesome, not sure how easy it would be to make one.

You can see on the picture below what I mean by pads and direct die.

Yes but even without the foam pads, using a shim is advised on these CPUs.
I still have one or two. And i'm using one on my XP3200+ (fragile and increasingly hard to find).
I think these can definitely be 3d printed, but probably more easily laser/plasma cut from a metal sheet.

Screenshot from 2023-02-27 11-36-49.png

Article about Socket A shims : https://www.legitreviews.com/what-is-a-shim-a … -you-use-one_64

Do you have a caliper that you'd be able to measure the thickness? I'd like to get a board and try a 3D printed version, though I wonder if there'd be enough interest to put in a order with SendCutSend or similar.

Today I got for cheap a ASUS A9550GE/TD/128M/A AGP8X video card, a ATI Radeon 9550 128MB, with active cooling. I wonder if it is the 128bit version.
Any opinions?

Munx wrote on 2023-02-27, 12:29:

With shims you have to keep in mind that not all S462 CPUs are the same height, so you can either end up not making good contact with the CPU die or the shim not taking any of the pressure. 3D printing them also has problems, considering that every printer has its own margin of error and making thickness consistent is practically impossible.

Consistent thickness on a single printer isn't overly difficult. It's just a matter of dialing in your printer with some calipers. I wonder if something like TPU would work and mitigate some of those issues, since it could act more like the foam pad.

RetroPC_King wrote on 2023-02-27, 12:56:

Today I got for cheap a ASUS A9550GE/TD/128M/A AGP8X video card, a ATI Radeon 9550 128MB, with active cooling. I wonder if it is the 128bit version.
Any opinions?

Should be 128 Bit since it has 8 x16 memory chips.
I have a 9550 (no-name), too, but haven't done very much with it yet. What I can say is that it is a bit slower than a Ti4200 in 3DMark 99 and 2000 when benched with an Athlon XP 2000+

Munx wrote on 2023-02-27, 12:29:

With shims you have to keep in mind that not all S462 CPUs are the same height, so you can either end up not making good contact with the CPU die or the shim not taking any of the pressure. 3D printing them also has problems, considering that every printer has its own margin of error and making thickness consistent is practically impossible.

It should be possible to tangent this when printed with flexible TPU material.
I haven't used this material for myself up to now, but in theory it should be flexible enough to let the cooler make good contact with some pressure applied while at the same time being a little bit higher than the die itself for its protection while mounting the cooler. At the same time evening out height tolerances from printing as well as different die heights should be possible.

Edit: Asininity was first bringing the TPU thought into the game. Must have overread this, sorry.

Asininity wrote on 2023-02-27, 12:52:

Do you have a caliper that you'd be able to measure the thickness? I'd like to get a board and try a 3D printed version, though I wonder if there'd be enough interest to put in a order with SendCutSend or similar.

I do, and i'm working from home today...

I'm reading 0.65mm (can someone confirm ?)

Plus several CPUs

No pictures from the side, but from what i see, the die is always slightly above the shim.

Yes, i had to empty my CPU box to find test candidates 😁

TrashPanda wrote on 2023-02-27, 11:52:

Should be possible to design ones that will work across a range of Athlon's/Pentium IIIs, then you only commission the model you need printed/cut. (I feel some printed ones would work well and certainly be cheap enough to warrant the effort)

It seems to work too. Here a PIII 1000.
The clips on the side are even the perfect size !

(Maybe for the topic in another thread ?)

acl wrote on 2023-02-27, 10:40:

Yes but even without the foam pads, using a shim is advised on these CPUs. I still have one or two. And i'm using one on my XP32 […]
Show full quote

TrashPanda wrote on 2023-02-27, 05:29:

The socket A CPUs don't have a IHS to protect the die so its a direct to die heatsink and AMD in their brilliance didn't really […]
Show full quote

Asininity wrote on 2023-02-27, 05:14:

What's the common cause of CPU death? I wonder if some 3D printer brackets could help with some of the cooler frustration.

I recently purchased an electric pump desoldering gun. It sounds like Socket A boards would be a good candidate for taking it through the paces.

The socket A CPUs don't have a IHS to protect the die so its a direct to die heatsink and AMD in their brilliance didn't really provide a great way to keep the cooler level or a way to prevent too much pressure when installing so one minor slip and it chips the die a bit too much pressure and it cracks it. now they did put soft felt pads on the CPU to help but they are woefully inadequate for the task. IIRC there were shims produced that keep the cooler level and from applying too much pressure but getting them nowadays is pretty much impossible.

a 3d printed bracket would be awesome, not sure how easy it would be to make one.

You can see on the picture below what I mean by pads and direct die.

Yes but even without the foam pads, using a shim is advised on these CPUs.
I still have one or two. And i'm using one on my XP3200+ (fragile and increasingly hard to find).
I think these can definitely be 3d printed, but probably more easily laser/plasma cut from a metal sheet.

Screenshot from 2023-02-27 11-36-49.png

Article about Socket A shims : https://www.legitreviews.com/what-is-a-shim-a … -you-use-one_64

AMD did actually toughen the core up a bit on XP and up, I skipped from K6-2 to palamino in the noughties, for my main rig, so on going backwards a year or two for cheap machines for fam, acquired spitfire durons and cracked those, derp, as I was XP calibrated.

Anyway, shims... they don't need to be metal, just dimensionally stable (This kinda contraindicates metal really as it expands on heating more than other materials) ... what I would do if I was good with a PCB layout package, would be to design a 4 piece to a board shim with perforations to crack them apart "PCB" and have JL or similar place make a bunch in the right thickness. No copper on the bottom or it might affect the bridges, or do some mirror thinking and precise measurement to see if you can align to bridges and connect them to do what you want.

acl wrote on 2023-02-27, 13:42:

I do, and i'm working from home today... […]
Show full quote

Asininity wrote on 2023-02-27, 12:52:

Do you have a caliper that you'd be able to measure the thickness? I'd like to get a board and try a 3D printed version, though I wonder if there'd be enough interest to put in a order with SendCutSend or similar.

I do, and i'm working from home today...

mesure.jpg

I'm reading 0.65mm (can someone confirm ?)

0.65 mm confirmed!

Yes, P III and Athlons have the same overall size.

What I can add:
Athlon XPs from Palomino up to Barton (Model 6, 8, 10) are spec'ed with a die height of 0.80-0.88mm. The original pads are 1.90mm max.
For P-III in FC-PGA package the die height is spec'ed at 0.787-0.889 mm.
So both almost match.
I'd say something with 1.00 to 1.10 mm made in TPU should be a good starting point.

acl wrote on 2023-02-27, 14:27:

(Maybe for the topic in another thread ?)

Yes, agree 😀

As someone who has messed around with rubber stick on feet and various other types of stick on foam for replacement bumpers... it's surprising how much force is required to compress even what seems to be low density stuff, when you have an area of more than about a square centimeter and apply the force with a flat surface.... i.e. stuff you can dent right down real easy with your thumbnail, may leave the heatsink hovering 0.2mm off your processor when the clamp is fully tensioned.

I would advise printing 0.1mm below the core for the bulk of the shim and just having a narrow ridge around sticking above it to compress.

BitWrangler has a point. Compressibility is significantly less when spread over a larger surface area; I hadn't considered this.

I visited the "Retro Computer Sale" which took place at the Centre for Computing History this Sunday (26-Feb-2023) in Cambridge. That's Cambridge, England BTW!

It's a small museum which has quite a number of exhibits relating to computer history in the UK. It is also "hands on" in that you can play games on a large range of home computers from the 1980's and consoles too. They hold events every now and again. Some years ago I attended a BBC Micro/Acorn Atom get together. Parking is limited. The museum is quite well hidden on a small industrial estate so it is easy to miss. Nevertheless Cambridge is a bit of a tourist hot spot in the UK, so if you are passing through you might want to plan a couple of hours for a visit. http://www.computinghistory.org.uk/. Entry was £10, the normal museum fee.

The area for sellers was quite small, but there was an impressive amount of retro stuff for sale, and prices were low. No Ebay prices here!

I was looking for old ISA cards and there were at least four sellers of old PC cards. More or less a quid each (I overpaid for the first 2 cards 😀 ).

I also bought an AOC brand CGA colour monitor for £50. What a bargain. They are really rare in the UK, don't appear often on Ebay, and when they do they are often damaged or not working, and cost a bomb. It works great !

Finally I bought a PC, again for £50. I didn't even look in the box. Came with an XT keyboard (DIN plug) which I really needed. I was alerted to what was in the unmarked box on the floor by a interested buyer, who only had £10 in cash on him. When he walked away I bought it. When I got it home I was surprised to see that it had a MFM hard disk, 384k plus I/O expansion card, original floppies and documentation (but was missing the MSDOS disk) and a VGA card. I discovered it was/is a Hyundai Super 16 8088. Interestingly the layout of my motherboard differs from those I have found online. HDD might be faulty. I have no 360k DOS boot disk. Interestingly the BIOS appears to be some hacked version to give compatibility with "CUB" monitors (a very well known brand of monitor used in UK schools in the 80's, usually with BBC micros).

I also got some other miscellaneous stuff: 2 mice; 8 old memory sticks; CF card to ISA adapters.

Other notable items seen: a Sharp MZ-80K in great condition (£120 at the start, reduced to £100 by lunchtime - what a steal? ) ; an Apple IIe in great condition for £100; a VT100 like terminal; seller of 3D printed joysticks/pads; UK101; PET; TI Color Computer; Atari 800, 400 and XL; plus other stuff I didn't even recognise. Ironically half of this stuff is already on display in the museum 😀

All in all, well worth the visit.

Am I totally wrong if I recall the Coppermine dies to be a lot let fragile than the Athlons?

Just picked up this Model 35 SX (386SX-20, 4MB, 40MBB HDD). I've been pretty much avoiding these "faster" PS/2 systems mainly because I'm not a masochist, but I couldn resist this one:

The thing that makes this interesting (and to my "taste") is that since it is "below 50", it has ISA bus and IDE HDD controller and regular molex connector from PSU. That means that it is almost as easy to expand as any other system, but this and model 40 are unfortunately relatively rare.

Only "IBM stuff" is pretty much the floppy drive. Incidentally, it is the only thing that needs repairs as it doesn't read disks currently. Hopefully recapping it and good clean helps. Amazingly, 40MB IBM HDD works just fine and some previous owner has already modded the Dallas RTC with coin battery already, so that is not an annoyance either.

TrashPanda wrote:

which was then ditched quickly by MS when Windows 7 arrived so the drivers had to be yet again changed.

Nope. Windows 7 was retro-compatible with the Vista model. I use Windows Vista graphics drivers on Windows 7 just fine. For example, on my T43 Thinkpad.

BitWrangler wrote on 2023-02-27, 14:56:

As someone who has messed around with rubber stick on feet and various other types of stick on foam for replacement bumpers... it's surprising how much force is required to compress even what seems to be low density stuff, when you have an area of more than about a square centimeter and apply the force with a flat surface.... i.e. stuff you can dent right down real easy with your thumbnail, may leave the heatsink hovering 0.2mm off your processor when the clamp is fully tensioned.

I would advise printing 0.1mm below the core for the bulk of the shim and just having a narrow ridge around sticking above it to compress.

It would probably be a lot easier to get consistent results thickness-wise if using a high resolution resin printer.

And with printed shims, you wouldn't have to worry about shorting anything out.

There are multiple versions of the metal shims as the bridges at least were in different locations on different models of the athlon CPUs

cyclone3d wrote on 2023-02-27, 19:07:

It would probably be a lot easier to get consistent results thickness-wise if using a high resolution resin printer. […]
Show full quote

BitWrangler wrote on 2023-02-27, 14:56:

As someone who has messed around with rubber stick on feet and various other types of stick on foam for replacement bumpers... it's surprising how much force is required to compress even what seems to be low density stuff, when you have an area of more than about a square centimeter and apply the force with a flat surface.... i.e. stuff you can dent right down real easy with your thumbnail, may leave the heatsink hovering 0.2mm off your processor when the clamp is fully tensioned.

I would advise printing 0.1mm below the core for the bulk of the shim and just having a narrow ridge around sticking above it to compress.

It would probably be a lot easier to get consistent results thickness-wise if using a high resolution resin printer.

And with printed shims, you wouldn't have to worry about shorting anything out.

There are multiple versions of the metal shims as the bridges at least were in different locations on different models of the athlon CPUs

I never tried to mod Athlon/Duron by altering the bridges, but i though the small contact points were protected by being a little under the surface, and covered with some coating.
Some sources about the famous "pencil trick" described some procedures to allow good contacts when closing a bridge. Generally by scraping the surface (more or less, not sure how to describe that). At least, this was not described as easy as bridging two soldering pads on a PCB.

Not sure that a metal shim would make contact and interact with the CPU.