Kahenraz wrote on 2025-01-30, 07:24:

It's really hard to separate those pressure fit pins. I destroyed a Pentium 2 cartridge trying to extract one once. Well done.

I much prefer the newer Coppermine design with the exposed core. Much easier to pop out the processor card for cleaning.

Yea its honestly a little sketchy, but I think once you can get the rhythm down it will be fine. There is probably a better tool out there than a flat head screw driver. RIP for your PII.

Been working on an old 8-Bit Microcomputer that I recently purchased - a Tangerine Oric-1. Sold as "untested" on eBay, this had a variety of nasty suprises such as no case screws and an unscrewed voltage regulator with it's heatsink and washers all loose inside the case.

After fixing the obvious problems as noted above, I thought it was still broken as it would only power on to a screen full of garbage. I tried everything I could think of, reseating chips, cleaning sockets and testing voltages. I was suspecting either bad RAM or some other bad chips here.

However, at one point whilst probing for voltages I accidentally shorted 2 pins on the CPU briefly and the machine sprung to life, booting into BASIC! The issue was the infamously bad reset circuit in the Oric. I discovered it will only boot to BASIC if you plug in the DC plug after already turning on the power adapter. If you turn on the power with the cable already connected, it's just a screen of garbage.

After getting it to boot up, I noticed immediately that the previous owner had upgraded the ROM to the Oric Atmos ROM, with version 1.1 of Extended BASIC.

There's 1 final issue that I haven't fully investigated - sometimes the keyboard misbehaves and/or certain keys fail to function. Pressing down hard on the keyboard seems to resolve it, until I knock the system or let it cool down. I'm guessing some cracked solder joints somewhere, I'll have to strip it all down and check it over.

But otherwise I've been loading a variety of games to test it all out, just .tap files converted to .wav via the cassette port for now. I have a Loci on the way for a more practical USB-based solution.

PTherapist wrote on 2025-01-30, 18:47:

Been working on an old 8-Bit Microcomputer that I recently purchased - a Tangerine Oric-1. Sold as "untested" on eBay, this had […]
Show full quote

Been working on an old 8-Bit Microcomputer that I recently purchased - a Tangerine Oric-1. Sold as "untested" on eBay, this had a variety of nasty suprises such as no case screws and an unscrewed voltage regulator with it's heatsink and washers all loose inside the case.

After fixing the obvious problems as noted above, I thought it was still broken as it would only power on to a screen full of garbage. I tried everything I could think of, reseating chips, cleaning sockets and testing voltages. I was suspecting either bad RAM or some other bad chips here.

However, at one point whilst probing for voltages I accidentally shorted 2 pins on the CPU briefly and the machine sprung to life, booting into BASIC! The issue was the infamously bad reset circuit in the Oric. I discovered it will only boot to BASIC if you plug in the DC plug after already turning on the power adapter. If you turn on the power with the cable already connected, it's just a screen of garbage.

After getting it to boot up, I noticed immediately that the previous owner had upgraded the ROM to the Oric Atmos ROM, with version 1.1 of Extended BASIC.

There's 1 final issue that I haven't fully investigated - sometimes the keyboard misbehaves and/or certain keys fail to function. Pressing down hard on the keyboard seems to resolve it, until I knock the system or let it cool down. I'm guessing some cracked solder joints somewhere, I'll have to strip it all down and check it over.

But otherwise I've been loading a variety of games to test it all out, just .tap files converted to .wav via the cassette port for now. I have a Loci on the way for a more practical USB-based solution.

Interesting! Sounds like a good rescue in progress to me! that thing with the plug and the timing of the power adapter is one of those strange things with some electronics where there should be no difference but some small factor makes all the difference.

so far though, you went from non functioning junk to quirky but functioning system. Hope the continued fixes go well from here

After finally finding an 8088 at the local recycler, I have managed to collect a majority of the processors I want to collect. At least for now. I ended up labeling all of them in an image I took for one of my friends, who was curious about what they all were.

AGP4LIfe? wrote on 2025-01-30, 07:00:

I decided to dive into the deep end and replace the thermal paste on a Slot A 600Mhz CPU. I really want to replace the factory […]
Show full quote

I decided to dive into the deep end and replace the thermal paste on a Slot A 600Mhz CPU. I really want to replace the factory TIM on my 950Mhz Thunderbird, so I used this 600 to "learn" the opening process..

I gotta say it doesn't feel great, those pins are tough and you really have apply a lot of pressure. However seeing the garage factory TIM on the inside of the 600 makes me wonder how hot these things really get, and further entices me to now replace the 950's.

I also found out my 600, has a 700 core. Interesting 🤔.

Everything went well and the CPU still works with no damage to the exterior case.Feels good. I think I'll practice one more time on a 650.

The attachment PXL_20250130_034728245~3.jpg is no longer available

The attachment PXL_20250130_034814466.jpg is no longer available

The attachment PXL_20250130_034746387.jpg is no longer available

What was the original TIM like? Was it still good enough for the cpu to cool properly or did it need replacing?

I have been trying to use the motherboard I fixed here, my Taken PCI400-4 which was working after a socket replacement and some loose pins around the southbridge resoldered: Re: Bought these (retro) hardware today
I was trying to test an ISA soundcard on it but it wouldn't boot reliably, getting stuck on / flipping between "C0" and "C1" post codes. Initially I thought it was another loose pin on one of the chipsets so spent ages reflowing the solder on those but to no avail.
Shoulda checked the basics at this point specifically clockspeed, but after some time figured out that the board would boot normally if heated around the CPU / northbridge area which is why I thought it was the northbridge. It would work sometimes when started up but it seemed whenever I left it for a while and it got cold, it would go back to being broken.

Using the hot air with a nozzle I was able to narrow it down to the area around the clock generator, the ARK 915A which has a bash mark on it:

Checking with the oscilloscope it was showing an unstable ~9mhz on the main cpu clock instead of the 33mhz it's supposed to be. Can't find a datasheet for the ARK 915A or Macronix MX8318 clock generators, but the MX8315 is pretty close. There was no clock on the 14.318mhz crystal so I thought perhaps it was the crystal that was bad, swapped that and no difference 😐

The conclusion seems to be that there's a broken bond wire inside the ARK 915A clock generator since borrowing the MX8318 from my Zida 4DPS has made the board reliable again.
When the ARK 915A clock generator is heated up then it will work with the crystal and do its job properly, but when cold the crystal oscillator is dead so I assume that the XO crystal output pin of the clock generator which drives the oscillator, only has a working connection when warm.
If I heat the ARK 915A then it works and it seems to keep working if it works when it starts up, so a heater atop the chip could work. Mechanical pressure also seems to work so I might try some kind of clip on top of the chip to keep pressure on it.
Or I could look at making some alternative clock generator work, I'd rather not do that and replacements for the ARK 915A or MX8318 both seem to be unavailable. Getting an alternative clock generator working would be a lot of work...

Anyways, now the board works again (albeit at the cost of the 4DPS which I'm not using right now) and it's set up for a Cyrix 5x86 processor I thought I'd try out my 5x86 chips - the IBM one works perfectly and seems to run very cool.
I have another one with the green heatsink and cyrix branding and that works just fine as well. Then there's the one I got on this M919 from a junk lot quite a while back: Re: Bought these (retro) hardware today
I've not tested that chip until now because pin R1 broke off when I was cleaning it, there was only the tiniest little dot of a connection poking through the ceramic. I tried soldering a pin directly to that and making it strong mechanically with superglue, but didn't want to test it because well... I don't think that's going to work.
I tried it this evening and the processor was dead, just "--" on the post card 🙁 As suspected, it did not work. Not possible to solder to something so tiny and have that stay connected to something as big as a cpu pin.
A couple of weeks ago though, while I was trying to go to sleep I was for some reason thinking about this and visualised a new way I could do that repair. I have some scrap PCBs that are 0.8mm thick with 2.54mm spaced through-holes, I used the mini grinder pen to clean off all the traces so now I've got some 6-pin segments of a PCB. Those slide over the pins at each corner of the CPU so that this modified CPU sits level in the socket.

I've soldered a long single strand of copper wire to pin R1's little silver dot of a connection and checked it's functioning by checking the resistance to ground of this missing R1 pin for Address line 28 and the adjacent address pin, both measure 17mega ohms to ground so the wire is working 😀
To get the PCB with the wire to sit flush there's a channel carved into the backside and a pin from a very dead Pentium 200MMX is soldered into the front-side of the pin repair PCB, which is better for this than a regular CPU pin because there's a larger peg that sits inside the Pentium's organic PCB. That peg gives a bit more mechanical strength and makes alignment easier.

The little PCB bits are held in place at the corners with superglue. The strand of wire is looped around from the outside of the CPU / pin repair PCB and is soldered to the pin at the front-side, so it now has no mechanical connection to the CPU pin but has an electrical connection.

Now the CPU works! 😁

edit: ugh now the floppy controller is giving me trouble... not this CPU that's causing it though!!

That's a fascinating way to repair a damaged CPU. Well done!

What was the original TIM like? Was it still good enough for the cpu to cool properly or did it need replacing?

Hard to say, I would say its borderline sufficient. It may or may not cause any problems, I don't have away of measuring core temp on these.
I had a blue screen while playing starcraft prior, but not yet post replacement. That could be almost anything though.

I know that I feel a lot better about gaming and 3DMark benching on it now that its Arctic silver 5.
If your going to overclock them I would 100% recommend replacement, or if its in a case with insufficient cooling in general.

The TIM wasn't broken down or anything, but it flake off like snow.

BetaC wrote on 2025-01-31, 21:22:

After finally finding an 8088 at the local recycler, I have managed to collect a majority of the processors I want to collect. At least for now. I ended up labeling all of them in an image I took for one of my friends, who was curious about what they all were.

The attachment IMG_4374 copy.jpg is no longer available

Nice collection!

Thermalwrong wrote on 2025-01-31, 21:52:

What was the original TIM like? Was it still good enough for the cpu to cool properly or did it need replacing? […]
Show full quote

AGP4LIfe? wrote on 2025-01-30, 07:00:

I decided to dive into the deep end and replace the thermal paste on a Slot A 600Mhz CPU. I really want to replace the factory […]
Show full quote

I decided to dive into the deep end and replace the thermal paste on a Slot A 600Mhz CPU. I really want to replace the factory TIM on my 950Mhz Thunderbird, so I used this 600 to "learn" the opening process..

I gotta say it doesn't feel great, those pins are tough and you really have apply a lot of pressure. However seeing the garage factory TIM on the inside of the 600 makes me wonder how hot these things really get, and further entices me to now replace the 950's.

I also found out my 600, has a 700 core. Interesting 🤔.

Everything went well and the CPU still works with no damage to the exterior case.Feels good. I think I'll practice one more time on a 650.

The attachment PXL_20250130_034728245~3.jpg is no longer available

The attachment PXL_20250130_034814466.jpg is no longer available

The attachment PXL_20250130_034746387.jpg is no longer available

What was the original TIM like? Was it still good enough for the cpu to cool properly or did it need replacing?

I have been trying to use the motherboard I fixed here, my Taken PCI400-4 which was working after a socket replacement and some loose pins around the southbridge resoldered: Re: Bought these (retro) hardware today
I was trying to test an ISA soundcard on it but it wouldn't boot reliably, getting stuck on / flipping between "C0" and "C1" post codes. Initially I thought it was another loose pin on one of the chipsets so spent ages reflowing the solder on those but to no avail.
Shoulda checked the basics at this point specifically clockspeed, but after some time figured out that the board would boot normally if heated around the CPU / northbridge area which is why I thought it was the northbridge. It would work sometimes when started up but it seemed whenever I left it for a while and it got cold, it would go back to being broken.

Using the hot air with a nozzle I was able to narrow it down to the area around the clock generator, the ARK 915A which has a bash mark on it:

The attachment IMG_5162 (Custom) (1).JPG is no longer available

Checking with the oscilloscope it was showing an unstable ~9mhz on the main cpu clock instead of the 33mhz it's supposed to be. Can't find a datasheet for the ARK 915A or Macronix MX8318 clock generators, but the MX8315 is pretty close. There was no clock on the 14.318mhz crystal so I thought perhaps it was the crystal that was bad, swapped that and no difference 😐

The conclusion seems to be that there's a broken bond wire inside the ARK 915A clock generator since borrowing the MX8318 from my Zida 4DPS has made the board reliable again.
When the ARK 915A clock generator is heated up then it will work with the crystal and do its job properly, but when cold the crystal oscillator is dead so I assume that the XO crystal output pin of the clock generator which drives the oscillator, only has a working connection when warm.
If I heat the ARK 915A then it works and it seems to keep working if it works when it starts up, so a heater atop the chip could work. Mechanical pressure also seems to work so I might try some kind of clip on top of the chip to keep pressure on it.
Or I could look at making some alternative clock generator work, I'd rather not do that and replacements for the ARK 915A or MX8318 both seem to be unavailable. Getting an alternative clock generator working would be a lot of work...

Anyways, now the board works again (albeit at the cost of the 4DPS which I'm not using right now) and it's set up for a Cyrix 5x86 processor I thought I'd try out my 5x86 chips - the IBM one works perfectly and seems to run very cool.
I have another one with the green heatsink and cyrix branding and that works just fine as well. Then there's the one I got on this M919 from a junk lot quite a while back: Re: Bought these (retro) hardware today
I've not tested that chip until now because pin R1 broke off when I was cleaning it, there was only the tiniest little dot of a connection poking through the ceramic. I tried soldering a pin directly to that and making it strong mechanically with superglue, but didn't want to test it because well... I don't think that's going to work.
I tried it this evening and the processor was dead, just "--" on the post card 🙁 As suspected, it did not work. Not possible to solder to something so tiny and have that stay connected to something as big as a cpu pin.
A couple of weeks ago though, while I was trying to go to sleep I was for some reason thinking about this and visualised a new way I could do that repair. I have some scrap PCBs that are 0.8mm thick with 2.54mm spaced through-holes, I used the mini grinder pen to clean off all the traces so now I've got some 6-pin segments of a PCB. Those slide over the pins at each corner of the CPU so that this modified CPU sits level in the socket.

The attachment 5x86_Prelim_199507-brokenpin.png is no longer available

I've soldered a long single strand of copper wire to pin R1's little silver dot of a connection and checked it's functioning by checking the resistance to ground of this missing R1 pin for Address line 28 and the adjacent address pin, both measure 17mega ohms to ground so the wire is working 😀
To get the PCB with the wire to sit flush there's a channel carved into the backside and a pin from a very dead Pentium 200MMX is soldered into the front-side of the pin repair PCB, which is better for this than a regular CPU pin because there's a larger peg that sits inside the Pentium's organic PCB. That peg gives a bit more mechanical strength and makes alignment easier.

The attachment cyrix-5x86-pin-repair.jpg is no longer available

The little PCB bits are held in place at the corners with superglue. The strand of wire is looped around from the outside of the CPU / pin repair PCB and is soldered to the pin at the front-side, so it now has no mechanical connection to the CPU pin but has an electrical connection.

The attachment 5x86 cpu and pci400 mainboard repaired.JPG is no longer available

Now the CPU works! 😁

edit: ugh now the floppy controller is giving me trouble... not this CPU that's causing it though!!

Professional work!

DarthSun wrote on 2025-02-01, 16:29:

BetaC wrote on 2025-01-31, 21:22:

After finally finding an 8088 at the local recycler, I have managed to collect a majority of the processors I want to collect. At least for now. I ended up labeling all of them in an image I took for one of my friends, who was curious about what they all were.

The attachment IMG_4374 copy.jpg is no longer available

Nice collection!

Thanks, makes it feel more worth the effort.

BetaC wrote on 2025-02-01, 20:58:

DarthSun wrote on 2025-02-01, 16:29:

BetaC wrote on 2025-01-31, 21:22:

After finally finding an 8088 at the local recycler, I have managed to collect a majority of the processors I want to collect. At least for now. I ended up labeling all of them in an image I took for one of my friends, who was curious about what they all were.

The attachment IMG_4374 copy.jpg is no longer available

Nice collection!

Thanks, makes it feel more worth the effort.

That's right. About 70% of my collection was on the table.

DarthSun wrote on 2025-02-01, 21:10:

BetaC wrote on 2025-02-01, 20:58:

DarthSun wrote on 2025-02-01, 16:29:

Nice collection!

Thanks, makes it feel more worth the effort.

That's right. About 70% of my collection was on the table.

The attachment akx5g_3.jpg is no longer available

Oh, wow! Nice collection 🙂.

Recapped another sound card 😁. These are all Elna caps except for two black 470uF Rubycon caps on the top-right. I'll install this on my 486.

dominusprog wrote on 2025-02-01, 21:27:

Oh, wow! Nice collection 🙂. […]
Show full quote

DarthSun wrote on 2025-02-01, 21:10:

BetaC wrote on 2025-02-01, 20:58:

Thanks, makes it feel more worth the effort.

That's right. About 70% of my collection was on the table.

The attachment akx5g_3.jpg is no longer available

Oh, wow! Nice collection 🙂.

Recapped another sound card 😁. These are all Elna caps except for two black 470uF Rubycon caps on the top-right. I'll install this on my 486.

The attachment IMG_20250201_154750.jpg is no longer available

It was a good work!
I also have such an Opti card, but I didn't swap capacitors on it.

Thermalwrong wrote on 2025-01-31, 21:52:

AGP4LIfe? wrote on 2025-01-30, 07:00:

I decided to dive into the deep end and replace the thermal paste on a Slot A 600Mhz CPU. I really want to replace the factory […]
Show full quote

I decided to dive into the deep end and replace the thermal paste on a Slot A 600Mhz CPU. I really want to replace the factory TIM on my 950Mhz Thunderbird, so I used this 600 to "learn" the opening process..

I gotta say it doesn't feel great, those pins are tough and you really have apply a lot of pressure. However seeing the garage factory TIM on the inside of the 600 makes me wonder how hot these things really get, and further entices me to now replace the 950's.

I also found out my 600, has a 700 core. Interesting 🤔.

Everything went well and the CPU still works with no damage to the exterior case.Feels good. I think I'll practice one more time on a 650.

The attachment PXL_20250130_034728245~3.jpg is no longer available

The attachment PXL_20250130_034814466.jpg is no longer available

The attachment PXL_20250130_034746387.jpg is no longer available

What was the original TIM like? Was it still good enough for the cpu to cool properly or did it need replacing?

Feeling a little braver after I completed the 600 Replacement I decided to step it up and replace the TIM on my K7-750
Here is a couple of pictures of the factory TIM as is before I removed it. Interestingly enough they use thermal paste for the Cache but a crappy thermal pad for the core?

I really need to get my self a Gold finger device, because this K7-750 Actually had a 900Mhz Core!!! That's crazy.
Since its underclocked 150Mhz. I could easily change the multiplier and be running a K7-900.
I am beginning to wonder if call these Slot A CPU's have higher than rated cores..

DarthSun wrote on 2025-02-01, 16:31:

Thermalwrong wrote on 2025-01-31, 21:52:

What was the original TIM like? Was it still good enough for the cpu to cool properly or did it need replacing? […]
Show full quote

AGP4LIfe? wrote on 2025-01-30, 07:00:

I decided to dive into the deep end and replace the thermal paste on a Slot A 600Mhz CPU. I really want to replace the factory […]
Show full quote

I decided to dive into the deep end and replace the thermal paste on a Slot A 600Mhz CPU. I really want to replace the factory TIM on my 950Mhz Thunderbird, so I used this 600 to "learn" the opening process..

I gotta say it doesn't feel great, those pins are tough and you really have apply a lot of pressure. However seeing the garage factory TIM on the inside of the 600 makes me wonder how hot these things really get, and further entices me to now replace the 950's.

I also found out my 600, has a 700 core. Interesting 🤔.

Everything went well and the CPU still works with no damage to the exterior case.Feels good. I think I'll practice one more time on a 650.

The attachment PXL_20250130_034728245~3.jpg is no longer available

The attachment PXL_20250130_034814466.jpg is no longer available

The attachment PXL_20250130_034746387.jpg is no longer available

What was the original TIM like? Was it still good enough for the cpu to cool properly or did it need replacing?

I have been trying to use the motherboard I fixed here, my Taken PCI400-4 which was working after a socket replacement and some loose pins around the southbridge resoldered: Re: Bought these (retro) hardware today
I was trying to test an ISA soundcard on it but it wouldn't boot reliably, getting stuck on / flipping between "C0" and "C1" post codes. Initially I thought it was another loose pin on one of the chipsets so spent ages reflowing the solder on those but to no avail.
Shoulda checked the basics at this point specifically clockspeed, but after some time figured out that the board would boot normally if heated around the CPU / northbridge area which is why I thought it was the northbridge. It would work sometimes when started up but it seemed whenever I left it for a while and it got cold, it would go back to being broken.

Using the hot air with a nozzle I was able to narrow it down to the area around the clock generator, the ARK 915A which has a bash mark on it:

The attachment IMG_5162 (Custom) (1).JPG is no longer available

Checking with the oscilloscope it was showing an unstable ~9mhz on the main cpu clock instead of the 33mhz it's supposed to be. Can't find a datasheet for the ARK 915A or Macronix MX8318 clock generators, but the MX8315 is pretty close. There was no clock on the 14.318mhz crystal so I thought perhaps it was the crystal that was bad, swapped that and no difference 😐

The conclusion seems to be that there's a broken bond wire inside the ARK 915A clock generator since borrowing the MX8318 from my Zida 4DPS has made the board reliable again.
When the ARK 915A clock generator is heated up then it will work with the crystal and do its job properly, but when cold the crystal oscillator is dead so I assume that the XO crystal output pin of the clock generator which drives the oscillator, only has a working connection when warm.
If I heat the ARK 915A then it works and it seems to keep working if it works when it starts up, so a heater atop the chip could work. Mechanical pressure also seems to work so I might try some kind of clip on top of the chip to keep pressure on it.
Or I could look at making some alternative clock generator work, I'd rather not do that and replacements for the ARK 915A or MX8318 both seem to be unavailable. Getting an alternative clock generator working would be a lot of work...

Anyways, now the board works again (albeit at the cost of the 4DPS which I'm not using right now) and it's set up for a Cyrix 5x86 processor I thought I'd try out my 5x86 chips - the IBM one works perfectly and seems to run very cool.
I have another one with the green heatsink and cyrix branding and that works just fine as well. Then there's the one I got on this M919 from a junk lot quite a while back: Re: Bought these (retro) hardware today
I've not tested that chip until now because pin R1 broke off when I was cleaning it, there was only the tiniest little dot of a connection poking through the ceramic. I tried soldering a pin directly to that and making it strong mechanically with superglue, but didn't want to test it because well... I don't think that's going to work.
I tried it this evening and the processor was dead, just "--" on the post card 🙁 As suspected, it did not work. Not possible to solder to something so tiny and have that stay connected to something as big as a cpu pin.
A couple of weeks ago though, while I was trying to go to sleep I was for some reason thinking about this and visualised a new way I could do that repair. I have some scrap PCBs that are 0.8mm thick with 2.54mm spaced through-holes, I used the mini grinder pen to clean off all the traces so now I've got some 6-pin segments of a PCB. Those slide over the pins at each corner of the CPU so that this modified CPU sits level in the socket.

The attachment 5x86_Prelim_199507-brokenpin.png is no longer available

I've soldered a long single strand of copper wire to pin R1's little silver dot of a connection and checked it's functioning by checking the resistance to ground of this missing R1 pin for Address line 28 and the adjacent address pin, both measure 17mega ohms to ground so the wire is working 😀
To get the PCB with the wire to sit flush there's a channel carved into the backside and a pin from a very dead Pentium 200MMX is soldered into the front-side of the pin repair PCB, which is better for this than a regular CPU pin because there's a larger peg that sits inside the Pentium's organic PCB. That peg gives a bit more mechanical strength and makes alignment easier.

The attachment cyrix-5x86-pin-repair.jpg is no longer available

The little PCB bits are held in place at the corners with superglue. The strand of wire is looped around from the outside of the CPU / pin repair PCB and is soldered to the pin at the front-side, so it now has no mechanical connection to the CPU pin but has an electrical connection.

The attachment 5x86 cpu and pci400 mainboard repaired.JPG is no longer available

Now the CPU works! 😁

edit: ugh now the floppy controller is giving me trouble... not this CPU that's causing it though!!

Professional work!

Thanks! It's been sitting on the to-do list for almost 3 years 😁

AGP4LIfe? wrote on 2025-02-02, 03:18:

Feeling a little braver after I completed the 600 Replacement I decided to step it up and replace the TIM on my K7-750 Here is […]
Show full quote

Thermalwrong wrote on 2025-01-31, 21:52:

AGP4LIfe? wrote on 2025-01-30, 07:00:

I decided to dive into the deep end and replace the thermal paste on a Slot A 600Mhz CPU. I really want to replace the factory […]
Show full quote

I decided to dive into the deep end and replace the thermal paste on a Slot A 600Mhz CPU. I really want to replace the factory TIM on my 950Mhz Thunderbird, so I used this 600 to "learn" the opening process..

I gotta say it doesn't feel great, those pins are tough and you really have apply a lot of pressure. However seeing the garage factory TIM on the inside of the 600 makes me wonder how hot these things really get, and further entices me to now replace the 950's.

I also found out my 600, has a 700 core. Interesting 🤔.

Everything went well and the CPU still works with no damage to the exterior case.Feels good. I think I'll practice one more time on a 650.

The attachment PXL_20250130_034728245~3.jpg is no longer available

The attachment PXL_20250130_034814466.jpg is no longer available

The attachment PXL_20250130_034746387.jpg is no longer available

What was the original TIM like? Was it still good enough for the cpu to cool properly or did it need replacing?

Feeling a little braver after I completed the 600 Replacement I decided to step it up and replace the TIM on my K7-750
Here is a couple of pictures of the factory TIM as is before I removed it. Interestingly enough they use thermal paste for the Cache but a crappy thermal pad for the core?

I really need to get my self a Gold finger device, because this K7-750 Actually had a 900Mhz Core!!! That's crazy.
Since its underclocked 150Mhz. I could easily change the multiplier and be running a K7-900.
I am beginning to wonder if call these Slot A CPU's have higher than rated cores..

The attachment PXL_20250202_011823307.jpg is no longer available

The attachment PXL_20250202_011827973.jpg is no longer available

The attachment PXL_20250202_013229595.jpg is no longer available

Hmm, that type of thermal pad doesn't look too bad to leave as-is. I have a couple of Slot-A CPUs but they're lower speed ones so I doubt I'd find a faster CPU in mine. That green core looks so reminiscent of the Thunderbird socket-A CPUs with the aluminium interconnects, like my T-bird 800 😀 Never managed to get a purple core (copper interconnects) so just skipped generations until barton from what I recall.
That thing must have some real overclocking potential, but do check what clock rating the L2 cache is good for, perhaps that's the limiting factor?

------
Today I resolved a problem I've been trying to fix for almost 5 years! I bought an Orchid Righteous 3D Voodoo 1 card for parts back in 2020 and had to learn a fair bit before I could fix it. I recall from the auction listing that they said it gave out mid-gaming-session where it would just crash or display a mess. This was caused by a bad / missing trace on the Voodoo 1's TMU chip. In the end I traced out and checked every pin, removed and replaced RAM, swapped the TMU RAM to the FBI spots and the fault was still with the TMU so it's not the RAM chips, it's the TMU...

Some time ago I fixed up this Voodoo 1: Re: What retro activity did you get up to today?
A few days back I got the proper SMD ferrite beads to repair to the power supply to that card's RAMDAC, but no matter what I did the Ti TVP3409 just always gives a noisy output. Then I put the card in the ultrasonic cleaner and it had new faults with the RAMDAC... then I tried to fix that and now the computer will not POST at all with this 3dfx card installed. It's a bridged pin on the FBI or the TMU, the PCB is such a state that now I give up on that card. It's a donor card now and will be used only for display purposes.

But that opens up an opportunity, the Orchid Righteous 3D needs a more functional TMU. Right now I have two cards that are faulty, I can make one good one though. So with the hot air I removed the TMU from the now quite broken 3d128 near-reference design 3dfx card and installed that onto the Orchid Righteous 3D.

Initially I thought I'd killed the card because it was giving no mojo output. Resoldered around the FBI in case that got some pins loose during the hot-air reflow, after doing that it could communicate properly but the TMU would not respond with its RAM count and the computer would crash when starting games. Looked close up and one corner of the chip had a bunch of loose pins, resoldered those without the scope late at night and then MOJO output was good with 2MB on FBI and 2MB on TMU, hooray!
Now the card can start games but there's black diamond shapes in tomb raider and party mode in unreal - but with the TMU transplant the card is working better than it ever did with the original! It can now detect its full 2MB of memory, it could only detect 1MB before now matter how much I tried changing.
Couldn't figure out what was causing the artifacting and blamed the RAM, which I can't hot-air when its late at night so went to bed.

Looked at the card again and within 30 seconds spotted a couple of pins that didn't look right on the area I was soldering at last night:

Once the bridge was cleared then the card started working properly 😁 At last!!! This took multiple years!!! Years of testing and swapping RAM and resoldering chips and checking connections and beeping out traces, it was the TMU that was bad all along!

Now the TMU has a heatsink - the 97xx datecode 3dfx chips get really hot, the TMU definitely benefits from a heatsink and perhaps it needs some active cooling. So glad it's fixed, at long last I have a relay-clicker 3dfx card and the RAMDAC output is perfect on this one. So 2x bad cards have now became 1x good card and 1x display card.

Oh also, I got some new flux - chipquik CQ4300 which is in a little tub and has to be applied with a stick rather than a syringe, it's good stuff - cleans right off with gentle brushing under sufficiently hot water so the card looks really clean now

I was watching arottenbit & Machinae Supremacy & MASTER BOOT RECORD @Helsinki last night. I think it's HARD enough, if you know what I mean. arottenbit with his Game Boy was a true killer. You could really feel the beat.

AGP4LIfe? wrote on 2025-02-02, 03:18:

I really need to get my self a Gold finger device, because this K7-750 Actually had a 900Mhz Core!!! That's crazy.
Since its underclocked 150Mhz. I could easily change the multiplier and be running a K7-900.
I am beginning to wonder if call these Slot A CPU's have higher than rated cores..

Will the goldfinger device also be able too change cache multiplier? (Well, divider)

Bought a Geforce 6800 Ultra from eBay in great condition. It came from some sort of old server farm.
It had a different bios installed on it, then the one that was listed (printed) on the back of the card.

It had the same brand but a down-clocked Geforce 6800 GT Bios installed on it. Idk why this would be done, but perhaps for better power efficiency and part longevity in the farm.

So I flashed it back to the 6800 Ultra bios and it runs just fine. Just a random oddity, I had never seen before.

H3nrik V! wrote on 2025-02-02, 13:00:

AGP4LIfe? wrote on 2025-02-02, 03:18:

I really need to get my self a Gold finger device, because this K7-750 Actually had a 900Mhz Core!!! That's crazy.
Since its underclocked 150Mhz. I could easily change the multiplier and be running a K7-900.
I am beginning to wonder if call these Slot A CPU's have higher than rated cores..

Will the goldfinger device also be able too change cache multiplier? (Well, divider)

mmm I don't think so, but im not super familiar with them.

Thermalwrong wrote on 2023-11-29, 03:08:

Heh I give up for now, this took so dang long. I lost so many of the smaller capacitors & resistors somehow and it took ages to […]
Show full quote

Heh I give up for now, this took so dang long. I lost so many of the smaller capacitors & resistors somehow and it took ages to find them, bend the legs, trim the legs. Check the values etc.

The attachment ct1350-rebuilt.JPG is no longer available

Here's how it fits in the slot of one of my mini SBC computers. I wanted to just use it in the big testing PC since that's got the remnants of a real case to check fitment, but for this initial testing I needed -5v and to be able to see power consumption

The attachment IMG_2566 (Custom).JPG is no longer available

Good news and bad news...... the ISA connector works though and it's not falling off. The glue is more of a space filler, those metal legs inside the fibreglass are giving all the important mechanical strength it needs, more now with the traces soldered on too.

A blast(er) from the past!

While avoiding some other projects I took another look at my Creative SoundBlaster CT1350 that came from a scrapper where they'd cut the ISA connector off to get the precious gold. Well over time I depopulated the card, derived a complete PCB design and schematic for it (still not got that fabbed yet), then put a replacement ISA connector onto this original 90s PCB and repopulated it.

I could not figure out why the card's Adlib / OPL2 side would work but the SoundBlaster DSP would not detect at all. Last time I was thinking about scoping / logic analysing the address or data lines, or swapping out the main CT1336 bus logic chip but those are all hard options so I've ignored this card for well over a year now...

Today I decided to start verifying the connections of the IRQ pins, the Address pins - I can use the boardview I created to confirm the start and end of any trace:

Starting with the IRQs I found that IRQ 3's trace was broken where the jumper connects to the PCB trace, must've broken when I removed the jumper and since it's underneath the jumper plastic I can't spot it at all.
The Address pins don't go to many places except the CT1336 and all the address traces were fine... So this time I next looked at the Jumpers to set the Address to "22x" or "24x" and this trace that connects the other end of that jumper is broken!

Because it couldn't listen on either 220 or 240, the soundblaster DSP audio and all communication with the soundblaster's DSP was broken.

It seems that when I pulled the jumper when depopulating the card the hole & trace got damaged, which was then hidden by the jumper plastic on top of it. Soldering a wire on top works well:

No way anyone could've spotted this from looking at pictures unfortunately.

And here it is working playing Doom just as a sound test on the motherboard that I fixed, with the CPU that I fixed 😁

Now I'm in better spirits about the CT1350 - that means the CT1336 is resilient since I thought that had failed. With this boardview I can make reproduction PCBs like the Snark Barker, if I finish up the solder mask, I may even upload the full kicad project for it at some point since it seems to be correct 😀
See this thread for more on the SoundBlaster CT1350 boardview: Taking a repair to the extreme - Remaking Soundblaster 2.0 CT1350B PCB