without heat spreader* oopsies!!!

The proof of the pudding is in the eating. I have a slightly chipped 600E which still works and loads of chipped dead Athlons.

One aspect (and probably the only one) the Willamete P4 outperformed the AMD Athlon was in running without a heatsink. A video back in 03 or 04 on Tom's Hardware went viral as the sacred smoke escaped the socketed Athlon with its heatsink removed while the P4 throttled back to an even slower pace than it's stock performance when the heatsink was popped off.

I've never seen anyone call the die a crystal before (although, I suppose, technically it is).

As long as the crack is only very close to the edge, it won't affect it. The functional part of a die never goes right to the edge, allowing minor variations in the cutting from the wafer.

flupke11 wrote on 2025-04-09, 16:59:

One aspect (and probably the only one) the Willamete P4 outperformed the AMD Athlon was in running without a heatsink. A video back in 03 or 04 on Tom's Hardware went viral as the sacred smoke escaped the socketed Athlon with its heatsink removed while the P4 throttled back to an even slower pace than it's stock performance when the heatsink was popped off.

Can confirm… Just burned yesterday an Athlon XP-M CPU, because I felt too lazy to actually secure the heatsink for a short test that I was doing. Even though the die/core wasn’t completely exposed, it had some contact with the heatsink through the Arctic MX-4 paste I was using, it still went up in smoke in a matter of 10 - 20 seconds after powering it on… Fortunately, I have a few more identical CPUs, but I still feel bad about it, since it was a very good overclocker. 🙁

bloodem wrote on 2025-04-09, 20:09:

flupke11 wrote on 2025-04-09, 16:59:

One aspect (and probably the only one) the Willamete P4 outperformed the AMD Athlon was in running without a heatsink. A video back in 03 or 04 on Tom's Hardware went viral as the sacred smoke escaped the socketed Athlon with its heatsink removed while the P4 throttled back to an even slower pace than it's stock performance when the heatsink was popped off.

Can confirm… Just burned yesterday an Athlon XP-M CPU, because I felt too lazy to actually secure the heatsink for a short test that I was doing. Even though the die/core wasn’t completely exposed, it had some contact with the heatsink through the Arctic MX-4 paste I was using, it still went up in smoke in a matter of 10 - 20 seconds after powering it on… Fortunately, I have a few more identical CPUs, but I still feel bad about it, since it was a very good overclocker. 🙁

that might be one of the saddest things I ever heard in computer disassembly, would be a bit mad if that happened...

I remember at the time of the Athlon Thunderbird CPUs release, there were heaps of people breaking their CPU die because they didn't carefully put their heatsink on.

Depends on your luck: Crimes Against Silicon: Scuffed CPU/GPU cores

bloodem wrote on 2025-04-09, 20:09:

flupke11 wrote on 2025-04-09, 16:59:

One aspect (and probably the only one) the Willamete P4 outperformed the AMD Athlon was in running without a heatsink. A video back in 03 or 04 on Tom's Hardware went viral as the sacred smoke escaped the socketed Athlon with its heatsink removed while the P4 throttled back to an even slower pace than it's stock performance when the heatsink was popped off.

Can confirm… Just burned yesterday an Athlon XP-M CPU, because I felt too lazy to actually secure the heatsink for a short test that I was doing. Even though the die/core wasn’t completely exposed, it had some contact with the heatsink through the Arctic MX-4 paste I was using, it still went up in smoke in a matter of 10 - 20 seconds after powering it on… Fortunately, I have a few more identical CPUs, but I still feel bad about it, since it was a very good overclocker. 🙁

Quite a common thing to have happen back in the day too, not many realised that the Athlons were that sensitive to improper cooling setups, P3s however were not and so long as they had some contact they would survive.

I cant remember if the P3 had thermal regulation on die to prevent overheating but I know for certain the Athlons didn't so they would happily run at full tilt with no or little cooling till they burned up.

Just checked and the P3 did have thermal throttling on die so that does explain their tendency of surviving, wonder why AMD decided Athlons didn't need it.

Trashbytes wrote on 2025-04-10, 10:56:

Quite a common thing to have happen back in the day too, not many realised that the Athlons were that sensitive to improper cooling setups, P3s however were not and so long as they had some contact they would survive.

Yeah... The problem is due to those four pesky black spacers located on each corner of the Athlon / Athlon XP CPUs. These spacers are actually quite a bit taller than the die itself, which is why, if you place the heatsink on top without properly securing it, the die will not have proper contact with the heatsink...

Trashbytes wrote on 2025-04-10, 10:56:

I cant remember if the P3 had thermal regulation on die to prevent overheating but I know for certain the Athlons didn't so they would happily run at full tilt with no or little cooling till they burned up.

Even the original Pentium 60 tends to survive extreme overheating (ask me how I know!), so I definitely expect all later Pentiums to also have some sort of thermal protection. In fact, in their famous video, Tom's Hardware also tested a Pentium 3 1 GHz, and it survived.

so I definitely expect all later Pentiums to also have some sort of thermal protection.

You can easily burn Pentium III Katmai without cooler. That's obviously hard to achieve unintentionally on Slot 1 CPUs though. Also, I wouldn't put my faith on anything but Intel chipset boards of that era, when it comes to thermal protection.

bloodem wrote on 2025-04-10, 15:25:

Trashbytes wrote on 2025-04-10, 10:56:

Quite a common thing to have happen back in the day too, not many realised that the Athlons were that sensitive to improper cooling setups, P3s however were not and so long as they had some contact they would survive.

Yeah... The problem is due to those four pesky black spacers located on each corner of the Athlon / Athlon XP CPUs. These spacers are actually quite a bit taller than the die itself, which is why, if you place the heatsink on top without properly securing it, the die will not have proper contact with the heatsink...

Those "pesky" black spacers are foam, so compress down when the hsf is attached properly. Their purpose was to help prevent damage to the die, by evening out pressure. Turn on a computer of any era after the 486 without properly attaching a heatsink and it's your own fault when it's damaged, unfortunately. People still managed, in some cases, to return them and get new replacements, as who can prove you mishandled it, deliberately or accidentally? The foam spacers were not on the earliest socket A CPUs, but were added in order to lessen the risk of damage.

debs3759 wrote on 2025-04-10, 20:00:

Those "pesky" black spacers are foam, so compress down when the hsf is attached properly. Their purpose was to help prevent damage to the die, by evening out pressure.

Uhm... I do know what they are and what their purpose is. In the past, when I wanted to do similar tests with socket A CPUs, I remembered that the spacers are there, and that they could potentially prevent good contact with the die. Unfortunately, this time I completely ignored them/forgot they exist...

debs3759 wrote on 2025-04-10, 20:00:

Turn on a computer of any era after the 486 without properly attaching a heatsink and it's your own fault when it's damaged, unfortunately.

That is not true at all. On most CPUs/sockets, gravity is enough to have a good cooler contact, especially when talking about 15 minute tests. I've done it with virtually all CPUs including, but not limited to, early Pentiums, the AMD K6-2+ (even with the heatsink removed!), Coppermine, Tualatin, Athlon 64, Core 2 Duo, etc.
Also, thanks for letting me know that it's my own fault, I really did not know this until you pointed it out. 😉 I thought it was the CPU's fault - it just suddenly decided to commit suicide. 😁

The Serpent Rider wrote on 2025-04-10, 17:14:

You can easily burn Pentium III Katmai without cooler. That's obviously hard to achieve unintentionally on Slot 1 CPUs though. Also, I wouldn't put my faith on anything but Intel chipset boards of that era, when it comes to thermal protection.

Yes, I think you are right - at least when it comes to the importance of the actual motherboard/chipset.
All Pentium 3 CPUs (including the Katmai) have an on-die thermal diode, but it's probably up to the motherboard to take action based on the readings from said diode.

The Athlon CPUs also have an on-die thermal diode, but it doesn't help them much, so this can be another clue pointing to the importance of the motherboard/chipset.

I have a 933 Pentium!!! which doesn't POST anymore, after the heat sink slipped on one side during mounting. I'm pretty sure it worked before, and there was some "glittery" dust around the edge, which I believe is silicon from the die.

I did use that 933, however since they apparently were pretty popular around here - it's the one speed I own most of ..

FC-PGA dies were a lot tougher (and/or had wider tolerances) than SoA dies, but core crunch was still a thing - I recall once losing a P3-700E (when those things were still worth significant money and I was a poor student 😦 ) due to slipping when mounting a particularly stubborn heatsink. I very much understand why heatspreaders came back into use afterwards - slightly worse thermal performance is in most cases a very acceptable tradeoff to virtually rule out accidents like that. Ricers can still de-lid and to get direct contact for their lN2 buckets 😉

Chips on the die was common on the Flip Chip Pentium 3s.

You might get lucky and it still be ok, sometimes it killed them stone dead. You dont know until it happens.

H3nrik V! wrote on 2025-04-11, 08:17:

I did use that 933, however since they apparently were pretty popular around here - it's the one speed I own most of ..

The price difference (at least in Europe) between the 1.0GHz and the 933 was significant, solely due to the cred factor of owning a giga chip. The 933 performed almost as well as the 1.0GHz and was often a better overclocker. Both were trumped (can we still use that?) by the Athlon being a superior chip, albeit with a few unfortunate drawbacks as we witnessed here.

We happily let ourselves fool by marketing ploys...

debs3759 wrote on 2025-04-09, 18:58:

I've never seen anyone call the die a crystal before (although, I suppose, technically it is).

Yeah it’s amusing to see all these SciFi shows showing crystal based technology where people or Aliens swap large crystals to load a new program or start an engine, and because they always look like large naturally occurring crystals it is easy to forget that our technology is largely based on silicon crystals too.

bloodem wrote on 2025-04-11, 05:03:

Yes, I think you are right - at least when it comes to the importance of the actual motherboard/chipset. All Pentium 3 CPUs (inc […]
Show full quote

The Serpent Rider wrote on 2025-04-10, 17:14:

You can easily burn Pentium III Katmai without cooler. That's obviously hard to achieve unintentionally on Slot 1 CPUs though. Also, I wouldn't put my faith on anything but Intel chipset boards of that era, when it comes to thermal protection.

Yes, I think you are right - at least when it comes to the importance of the actual motherboard/chipset.
All Pentium 3 CPUs (including the Katmai) have an on-die thermal diode, but it's probably up to the motherboard to take action based on the readings from said diode.

The Athlon CPUs also have an on-die thermal diode, but it doesn't help them much, so this can be another clue pointing to the importance of the motherboard/chipset.

So, I did a bit of digging, and the issue is best described by Anand here and here.

The relevant quotes from the articles:

The K7VTA3 does not support reading temperatures directly off of the Athlon XP's die; instead it resorts to an external thermistor like most motherboards.

Winbond 83782D The 83782D is the success to the 83781D and offers one key advantage - the ability to read CPU temperature from t […]
Show full quote

Winbond 83782D
The 83782D is the success to the 83781D and offers one key advantage - the ability to read CPU temperature from the on-die thermal diode of any 0.25 micron or 0.18 micron Intel CPU. Reading from the on-die thermal diode offers the most accurate CPU temperature possible.
Otherwise, the 83782D offers similar specs - monitoring for 9 voltages, 3 temperatures, and 3 fan speeds. Case intrusion is still supported, as is the CPU VID detection of the correct Vcore. Those 2 additional voltages allow for monitoring of just about every voltage in the system, usually +/-5V, +/-12V, +3.3V, Vcore, VTT, +5Vsb, and Vbat.
Most motherboard manufacturers that choose the 83782D are wise enough to read the CPU temperature from the on-die thermal diode. However, a surprising number still use this in conjunction with an external thermistor near the CPU slot/socket, which leads to slower reaction time for temperature changes.

So, as far back as the early year 2000 (probably even earlier than that), there were monitoring ICs (such as the Winbond 83782D), that were capable of reading the on-die thermal diode of most Intel CPUs of the time.
However, even during the KT333 days, most VIA motherboards (including the one I was using when I cooked my chip), were relying on an external thermistor right below the CPU. Obviously, this thermistor will never sense the rapid, destructive change in temperature inside the die (which takes mere seconds) when a cooler is missing or improperly mounted.

Conclusion: these CPUs cooked themselves to death through no fault of their own.

Anyway, I just noticed this thread became pretty big! Anyway, wanted to say that the PIII I had should be atleast booting if I get the RAM and GPU, the motherboard D-LED recognizes it. I have a picture of the die before I reassembled my heatsink back on, I can send it later (not very high quality tho, since I didn't really photoed it upclose back then).