momaka wrote on 2025-06-19, 20:51:

That said, try re-inserting and re-testing each CPU several times into the socket. Also, does the behavior change if you apply pressure to the CPU (without the heatsink) and power on the board?

Without the heat sink? Pretty sure that would fry the core in no time..

Powering this on without the heatsink would be a very bad idea IMO. No heatspredder and i am not sure P3 even had overheat protection/throttling sufficient to prevent damage.

momaka wrote on 2025-06-19, 20:51:

I do that all the time (handle electronic boards by touching wherever) and have never killed a single one by static discharge - […]
Show full quote

I do that all the time (handle electronic boards by touching wherever) and have never killed a single one by static discharge - even when I'm often walking at home with my wool socks (in the winter) on the carpet.
It really takes bad luck to kill modern stuff by way of ESD.
I do agree that 3 dead CPUs in a row is too much, though - especially P3's. I've also not seen P3's with cracked/chipped dies as much as Athlon/XP's. So all in all, I don't think this is the issue either.
Probably like you said - either really bad luck or maybe O/P or O/P's hardware is doing something "funny" to make the CPUs not work.

Sometimes the problem is in the quality of packaging, although usually not the case with CPUs. I bought a GeForce 780 and the seller basically killed it by packiging it poorly in a soft thin box. I bought DDR 1GB with head spreader and it arrived dead (how can you even kill memory module like that). My survival rate of received functional semi-modern GPUs (GeForce 280 or newer) is about 30%. Most of cards arrive dead. Sellers either don't test them before sending or package them improperly in soft, thin boxes. That survival rate is skewed by about 5 cards I picked up in person, all of which work, otherwise it would have been maybe 15%. It is so low it is often not worth it to ship any GPU.

I never buy hardware without warranty as people tend to mishandle it too frequently.

stealthjoe wrote on 2025-06-20, 02:50:

The PSU I am using is Gigabyte P450B which is supposed to be a decent and safe one.

Ah, one of those. FWIW, these were also from the affected PSUs from the Gigabyte scandal (about their PSUs blowing up when overloaded.)
I fixed one a few years ago that I got for really cheap. Was lucky that mine was just a DOA unit from a bad 5VSB rectifying diode.

Anyways, I wouldn't call these PSUs exactly safe, but I wouldn't call them unsafe either.
What they are probably not good for is powering 5V-based PCs... though I haven't ran mine through my usual crossloads to check. It's a group-regulated design and ATX v2.2 compliant, so I suspect it won't do well with any system that pulls more from its 5V rail and its 12V rail. That's not to say I think this is the source of your issues... but certainly something to keep in mind if you use this PSU for testing retro hardware, specifically boards that power the CPU from 5V rail. In the case of P3, luckily they are not power-hungry at all (30 Watts TDP for the higher end ones, lower for lower models like yours), so I think your PSU probably just scuffs by.

H3nrik V! wrote on 2025-06-20, 04:39:

Without the heat sink? Pretty sure that would fry the core in no time..

Socket 462 CPUs - yes, absolutely.
Socket 370 CPUs - they just crash, as demonstrated by that Tom's Hardware video from way back in the days.
If worried, just put a small chipset heatsink on top with whatever scrap thermal pad or thermal compound you got - should keep the CPU temps OK for a quick <30 seconds test.

AlexZ wrote on 2025-06-20, 14:37:

Sometimes the problem is in the quality of packaging, although usually not the case with CPUs.

Yeah, it's hard to badly pack a CPU... though not impossible, certainly. 😁

Had a motherboard with CPU and RAM shipped to me once in a box with no padding whatsoever inside. Best part (or worst, depending on with how little humor you want to look at it 🤣 ) was the CPU and RAM were detached from the motherboard. So everything was all flopping around in there losely. It was too cheap to pass, though, so no regrets. The mobo was listed for parts, and as I saw from the listing, had bad caps (which is likely why it wasn't working.) Miraculously, after recapping the board and straightening pins on the CPU (socket 939 Athlon 64), everything ended up working.

AlexZ wrote on 2025-06-20, 14:37:

I never buy hardware without warranty as people tend to mishandle it too frequently.

Yeah, but even with warranty, dealing with having to ship stuff back for some of the sellers really puts me off.
That's why most of the time, I prefer to buy the "as-is, no warranty, untested" items and assume it really won't be working... that is, if I think I might be able to fix the item. Usually it's much cheaper this way too (up to a point anyways.) In the case of semi-modern GPUs, that's rarely the case.
Now what I *don't* buy are stuff from no-effort low-ball sellers that literally pick obviously broken stuff from the dumpster and try to sell it as -just- "untested", sometimes at almost premium prices too.

momaka wrote on 2025-06-22, 08:44:

Socket 462 CPUs - yes, absolutely. Socket 370 CPUs - they just crash, as demonstrated by that Tom's Hardware video from way back […]
Show full quote

H3nrik V! wrote on 2025-06-20, 04:39:

Without the heat sink? Pretty sure that would fry the core in no time..

Socket 462 CPUs - yes, absolutely.
Socket 370 CPUs - they just crash, as demonstrated by that Tom's Hardware video from way back in the days.
If worried, just put a small chipset heatsink on top with whatever scrap thermal pad or thermal compound you got - should keep the CPU temps OK for a quick <30 seconds test.

Wasn't that the Pentium 4, which had a built-in thermal monitor diode? (And a heat spreader, that probably gave a lot of extra time to react)? Don't recall ever seeing it with FCPGA cpus. Maybe FCPGA2 which also has a heat spreader?

I can agree with a small heatsink to run for a short period of time though.

H3nrik V! wrote on 2025-06-22, 14:08:

momaka wrote on 2025-06-22, 08:44:

Socket 462 CPUs - yes, absolutely. Socket 370 CPUs - they just crash, as demonstrated by that Tom's Hardware video from way back […]
Show full quote

H3nrik V! wrote on 2025-06-20, 04:39:

Without the heat sink? Pretty sure that would fry the core in no time..

Socket 462 CPUs - yes, absolutely.
Socket 370 CPUs - they just crash, as demonstrated by that Tom's Hardware video from way back in the days.
If worried, just put a small chipset heatsink on top with whatever scrap thermal pad or thermal compound you got - should keep the CPU temps OK for a quick <30 seconds test.

Wasn't that the Pentium 4, which had a built-in thermal monitor diode? (And a heat spreader, that probably gave a lot of extra time to react)? Don't recall ever seeing it with FCPGA cpus. Maybe FCPGA2 which also has a heat spreader?

I can agree with a small heatsink to run for a short period of time though.

You can also just put your finger on the die, although socket a AMD processors get quite hot.

H3nrik V! wrote on 2025-06-22, 14:08:

Wasn't that the Pentium 4, which had a built-in thermal monitor diode? (And a heat spreader, that probably gave a lot of extra time to react)? Don't recall ever seeing it with FCPGA cpus. Maybe FCPGA2 which also has a heat spreader?

The P4 definitely had it.
P3 FCPGA2 (Tualatin) also had it, IIRC... as well as a heatspreader, as you noted.
Now I don't remember if the P3 had it... but looks like it does. Either way, I do remember that Tom's Hardware video showing a "naked core" P3 and it just crashed when the HS was off, but that was that. No damage to CPU.
I *think* for the FCPGA P3, there was a thermal diode inside, but just not visible to the rest of the system and only used to shut down the CPU in case of an overheat condition. Not 100% sure, though. That said, most old Dell Latitude laptops with P3's (like my C600 for example) are set to make the cooling/CPU fan only turn on when the CPU temps go past something like 65 or 70C. I can confirm this is the case, because I've inserted a thermocouple in mine onto its heatsink, and it was running close to 70-73C before the fan kicked in. Seems kind of high, but those P3's are pretty tough. That laptop is still working FWIW. I used it almost daily from 2010 through 2018 as a main laptop and a machine to type on. Before that (2001 through approx. 2008-2009), it was used as an office laptop by two different organizations. So, it's close to 20 years of regular use. I'll take that as a show for P3's durability.

pete8475 wrote on 2025-06-22, 19:03:

You can also just put your finger on the die, although socket a AMD processors get quite hot.

DON'T!
Even the thermal diodes in latter AMD socket 462 CPUs are not quick enough to catch the fast temperature rise of the CPU die, thus it's pretty much instant death if you fire up one of those without a heatsink. A finger on top will just leave you with a burn... and if you're quick enough to pull it away, by the time you do that and reach for the PSU's power button to cut power to the system, the CPU would be already overheated and dead.

And even for AMD s462 CPUs with lower TDP of 40-ish Watts, you can still expect the die to heat up to 150+ C in just a few seconds. It's about the same as soldering tips on my T12 soldering iron when running with a 20V power adapter instead of its own 24V AC supply - takes just a few seconds @ 40W power for the tip to reach 200+ C. You do not want to hold onto one of those when they heat up (I know from experience here 🤣 ) It would be the same with the s462 CPUs - you'll end up with a burned CPU and possibly a burned finger skin too.

momaka wrote on 2025-06-23, 07:17:

DON'T! Even the thermal diodes in latter AMD socket 462 CPUs are not quick enough to catch the fast temperature rise of the CPU […]
Show full quote

pete8475 wrote on 2025-06-22, 19:03:

You can also just put your finger on the die, although socket a AMD processors get quite hot.

DON'T!
Even the thermal diodes in latter AMD socket 462 CPUs are not quick enough to catch the fast temperature rise of the CPU die, thus it's pretty much instant death if you fire up one of those without a heatsink. A finger on top will just leave you with a burn... and if you're quick enough to pull it away, by the time you do that and reach for the PSU's power button to cut power to the system, the CPU would be already overheated and dead.

And even for AMD s462 CPUs with lower TDP of 40-ish Watts, you can still expect the die to heat up to 150+ C in just a few seconds. It's about the same as soldering tips on my T12 soldering iron when running with a 20V power adapter instead of its own 24V AC supply - takes just a few seconds @ 40W power for the tip to reach 200+ C. You do not want to hold onto one of those when they heat up (I know from experience here 🤣 ) It would be the same with the s462 CPUs - you'll end up with a burned CPU and possibly a burned finger skin too.

There was a joke floating around back in the day about this being the fastest method to get AMD logo etched into a finger.

The result is definitely a dead CPU and likely a nasty burn. Though if you expect it you may be able to remove the finger fast enough. I've touched 200+C things enough times to know that it's possible without causing burns.

That said i've had an accident once with heatsink popping off while the system was running. Somehow motherboard was able to catch that and shutdown in time (pretty much instantly), but that was an advertised feature of that motherboard using external sensor, probably functioning based on rapid temperature raise, not actual temperatures.

Honestly i would not try it with pentium3 either. It may survive, but why risk it?

For people who worked with PC hardware back then even the idea of powering the system on without heatsink is scary, i've had a guy at work run up to me and pull the cord before saying anything when he saw me doing it with something modern-ish (LGA1155 IIRC). Said he thought i just forgot and was going to destroy the CPU, was pretty amazed that it can run like this, go into bios etc.

Archer57 wrote on 2025-06-23, 09:15:

For people who worked with PC hardware back then even the idea of powering the system on without heatsink is scary...

You know, I'm actually this way too - I pretty much never power anything without a heatsink, unless I am absolutely positively sure it won't matter. And yes, even modern systems I don't power on without a heatsink. In my case, it was the reverse when I worked for a large PC repair place. I was tasked with build a system with all new components (circa 2017 hardware, so definitely not critical if I didn't have the CPU cooler on). I put the motherboard, RAM, CPU, and its cooler for a test before putting it all in the case. My supervisor was passing by and asked, "why are you testing the board with the CPU cooler on?" and I was a bit baffled. Why should I not be testing it with the cooler on. He then said you don't need to for a quick BIOS test. Well I know that! So he continued and said it could save me time if the board turned out to be defective (so I won't have to unmount the CPU cooler.) Valid point, I suppose, but what are the chances, right? And as it turned out, right as we were working then, we saw 6th and 7th gen Intel CPUs later that year come back dead due to either not having a cooler put on properly or if they have ran way too hot for too long. So when the discussion came again, I told him this is why I never do a board test without the CPU cooler on. Sure modern CPUs can survive it... but letting them overheat even once on day 1 of their life is a bad idea IMO for their overall lifespan.

So in short, I totally understand and I'm much like your older colleagues in that regard. But every once in a while, I do perform cooler-less tests if the CPU seems to be not heating up much or at all (due to the board being faulty and whatnot.) Though again, I make no exception for socket 462 - those I always run with a heatsink no matter what.

Archer57 wrote on 2025-06-23, 09:15:

Honestly i would not try it with pentium3 either. It may survive, but why risk it?

Agreed... but in the case of this thread, if the CPU is already suspected dead, it may not be that much of a risk.

Otherwise, I completely agree.

Archer57 wrote on 2025-06-23, 09:15:

I've touched 200+C things enough times to know that it's possible without causing burns.

Same.

momaka wrote on 2025-07-08, 17:00:

Valid point, I suppose, but what are the chances, right? And as it turned out, right as we were working then, we saw 6th and 7th gen Intel CPUs later that year come back dead due to either not having a cooler put on properly or if they have ran way too hot for too long. So when the discussion came again, I told him this is why I never do a board test without the CPU cooler on. Sure modern CPUs can survive it... but letting them overheat even once on day 1 of their life is a bad idea IMO for their overall lifespan.

Yeah, overheating is not good, even briefly.

Impressive thing about modern CPUs is that they will not overheat. You can power on the system, go into bios and see like 50-60C. You can also hold a finger on top of it and it will still be just warm, not even hot. Those 50-60C are inside the die...

So if care is taken this systems can be powered on to verify if it passes POST at least, without heatsink, safely. And it can, indeed, save a lot of time. Mostly in cases when troubleshooting something. For new hardware which theoretically should work with no issues tests like this probably are still not justified...

And situations you describe are also one reason i always preferred and still prefer to build stuff myself. This way i know how the hardware was handled. Who knows how someone who built the system handled the hardware otherwise? He might have been careful, or he might have powered it on and left to have a break while it was sitting at 100C for an hour...

I also never do this to things with bare die. S462 is extreme example, but even modern laptop CPUs and such. That heat spreader makes much larger difference than it might seem. It is, after all, a piece of copper and it has pretty significant heat capacity, preventing too fast or uneven heating of the die, which otherwise may do damage even with all the modern safety features...

In laptops just putting a piece of aluminum on top (like old socket AM2 stock cooler with fan removed) is usually sufficient for a short test, may be this is a good compromise for desktop systems too - saves time installing the cooler, definitely prevents overheating.