Based on what I have seen in previous similar threads on VOGONS, most will probably disagree with me here... but I really think i5 and i7 CPUs are an unnecessary overkill for a GOOD XP build. With that said, my personal recommendation is towards the i3's and Pentium G / dual core. Of course, it will all largely depend on what you intend to use the system for, too. If running really demanding games like Crysis is up high on your list, then the i5 or i7 might become a more sensible choice. Otherwise for just about anything else form the XP era, a high[er]-clocked i3 or Pentium G/dual-core should do just as well, since most XP era games rarely use more than 1 or 2 CPU cores. The i3 and P D-C/G also usually have much lower TDP (so will require less energy to run and run cooler too and/or not require any large/special/expensive coolers/heatsinks) and are cheap (can buy a few/many spares for the price of one i5 or i7.) Moreover, if your motherboard supports OC-ing, you might be able to get some really high core clocks... and again, without having to resort to expensive or large CPU coolers.
st31276a wrote on 2024-10-22, 10:48:
Is thermal compound drying out and affecting performance really a thing?
Yes.
But whether it will happen or not (or rather, how quickly) depends on a number of factors.
Prime factor is really the running temperature: the higher the temps at which the TC (thermal compound) is exposed to, the quicker it will degrade and/or dry out.
Of course, the amount of run time is also a factor, particularly run time spent at elevated/high temperatures.
Lastly, the heat differential between the two mediums that the TC joins can also be a factor. For example, if you have two CPU dies with the same TDP (thermal design power... i.e. heat output), but one has a die with twice the surface area of the other, then the CPU with the larger CPU die surface area will strain the TC less, because the amount of heat per unit of surface area will be smaller. This means smaller heat differential per surface area, so the TC will have "less work to do".
Like you, I also have systems where I never ever changed the thermal compound, some of which are over 20 years old now and with many hours of run time. But these are typically either low-TDP systems (Pentium 3 or Athlon 64) or higher-TDP systems with large coolers and fast running fans. So the end result is the same: the temperature of the heatsink (and CPU underneath) is kept low, so the thermal compound didn't dry out or age too badly. By low temperature, that would typically be under 60C at all times (preferably under 50-55C). And with my P3 systems, drying compound is never really a problem, because these typically operate at under 40-45C max... and that's with their fans running at reduced speeds to keep them a little more quiet.
st31276a wrote on 2024-10-22, 10:48:
My understanding is that the compound must only fill some very tiny imperfections between the cpu top and the heatsink bottom so that they touch perfectly, most gets squeezed out and what remains stays there, trapped. It only needs to flow when it is installed, thereafter it remains static.
Well, the thing is, nothing is "static", even if it appears to be so.
In case of CPU's, the heating and cooling causes all of the various materials the CPU is made from to expand and contract at slightly different rates. So while everything does appear to be static, there will always be micro-movements between various materials. In the case of thermal compound, this will mostly lead to the "pumping" effect, where any liquid substance in the compound may be pushed out while the "solids" remain stuck to the various surfaces. Over time, this is (IIRC) what causes TC to dry out.
st31276a wrote on 2024-10-22, 11:00:
Yes I understand, I was only wondering if it would ever realistically become an issue.
It has.
There was a thread here on Vogons maybe a few months back (in the summer I think) where a user had an issue with an Athlon 64 CPU running at really high temperatures. Several different stock and non-stock coolers were tested, all yielding poor temperatures. The issue turned out to be a mix of a faulty core and bad TC between the CPU die and CPU IHS (probably the latter led to the former issue... i.e. the failed TC overheated the die, and that permanently damaged the internal temperature diode to always read higher than normal.) When the user delidded the CPU and changed the TC, the temperatures improved quite a bit (dropped around 10-15C IIRC), but were still overall high (~50C idle, which is a bit too much for an A64 Venice core CPU.) I'll see if I can dig the thread a bit later.
This hasn't been the only CPU I've seen with such issue. In fact, a recent encounter with this was over the summer when I built a retro PC with my nephew. We got a whole used PC as a base for the project. It was a Core 2 Quad -based build. During testing, we saw core #1 & 2 with the temperatures always being higher than cores 3 & 4, especially under load (almost 10-12 degrees Celsius worth of difference with larger CPU loads). Most likely the TC between the CPU dies/cores (1&2) and IHS was going dry... and I can't say I would be surprised if that really was the case, because whoever built the computer originally used one of those "low profile" Intel stock pushpin coolers that's really meant for 45nm Core 2 Duo's. It was obviously undersized for the TDP of the C2Q (8300) in the system, making it run hot all the time (~45-50C idle, 65-70C under heavy but not 100% load, IIRC). My theory is that this eventually lead to TC between CPU's IHS and cores 1&2 to dry out (the TC between the cooler and CPU looked like it was changed once or twice before... probably someone trying to figure out why the temperatures were still high.) We didn't de-lid the CPU to change the TC there due to time constraint, but I did give him a much larger cooler (3 heatpipe Arctic Freezer, IIRC). So with the new cooler, at least we were able to drop the overall temperatures of the entire CPU, which translated to cores 1&2 to also run a little cooler (though still consistently higher than 3&4). But at least no core(s) were at the verge of throttling under load anymore. Actually, cores 3&4 rarely broke the mid-50's Celsius, so that was a nice upgrade.
