Wikipedia:Reference desk/Archives/Computing/2021 November 29

Welcome to the Wikipedia Computing Reference Desk Archives
The page you are currently viewing is a transcluded archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages.

November 29

RAM Clean

Hello! SO I'm curious as to whether removing the RAM from it's socket (with the computer off obviously) and cleaning it and its socket would improve PC performance since my PC was rather dusty and I'm wondering if some dust got in and might be affecting performance. ― Blaze The Wolf^Talk_{Blaze Wolf#6545} 15:23, 29 November 2021 (UTC)

It is profoundly unlikely that accumulated dust has any effect on your system performance. However, "profoundly unlikely" is not the same as "impossible."

Here's a very nice summary article about signal integrity in current-generation consumer-grade computer memory. I could be convinced that "accumulated dust" might alter signal integrity in a manner that could be measured using specialist laboratory equipment. I could even be convinced that this could manifest as a measurable performance degradation, as observed using some high-level system performance benchmark. If we wanted to spend lots of time and money to use special laboratory equipment and expert technician time, we might be able to definitively answer whether your specific instance of dust actually had any meaningful performance impact on your system. In practice, the effect is extraordinarily small - almost zero - and almost certainly smaller than most other random fluctuations in system performance caused by other complicated system interactions. It is such a small effect that it would be excruciatingly improbable to convincingly demonstrate a causal relationship.

And yet - if you spend any time studying modern computer performance benchmarks - there are a lot of factors that are subject to apparent randomness, and for which it is difficult to demonstrate a convincing causal relationship. It sort of devolves into ... a matter of "art" and experience - "I assert that dust is not relevant to computer performance, and I further assert that it would be difficult to prove scientifically." Fascinatingly, when we design and implement, say, a new cache prediction algorithm, this is the very same degree of scientific rigor that we employ: an argument from authority, credited with a bit of experience and intuition. We'd like to be methodical and prove whether we see a performance improvement or degradation - but the reality is, modern computers are quite mature and the easy-to-prove stuff was proved-beyond-doubt many decades ago. But we still have to innovate somehow!

The art of computing is not totally defeated by the engineering and science of computing!

Nimur (talk) 19:10, 29 November 2021 (UTC)

Alright. I do think the dust at least somewhat attributed to worsened performance due to the dust clogging up the CPU fan (you would know the dust made a difference because before I cleaned it, whenever I launched a game you could hear the fan revving up, and after I cleaned it, silence). I figured that maybe the dust affected performance due to it possible inhibiting how well the electricity can flow from the pins on the RAM stick to the motherboard. ― Blaze The Wolf^Talk_{Blaze Wolf#6545} 19:13, 29 November 2021 (UTC)

Well, that would be a confounding variable - the dust-in-the-RAM is associated with the dust-in-the-fan, and I could be more easily convinced that dust-in-the-fan has a causal relationship to observable system performance (by way of fan speed, by way of system thermal relief, by way of closed-loop-thermal-control-system). Still difficult to prove - but at least our methodology is becoming holistically analytic! Nimur (talk) 19:18, 29 November 2021 (UTC)

I have some experience with home built computers and so also I have some knowledge of both AMD and Intel CPUs, and also both AMD and Nvidia discrete GPUs. With a recentish CPU and GPU, it's fairly unsurprising that cleaning out the heatsink may improve performance. Boost clocks depend on various limits imposed by the manufacturer sometimes partially adjustable in your EFI settings (for CPUs), but generally temperature is one of them. BTW by recentish I mean ~10 years. Prior to that CPUs did tend to have thermal throttling so you could still see an effect but arguably a well made desktop especially one that isn't aiming for extreme compactness should generally have been far from thermal throttling so it's something you'd have been less likely to see. I'd say both Intel and AMD have gotten more aggressive in their boost offerings so in some ways it's even more likely with more recent systems e.g. 5 years.

In fact the recently launched Alder Lake CPUs at least of the K variety are known for being particularly aggressive removing a default time limit (tau) for the boost clocks. Meaning that even with stock settings the 12900K may operate with the Maximum Turbo Power of "241W" continously. Unsurprisingly many cooling solutions cannot handle this and I think many feel it's not unreasonable to design your system without intending to. So even without dust your performance may be limited by your cooling system. Although I'd note that this is something only likely to affect extreme benchmarks capable of significantly using most or all the cores of the system e.g. the in/famous Cinebench. It's something that won't affect gaming much at all unless your cooling system is woefully undersized so it can barely or maybe can't even handle the Processor Base Power of "125W". However since this is something that isn't that hard to benchmark, you can tests showing it (via differing cooling systems or cases) both on Alder Lake and other CPUs.

Getting back to the question, while too much dust can negatively affect performance I wouldn't assume it's the case here. While I'd definitely encourage cleaning out dust, the fact that the system is fairly silent now but needed to ramp up a bit before doesn't really tell us whether performance suffered before. If it's silent now this would suggest to me the cooling isn't being particularly taxed. So while it's good to clean out the dusk there's a fair chance it was still working well enough before even if it was noiser. Especially if this is a moderately sized desktop not a notebook. Considering the placebo effect in this case not only from the act of cleaning but from the lower noise, "it seems faster" is no susbtitute to benchmarks. (Even reliably benchmarking something can be difficult.)

As for the RAM issue, I'd say the chance it can affect performance especially if the system doesn't have ECC and most do not, is so low it's simply not worth considering. I'd also note that while dust can get into surprisingly places, the chance it will get into the RAM socket when the RAM is already seated is very slim. So in fact removing the RAM if you haven't sufficiently cleaned the system of dust is probably a bigger risk.

Note that I'm explictly not saying dirt or dust on the RAM stick or socket isn't a problem. It can be but the effect will generally be crashes or other errors or even the system not POSTing. It won't be lower performance. Actually reseating components is a fairly standard part of troubleshooting. This is more likely just poor contact for some other reason, however cleaning the contacts and socket would be another step especially if you're fairly sure the problem is the RAM (e.g. memtest errors). Although in terms of dirt I'd note again this is something far more likely to happen during assembly or re-assembly that something than over time. (However if the dust or dirt was already there, it may move around over time especially with handling so could become a problem when it wasn't before.)

Nil Einne (talk) 12:55, 3 December 2021 (UTC)