Discussion Ryzen 7000 series CPU Lapping

Oct 2, 2022
4
0
10
Lapping:
From what I heard the new Ryzen 7000 series has an extra 2mm IHS thickness for cooler compatibility; it seems that you would be able to lap the cpu 2mm to achieve much better thermals. This would serve as an alternative to delidding as that is quite a risky endeavor even with the correct tools. Though I am unsure about the best way to lap the cpu 2mm. Doing it by hand would take absolutely forever even with a low grit sandpaper. While using a machine would be efficient, it won't necessarily provide the flattest lap if you aren't careful; there would also be problems with holding the cpu in place and making sure not to overheat the cpu from the material being removed.

The cooling problem:
With the 2 mm difference most coolers would then not seem to fit (hence why AMD seemingly made the IHS so thick for am5). It seems that any cooler that uses standoffs to modify the height of the cooler would be able to make up the difference by simply using different standoffs. Most AIOs use standoffs and you should be able to simply get standoffs that are 2 mm shorter than the ones provided or modify the included standoffs by removing 2mm of material. Though this might become less of a problem with the am5 bracket adjustments that will be released for the majority am4 compatible coolers.

Reason behind this:
I am planning to build a mini-itx pc with the new 7700x but I would rather not have the processor hit 95c at all, especially when I might have cpu heavy tasks open for 4+ hours at a time. The high temperatures have a ton of disadvantages: higher power draw, possible effect on the longevity of the cpu and less overclocking potential. With this simple fix, the processor could drop up to 15c. Combine this with undervolting and you get the same performance with the cpu at probably about 50c (estimate based on 7000s undervolting video, which demostrated that the cpu dropped to under 70c while staying at the same performance level). Also helps with the hot air output from the case making it less of a heater and more of a computer.

Hopefully this is useful for anyone else hoping to improve temps without the danger of killing your cpu with delidding.
 
Even if you shave off 2mm, there's a question of if the cooler will actually make contact with the IHS at that point. Anything that clips on is very likely not going to work. If you have to screw it in, the cooler manufacturer would've likely designed how far the cooler goes in based on the IHS dimensions so you may have a significant gap when the screws go in as far as they can.

I think just doing a power limit and undervolt is enough. If the processor is designed to work with a max core temp of 95C, then even dropping it to 70C would supposedly quadruple the lifespan of the part... which is already an unknown anyway. In addition, it looks like you're linearly adding cooling results. That isn't really how it works.
 
  • Like
Reactions: Why_Me
Oct 2, 2022
4
0
10
Even if you shave off 2mm, there's a question of if the cooler will actually make contact with the IHS at that point. Anything that clips on is very likely not going to work. If you have to screw it in, the cooler manufacturer would've likely designed how far the cooler goes in based on the IHS dimensions so you may have a significant gap when the screws go in as far as they can.

I think just doing a power limit and undervolt is enough. If the processor is designed to work with a max core temp of 95C, then even dropping it to 70C would supposedly quadruple the lifespan of the part... which is already an unknown anyway. In addition, it looks like you're linearly adding cooling results. That isn't really how it works.
Well in my post, under "The cooling problem:" I mentioned how most of the coolers would no longer fit; but by using coolers that utilize standoffs for their attachment method could be modified to fit the cpu by simply modifying the standoffs or buying shorter ones. Though I agree that coolers that rely on the basic clips would no longer work as they are designed with the exact cpu height in mind (though most coolers that use screws are better anyways because of better spread).

I would like to say that I am not adding them linearly. In this video, it was shown that you could get the temperature to about 61c with undervolting and power limiting. In this video der8aur demonstrates on a 7950x that you can get a 21c drop in temperatures with delidding. While delidding of Intel chips (with a normal IHS) has shown about a 10-15c improvement on average. So if removing the IHS (about 4-5mm) improves the temps on Intels cpus by about 10-15c. It isn't inconceivable that you could drop the temperatures of the cpu by an extra 5-7c when undervolting and power limiting. I do agree that my previous estimate of an up to 15c drop from purely removing 2mm is way out of reasonable range and probably only achievable with delidding.
 
Actually let's step back because this is part of the reason why I hate what AMD is doing and how AMD's temperature reporting system keeps causes no amount of endless confusion and panic.

Unless you're using HWiNFO, what you're seeing on any Ryzen CPU temperature reading is the hottest part of the processor. It doesn't mean the whole entire processor is that hot. And looking at Gamers' Nexus's reviews on the part, only one core actually hits the thermal limit. The rest are about 10-15C cooler than the hottest core.

The second thing to note is why are you doing this? If it's to get the CPU to dump less heat, then temperature is not the value you should be looking at. You should be looking at watts. Energy is expressed in joules, watts is joules per second. Therefore, the fewer watts the CPU is pulling, the less energy is dissipating and thus, the less heat gets to go elsewhere. My phone's hardware gets up to 50-50C sometimes, as hot as my video card, but my phone isn't pulling 120W. If it's for longevity, then I'd argue it's as pointless of an endevour as trying to find the perfect formula to game the stock market. There's no way to determine how long the part will last. We still have people coming in with system specs from the early 2010s. I think people may have Pentium 4's still floating around and those always ran hot. I would be confident that despite parts of the CPU hitting 90C, 99.99% of them will last for at least 5 years. If we went by the rule of thumb of dropping 10C doubles the life expectancy of the part, then you're looking at 20 years. I mean, that's pretty good, but are you really going stick around with a processor for 20 years?

So is doing something that sounds like you have little to no experience with, will mutilate the processor, will void its warranty, will likely cost a non-trivial amount of money, time, and effort, among other possible complications, all just to have it last for an indeterminate amount of time longer worth it for you?
 
Oct 2, 2022
4
0
10
Actually let's step back because this is part of the reason why I hate what AMD is doing and how AMD's temperature reporting system keeps causes no amount of endless confusion and panic.

Unless you're using HWiNFO, what you're seeing on any Ryzen CPU temperature reading is the hottest part of the processor. It doesn't mean the whole entire processor is that hot. And looking at Gamers' Nexus's reviews on the part, only one core actually hits the thermal limit. The rest are about 10-15C cooler than the hottest core.

The second thing to note is why are you doing this? If it's to get the CPU to dump less heat, then temperature is not the value you should be looking at. You should be looking at watts. Energy is expressed in joules, watts is joules per second. Therefore, the fewer watts the CPU is pulling, the less energy is dissipating and thus, the less heat gets to go elsewhere. My phone's hardware gets up to 50-50C sometimes, as hot as my video card, but my phone isn't pulling 120W. If it's for longevity, then I'd argue it's as pointless of an endevour as trying to find the perfect formula to game the stock market. There's no way to determine how long the part will last. We still have people coming in with system specs from the early 2010s. I think people may have Pentium 4's still floating around and those always ran hot. I would be confident that despite parts of the CPU hitting 90C, 99.99% of them will last for at least 5 years. If we went by the rule of thumb of dropping 10C doubles the life expectancy of the part, then you're looking at 20 years. I mean, that's pretty good, but are you really going stick around with a processor for 20 years?

So is doing something that sounds like you have little to no experience with, will mutilate the processor, will void its warranty, will likely cost a non-trivial amount of money, time, and effort, among other possible complications, all just to have it last for an indeterminate amount of time longer worth it for you?
"It doesn't mean the whole entire processor is that hot. And looking at Gamers' Nexus's reviews on the part, only one core actually hits the thermal limit. The rest are about 10-15C cooler than the hottest core." Still means that if you lower the temperature the cpu in general could boost higher.

If my goal was to get the process to "dump less heat" I would not be posting a discussion post about lowering cpu temperatures with lapping. Undervolting and power limiting can only do so much when the cpu is naturally held back by the increased heat transfer resistance that is caused by the increase in IHS thickness. By removing the extra material put on the IHS you can move heat much quicker from the cpu.

"The second thing to note is why are you doing this?" to be able to get the most of the part that I paid for? In the same video by der8aur the cpu was able to boost much higher without any other modifications because of the much lower temperatures. I want my technology to last, maybe in 5 years the cpus that ran at 95c will have much faster degradation. "There's no way to determine how long the part will last." does not mean that you shouldn't try to actively prolong the longevity of it.

"If we went by the rule of thumb of dropping 10C doubles the life expectancy of the part, then you're looking at 20 years. I mean, that's pretty good, but are you really going stick around with a processor for 20 years?" No, but an older system will always have a use. Be it for hosting a Minecraft server or a system for dumping all your pictures and videos on.

"I would be confident that despite parts of the CPU hitting 90C, 99.99% of them will last for at least 5 years." sure but what if it doesn't? I am not just going to take a manufacturers word because its good for them if their product breaks and I end up paying for a more expensive one later on. Mind you AMD products only have a 3 year warranty. All they care about is that the cpu outlives the warranty(they are a for profit business after all). As mentioned above degradation is a thing and I highly doubt that it won't degrade a much quicker pace. The traces haven't changed, the material that the processor is made from has not changed either. I see no reason why it is okay to run the it at 95c now compared to 2 years ago.

"So is doing something that sounds like you have little to no experience with, will mutilate the processor, will void its warranty, will likely cost a non-trivial amount of money, time, and effort, among other possible complications, all just to have it last for an indeterminate amount of time longer worth it for you?" People get experience by trying things, I made this post to get input of how I should approach this project and if I missed anything, and instead I got a reply saying that its pointless and don't do it.

To say that lapping is mutilating the processor is just delusional. Lapping has been done since the time cpus became a thing. Voiding the warranty of the processor is kind of obvious.

Time and effort is what is required to get proficient at something, I do not mind spending my time and my effort to accomplish something I think would bring benefit to me and allow me to get more experience at custom modification to parts.

Though I agree that it is basically a shot in the dark if it would improve the longevity of the processor, I would assume it would. Though longevity is part of the goal you are correct that there isn't enough proof to say that lower temps for sure increase product life and lessen degradation but for me it seems that even if it doesn't there are enough benefits to outweigh the cons.

I think your input has been helpful to a degree, so I thank you for that.
 
Oct 2, 2022
4
0
10
I wish for return of direct die, I never chipped or scratched any. At best they could make just a window(s) for die(s) and leave IHS at same height for support
There would be people that would find a way to screw up the die . So I don't think direct die will ever be mainstream because neither Intel or AMD are willing to take the risk of releasing a cpu with no IHS.
 

TRENDING THREADS