CPU fatigue: myth or real?

Cerulean

Cerulean

[H]F Junkie
Joined
Jul 27, 2006
Messages
9,477
As there is not a single article on the internet on this, I thought I would start a thread to see if others have been making similar observations.

I have a Lenovo ThinkPad X220 with an i5 CPU. This laptop is at least 10 years old if not somewhere within the 15 year ballpark. When I first got this laptop several years ago, I became familiar with it's performance and had upgraded it to 16GB RAM and Samsung Pro SSD. Over the years I've come to observe a strange slowness that got worse with time regardless of reimaging/reloading the OS and regardless of what SSD I used. It's not producing any errors or showing any clear sign, but it's obvious that with the amount of RAM available, great SSD storage, thermal paste that has been replaced with Grizzly Kryonaut, disassembled and rebuilt several times, dust cleaned out regularly, batteries replaced with genuine Lenovo, using genuine Lenovo model-original chargers, great CPU model for basic tasks ... something about the CPU is unusual. It's difficult to be concrete in description.

Another example is a server I dealt with a couple years ago. This server had 48GB+ RAM, a RAID10 disk configuration (albeit spinning), and dual quad-core Xeon E3 Xeon CPUs. When it came to booting VMs or getting logged in, it was just doggone dirt slow. I've done tests on the storage and found nothing wrong. Over time as I would do testing on MDT images with blank VMs and interact with the MDT server, I would find things like applications are increasingly taking longer and longer to launch and load despite how fresh of a Windows Server install there was and how basic and simple the OS configuration was. It grew on me through testing NICs, RAMDisk tests, and other examinations that I became highly suspect of the CPUs dying. I couldn't do CPU tests on the host, unfortunately, due to lack of LOM and physical access. I recall how of all apps, Google Chrome seemed to be one of the worst (I suspect it has something to do with multithreading). When I first worked with this server several years ago, its performance was very adequate and good. I didn't see any of these issues.

Another example, my brother had a laptop that was an engineering-grade Dell Precision. Heavy, bulky, but quiet powerful. It had a quad-core i7, 32-64GB RAM, SSD, and was a relatively modern laptop (manufactured sometime within the last 5-7 years). The heck was used out of it, but it evidently its multicore performance dropped over time until it was so rubbish the laptop had to be replaced. My brother described that the CPU did everything well, except for when you threw very specific "problems" at it to solve that used a certain core or function of the CPU that used to do well but now performs poorly. There aren't any tests you can run because the CPU will still do the work, just with significant performance loss compared to when it was new. No errors.

Has anyone else in their experience encountered and grew to strongly suspect "CPU fatigue"? Care to share your experiences?
 
Last edited:
I normally find the os is mores suspect then the cpu ever is. Unforcuantly your examples are kinda pointless unless you throw a similar chip in the rig and observe the results. Additionally laptop and server workloads do not stress the chip at all under acceptable thermals
 
Last edited:
Cerulean said:
As there is not a single article on the internet on this, I thought I would start a thread to see if others have been making similar observations.

I have a Lenovo ThinkPad X220 with an i5 CPU. This laptop is at least 10 years old if not somewhere within the 15 year ballpark. When I first got this laptop several years ago, I became familiar with it's performance and had upgraded it to 16GB RAM and Samsung Pro SSD. Over the years I've come to observe a strange slowness that got worse with time regardless of reimaging/reloading the OS and regardless of what SSD I used. It's not producing any errors or showing any clear sign, but it's obvious that with the amount of RAM available, great SSD storage, thermal paste that has been replaced with Grizzly Kryonaut, disassembled and rebuilt several times, dust cleaned out regularly, batteries replaced with genuine Lenovo, using genuine Lenovo model-original chargers, great CPU model for basic tasks ... something about the CPU is unusually sluggish. It's difficult to be concrete in description.

Another example is a server I dealt with a couple years ago. This server had 48GB+ RAM, a RAID10 disk configuration (albeit spinning), and dual quad-core Xeon E3 Xeon CPUs. When it came to booting VMs or getting logged in, it was just doggone dirt slow. I've done tests on the storage and found nothing wrong. Over time as I would do testing on MDT images with blank VMs and interact with the MDT server, I would find things like applications are increasingly taking longer and longer to launch and load despite how fresh of a Windows Server install there was and how basic and simple the OS configuration was. It grew on me through testing NICs, RAMDisk tests, and other examinations that I became highly suspect of the CPUs dying. I couldn't do CPU tests on the host, unfortunately, due to lack of LOM and physical access. I recall how of all apps, Google Chrome seemed to be one of the worst (I suspect it has something to do with multithreading).

Another example, my brother had a laptop that was an engineering-grade Dell Precision. Heavy, bulky, but quiet powerful. It had a quad-core i7, 32-64GB RAM, SSD, and was a relatively modern laptop (manufactured sometime within the last 5-7 years). The heck was used out of it, but it evidently its multicore performance dropped over time until it was so rubbish the laptop had to be replaced.

Has anyone else in their experience encountered and grew to strongly suspect "CPU fatigue"? Care to share your experiences?
Click to expand...
More then likely it needs a cleaning and repasting the cpu. The temps are more then likely higher then it was new so it would not boost as high as before. That is if nothing software wise chanced. You could have more programs in the background running or newer versions that might be more demanding. The cpu normally doesn't slow down if it starts going bad. You more likely encounter crashes then anything. Be it just a program or the entire system.
 
I think you have 10 years of added bloat to the OS. As in, new performance robbing "features" introduced with new patches, etc. Not to mention bios patches for vulnerabilities.
 
FYI X220 was released in 2011, so barely 10 years old (assuming it was bought when it first released), my wife's using the 2012 X230 still with no problems but I've maintained cleaning the internals and replacing the paste every few years as well as upgraded the OS to Win10 a few years ago.
That said you've listed two laptop and one server as examples, all 3 of which can/often are subject to much higher heat loads and harsh environmentals, especially the laptops with already choked cooling.
There are many factors especially around the spinning storage that I would find much more likely to be an issue before the CPU.
That said as operating systems get more and more advanced and have more complexity/security patches to them they'd tend to run slower in return as well.

I agree with Vegeta's sentiment you're more likely to see BSODs or crashes than just sluggishness unless its thermal throttling.
Is fatigue possible? sure, is it likely to have a major impact like you're describing? not likely is my two cents
 
Thermal paste doesn't need to be reapplied because it's "dried out." It appears that way due to something called the "pump out effect." Thermal paste is designed to operate for a decade or more. The issue here is that the OS becomes more bloated and convoluted as features and security patches are added. Also, Intel CPU's have specifically had cumulative security mitigations added to the OS over time which will rob you of performance. CPU's don't "fatigue" they degrade or outright fail. If they degrade, their clock speeds will have to be dropped, they'll crash with specific types of exceptions. In some instances they'll require more voltage than they used to and things like that.

I've got a pair of Pentium Pro 180's that have been overclocked to 200MHz every day of their lives. I used that box for almost 10 years as a server. Those CPU's still work to this day.
 
I had this happen with a WinXP computer once, got a new HDD and reloaded the OS, same problem. Then I installed Linux and the problems all went away.
 
Never ran into CPU fattique
Prim95 stable from build date to end day. if you run into slow downs is more likely software issues or thermal throtling from bad cooling
 
You will get some silicon degradation after much use, sometimes to the point it needs a volt bump. Also caps and other components with finite lifetimes.

But mostly bloat and dusty coolers, pumped out paste etc.
 
N4CR said:
You will get some silicon degradation after much use, sometimes to the point it needs a volt bump. Also caps and other components with finite lifetimes.

But mostly bloat and dusty coolers, pumped out paste etc.
Click to expand...
Generally only if you're overclocking, the auto voltage takes care of that in most standard use cases.
 
electro-migration is a thing, when intel had a hack a few months back some documents related to testing theyve done on comet lake and electro-migration was leaked. and it should be fairly translatable to asfar back as skylake. and by extension not too different for most of the entire CORE line back to 2011 or so whenever it came out. im sorry i dont remember the exact numbers i saw. but it was pretty negligible even with pretty significant overvolting. it shouldnt make that much of a difference in user experience.

i will say these older CPUs are really starting to feel slow now. ive played around with quite a bit of hardware that age and it all feels really dated now especially compared to the top tier stuff now. i have one of those 3500x 6c/6t china only CPUs and even that feels kinda slow compared to the 8c/16t offerings (or more) of recent from red and blue teams. i know im for damn sure never buying a less than 8c/16t CPU for myself everagain (except for mining rigs) ..
 
would the bios be affected with that intel slowdowns to help mitigate all those firmware lvl attacks i know GN did videos on the bios updates to skylake and others cpus.

reffering to specter and meltdown patches,
 
Shadowarez said:
would the bios be affected with that intel slowdowns to help mitigate all those firmware lvl attacks i know GN did videos on the bios updates to skylake and others cpus.

reffering to specter and meltdown patches,
Click to expand...
lol forgot about those, yeah it could be. isnt it up to like ~20% degradation after all the patches?!
 
vegeta535 said:
Maybe 5%. People made such a huge deal about it but now seems people forgot about it.
Click to expand...

Because people on forum has a tendency to focus on sounding smart rather being smart. so since its not the popular thing around it they discharge the knowledge and move on to what is the new focus

The windows OS had a big impact on these patches affected on performance. Windows 7 did font rendering in a deeper ring than windows 10 and thereby had to switch more cause a bigger performance drop than windows 10.
Also older CPU's had a bigger performance drop from the updates.
Bottomline: The big issues system are probable existing in way less numbers today, so less people are effect leads to less people talking about it

Beside most of the security issue these patches was not an issue for gaming computer, since the security they breached was not even in place on those systems. However for server farms like AWS it was a huge bad thing security wise
 
Dan_D said:
Thermal paste doesn't need to be reapplied because it's "dried out." It appears that way due to something called the "pump out effect." Thermal paste is designed to operate for a decade or more. The issue here is that the OS becomes more bloated and convoluted as features and security patches are added. Also, Intel CPU's have specifically had cumulative security mitigations added to the OS over time which will rob you of performance. CPU's don't "fatigue" they degrade or outright fail. If they degrade, their clock speeds will have to be dropped, they'll crash with specific types of exceptions. In some instances they'll require more voltage than they used to and things like that.

I've got a pair of Pentium Pro 180's that have been overclocked to 200MHz every day of their lives. I used that box for almost 10 years as a server. Those CPU's still work to this day.
Click to expand...
Those CPU's had transistors you could draw with a pencil. Old tech is often more resilient because of more robust construction. In this case I would almost certainly agree the issue is the OS. Try Linux.
 
You must log in or register to reply here.
Back
Top