AMD Phenom II Overheat - Tried Everything

[lkmellone]

Hey guys, it's me again.
Well, some of you might have read my old topic (here: http://www.tomshardware.com/answers/id-2445671/thermal-compound-break.html) which I tried to change my TIM in order to lower CPU Temperatures.
Well, now I tried to buy a aftermarket cooler (the only one I can afford by this time), and some other fans to put into my case.
Anyway, the temperatures didn't drop as much as I expected.
My system is cool, everything is very cool, but the processor's still overheating (I think).

On idle, I'm getting 39~40°C on CPU's temps.
Ambient temperature is around 34°C (Brazillian Summer hehe)

My pc config is the following:
AMD Phenom II X4 955 BE 3.2GHz
16GB DDR3 1333MHz (Dual Channel)
EVGA GTX 460 1GB
Gigabyte GA-790XTA-UD4
SATA II Seagate 500GB 7200RPM

CPU Cooler: Xigmatek Loki II
Case Fans: Cougar Vortex
TIM: MX-4

 

[lkmellone]

Underclocking wouldn't really lower the heat by itself. If VCore/Voltage is an auto it may lower somewhat by itself if downclocked but in the case that it doesn't a 2.4GHz CPU with the same voltage as a 4GHz CPU would still be receiving the same amount of voltage/creating about the same heat.

That being said undervolting can do wonders but you need to make sure the settings you use is stable. I undervolted my old Core 2 Duo laptop after 10+ years and I was amazed at how quiet the fan now runs and how less hot it is.

So if your OCing and can't get a pretty good aftermarket cooler I would go back to stock clocks/settings. From there I would try to undervolt the CPU to the lowest possible stable settings. If you don't really have any case fans I would add some to create better airflow and tidy up the inside of the case (if it's dirty/full of wires). Managing your cables well will obstruct the airflow less. If that's not enough then this is where you begin to underclock the CPU which will allow you to lower you VCore/voltages even more.

This is a damn good advice.
Anyway, my processor is running at stock 3.2GHz, and I've never OCd it. I know my CPU Cooler is not that good for OCing, and I should invest more, but I thought it would decrease a little more in terms of temperature.
What I think, is that my CPU is having a high output temperature, because of Vcore voltages.

But, how can I undervolt it, without a good BIOS assistance?
Also, what can I do if I undervolt and run stress tests, and the CPU temp goes too high? Lower the voltage again and try again?

Assuming you have a "standard" Blue bios screen un navigable by mouse, there is a chance that you actually cannot do any over/under voting or over/under clocking on your computer. But looking at Gigabyte's specifications for your motherboard, it seems as you can, I believe that your oc'ing and ob'ing (and counterparts) are usually under the CPU tab on the top. Check there and if you see things like clock multiplier and voltage, you are in the right place

Hey, this is my BIOS:

These are the only Options I have for Volting. If I change System Volting Control from Auto to Manual, I have to set manually a voltage for everything.

 

[slyu9213]

Underclocking wouldn't really lower the heat by itself. If VCore/Voltage is an auto it may lower somewhat by itself if downclocked but in the case that it doesn't a 2.4GHz CPU with the same voltage as a 4GHz CPU would still be receiving the same amount of voltage/creating about the same heat.

That being said undervolting can do wonders but you need to make sure the settings you use is stable. I undervolted my old Core 2 Duo laptop after 10+ years and I was amazed at how quiet the fan now runs and how less hot it is.

So if your OCing and can't get a pretty good aftermarket cooler I would go back to stock clocks/settings. From there I would try to undervolt the CPU to the lowest possible stable settings. If you don't really have any case fans I would add some to create better airflow and tidy up the inside of the case (if it's dirty/full of wires). Managing your cables well will obstruct the airflow less. If that's not enough then this is where you begin to underclock the CPU which will allow you to lower you VCore/voltages even more.

This is a damn good advice.
Anyway, my processor is running at stock 3.2GHz, and I've never OCd it. I know my CPU Cooler is not that good for OCing, and I should invest more, but I thought it would decrease a little more in terms of temperature.
What I think, is that my CPU is having a high output temperature, because of Vcore voltages.

But, how can I undervolt it, without a good BIOS assistance?
Also, what can I do if I undervolt and run stress tests, and the CPU temp goes too high? Lower the voltage again and try again?

The advice makes sense but it's also a paint to do sometimes because some motherboards do not support messing with Vcore at all or only allow you to increase them. I've seen 1 thread of a person with the same motherboard not able to alter the VCore in the BIOS. But when I read the Manual it seems like you should be able to alter the BIOS manually. The CPU Vcore setting will be in the MB Intelligent Tweaker section in the BIOS. From there you will have to change System Voltage Control to Manual (from auto) and then turn CPU Voltage Control to Manual. Once you do that the BIOS should allow you to change the VCore. I'm not sure if you will be allowed to lower the voltage though, you'll have to check yourself.

Undervolting should never make the CPU temps run higher than the CPU at higher VCore. First thing is that you want to find a stable VCore value. So lower the VCore in a small increment (if possible) and then stress the CPU for about 10 minutes with a program like Prime95. If it passes 10 minutes without crashing try lowering a little more and run test for stability again. Continue this until your PC does not boot or crashes during the stability test. At this point you should set the vcore to the last stable option and then test for a good 2-12 hours. If it's stable for that long without crashing cores/throttling then you probably have a stable undervolt. While you do this though you want to keep looking at the temps t make sure it doesn't go over the maximum safe/recommended temps for the processor. I understand you live in Brazil with the warm/hot weather so outcomes may be different from what could be expected in a cooler location.

.:edit:.

Even if you turn System Voltage Control to manual I believe you can keep everthing else on auto except the CPU Voltage which needs to be manual. This is the case with my GA 770 UD3. It's running a Sempron 145 unlocked to a Dual-Core. Surprisingly I realized that the CPU was running on lower voltage than stock while I was reconfiguring that PC so I think this will all work for you.

 

[lkmellone]

Slyu, thanks for the advice. I'll try and report the results or any questions later!

UPDATE::
This is what I get when I set Voltages to normal:

Are they really normal?
Also, my CPU VCore Voltage range is 0.944V to 1.325V. If I set it to manual, my MB won't tweak it according to my usage, will it?

Like my Voltage on idle is 0.944V, and on full load is 1.32V. If I change to manual, will it be stuck on the Voltage I set?

 

[slyu9213]

Slyu, thanks for the advice. I'll try and report the results or any questions later!

UPDATE::
This is what I get when I set Voltages to normal:

Are they really normal?
Also, my CPU VCore Voltage range is 0.944V to 1.325V. If I set it to manual, my MB won't tweak it according to my usage, will it?

Like my Voltage on idle is 0.944V, and on full load is 1.32V. If I change to manual, will it be stuck on the Voltage I set?

All the settings at normal seem to be the same as other motherboards for the Deneb Phenom II X4 CPUs. As long as you have all your power saving features enabled in the BIOS the voltage should drop lower than what the VCore is set to. One easy way to test this is to set a lower VCore and see what kind of vcore is being used when idle through CPU-Z. Could you change CPU Voltage Control to Manual and then take a picture of how the BIOS looks like after that?

 

[lkmellone]

Did it, and my idle Voltage is now 0.912V and Load is 1.300V
Note: Didn't run the stress test at this point.

Screenshot as required:

 

[slyu9213]

Did it, and my idle Voltage is now 0.912V and Load is 1.300V
Note: Didn't run the stress test at this point.

Screenshot as required:

PERFECT. Now you have the tool/knowledge to look for a stable undervolt. You know how to undevolt in the BIOS, test for stability and know that at idle the CPU will be at .912V so everything sounds good.

I'll leave this link for you. Tomshardware made an article about undervolting a Phenom II X4 955 BE in 2009. It should give you comparable data on what kind of undervolt you can probably expect. I'll allow you to read the article yourself but Tomshardware succesfully undervolted to 1.175V which lowered the peak system power from 216W to 179W. There is no temperture numbers but seeing that the CPU uses less power it might give you a nicer temps. You may not get stable as low as 1.175V but I can see you being able to undervolt a bit.

http://www.tomshardware.com/reviews/undervolt-cpu-phenom,2348-4.html

 

[lkmellone]

Did it, and my idle Voltage is now 0.912V and Load is 1.300V
Note: Didn't run the stress test at this point.

Screenshot as required:

PERFECT. Now you have the tool/knowledge to look for a stable undervolt. You know how to undevolt in the BIOS, test for stability and know that at idle the CPU will be at .912V so everything sounds good. I'll leave this link for you. Tomshardware made an article about undervolting a Phenom II X4 955 BE in 2009. It should give you comparable data on what kind of undervolt you can probably expect. I'll allow you to read the article yourself but Tomshardware succesfully undervolted to 1.175V which lowered the peak system power from 216W to 179W. There is no temperture numbers but seeing that the CPU uses less power it might give you a nicer temps

http://www.tomshardware.com/reviews/undervolt-cpu-phenom,2348-4.html

Pretty thanks! I'll read the article and run some test after.
Just another question: Will my Idle VCore Voltage go down everytime I down that VCore Voltage?
Is there any risk?
Also, if my pc doesn't turn on, I'll have to reset the CMOS, right?

 

[slyu9213]

Did it, and my idle Voltage is now 0.912V and Load is 1.300V
Note: Didn't run the stress test at this point.

Screenshot as required:

PERFECT. Now you have the tool/knowledge to look for a stable undervolt. You know how to undevolt in the BIOS, test for stability and know that at idle the CPU will be at .912V so everything sounds good. I'll leave this link for you. Tomshardware made an article about undervolting a Phenom II X4 955 BE in 2009. It should give you comparable data on what kind of undervolt you can probably expect. I'll allow you to read the article yourself but Tomshardware succesfully undervolted to 1.175V which lowered the peak system power from 216W to 179W. There is no temperture numbers but seeing that the CPU uses less power it might give you a nicer temps

http://www.tomshardware.com/reviews/undervolt-cpu-phenom,2348-4.html

Pretty thanks! I'll read the article and run some test after.
Just another question: Will my Idle VCore Voltage go down everytime I down that VCore Voltage?
Is there any risk?
Also, if my pc doesn't turn on, I'll have to reset the CMOS, right?

No, from what I've read the Idle VCore does not change when you undervolt. Cool and Quiet is what makes the CPU idle at ~800MHz and the low VCore. Tomshardware's 955 had a idle VCore of .960 compared to your .912. But CPUs have different voltages so it seems normal. Like how you tested to see if manually setting VCore would not allow it to drop to .912v at idle you can check if the idle vcore drops when you lower the vcore by checking it's vcore at idle periods through CPU-Z. If let's say you drop your VCore to 1.2V and idle is still at .912v then you will have no issues.

And yes if your PC does not turn on because it needs more VCore than what you selected a cmos/bios reset will fix everything

 

[lkmellone]

Did it, and my idle Voltage is now 0.912V and Load is 1.300V
Note: Didn't run the stress test at this point.

Screenshot as required:

PERFECT. Now you have the tool/knowledge to look for a stable undervolt. You know how to undevolt in the BIOS, test for stability and know that at idle the CPU will be at .912V so everything sounds good. I'll leave this link for you. Tomshardware made an article about undervolting a Phenom II X4 955 BE in 2009. It should give you comparable data on what kind of undervolt you can probably expect. I'll allow you to read the article yourself but Tomshardware succesfully undervolted to 1.175V which lowered the peak system power from 216W to 179W. There is no temperture numbers but seeing that the CPU uses less power it might give you a nicer temps

http://www.tomshardware.com/reviews/undervolt-cpu-phenom,2348-4.html

Pretty thanks! I'll read the article and run some test after.
Just another question: Will my Idle VCore Voltage go down everytime I down that VCore Voltage?
Is there any risk?
Also, if my pc doesn't turn on, I'll have to reset the CMOS, right?

No, from what I've read the Idle VCore does not change when you undervolt. Cool and Quiet is what makes the CPU idle at ~800MHz and the low VCore. Tomshardware's 955 had a idle VCore of .960 compared to your .912. But CPUs have different voltages so it seems normal. Like how you tested to see if manually setting VCore would not allow it to drop to .912v at idle you can check if the idle vcore drops when you lower the vcore by checking it's vcore at idle periods through CPU-Z. If let's say you drop your VCore to 1.2V and idle is still at .912v then you will have no issues.

And yes if your PC does not turn on because it needs more VCore than what you selected a cmos/bios reset will fix everything

Ran some stress test for 1.3V and it's normal. Now, I changed to 1.275V and my Idle Voltage went to 0.896V.
For some reason, this is scaring me...
Shouldn't I try to change the CPU NB VID Voltage instead? How can I make the Idle voltage stop changing?

 

[slyu9213]

Did it, and my idle Voltage is now 0.912V and Load is 1.300V
Note: Didn't run the stress test at this point.

Screenshot as required:

PERFECT. Now you have the tool/knowledge to look for a stable undervolt. You know how to undevolt in the BIOS, test for stability and know that at idle the CPU will be at .912V so everything sounds good. I'll leave this link for you. Tomshardware made an article about undervolting a Phenom II X4 955 BE in 2009. It should give you comparable data on what kind of undervolt you can probably expect. I'll allow you to read the article yourself but Tomshardware succesfully undervolted to 1.175V which lowered the peak system power from 216W to 179W. There is no temperture numbers but seeing that the CPU uses less power it might give you a nicer temps

http://www.tomshardware.com/reviews/undervolt-cpu-phenom,2348-4.html

Pretty thanks! I'll read the article and run some test after.
Just another question: Will my Idle VCore Voltage go down everytime I down that VCore Voltage?
Is there any risk?
Also, if my pc doesn't turn on, I'll have to reset the CMOS, right?

No, from what I've read the Idle VCore does not change when you undervolt. Cool and Quiet is what makes the CPU idle at ~800MHz and the low VCore. Tomshardware's 955 had a idle VCore of .960 compared to your .912. But CPUs have different voltages so it seems normal. Like how you tested to see if manually setting VCore would not allow it to drop to .912v at idle you can check if the idle vcore drops when you lower the vcore by checking it's vcore at idle periods through CPU-Z. If let's say you drop your VCore to 1.2V and idle is still at .912v then you will have no issues.

And yes if your PC does not turn on because it needs more VCore than what you selected a cmos/bios reset will fix everything

Ran some stress test for 1.3V and it's normal. Now, I changed to 1.275V and my Idle Voltage went to 0.896V.
For some reason, this is scaring me...
Shouldn't I try to change the CPU NB VID Voltage instead? How can I make the Idle voltage stop changing?

Is Cool'n'Quiet enabled? Tomshardware's article said that Cool'n'Quiet managed the effective idle vcore to .960 no matter what they lowered the vcore too. Lowering CPU/NB Vcore will lower temps too but that has more to do wit the North Bridge which effects the performance your memory. I honestly can't find much information about this. The good thing here is that as long as the lower idle vcore is stable at that speed then it is okay. Damage wouldn't happen to the CPU. You might try to lock the CPU at 800MHz either through the BIOS or Windows and then see if running a stability test will have the CPU run at 800MHz with idle vcore. If it does you should see if that is stable also.

One thread I read a person that said that they could go as low as .6V when idle but not any lower than that because below .6V is too low to have the CPU run.

 

[lkmellone]

I tried to have more information on this, but I couldn't.
Cool'n'Quiet is enabled, but anyway, I'm having this .896 Idle... I don't know how inteligent Cool'n'Quiet is to stop lowering the idle voltage.
I ran another 30-mins test @1.275V and it's stable so far.

About locking the clock frequency, how can I do that?

 

[slyu9213]

Well in the BIOS there is probably a setting called CPU Multiplier or something of that sort. The FSB is 200 so to get 3200MHz the mutliplier would need to be 16. So look for a setting that seems to be CPU Frequency/Multiplier or something with x16/16x. So to get 800MHz through BIOS you would need to set the multiplier to 4. After doing that try running a stability test. If Cool'n'Quiet is smart they will probably set the VCore to the idle one but again the CPU is technically idle as there will be 100% load on it so I'm not sure if the CPU will be using .960V/.896V or the full 1.275V. For Windows you could go to the Windows Power option and configure any of the plans to run from 5% minimum to 25% maximum.

Tomshardware has gone as low as 1.175V but actual users tend to find the 1.2-1.275V the most stable for undervolting Deneb or at least the X4 955 BE. It would be safe to continue your undervolt till it becomes unstable which may come at the lower 1.2XV mark.

I found another thread where another user or two can run lower than the .960V idle vcore. The first user can do 780MHz at .7V or even .6V. Also the user can run 2GHz at .95V which is less VCore than what a 955 does idle at 800MHz.

 

[lkmellone]

Well in the BIOS there is probably a setting called CPU Multiplier or something of that sort. The FSB is 200 so to get 3200MHz the mutliplier would need to be 16. So look for a setting that seems to be CPU Frequency/Multiplier or something with x16/16x. So to get 800MHz through BIOS you would need to set the multiplier to 4. After doing that try running a stability test. If Cool'n'Quiet is smart they will probably set the VCore to the idle one but again the CPU is technically idle as there will be 100% load on it so I'm not sure if the CPU will be using .960V/.896V or the full 1.275V. For Windows you could go to the Windows Power option and configure any of the plans to run from 5% minimum to 25% maximum.

Tomshardware has gone as low as 1.175V but actual users tend to find the 1.2-1.275V the most stable for undervolting Deneb or at least the X4 955 BE. It would be safe to continue your undervolt till it becomes unstable which may come at the lower 1.2XV mark.

I found another thread where another user or two can run lower than the .960V idle vcore. The first user can do 780MHz at .7V or even .6V. Also the user can run 2GHz at .95V which is less VCore than what a 955 does idle at 800MHz.

Sorry, I didn't understand this well. Couldn't I just test idle idling the pc?
I'm afraid of changing Multipliers and f**k things up.

I'm running idle now @0.864V and Load @1.250V
Ran a prime95 for 40 minutes and load is stable so far...

 

[slyu9213]

Well in the BIOS there is probably a setting called CPU Multiplier or something of that sort. The FSB is 200 so to get 3200MHz the mutliplier would need to be 16. So look for a setting that seems to be CPU Frequency/Multiplier or something with x16/16x. So to get 800MHz through BIOS you would need to set the multiplier to 4. After doing that try running a stability test. If Cool'n'Quiet is smart they will probably set the VCore to the idle one but again the CPU is technically idle as there will be 100% load on it so I'm not sure if the CPU will be using .960V/.896V or the full 1.275V. For Windows you could go to the Windows Power option and configure any of the plans to run from 5% minimum to 25% maximum.

Tomshardware has gone as low as 1.175V but actual users tend to find the 1.2-1.275V the most stable for undervolting Deneb or at least the X4 955 BE. It would be safe to continue your undervolt till it becomes unstable which may come at the lower 1.2XV mark.

I found another thread where another user or two can run lower than the .960V idle vcore. The first user can do 780MHz at .7V or even .6V. Also the user can run 2GHz at .95V which is less VCore than what a 955 does idle at 800MHz.

Sorry, I didn't understand this well. Couldn't I just test idle idling the pc?
I'm afraid of changing Multipliers and f**k things up.

I'm running idle now @0.864V and Load @1.250V
Ran a prime95 for 40 minutes and load is stable so far...

Lol yes you are correct. As soon as the CPU has to work the CPU will go up to 3.2GHz anyway. I guess you can test the idle stability by leaving your PC at idle for a long time without it crashing/freezing. Post back when you either find an unstable voltage or when you find the lowest stable CPU VCore. Are you taking a look at the temps while checking the CPU for stability?

 

[lkmellone]

Well in the BIOS there is probably a setting called CPU Multiplier or something of that sort. The FSB is 200 so to get 3200MHz the mutliplier would need to be 16. So look for a setting that seems to be CPU Frequency/Multiplier or something with x16/16x. So to get 800MHz through BIOS you would need to set the multiplier to 4. After doing that try running a stability test. If Cool'n'Quiet is smart they will probably set the VCore to the idle one but again the CPU is technically idle as there will be 100% load on it so I'm not sure if the CPU will be using .960V/.896V or the full 1.275V. For Windows you could go to the Windows Power option and configure any of the plans to run from 5% minimum to 25% maximum.

Tomshardware has gone as low as 1.175V but actual users tend to find the 1.2-1.275V the most stable for undervolting Deneb or at least the X4 955 BE. It would be safe to continue your undervolt till it becomes unstable which may come at the lower 1.2XV mark.

I found another thread where another user or two can run lower than the .960V idle vcore. The first user can do 780MHz at .7V or even .6V. Also the user can run 2GHz at .95V which is less VCore than what a 955 does idle at 800MHz.

Sorry, I didn't understand this well. Couldn't I just test idle idling the pc?
I'm afraid of changing Multipliers and f**k things up.

I'm running idle now @0.864V and Load @1.250V
Ran a prime95 for 40 minutes and load is stable so far...

Lol yes you are correct. As soon as the CPU has to work the CPU will go up to 3.2GHz anyway. I guess you can test the idle stability by leaving your PC at idle for a long time without it crashing/freezing. Post back when you either find an unstable voltage or when you find the lowest stable CPU VCore. Are you taking a look at the temps while checking the CPU for stability?

haha I'll do this then. I think tomorrow I'll down the Vcore a little bit and do a long stress test.
Yeah, constantly. Now my CPU is running @61°C - Full Load / 38°C - Idle.
Before undervolting, it was @67°C - Full Load / 40°C Idle.

I also think that I can play some games to do some stress, right? :b

 

[slyu9213]

Well in the BIOS there is probably a setting called CPU Multiplier or something of that sort. The FSB is 200 so to get 3200MHz the mutliplier would need to be 16. So look for a setting that seems to be CPU Frequency/Multiplier or something with x16/16x. So to get 800MHz through BIOS you would need to set the multiplier to 4. After doing that try running a stability test. If Cool'n'Quiet is smart they will probably set the VCore to the idle one but again the CPU is technically idle as there will be 100% load on it so I'm not sure if the CPU will be using .960V/.896V or the full 1.275V. For Windows you could go to the Windows Power option and configure any of the plans to run from 5% minimum to 25% maximum.

Tomshardware has gone as low as 1.175V but actual users tend to find the 1.2-1.275V the most stable for undervolting Deneb or at least the X4 955 BE. It would be safe to continue your undervolt till it becomes unstable which may come at the lower 1.2XV mark.

I found another thread where another user or two can run lower than the .960V idle vcore. The first user can do 780MHz at .7V or even .6V. Also the user can run 2GHz at .95V which is less VCore than what a 955 does idle at 800MHz.

Sorry, I didn't understand this well. Couldn't I just test idle idling the pc?
I'm afraid of changing Multipliers and f**k things up.

I'm running idle now @0.864V and Load @1.250V
Ran a prime95 for 40 minutes and load is stable so far...

Lol yes you are correct. As soon as the CPU has to work the CPU will go up to 3.2GHz anyway. I guess you can test the idle stability by leaving your PC at idle for a long time without it crashing/freezing. Post back when you either find an unstable voltage or when you find the lowest stable CPU VCore. Are you taking a look at the temps while checking the CPU for stability?

haha I'll do this then. I think tomorrow I'll down the Vcore a little bit and do a long stress test.
Yeah, constantly. Now my CPU is running @61°C - Full Load / 38°C - Idle.
Before undervolting, it was @67°C - Full Load / 40°C Idle.

I also think that I can play some games to do some stress, right? :b

Yeah, you could do that also. It won't be as thorough as a program like Prime95 but it will stress the CPU. It's always recommended to get a CPU stable on CPU stressing programs but technically getting it stable with the most heaviest load you do in normal day use is a choice also. You can stress the PC with gaming and then stress with a stressing/stability program when you have to do something else for a long period of time (sleep, clean, cook, eat, homework, work, etc). Gaming is less intensive compared to Prime95 so it will put less load on the CPU and meaning it's less likely to hit the 61C you get while stressing in games. I think 62C is the highest temp you want your CPU to be for Deneb processors. You're already below it at 61C so it would be lower while gaming. Looking good so far. Can't wait to see if it can go any lower and be stable.

 

[lkmellone]

Hey guys,
The results so far are pretty straight foward. I'm having full load temps @56°C-60°C with 1.225V (some changes are based on room temp).
I ran a 40-min test and some games for a couple of hours and it's stable so far. I'll try to lower a bit more and run a 6hours test.
I'll report the further results.
Thanks for everyone! You guys helped me a lot! A special thanks for slyu9213 hehe

 

[slyu9213]

Once again no problem. I have no problems with the usual topics asking AMD vs Intel and recommendations on computer builds. The most fun part is when you help someone who is having an issue with a computer. The real fun here is dropping the temperature without using money while keeping the same performance. I'll be keeping an eye on this thread to see any updates/results.

There is another thing you can do to drop temps a bit but it's a long process. It's called lapping the CPU/Heatsink and what you basically do is use sand paper to get rid of any unevenness on the CPU IHS (metal slab with CPU info on the CPU) and the bottom of the heatsink that touches the thermal compound and CPU. You use multiple grit levels of sandpaper and while straightening the CPU it removes the nickel from the IHS and shows the actual material which is copper. During the process the CPU IHS gets mirror shiny too. Doing this shaves off CPU temps ~5-8C. I don't recommend anyone doing this unless they plan on keeping the CPU (not sell) and after the warranty of the CPU is over. Even I haven't done it yet but I may later this year with the CPUs I have laying around for practice. I don't know how much the different level of sandpaper costs so not sure if it's worth doing.