- Joined
- Sep 16, 2003
- Location
- Knoxville, TN
By popular demand I figured I'd go ahead and start a proper thread concerning this matter. Alot of people see Burning-In CPUs and the like is a myth. Frankly there is evidence to support either or. With this in mind Ill go ahead and list the process.
The Thought Process:
The good ole, "that which does not kill me only makes me stronger". Much like an athlete you cannot take an average schmo off the streets and make him run a marathon. Sure, some people can do things better than others, but training is key to success.
Once you start the process you gotta be pretty hard with the CPU. Never allow it to back down, basically FORCE it to do as you say instead of apeasing to what it wants to do.
What we want to do is starve the CPU, lean off the fat, and train the chip to do what WE want, not what it wants.
The Method:
As per the discussion this method revolves around stariving your parts of power, thus making them more responsive to it. The beauty of this method is that it applies to ANYTHING. You can use this for RAM, CPUs, even GPU! I use this method on everything I own.
What you want to do is go to your maximum stable overclock. Take the usual volts and reduce them below what would normally be stable. However You want the voltage high enough that the PC will not crash for atleast a few hours during each session.
A great way to get a grip on "the sweet spot" is to use a program like Prime95 to get the stability down from 100% to about a couple minutes / 1 hour's worth. Also be sure write down the length of this "control / base" session. That way it will provide a solid marker to show what progress has or has not been made.
Here is an example of what I did with my 2600+M
Example (my 2600+)
261 x 10 = Stable at 1.875v
261 x 10 = Stable at 1.80v (for the burn in test)
From here you force the CPU to 100% load by any means you have. Good programs for doing this are: CPU Burn, Toast, F@H, S&M, SiSandra Burn In Utilitiy.
Be warned that you dont something that can error out. You want to force it to run for several hours at a time, REGUARDLESS OF ERRORS. Make sure all important data is backed up and stored off the primary HDD. There is a serious risk of ruining your OS from this, so be warned.
Also, keep in mind that if the PC is too unstable each session will be interupted by either a BSOD or a reboot. This is something you want to avoid *the very best you phsyically can*. If you are getting BSODs or Random Reboots you are pushing too hard and it will only hurt this process, not help.
[EDIT]
Generally speaking this method takes a long time. You normally want each session to last several hours at a time. Most people doing this (including myself) start to see results after the first week of burning in.
**Note that this is not consecutive session, but rather sporatic sessions done during the day (that are several hours in length).**
Good times for sessions are:
Before School
Before Going To Bed
Before Going to Work
I tend to leave it running anywhere from 4-8 hours at a time. If possible. Naturally the longer its left alone the faster the process will occur.
From here you move to phase 2 of my method.
Once you feel you have made serious progress towards burning your CPU in, verify it by comparing the new "control / base" stability run against your previous SuperPI or Prime95 session. If it appears to be significantally more stable than it did previously you can either choose to continue with what you have already and burn it in again with even lower vcore, more sessions at the same vcore or move to phase 3.
Phase 3 is what makes my method so popular. In this phase you return the newly burned CPU to its *original* vcore levels prior to burning it in. Depending on how much vcore you have shaved off, and how much more stable it was compared to the initial levels, you will see anywhere from 50-350mhz.
Progress Notes:
What usually happens is that the first couple of times/hours, nothing happens. Shortly there after there is a short phase which it will appear that your CPU is actually losing stability. THIS is the point where you want to push it the hardest. Once that subsides there is usually a small gain in speed which quickly gets larger, and larger, and larger.
Generally speaking those who have tried this have either:
A) Shaved .05v or more from their normal overclocks
B) Gained about 150-300mhz on their overclock using the old voltage
C) Very little gain, almost negligable
D) Nothing at all.
Final Notes:
It usually takes a couple times to get the hang of it, but keep with it. While this is not GARUNTEED to work it does to some extent on any CPU reguardless of how old it is. As always your milliage *will* vary.
Member Results:
AthlonXP:
Sentential: 1800XP+ (2272mhz @ 1.82v -----> 2370 @ 1.75v)
Sentential: 2400XP+M (2412mhz @ 1.875v -----> 2460 @ 1.8v)
Sentential: 2600XP+M (2610mhz @ 1.825v -----> 2630 @ 1.75v)
Athlon64 (S754):
Athlon64 (S939):
Sentential: 3200+ (2415mhz @ 1.54v -----> 2445 @ 1.49v)
__________________________________
Pentium4 (S478):
Pentium4 (LGA775):
Sentential: 640 (3968mhz @ 1.44v -----> 4048 @ 1.42v)
The Thought Process:
The good ole, "that which does not kill me only makes me stronger". Much like an athlete you cannot take an average schmo off the streets and make him run a marathon. Sure, some people can do things better than others, but training is key to success.
Once you start the process you gotta be pretty hard with the CPU. Never allow it to back down, basically FORCE it to do as you say instead of apeasing to what it wants to do.
What we want to do is starve the CPU, lean off the fat, and train the chip to do what WE want, not what it wants.
The Method:
As per the discussion this method revolves around stariving your parts of power, thus making them more responsive to it. The beauty of this method is that it applies to ANYTHING. You can use this for RAM, CPUs, even GPU! I use this method on everything I own.
What you want to do is go to your maximum stable overclock. Take the usual volts and reduce them below what would normally be stable. However You want the voltage high enough that the PC will not crash for atleast a few hours during each session.
A great way to get a grip on "the sweet spot" is to use a program like Prime95 to get the stability down from 100% to about a couple minutes / 1 hour's worth. Also be sure write down the length of this "control / base" session. That way it will provide a solid marker to show what progress has or has not been made.
Here is an example of what I did with my 2600+M
Example (my 2600+)
261 x 10 = Stable at 1.875v
261 x 10 = Stable at 1.80v (for the burn in test)
From here you force the CPU to 100% load by any means you have. Good programs for doing this are: CPU Burn, Toast, F@H, S&M, SiSandra Burn In Utilitiy.
Be warned that you dont something that can error out. You want to force it to run for several hours at a time, REGUARDLESS OF ERRORS. Make sure all important data is backed up and stored off the primary HDD. There is a serious risk of ruining your OS from this, so be warned.
Also, keep in mind that if the PC is too unstable each session will be interupted by either a BSOD or a reboot. This is something you want to avoid *the very best you phsyically can*. If you are getting BSODs or Random Reboots you are pushing too hard and it will only hurt this process, not help.
[EDIT]
gvblake said:
Generally speaking this method takes a long time. You normally want each session to last several hours at a time. Most people doing this (including myself) start to see results after the first week of burning in.
**Note that this is not consecutive session, but rather sporatic sessions done during the day (that are several hours in length).**
Good times for sessions are:
Before School
Before Going To Bed
Before Going to Work
I tend to leave it running anywhere from 4-8 hours at a time. If possible. Naturally the longer its left alone the faster the process will occur.
From here you move to phase 2 of my method.
Once you feel you have made serious progress towards burning your CPU in, verify it by comparing the new "control / base" stability run against your previous SuperPI or Prime95 session. If it appears to be significantally more stable than it did previously you can either choose to continue with what you have already and burn it in again with even lower vcore, more sessions at the same vcore or move to phase 3.
Phase 3 is what makes my method so popular. In this phase you return the newly burned CPU to its *original* vcore levels prior to burning it in. Depending on how much vcore you have shaved off, and how much more stable it was compared to the initial levels, you will see anywhere from 50-350mhz.
Progress Notes:
What usually happens is that the first couple of times/hours, nothing happens. Shortly there after there is a short phase which it will appear that your CPU is actually losing stability. THIS is the point where you want to push it the hardest. Once that subsides there is usually a small gain in speed which quickly gets larger, and larger, and larger.
Generally speaking those who have tried this have either:
A) Shaved .05v or more from their normal overclocks
B) Gained about 150-300mhz on their overclock using the old voltage
C) Very little gain, almost negligable
D) Nothing at all.
Final Notes:
It usually takes a couple times to get the hang of it, but keep with it. While this is not GARUNTEED to work it does to some extent on any CPU reguardless of how old it is. As always your milliage *will* vary.
Member Results:
AthlonXP:
Sentential: 1800XP+ (2272mhz @ 1.82v -----> 2370 @ 1.75v)
Sentential: 2400XP+M (2412mhz @ 1.875v -----> 2460 @ 1.8v)
Sentential: 2600XP+M (2610mhz @ 1.825v -----> 2630 @ 1.75v)
Athlon64 (S754):
Athlon64 (S939):
Sentential: 3200+ (2415mhz @ 1.54v -----> 2445 @ 1.49v)
__________________________________
Pentium4 (S478):
Pentium4 (LGA775):
Sentential: 640 (3968mhz @ 1.44v -----> 4048 @ 1.42v)
Last edited:
![]-[itman](/forums/data/avatars/s/9/9309.jpg?1650325391){kind=avatar}
