Prime is melting my 4770k, oh the humanity. Other options?

[StoleMyOwnCar]

I believe it was just mentioned earlier that the safe voltage is 1.3, so you're well over it.

 

[Hurin]

I believe it was just mentioned earlier that the safe voltage is 1.3, so you're well over it.

I believe you'll find that the stated 1.3v is not the only opinion on the matter.

Though, I do see that it was Dan that said that above 1.3v might degrade the CPU (he did not say it was conclusive, however). And I trust his judgement. May back things off.

Here's what he said (and he was apparently quite careful in how he worded it):

One final note, be careful with your voltages and thermals. We've seen CPUs that seem to degrade in a short time. They'll run with DDR3 1866MHz memory speeds or so, or hit 4.8GHz at first and then over time they'll do less and less. Our sample size isn't large enough to state this conclusively, but it's something that we've "noticed" on one or more CPUs. There are other variables possibly at play, so again your mileage may vary. I'd hate to see people roast their CPUs when they don't have the luxury of simply having a new one sent to them. So be careful. Remember, at 4.4GHz Haswell is roughly as fast as Ivy Bridge was at 4.6GHz and Sandy Bridge at 4.8GHz.

Reading that very carefully, I think I'm willing to risk 1.34 vcore. If worse comes to worse, I can scrounge up the money for another CPU. I really just want a decent amount of distance between my old i920 at 4.1GHz. And 4.5GHz (or less) feels too much like a "sidegrade."

Edit: Sorry if that sounded too contentious and/or argumentative at the outset. I guess I got a little pissy that all those failed stress tests had potentially led me to naught.

 

[Riccochet]

I went from a 920 at 4.2 to my 4670k at 4.6. No where near a "sidegrade".

 

[Hurin]

I went from a 920 at 4.2 to my 4670k at 4.6. No where near a "sidegrade".

With respect, I momentarily clocked my 4770K to the same clock speed as my old 920. . . 4.1GHz.

The difference in two CPU-bound games (Rome Total War 2 and Company of Heroes 2) was only a few FPS (edited/corrected). I keep meaning to update my 920 thread with the actual numbers.

Our 920s were still quite capable. So we need to put some distance between them via OC MHz in order to realize any tangible benefit in games.

Of course, there is the added benefit of 6Gb/s SATA and USB 3. And it's nice to OC via multiplier instead of via BCLK. But those don't make my game perform any better when comparing my 920 at 4.1GHz and my 4770K had I kept it clocked at 4.3-4.5GHz.

But we digress. . .

 

[Justintoxicated]

Good luck!

I'd like to delid but at my current OC I never break 70C gaming so I feel it would be taking an unnecessary risk.

I have mixed thoughts on this, with BF4 I have yet to even hit 50c... I gotta figure out how to turn adaptive voltage off in my bios, it works from the ASUS software but it still seems to add voltage when I OC in the bios.

 

[WinSomeLoseNone]

There will be a setting for adaptive/static for the VCore & Ring voltage in the BIOS.

I have mixed thoughts on this, with BF4 I have yet to even hit 50c... I gotta figure out how to turn adaptive voltage off in my bios, it works from the ASUS software but it still seems to add voltage when I OC in the bios.

I just beat the Crysis 3 campaign last night without a single stability issue at my current OC. I've also logged about 15 hours in BF4 without a single crash over the past week.

Intel stress test has been run for a total of 10 hours with max temps @ 81C

Prime is stable for about 15 minutes (static volt) but I could give a shit at this point.

I'm calling it stable.

 

[AMD T-type]

With respect, I momentarily clocked my 4770K to the same clock speed as my old 920. . . 4.1GHz.

The difference in two CPU-bound games (Rome Total War 2 and Company of Heroes 2) was only a few FPS (edited/corrected). I keep meaning to update my 920 thread with the actual numbers.

Our 920s were still quite capable. So we need to put some distance between them via OC MHz in order to realize any tangible benefit in games.

Of course, there is the added benefit of 6Gb/s SATA and USB 3. And it's nice to OC via multiplier instead of via BCLK. But those don't make my game perform any better when comparing my 920 at 4.1GHz and my 4770K had I kept it clocked at 4.3-4.5GHz.

But we digress. . .

If all you are doing is playing video games sure...

But what if you are compressing/encoding, running a VM or two, etc, etc, there is most definitely a major performance increase if you are using it for the right reasons.
You can't just say "Eh, i only gained 5FPS in blah blah, def not worth the upgrade"

 

[Hurin]

If all you are doing is playing video games sure...

But what if you are compressing/encoding, running a VM or two, etc, etc, there is most definitely a major performance increase if you are using it for the right reasons.
You can't just say "Eh, i only gained 5FPS in blah blah, def not worth the upgrade"

The "right" reasons? I didn't realize such a concept exists.

As for VMs. . . I do run several VMs. . . and 4.1GHz on my i920 is passed along by the hypervisor just as well as 4.1GHz on my 4770K.

If by "right" reason, you mean video encoding and other more "niche" uses, then you have a good argument. But day-to-day usage by the average user (and gamer) is going to be largely unaffected.

That is, those who are somehow using it "wrong". . . I guess.



P.S. There's a reason why the benchmarks of CPUs within the reviews on this site are game-oriented.

 

[StoleMyOwnCar]

I believe you'll find that the stated 1.3v is not the only opinion on the matter.

Though, I do see that it was Dan that said that above 1.3v might degrade the CPU (he did not say it was conclusive, however). And I trust his judgement. May back things off.

Here's what he said (and he was apparently quite careful in how he worded it):

Reading that very carefully, I think I'm willing to risk 1.34 vcore. If worse comes to worse, I can scrounge up the money for another CPU. I really just want a decent amount of distance between my old i920 at 4.1GHz. And 4.5GHz (or less) feels too much like a "sidegrade."

Edit: Sorry if that sounded too contentious and/or argumentative at the outset. I guess I got a little pissy that all those failed stress tests had potentially led me to naught.

I honestly have no idea what you're talking about. This thing at stock was a huge improvement over my i5 750 for almost everything I tried (granted that's not yours but whatever); things felt much more snappy. Plus, I think these 780's in SLI... well I don't know how they'd work with a 750 over a 4770k. Who knows?

As for the voltage thing... your money is yours to gamble with as you please; just note the official docs and MSI are against that opinion. Supposing we aren't in some big smoldering crater about 5-10 years down the line, go ahead and tell us how using 1.34 vcore worked out.

 

[Hurin]

I honestly have no idea what you're talking about. This thing at stock was a huge improvement over my i5 750 for almost everything I tried (granted that's not yours but whatever); things felt much more snappy.

Placebo? Was your i5 overclocked?

The point is that there is no vast difference in everyday usage between a 4.1GHz haswell and a 4.1GHz Bloomfield (i7-920). I'm not sure why this is suddenly so controversial since it's been stated again and again by many people who have made the switch.

Plus, I think these 780's in SLI... well I don't know how they'd work with a 750 over a 4770k. Who knows?

There's a review on Anandtech that looks at CPU scaling on multi-GPU rigs. The answer is: Not much difference but the difference does increase somewhat along with resolution.

As for the voltage thing... your money is yours to gamble with as you please; just note the official docs and MSI are against that opinion. Supposing we aren't in some big smoldering crater about 5-10 years down the line, go ahead and tell us how using 1.34 vcore worked out.

I think the point was that it was declared that anything above 1.30v is unsafe based upon a statement by another person. Yet what was actually said by that cited source was that there were unconfirmed observations of "one or more" CPUs possibly degrading due to "volts and thermals" with many caveats thrown in about how this was not conclusive and there may be other factors at play.

 

[Justintoxicated]

There will be a setting for adaptive/static for the VCore & Ring voltage in the BIOS.



I just beat the Crysis 3 campaign last night without a single stability issue at my current OC. I've also logged about 15 hours in BF4 without a single crash over the past week.

Intel stress test has been run for a total of 10 hours with max temps @ 81C

Prime is stable for about 15 minutes (static volt) but I could give a shit at this point.

I'm calling it stable.

I wish I was as easy to please as you. Since my rig rebooted after 12 hours of small FFT with prime 28.1 I am going to consider it not stable, Granted I haven't had any reboots playing BF or the like, only when stress testing with small FFT prime which even with a massive watercooling setup will still bring the temp to 91c+...

At this point I think it might be necessary to delid haswell if you want to be able to run prime95 small FFT on these CPU's even as intel designed them with stock speeds and stock cooler. I think Intel screwed these CPUs up...

 

[StoleMyOwnCar]

Placebo? Was your i5 overclocked?

The point is that there is no vast difference in everyday usage between a 4.1GHz haswell and a 4.1GHz Bloomfield (i7-920). I'm not sure why this is suddenly so controversial since it's been stated again and again by many people who have made the switch.

I felt a pretty big difference doing everything, especially Civ 5. I calculate turns pretty fast now.

There's a review on Anandtech that looks at CPU scaling on multi-GPU rigs. The answer is: Not much difference but the difference does increase somewhat along with resolution.

And I happen to be playing on 1440p.

I think the point was that it was declared that anything above 1.30v is unsafe based upon a statement by another person. Yet what was actually said by that cited source was that there were unconfirmed observations of "one or more" CPUs possibly degrading due to "volts and thermals" with many caveats thrown in about how this was not conclusive and there may be other factors at play.

I think my point was that you're still gambling. This is objectively true. Good luck.

 

[Hurin]

I tend to follow that ("Justin's") thinking while OCing.

If I can make it crash in any way, then it's not stable. I want it to be able to pass the same torture test while OCed as it does when at stock.

To me, the whole point of OCing is to get free performance with zero drawbacks. And while I can't prove a negative (that there is no instability), passing the same stress tests at stock and OC gives me some reassurance that my OC is problem-free with no potential for issues down the road when I happen to start stressing it in other ways (video encoding, etc).

 

[Hurin]

I think my point was that you're still gambling. This is objectively true. Good luck.

That may be your point now. But your point prior was that we had been told that above 1.3v is unsafe. Which has not in any way been demonstrated to be true even by the person who carefully worded their initial, limited findings.

"I believe it was just mentioned earlier that the safe voltage is 1.3, so you're well over it."
​

 

[StoleMyOwnCar]

That may be your point now. But your point prior was that we had been told that above 1.3v is unsafe. Which has not in any way been demonstrated to be true even by the person who carefully worded their initial, limited findings.

"I believe it was just mentioned earlier that the safe voltage is 1.3, so you're well over it."
​

Nor has it been proven safe. There is generally a higher probability to have be at an unsafe voltage the further away it is from the spec that Intel has intended for it to run at. Right now, there is simply smoke. That is, talks regarding the 1.3 volt limit. There may not necessarily be a fire, but it's not implausible, and what you're doing is being one of the test cases for said assertion; you'll just be part of a statistic later on. I wanted to let you know that.

Also, it's not just those people. Also like I said, MSI MPower motherboard actually warns you to keep the vcore under 1.3V. Just a bit more smoke to add to that possible fire, there.

What you want to gamble with is your prerogative nonetheless though. I also wish to ask you, since you seem to think that you "sidegraded" to the 4770k: how noticeable is the difference between 3.9Ghz and 4.5Ghz, then, since the difference between Haswell and Bloomfield for you is so low (especially considering the huge performance gap between them in most tests)? How about 4.2 and 4.5?

 

[Hurin]

You said that something had just been said that was never actually said.

When you say that a given voltage is "unsafe" by a significant margin and then cite an authority, maybe it's just me, but I would think that you would want that cited authority to actually substantiate what you asserted.

The internet is replete with asserted facts that are seldom checked. But upon checking, turn out to be based on very shabby "evidence." The un-nuanced assertion that >1.3v is unsafe was on its way to becoming one of those dubious and yet accepted-at-face-value facts (at least in this thread). Someone checking to see if their volts are safe could read what you wrote and. . . well. . . believe you. As I did, before I went and actually checked your cited source.

Of course it's a fact that it hasn't been proven "safe" either. Since you can't prove a negative, you can't prove that 1.34 won't harm any particular CPU.

I really sorta grow tired of your sophistry on this. You said "I believe it was just mentioned earlier that the safe voltage is 1.3, so you're well over it." That sentence is untrue. Why not just show some integrity --which you seem to have-- and admit that you overstated what Dan had written?

All I ask is that people check their facts before asserting things as fact. And if they fail to do so, even via an honest mistake, they forthrightly retract what they said instead of engaging in rhetorical shenanigans to try to muddy the issue.

I also wish to ask you, since you seem to think that you "sidegraded" to the 4770k: how noticeable is the difference between 3.9Ghz and 4.5Ghz, then, since the difference between Haswell and Bloomfield for you is so low (especially considering the huge performance gap between them in most tests)? How about 4.2 and 4.5?

I wouldn't know, because as I've stated repeatedly, I'm at 4.7GHz.

Edit: It is not surprising that motherboard companies (Asus does as well) cite very conservative voltages as safe. Asus doesn't recommend anything over 1.25 unless you have a custom water cooling solution. That's not exactly compelling.

 

[StoleMyOwnCar]

You said that something had just been said that was never actually said.

When you say that a given voltage is "unsafe" by a significant margin and then cite an authority, maybe it's just me, but I would think that you would want that cited authority to actually substantiate what you asserted.

The internet is replete with asserted facts that are seldom checked. But upon checking, turn out to be based on very shabby "evidence." The un-nuanced assertion that >1.3v is unsafe was on its way to becoming one of those dubious and yet accepted-at-face-value facts (at least in this thread). Someone checking to see if their volts are safe could read what you wrote and. . . well. . . believe you. As I did, before I went and actually checked your cited source.

Of course it's a fact that it hasn't been proven "safe" either. Since you can't prove a negative, you can't prove that 1.34 won't harm any particular CPU.

I really sorta grow tired of your sophistry on this. You said "I believe it was just mentioned earlier that the safe voltage is 1.3, so you're well over it." That sentence is untrue. Why not just show some integrity --which you seem to have-- and admit that you overstated what Dan had written?

All I ask is that people check their facts before asserting things as fact. And if they fail to do so, even via an honest mistake, they forthrightly retract what they said instead of engaging in rhetorical shenanigans to try to muddy the issue.

I'm getting just as tired of you. Your argument against me is essentially that "I'm not sticking with my original point". That wasn't even my "point", that was just a comment based on our (everyone in this thread) interpretation of what Dan said, and this link:
http://www.tomshardware.com/answers/id-1694551/safe-voltage-temp-4770k.html#11533981

Along with what my motherboard is telling, and I doubt MSI would put a conservative voltage limit on an MPower MB. My point is simply that if you follow that interpretation or not... that's up to you. However, considering you're further away from the stock voltages than us (which are almost assured to be "safest"), you're being a test dummy for that extreme of an overclock. It's in no way an assertion that your voltages are soundly unsafe. I'm simply saying you are taking more of a gamble. Which is something you can't seem to understand.

Furthermore:

I wouldn't know, because as I've stated repeatedly, I'm at 4.7GHz.

Ugh, you don't understand what I'm getting at here, either. I'm essentially asking you how much "less of a sidegrade" that 800mhz overclock is compared to the stock. Apparently you think that [email protected] is a "sidegrade" compared to your old i7, but "Haswell overclocked by 800mhz isn't". So what exactly makes this 800mhz overclock so perceptibly less of a sidegrade compared to the 3.9Ghz stock?

 

[Hurin]

It's in no way an assertion that your voltages are soundly unsafe.

And yet this is. . .

I believe it was just mentioned earlier that the safe voltage is 1.3, so you're well over it.

The mind boggles.

Ugh, you don't understand what I'm getting at here, either. I'm essentially asking you how much "less of a sidegrade" that 800mhz overclock is compared to the stock. Apparently you think that [email protected] is a "sidegrade" compared to your old i7, but "Haswell overclocked by 800mhz isn't". So what exactly makes this 800mhz overclock so perceptibly less of a sidegrade compared to the 3.9Ghz stock?

I understand fully what you're "getting at". . . I just don't find your point compelling. It should seem obvious (or "objectively true" as you might put it) that a 4770k operating at 4.7GHz is going to be a more appealing upgrade from an i7-920 @ 4.1GHz than a 4770K operating at 3.9GHz.

Look, people who spend a lot of money on something and expect it to be faster will likely perceive an increase in speed. If you never bother to check via reliable benchmarks, you're going to be happy with your upgrade. It's all in the eye of the beholder. And I'm happy that you're happy. I'd caution you not to look into it any further though lest that happiness evaporate.

Yet, objectively, some i7-920 owners with substantial overclocks have actually tested and documented the performance difference (or lack thereof) on Haswell and expressed their disappointment.

I'll take the latter, demonstrable numbers over the subjective, likely placebo-driven "impressions" of someone who just says: "My computer feels much faster. I love my new CPU!" And yet I don't begrudge them their happiness. I do begrudge them their assertions that I'm "crazy" for not agreeing with them even when my "impressions" are actually backed up by my own benchmarks and those of others.

 

[Hurin]

Along with what my motherboard is telling, and I doubt MSI would put a conservative voltage limit on an MPower MB.

As long as we're citing motherboard manufacturers. . .

Again, I don't hold motherboard manufacturers to be an authority either way. They tend to be conservative lest they be blamed for the fried CPUs of their customers.

And yet, here's MSI saying that 1.3v-1.5v is "fun" (just above normal) and "danger" begins at 1.5v+.

So, our digression continues. But it only continues because it was necessary to point out that what you stated, that Dan had told us that >1.30v is not safe, was in fact not actually stated by Dan. He in fact went to great pains to avoid saying that and loaded the post with caveats.

 

[Riccochet]

No, Dan D did not say specifically that more than 1.3v would cause degradation. What he did say is that in the small sample of processors they've encountered they've seen degradation due to the stresses they put on the chips. You can certainly go back through ALL the tests they've done with their chips and see what stresses those are. 1.3v, well, they used more than that.

 

[Hurin]

No, Dan D did not say specifically that more than 1.3v would cause degradation. What he did say is that in the small sample of processors they've encountered they've seen degradation due to the stresses they put on the chips. You can certainly go back through ALL the tests they've done with their chips and see what stresses those are. 1.3v, well, they used more than that.

And, to be clear, they even expressed uncertainty about whether this happened to more than one CPU. . .

Our sample size isn't large enough to state this conclusively, but it's something that we've "noticed" on one or more CPUs. There are other variables possibly at play, so again your mileage may vary.
​

They are careful there to make it clear that they can't even say it happened more than once (perhaps more than one person noticed it, or it was noticed on two different systems, but they're not sure if the same CPU was used in both, etc.). Along with stating that there are other potential causes that could be responsible for the behavior they observed. And, to say nothing of the fact that the only exhibited behavior is a decrease in the RAM speed the chip would run, not BSODs (unrelated to memory speed) or other classic symptoms of "electromigration" usually attributed to heat and voltage.

You simply can't get from that to "it was just mentioned that the safe voltage is 1.3."

 

[StoleMyOwnCar]

And yet this is. . .



The mind boggles.

Why, because you can't read "believe" and "mentioned" tidbits? I was just essentially acknowledging that these degradation articles aren't necessarily "facts" (though you've yet to explain that TomsHardware forum link). But that they're likely not necessarily coming out of nowhere. You seem to also really fricken love to argue over trivialities and semantics. Fact of the matter is, unlike many past Intel procs, we don't know what the safe voltages for these are. The few sources that we have seen any kind of evidence from, reputable or not, say 1.3... in this thread. That's all I was saying. I don't give a crap what the hell you do to your processor anymore, so you can take that or leave it, I'll stop arguing about it.

I understand fully what you're "getting at". . . I just don't find your point compelling. It should seem obvious (or "objectively true" as you might put it) that a 4770k operating at 4.7GHz is going to be a more appealing upgrade from an i7-920 @ 4.1GHz than a 4770K operating at 3.9GHz.

http://www.anandtech.com/bench/product/47?vs=836
In all of those tests, the 4770k is 80-100%+ faster at stock. That means this CPU is roughly 85% (let's say) on average faster than your old one, or almost twice as fast. Now if we do the math though, obviously a 4.1Ghz overclock on the 920 is likely going to bring it much closer to the 4770k.

3.9Ghz->4.7Ghz is a 20% increase in core clocks, so you're essentially saying 1.2*85%=102% or an 17% delta over your previous increase over your previous processor is somehow suddenly acceptable perceptively compared to the previous increase over your previous processor, and that's what you need to not make this a "sidegrade". That just didn't make much sense to me.

As for me, no, for me it isn't an imperceptible difference at all, especially not in Civ 5 and some of the other things I do.
http://www.anandtech.com/bench/product/109?vs=836

This thing archives more than twice as fast as my old CPU, converts my songs much faster, and does everything else faster. This is keeping the same Windows install, too. Of course I did not overclock my i5 750 like you did your 920.

Do what you want, though, I was just trying to give you some words of caution and you are quite frankly the cattiest person I've ever met regarding that. If you value your few hundred extra mhz that much then by all means do it. Tell us how it goes down the road.

 

[Hurin]

Why, because you can't read "believe" and "mentioned" tidbits? I was just essentially acknowledging that these degradation articles aren't necessarily "facts"

Those "tidbits" don't provide the qualifiers that you seem to think they do. But, if they did, why on earth wouldn't you more graciously accept the correction that you claim those qualifiers allow for?

(though you've yet to explain that TomsHardware forum link)

Someone on a forum somewhere said 1.3v. Yay! I've seen others on other forums just as authoritatively say 1.4v or even higher. Just as you said authoritatively (though incorrectly) here that we were just told that it's 1.3v. To say nothing of the MSI image I posted.

But that they're likely not necessarily coming out of nowhere. You seem to also really fricken love to argue over trivialities and semantics.

Guilty. But I'd characterize it as pointing out where people draw inaccurate/wrong conclusions due to a faulty memory, or inability to read carefully and critically. As I said, you can't get to the conclusion you drew from what you read if you actually read it carefully.

Do what you want, though, I was just trying to give you some words of caution and you are quite frankly the cattiest person I've ever met regarding that. If you value your few hundred extra mhz that much then by all means do it. Tell us how it goes down the road.

Will do. My apologies for the cattiness. But would it have killed you to just freakin' say that you overstated and/or misread Dan's observation? Or at the very least should not have stated it so flatly without any (real) qualifications beyond what amounts to "I think I read. . ."?

 

[Gadfly]

Hey all..

I think there is a fundamental misunderstanding about semi-conductors in this thread. They get old, they start to "bleed" electrons, and will run hotter and hotter as they age, until eventually they start to fail. Most CPU's will be thrown out long before they fail, but most will start to run hotter even after just a few years.

The more voltage you run, and hotter you run the chip, the faster this aging process will occur. If you over volt/temp and and blow (arc) a trace, the chip is done.

OC'ing has that negative effect on the CPU, It will degrade faster; but if you keep it's temps under control with a reasonable voltage bump, it will most likely remain stable much longer than you will own it. How much voltage is not a question that can be applied universally to a model or family of chips. Each batch, each wafer will be different. Just as why some will clock at 4.3, and others at 4.8Ghz; some might outright fry at 1.35v and another will be fine at 1.5v, it really just depends on the quality, thickness, and cleanliness of the traces; but ALL CPU's will degrade with age and use, how much and how fast is simply not fixed.

That said... the prime 95 debate is as old as the program itself; as it really is not more than a bragging tool. If you can overclock you cpu and it is stable for your usage (playing games all day, compiling code, or doing some renders). It is stable, if it fails BSOD's after 2 hours of prime95 who cares?

It is stable in the real world, which is all that matters.

 

[Hurin]

That said... the prime 95 debate is as old as the program itself; as it really is not more than a bragging tool. If you can overclock you cpu and it is stable for your usage (playing games all day, compiling code, or doing some renders). It is stable, if it fails BSOD's after 2 hours of prime95 who cares?

It is stable in the real world, which is all that matters.

With great respect for what you said right up to that point. . . I don't Prime95 for 24 hours for bragging rights. I do it because time and time again I've thought a system was stable after many hours of testing and using it. . . but there would be nagging behavior that would give me pause. An application would hang on closing (but the OS would still be fine), or an ISO would fail to mount (but the second try would work). These things would happen just often enough to make me doubt the overclock, but not so often that I could be sure they weren't just a fluke or a bug in the application.

In those situations, I have found a 24-hour run of Prime95 w/ Looping 3D demo (now using Valley) will 99% of the time demonstrate an instability. And then once I can pass that torture test, the rest of the inconclusive symptoms and behavior disappear.

I think there may be a lot of people "calling it stable" who are perhaps attributing everyday symptoms of instability with just flukey behavior, badly coded apps, or sunspots (heh).

I do a prime95 24-hour test to satisfy myself that there's likely not instability in the system I depend on in multiple ways. I couldn't really give a shit how stable other people consider my computer. Heh, and yet I took the time to write this.

 

[Justintoxicated]

Hey all..

I think there is a fundamental misunderstanding about semi-conductors in this thread. They get old, they start to "bleed" electrons, and will run hotter and hotter as they age, until eventually they start to fail. Most CPU's will be thrown out long before they fail, but most will start to run hotter even after just a few years.

The more voltage you run, and hotter you run the chip, the faster this aging process will occur. If you over volt/temp and and blow (arc) a trace, the chip is done.

OC'ing has that negative effect on the CPU, It will degrade faster; but if you keep it's temps under control with a reasonable voltage bump, it will most likely remain stable much longer than you will own it. How much voltage is not a question that can be applied universally to a model or family of chips. Each batch, each wafer will be different. Just as why some will clock at 4.3, and others at 4.8Ghz; some might outright fry at 1.35v and another will be fine at 1.5v, it really just depends on the quality, thickness, and cleanliness of the traces; but ALL CPU's will degrade with age and use, how much and how fast is simply not fixed.

That said... the prime 95 debate is as old as the program itself; as it really is not more than a bragging tool. If you can overclock you cpu and it is stable for your usage (playing games all day, compiling code, or doing some renders). It is stable, if it fails BSOD's after 2 hours of prime95 who cares?

It is stable in the real world, which is all that matters.

Well prime 95 v28.1 was just released to test haswell, so it's not as old as your thinking? I mean it is an old program but the stress it puts out is different than the old versions Also if my system can run prime95 for 24 hours and yours can't why is that? Might it be because your computer is not entirely stable? I'll take stability over a couple hundred Mhz anyday! Now if all haswells could not run prime95 v28.1 that would be a little different.

 

[Gadfly]

Well prime 95 v28.1 was just released to test haswell, so it's not as old as your thinking? I mean it is an old program but the stress it puts out is different than the old versions Also if my system can run prime95 for 24 hours and yours can't why is that? Might it be because your computer is not entirely stable? I'll take stability over a couple hundred Mhz anyday! Now if all haswells could not run prime95 v28.1 that would be a little different.

The application itself is very old, and at it's core, it is running the same stresses on the CPU by running prime number calculations. It was been used by overclockers as a stability test since at least the mid 90's as that is when I started using it.

There are many reasons why a 24 hour loop at 100% load may fail, and only a few have to do with an overclock, but I will completely agree that stability > a few hundred Mhz, but absolute stability and overclocking are counter productive by nature.

Prime95 is a useful tool to have in your tool kit, it is just not really an effective demonstration of real world stability, nor is it the only measure of it.

 

[Gadfly]

With great respect for what you said right up to that point. . . I don't Prime95 for 24 hours for bragging rights. I do it because time and time again I've thought a system was stable after many hours of testing and using it. . . but there would be nagging behavior that would give me pause. An application would hang on closing (but the OS would still be fine), or an ISO would fail to mount (but the second try would work). These things would happen just often enough to make me doubt the overclock, but not so often that I could be sure they weren't just a fluke or a bug in the application.

In those situations, I have found a 24-hour run of Prime95 w/ Looping 3D demo (now using Valley) will 99% of the time demonstrate an instability. And then once I can pass that torture test, the rest of the inconclusive symptoms and behavior disappear.

I think there may be a lot of people "calling it stable" who are perhaps attributing everyday symptoms of instability with just flukey behavior, badly coded apps, or sunspots (heh).

I do a prime95 24-hour test to satisfy myself that there's likely not instability in the system I depend on in multiple ways. I couldn't really give a shit how stable other people consider my computer. Heh, and yet I took the time to write this.

Simply running prime95 for 24 hours proves little in terms of real world stability. If you use it as a tool to make yourself feel better, then fair enough; but it is hardly the be all end all tool for determining stability. It has been known for many many moons that the prime95 torture test is just that.. a torture test, and though it will stress your cooling and cpu to it's limits, there are many many causes of loop failures which have nothing to do with an overclock, or it's stability.

 

[Justintoxicated]

Simply running prime95 for 24 hours proves little in terms of real world stability. If you use it as a tool to make yourself feel better, then fair enough; but it is hardly the be all end all tool for determining stability. It has been known for many many moons that the prime95 torture test is just that.. a torture test, and though it will stress your cooling and cpu to it's limits, there are many many causes of loop failures which have nothing to do with an overclock, or it's stability.

What are these causes then? Mine passes 24 hours 100% no errors, and so has every stable computer I have ever tested. Only overclocked computers ever seem to fail for me. A CPU that can't run 100% load for 24 hours seems unstable to me. Now on a 110 degree environment I'm sure I might loose some stability so I wouldn't say it is absolute, but it is "stable" since I do not get errors. Overclocking and being stable isn't counter productive at all IMO. Overlcocking is all about tweaking your PC to run optimal, and optimal to many means free MHZ while still being stable (functioning correctly at all times).

 

[Hurin]

I think what he means is that if Prime95 fails, you can't be sure it's not a fault in Prime95, or even a motherboard issue, or a driver issue. . . etc.

But, here's the thing. . . if you're being meticulous, you've already established that your computer can run Prime95 for 24 hours at stock speed. If it can do that, but can't do that when you're overclocked, that strongly indicates that there is something about your overclock that isn't entirely stable.

By validating your hardware at the outset at stock speeds, you establish to a great degree of certainty that your testing methodology is valid.

--H