Rumour: Bulldozer 50% Faster than Core i7 and Phenom II.

[AtenRa]

I have not seen any evidence presented by you or the countless other OCers in the well over a decade since I've first engaged in these types of OC vs. No OC debates that it does not shorten service life. If OCers were so convinced that it was a non-issue why has no one ever (to the best of my knowledge) set up this experiment:

Purchase three identical computers

Leave one box stock

Set up one at a MILD OC

Set up one at a KILLER OC but still on air

Set each up at the same day to run Folding (or some other high steady load) 24/7

Keep them running for however many years it will take until they die.

Publish the results

Until then it's just your word that it's safe. Well, I don't know you, and although I respect your opinions, that does not mean that I'm going to clone them for myself.

Stock box will last 20 years

mild OC PC will last 18 years

Killer OC on air will last 10-15 years

Have you ever used a CPU for more than 5-6 years for work ??? I have used in the past an OCed CPU for SETI running 24/7 for a year until i replaced the entire PC with a new dual core, that cpu still works.

I will say again, if you know what you doing OC is not dangerous and nobody forcing you to OC your hardware but,
try and learn from people that they have done it countless times with countless different hardware.

Trying OC is not the same as having a dick in you ass (sorry but i had to say that)

 

[LOL_Wut_Axel]

Frankly I'm not worried about service life.

As mentioned earlier, silent corruption is the real killer. Bad data in a business? Lights out.

Which you have yet to prove, BTW. I have provided facts of the contrary. Only documented effects of doing a small OC on stock voltage is tiny bit higher temps, which leads to a theoretical increase of degradation. Academic at best, anyway, since the chip will still last way beyond its service life, whether it be five or ten years, or twenty years.

When you can prove this supposed corruption exists for OCing a newer CPU 300MHz get back to me. I've already posted facts as to otherwise.

Frankly, this whole thread except for a small amount of posters wreaks of failure. Presented with objective, empirical data, they'll still be paranoid and others make sure they remain so. Then they'll tell you to prove something when the burden of proof is on them.

 

[Cogman]

Frankly I'm not worried about service life.

As mentioned earlier, silent corruption is the real killer. Bad data in a business? Lights out.

A system will crash pretty dang fast once the CPU starts putting out incorrect data. The worry of the CPU corrupting large swaths of data is pretty unfounded. (A Mb of data at most is going to be corrupted before the whole system goes down).

Silent corruption from the CPU just isn't likely. Now, silent corruption from the Memory or Hd, that is a real possibility.

That being said, you should NEVER overclock a system meant for business work. Stability is way more important than speed.

 

[kevinsbane]

A system will crash pretty dang fast once the CPU starts putting out incorrect data. The worry of the CPU corrupting large swaths of data is pretty unfounded. (A Mb of data at most is going to be corrupted before the whole system goes down).

Silent corruption from the CPU just isn't likely. Now, silent corruption from the Memory or Hd, that is a real possibility.

That being said, you should NEVER overclock a system meant for business work. Stability is way more important than speed.

This is why you do not overclock business machines. An overclock, however trivial, will increase the instability of any given system by a given non-zero increment.

OCing a business machine: is the benefit (do things marginally faster) worth the cost (time spent overclocking and stability testing) and risks (instability, crashes, potential lost time & money, The Call (R) from the boss about why his machine crashed while presenting a million dollar proposal to a client)? Running a machine at stock, you can blame Intel or Dell or whoever. Running a machine out of spec (overclocking...), you're on the hook.

 

[bridito]

You can't draw any scientific conclusions from just 3 machines. In fact, you would need several 1000 machines before you can even think of drawing a conclusion. This is why there are no published results. Nobody is going to spend the millions it takes to verify that a machine being run out of spec has comparable service life.

I'm not so sure. I don't think that Intel or AMD would set its warranty limits as they are without having tested the living daylights out of various CPUs at various clocks. These guys have billions of dollars in R&D budgets and they know a lot more about CPUs than they publicly state.

a CPU pushed to its limits will eventually slowly degrade and be unable to maintain its high clock, especially if it is being overvolted. This is a known phenomena that is pretty easy to observe. That being said, it usually takes a couple of months to maybe even a year before the system degrades into an unstable state.

OK, so I'm almost at 3 years with my i7 940. Had I OC'd it on Day One and kept it there, would I be corrupting data or showing other signs of instability? Again, if there is the minutest even remotely feasible chance, count me out!

That being said, you should NEVER overclock a system meant for business work. Stability is way more important than speed.

You have elegantly and succintly summarized my opinion. Thank you.

Stock box will last 20 years

mild OC PC will last 18 years

Killer OC on air will last 10-15 years

Why does that sound remarkably like "BD IS 4X FASTER THAN 2600K!" I'll believe it... when I see the indy benchys. I'm sure you're a perfectly nice and intelligent person but I have placed my trust firmly in benchys, benchys and only benchys. Until the real, true, verifiable, duplicatable benchys are available it's just some guy spouting fanciful illusions on the internet... Just like this guy:

http://articles.chicagotribune.com/...10704_1_modern-church-universe-splinter-group

Hey, I'm just as willing to believe the Earth is at the center of the universe... show me the evidence and I'll not only agree, but evangelize it!

Trying OC is not the same as having a d**k in you ass (sorry but i had to say that)

I assure you neither is going to happen... at least while I'm still alive!

 

[bridito]

This is why you do not overclock business machines. An overclock, however trivial, will increase the instability of any given system by a given non-zero increment.

OCing a business machine: is the benefit (do things marginally faster) worth the cost (time spent overclocking and stability testing) and risks (instability, crashes, potential lost time & money, The Call (R) from the boss about why his machine crashed while presenting a million dollar proposal to a client)?

Completely agreed. I've had days where a deadline right to the minute results in the issuance of a $5,000 invoice. Am I supposed to risk a BSOD or other catastrophe because I was too damn cheap to shell out another couple of hundred bucks on a faster CPU so I OCd a cheaper one? That can be the conclusion only of the graduates from the JWoww & Snooki Business School.

 

[LOL_Wut_Axel]

This is why you do not overclock business machines. An overclock, however trivial, will increase the instability of any given system by a given non-zero increment.

OCing a business machine: is the benefit (do things marginally faster) worth the cost (time spent overclocking and stability testing) and risks (instability, crashes, potential lost time & money, The Call (R) from the boss about why his machine crashed while presenting a million dollar proposal to a client)?

BS.

Prove to me that a Core i5 2500 at 3.6GHz on stock voltage is any less stable than one at 3.3GHz. You don't have to spend time stability testing any more than the first time you do the OC. Prove these risks.

Your proposal example is just hilarious bull crap taken directly out of your ass and based on nothing but your own bias.


OK, You were warned. Vacation time.
Markfw900
Anandtech Moderator

 

[VirtualLarry]

Which you have yet to prove, BTW. I have provided facts of the contrary. Only documented effects of doing a small OC on stock voltage is tiny bit higher temps, which leads to a theoretical increase of degradation. Academic at best, anyway, since the chip will still last way beyond its service life, whether it be five or ten years, or twenty years.

When you can prove this supposed corruption exists for OCing a newer CPU 300MHz get back to me. I've already posted facts as to otherwise.

Well, if I experienced it at 500Mhz OC, and every chip is mfg with varying tolerances, don't you think that there is a possibility that some chips might be unstable with a 300Mhz stock-voltage OC?

Don't you think that you personally drawing a line at 300Mhz, and excluding my example because it is 500Mhz, is a bit arbitrary? Since nearly all E5200 (and derivatives, like the Q9300 - two E5200s in one chip) can usually handle a stock voltage OC from 2.5 to 3.0. But the thing is, not ALL chips can do that. I had to bump up the vcore a trivial amount to avoid crashing OCCT on one of my desktops.

Shall I remove the vcore boost, and try OC it 300Mhz, and then show you that your statement is bunk?

 

[LOL_Wut_Axel]

Well, if I experienced it at 500Mhz OC, and every chip is mfg with varying tolerances, don't you think that there is a possibility that some chips might be unstable with a 300Mhz stock-voltage OC?

Don't you think that you personally drawing a line at 300Mhz, and excluding my example because it is 500Mhz, is a bit arbitrary? Since nearly all E5200 (and derivatives, like the Q9300 - two E5200s in one chip) can usually handle a stock voltage OC from 2.5 to 3.0. But the thing is, not ALL chips can do that. I had to bump up the vcore a trivial amount to avoid crashing OCCT on one of my desktops.

Shall I remove the vcore boost, and try OC it 300Mhz, and then show you that your statement is bunk?

All new chips should handle a 300MHz OC just fine. And no, once you get a certain point, you get to a limit to where you're reaching the point where it starts becoming unstable. At many times that is at 500MHz.

You doing this won't prove anything, anyway. Since you over-volted the chip it's already degraded and may not handle something it would've handled just fine some months ago. I'm arguing over a chip in always stock voltage.

 

[Cogman]

I'm not so sure. I don't think that Intel or AMD would set its warranty limits as they are without having tested the living daylights out of various CPUs at various clocks. These guys have billions of dollars in R&D budgets and they know a lot more about CPUs than they publicly state.

Intel and AMD often don't work with the notion of "lets see just how fast we can push this" rather, they say "Ok, this is the speed requirement, lets make sure we hit it."

In otherwords, they usually don't test the case of overclocking, rather, they design to a target. (and yes, they test, and test, and test, and test their systems 1000x over. And after they finish testing, they give the CPU to systems manufactures such as HP or Dell, who then test, and test, and test the CPUs like crazy.)

OK, so I'm almost at 3 years with my i7 940. Had I OC'd it on Day One and kept it there, would I be corrupting data or showing other signs of instability? Again, if there is the minutest even remotely feasible chance, count me out!

Data corruption (beyond the norm) is pretty unlikely, impossible even. The fact is, you're system will crash just as soon as data corruption begins. It is sort of its own correcting system.

Now, if you overclocked from day one, would there be an increased chance of instability? Well, it depends on how you are overclocking. If you haven't move the voltage from stock, then I would say no, your system will be just as stable as it was before. If you have, then yes, your chances of instability have increased (by how far depends on how much you've increased your voltage.)

If there was any instability, it more than likely could be fixed by reducing how much you are overclocking the CPU.

In other words, it depends. Your system could be 100% stable after 3 years, if you are overclocking by 100Mhz, you haven't done any damage.

 

[Cogman]

All new chips should handle a 300MHz OC just fine. And no, once you get a certain point, you get to a limit to where you're reaching the point where it starts becoming unstable. At many times that is at 500MHz.

You doing this won't prove anything, anyway. Since you over-volted the chip it's already degraded and may not handle something it would've handled just fine some months ago. I'm arguing over a chip in always stock voltage.

I've personally seen chips that everyone says "should" be able to do X at stock voltage fall flat on their face. The most recent one was an e6600.

OCing is a pot shot. You have no guarantee that you're system will be 100% stable 1 Mhz beyond stock frequency.

 

[kevinsbane]

BS.

Prove to me that a Core i5 2500 at 3.6GHz on stock voltage is any less stable than one at 3.3GHz.

Prove to me that a Core i5 2500 at 3.6 GHz on stock is 100% as stable as one at 3.3 GHz. The onus isn't on me to prove that it isn't. I know an i5 2500K @ 4.0 on stock voltage is less stable than one at 3.3. It seems obvious to me that the faster you push something, the more unstable it gets. Why then should I not make the very reasonable assumption that going to 3.6 will introduce a non-zero increase in stability? If it truly is a zero increase in instability, why do stability testing at all?

You don't have to spend time stability testing any more than the first time you do the OC. Prove these risks.

How long does that take? How much time (and thus, money) should I spend doing the stability tests to ensure that I get 100% the same stability as stock? Does my investment of time and money (this is a work machine, I get paid for my time) mean the business can make more money when I overclock the machines? More than what was invested, and enough to take the risks?

This is NOT an issue of whether overclocking is safe or not. We know it can be done safely. This is an issue of whether overclocking is worth it or not in a business environment.

 

[bridito]

Intel and AMD often don't work with the notion of "lets see just how fast we can push this" rather, they say "Ok, this is the speed requirement, lets make sure we hit it."

In otherwords, they usually don't test the case of overclocking, rather, they design to a target. (and yes, they test, and test, and test, and test their systems 1000x over. And after they finish testing, they give the CPU to systems manufactures such as HP or Dell, who then test, and test, and test the CPUs like crazy.)

I see your point, I'm just disputing the rationale behind "designing to a target." If a particular architecture was designed for 3.4 GHz and it was found to be rock steady, reliable, etc. at 4.2 GHz with the accompanying across the board speed gain, why would Intel/AMD go pour a few billion more dollars into designing a completely new CPU? They'd just add a digit to the nomenclature, pop the MSRP up by $300 and call it an all new chip. It boggles my imagination to think that they wouldn't even push the suckers to see how fast they could run just to make sure.

In other words, it depends. Your system could be 100% stable after 3 years, if you are overclocking by 100Mhz, you haven't done any damage.

Unfortunately "it depends" and "could be" is not sufficient to pacify me. I'll stick to my box stock CPU run firmly within warrantiable limits and at least I'll be sure that if I blow a deadline and have to eat a $5,000 invoice, at least I'll only have my own sloth to blame, and not the fact that I could have spent a few extra bucks for a faster CPU instead of taking the cheap way out to get some more vroom vroom.

This is an issue of whether overclocking is worth it or not in a business environment.

100% agreed.

Anyway, it's 9 am here and I owe I owe so off to work I go. In the next minute, this PC is now going to turn from an enthusiast's system which could be pushed to its limits just for the lolz with no negative impact with the exception of maybe buying replacement components to a business system which either works impeccably as long as I need it to or I'm going to be living under a freeway overpass in a large cardboard box. Later!

 

[Phynaz]

Prove to me that a Core i5 2500 at 3.6GHz on stock voltage is any less stable than one at 3.3GHz. You don't have to spend time stability testing any more than the first time you do the OC. Prove these risks.

Your proposal example is just hilarious bull crap taken directly out of your ass and based on nothing but your own bias.

You're asking somebody to prove a negative. How about you prove the positive.

Prove every Core i5 2500 at 3.6GHz on stock voltage is just as stable as one at 3.3GHz

 

[formulav8]

you people are boring

 

[Phynaz]

you people are boring

It happens when you reach middle age :biggrin:

At least that's my excuse.

 

[Cogman]

I see your point, I'm just disputing the rationale behind "designing to a target." If a particular architecture was designed for 3.4 GHz and it was found to be rock steady, reliable, etc. at 4.2 GHz with the accompanying across the board speed gain, why would Intel/AMD go pour a few billion more dollars into designing a completely new CPU? They'd just add a digit to the nomenclature, pop the MSRP up by $300 and call it an all new chip. It boggles my imagination to think that they wouldn't even push the suckers to see how fast they could run just to make sure.

Because they have more targets than just "how fast can it go." One big one that limits them from pushing all their chips to 4.2 Ghz is power consumption. When you overclock, you raise power consumption. Anything beyond 100Ws of power consumption is going to be a non-stater for both companies, yet overclocking frequently passes that powerconsumption level.

Unfortunately "it depends" and "could be" is not sufficient to pacify me. I'll stick to my box stock CPU run firmly within warrantiable limits and at least I'll be sure that if I blow a deadline and have to eat a $5,000 invoice, at least I'll only have my own sloth to blame, and not the fact that I could have spent a few extra bucks for a faster CPU instead of taking the cheap way out to get some more vroom vroom.

What it depends on are things that you control. It depends on how far you push your cpu. That was my point. If you are barely pushing your CPU, then it will be 100% stable.

Now, barely pushing is going to be a CPU to CPU dependent thing. However, if you push you're CPU to the max (without increasing voltage), test for stability, then take half that frequency, I can guarantee that your CPU will be 100% stable after 3 years.

But again, if this is for business, I wouldn't do it. Just because chances are pretty high that you really don't need a faster CPU. So going through the whole rigmarole of overclocking and testing is quite worthless.

 

[LOL_Wut_Axel]

Prove to me that a Core i5 2500 at 3.6 GHz on stock is 100% as stable as one at 3.3 GHz. The onus isn't on me to prove that it isn't. I know an i5 2500K @ 4.0 on stock voltage is less stable than one at 3.3. It seems obvious to me that the faster you push something, the more unstable it gets. Why then should I not make the very reasonable assumption that going to 3.6 will introduce a non-zero increase in stability? If it truly is a zero increase in instability, why do stability testing at all?

How long does that take? How much time (and thus, money) should I spend doing the stability tests to ensure that I get 100% the same stability as stock? Does my investment of time and money (this is a work machine, I get paid for my time) mean the business can make more money when I overclock the machines? More than what was invested, and enough to take the risks?

This is NOT an issue of whether overclocking is safe or not. We know it can be done safely. This is an issue of whether overclocking is worth it or not in a business environment.

I already addressed that. Look back into the thread. You're the ones making wild claims with no proof to support them. All your arguments have been based on unproven things.

A Core i5 at 4GHz on stock voltage is less than stable than one at 3.6 or 3.3 because chips have headroom as to how much they can go before they start becoming unstable on stock and raised voltage. At that speed it hasn't reached it.

I already said it could be valuable to a home or home business user because it increases efficiency. I also debunked the degradation BS that was spreading here.

There's no point in arguing this, at the end of the day. You morons will keep arguing even when evidence is presented to you several times that you're wrong and you keep spreading BS.

 

[Cogman]

I already said it could be valuable to a home or home business user because it increases efficiency. I also debunked the degradation BS that was spreading here.

There's no point in arguing this, at the end of the day. You morons will keep arguing even when evidence is presented to you several times that you're wrong and you keep spreading BS.

You've debunked nothing. I've seen system degradation from overclocking first hand. It is known and proven. There is physics to back it up. You've provided no evidence.

 

[LOL_Wut_Axel]

You've debunked nothing. I've seen system degradation from overclocking first hand. It is known and proven. There is physics to back it up. You've provided no evidence.

Yes, I have. Look at the study Anandtech did on Core 2. It's linked around one page back. Degradation from over-clocking on stock voltage can only come from higher temperatures, though that's easy to eliminate.

Now shut up.

 

[busydude]

Now shut up.

Oh, that is real classy.

 

[dali71]

Yes, I have. Look at the study Anandtech did on Core 2. It's linked around one page back. Degradation from over-clocking on stock voltage can only come from higher temperatures, though that's easy to eliminate.

Now shut up.

Oh, that is real classy.

New guy is just making a name for himself. Not a good name, but apparently he's not concerned about that.

 

[kevinsbane]

I already addressed that. Look back into the thread.

Very insufficiently. See the below quotes, by you.

From what I'm reading all this sounds to me just like panicking over a 0.0001% chance of something going wrong. I'll believe a 300MHz OC on a modern CPU at stock voltage can cause havok on hardware when I see it.

Here's the problem with your argument: there's a lot less "risk" of OCing a CPU 300MHz on stock voltage than from OCing a CPU 800MHz+ on a moderate over-volt.

You haven't proved it's a problem or that it may become one. Several stability tests will tell you otherwise. Where's one of them that proves all them wrong? ...

Your only proof that you presented that I could see was supposed to prove that a 300 mhz OC is completely reliable is the last quote... about how some stability tests (I am guessing Prime 95, LinX, etc...) will tell you it's stable. Nowhere have you made a case that there is zero increase in instability with a mild overclock. It's just that you feel that the risk is very low. Which is the same position I'm taking. Where we differ is that I feel that the benefit of mild overclocks are not worth the costs and risks for business (ie, you're making money off it), including home businesses.

I already said it could be valuable to a home or home business user because it increases efficiency. I also debunked the degradation BS that was spreading here.

I'm going to make a generalization. No one here really cares about degradation. It is as you say, it doesn't matter.

How is does this increase in efficiency make me money? Why shouldn't I just buy more efficient parts, so I don't even have to put in the time to support my own overclock?

Why don't we go the other way, and UNDERvolt our CPU's? I can say that the risk is small that way too, and would be far more beneficial in efficiency improvements than a mild overclock. Not that I would, for the exact same reasons I would not overclock a business computer. Undervolting reduces the stability, reducing reliability.

Go through the math. In 99% of the cases, it simply isn't worth it, even if there is 0 risk. Say it takes me 1 hour to overclock and validate the overclock so that the risk increase is 0. How much money could I have charged a client in that hour? $50? $100? $150? What benefit do I get in effiency? $2-3 a month in energy savings? And that's a best-case, ideal scenario. It would be more like spend 2 hours now, and 2 hours in the next 5 years related to OCing troubleshooting.

A Core i5 at 4GHz on stock voltage is less than stable than one at 3.6 or 3.3 because chips have headroom as to how much they can go before they start becoming unstable on stock and raised voltage. At that speed it hasn't reached it.

I see. How did you find out that 3.6 GHz on an i5 2500@stock volts was inside the headroom? What is the headroom?

 

[Markfw]

With all due respect and appreciation for your perspective could you please tell me why I have to be pressed into adopting your viewpoint and insulted when I do not? You are making arguments which are no more compelling than those made by my homosexual friends in that I really should "try it". I'm not going to engage in either of these pursuits. Homosexuality because I like chicks ONLY... and OCing because I have not seen any evidence presented by you or the countless other OCers in the well over a decade since I've first engaged in these types of OC vs. No OC debates that it does not shorten service life. If OCers were so convinced that it was a non-issue why has no one ever (to the best of my knowledge) set up this experiment:

Purchase three identical computers

Leave one box stock

Set up one at a MILD OC

Set up one at a KILLER OC but still on air

Set each up at the same day to run Folding (or some other high steady load) 24/7

Keep them running for however many years it will take until they die.

Publish the results

Until then it's just your word that it's safe. Well, I don't know you, and although I respect your opinions, that does not mean that I'm going to clone them for myself.

I have sort of done that. The problem is that they become obsolete in 3-5 years, and the last CPU I had die was an AMD K7 XP2400.

If they last 5 years, thats all I care about.

 

[podspi]

I have sort of done that. The problem is that they become obsolete in 3-5 years, and the last CPU I had die was an AMD K7 XP2400.

If they last 5 years, thats all I care about.

I think my old T-bird died on me. I say think because when it stopped working, I had enough extra parts to test everything but the CPU and motherboard. IIRC, it ran ~60C+ constantly. At that time in my life I had no idea what I was doing, so it is certainly possible that was my fault...