Actually, I'd say the danger of electromigration is greatest. The risk does increase with heat, but it increases more so with voltage (which also happens to increase heat greatly as well). Pump the voltage to high, and it's a when, not if, the electrons will break down the insulated pathways and form their own.
-
We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.
Overclocking is the Dumbest thing you can do!
Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
- Status
Actually, I'd say the danger of electromigration is greatest. The risk does increase with heat, but it increases more so with voltage (which also happens to increase heat greatly as well). Pump the voltage to high, and it's a when, not if, the electrons will break down the insulated pathways and form their own.
ultimahwhat
Member
overclocking to me is like eating dairy, despite the fact that i'm lactose intolerant. I know there's a good chance for adverse effects (e.g. diarrhea and gas, or in the case of the cpu, shorter lifespan and spontaneous combustion), but i don't care because it just tastes/feels so good...
Not being argumentative but such a test does exist...how do you think AMD and Intel determine the maximum safe clockspeed for any given chip when they bin them?
But for us enthusiasts who want everything for free, including that extra MHz we get from overclocking, it is not really practical to buy a $250k tester and hire the necessary Intel/AMD test program engineers to get the job done 😀
So we could certainly say is it possible, just not practical. 😛
And even then, sometimes the big boys get it wrong too...remember the 1.13GHz P3?
BonzaiDuck
Lifer
Somewhere here and several times through '07 and before March, '08, I'd posted some guesses about the chip-making process, their quality-control, binning and testing. And I ventured to speculate about the economic imperatives for producing a product-differentiated line of processors. I posed an implicit question: "How do they choose the retail-box maximum voltage?" and supplied a speculative conclusion that failure rates over a range of voltages, and failure rates over extended time, were exponential probability distributions. I said that the accounting-side or marketing-side of the house would influence a choice to pick those specs as an assurance of near-0% probability of RMA warranty returns within the 3-year warranty period.
A March, 2008 Anandtech article pretty much validates my guesswork and deductions.
That being said, their choice to spec a CPU as "1,333 Mhz FSB" would probably follow the same logic. They're capturing certainties about existing technology, and as motherboard makers, together with their competitors ASUS, MSI etc. and the chipset designers, they've simply spec'd to a standard which is, again, based on finding a line that defines a "near-0% probability."
They don't HAVE to say "1,333 FSB" or "1,066 FSB." So there would also be a range of operability around those choices as well.
I won't go into detail about the "serious stuff" I'm doing with a mildly over-clocked Q6600 -- only to say that the data files are stored on another stock-clocked networked system with a RAID5 array and two levels of firewalls and other security features. But I have broken -- slightly -- the rule that I would insist be followed if you were doing something like real-time patient-monitoring in a hospital, or controlling the Diablo Canyon or San Onofre nuclear power plants.
Personally, I think if you're OC'ing a processor in a family of processors for which the flagship model is spec'd to run at your own choice of an over-clock setting, your worries are speculative and probably exaggerated. Your worries about losing the support of the manufacturer or losing some court-case because something went wrong -- are worries about dead certainty. Besides, most of us aren't using ECC memory, and there's the occasional cosmic ray that might change a bit-setting.
I actually think I "know" Person1 with his EEE degree. He thinks his mother is trying to poison him; he bought his house for $85,000, and has now run up his mortgage debt by refinancing it to a 2006 $350,000 market-value, which is now, in turn, in the toilet. He whines that Vietnamese immigrants in the electronics industry have underpriced his labor and taken away his livelihood. And several of his friends have cut him off, because "That boy has problems! He may even be . . . . dangerous!!"
Nope. I'll say this, though. My favorite Anthony Hopkins movie is "The World's Fastest Indian."
[Now, to push that CPU_FSB up another 5 Mhz . . . . . . . Hey! get your hands off the shutdown switch to my reactor! I'M running things here!! . . . . ]
A March, 2008 Anandtech article pretty much validates my guesswork and deductions.
That being said, their choice to spec a CPU as "1,333 Mhz FSB" would probably follow the same logic. They're capturing certainties about existing technology, and as motherboard makers, together with their competitors ASUS, MSI etc. and the chipset designers, they've simply spec'd to a standard which is, again, based on finding a line that defines a "near-0% probability."
They don't HAVE to say "1,333 FSB" or "1,066 FSB." So there would also be a range of operability around those choices as well.
I won't go into detail about the "serious stuff" I'm doing with a mildly over-clocked Q6600 -- only to say that the data files are stored on another stock-clocked networked system with a RAID5 array and two levels of firewalls and other security features. But I have broken -- slightly -- the rule that I would insist be followed if you were doing something like real-time patient-monitoring in a hospital, or controlling the Diablo Canyon or San Onofre nuclear power plants.
Personally, I think if you're OC'ing a processor in a family of processors for which the flagship model is spec'd to run at your own choice of an over-clock setting, your worries are speculative and probably exaggerated. Your worries about losing the support of the manufacturer or losing some court-case because something went wrong -- are worries about dead certainty. Besides, most of us aren't using ECC memory, and there's the occasional cosmic ray that might change a bit-setting.
I actually think I "know" Person1 with his EEE degree. He thinks his mother is trying to poison him; he bought his house for $85,000, and has now run up his mortgage debt by refinancing it to a 2006 $350,000 market-value, which is now, in turn, in the toilet. He whines that Vietnamese immigrants in the electronics industry have underpriced his labor and taken away his livelihood. And several of his friends have cut him off, because "That boy has problems! He may even be . . . . dangerous!!"
Nope. I'll say this, though. My favorite Anthony Hopkins movie is "The World's Fastest Indian."
[Now, to push that CPU_FSB up another 5 Mhz . . . . . . . Hey! get your hands off the shutdown switch to my reactor! I'M running things here!! . . . . ]
Rubycon
Madame President
Of course it exists but not for the end user hence my point.
The 1.13GHz fiasco was clearly Blue vs. Green because Intel did not want to be dethroned (well they were in IPC but like megapixels to cameras the MHz race was clearly ON with the consumer!) My Athlon Slot-A 850 at 1.06GHz (SMD resistor modded!) tromped all over my coppermine 700 at 933. Of course I had to deal with the quirks of VIA (Abit KA7) but that's nothing new. 😉
This is also what essentially killed the pretested chip biz with enthusiasts.
It would be VERY nice to have a "CHIP TESTER" in drugstores. (for those of you old enough to remember tube testers!)
A little OT - how do those memory testers work so fast? It takes hours with Memtest yet you can slap a DIMM in a bench tester and push a button and get a result in seconds? Or is it about as useful (not) as a motherboard BIOS self test?
BonzaiDuck
Lifer
I can't answer RubyCon's last question, but I need to play "I got dis college-degree . . . . and I know sumpin" game.
Per my remarks above per how the company chooses its spec limits, how it manufactures their product(s).
Bear in mind that, while not a monopolist, Intel fits into either a duopoly model with AMD, or a dominant-firm description. In any of those "imperfect competition" models, there is plenty of room for "price discrimination."
Let me explain.
General Practitioner MD's used to be cited as a benevolent example of price discrimination. They could give a wealthy citizen a gall-bladder operation for $1,000, then turn around and charge the poor widow with six kids $100 for the same thing.
Back in the '90s, the discovery of "counterfeit" Pentium II CPUs which had been modified in the Phillipines turned up indications about how the Intel production process works. Apparently Intel was making a single specification of the Pentium II. In the marketplace, there were some four flavors: 266 Mhz, 300 Mhz, 333 Mhz, and 400 Mhz. The price-scale followed the processor-speed. Apparently, they were manufacturing one processor to a single spec: the 400 Mhz item, and at the end of the assembly line, they'd disable some quantity of product to run at the successively lower speeds. This then provided an opportunity to counterfeiters, who bought up the 266- or 300 Mhz units, and re-enabled them to run at higher speeds. The Intel Product Distributor network had a minimum price of $525 for their 333 Mhz item in 1998, but the counterfeiter would buy up the low-end item and then sell it for just below $500 after performing the alteration.
We do not know precisely what they are doing today, but it is not just the QC and test-results that figure in the binning of these processors. Intel has some idea of how many processors of each model in a product line that it wants to sell -- in order to maximize sales revenue.
So "engineering?" Yes. Technical limitations? Yes. Statistical probability distributions for failure over voltage or speed ranges and time? Yes. Marketing strategies, product-differentiation and cost reduction in production? Yes -- and don't kid yourself.
Per my remarks above per how the company chooses its spec limits, how it manufactures their product(s).
Bear in mind that, while not a monopolist, Intel fits into either a duopoly model with AMD, or a dominant-firm description. In any of those "imperfect competition" models, there is plenty of room for "price discrimination."
Let me explain.
General Practitioner MD's used to be cited as a benevolent example of price discrimination. They could give a wealthy citizen a gall-bladder operation for $1,000, then turn around and charge the poor widow with six kids $100 for the same thing.
Back in the '90s, the discovery of "counterfeit" Pentium II CPUs which had been modified in the Phillipines turned up indications about how the Intel production process works. Apparently Intel was making a single specification of the Pentium II. In the marketplace, there were some four flavors: 266 Mhz, 300 Mhz, 333 Mhz, and 400 Mhz. The price-scale followed the processor-speed. Apparently, they were manufacturing one processor to a single spec: the 400 Mhz item, and at the end of the assembly line, they'd disable some quantity of product to run at the successively lower speeds. This then provided an opportunity to counterfeiters, who bought up the 266- or 300 Mhz units, and re-enabled them to run at higher speeds. The Intel Product Distributor network had a minimum price of $525 for their 333 Mhz item in 1998, but the counterfeiter would buy up the low-end item and then sell it for just below $500 after performing the alteration.
We do not know precisely what they are doing today, but it is not just the QC and test-results that figure in the binning of these processors. Intel has some idea of how many processors of each model in a product line that it wants to sell -- in order to maximize sales revenue.
So "engineering?" Yes. Technical limitations? Yes. Statistical probability distributions for failure over voltage or speed ranges and time? Yes. Marketing strategies, product-differentiation and cost reduction in production? Yes -- and don't kid yourself.
Goldfish4209
Member
OMGZ, my 6750 @ 3.2Ghz will only last 5 upgrade cycles now! Run and hide!
BonzaiDuck
Lifer
Oh, yeah, Goldfish!! In all my graduate wisdom, I forgot about planned obsolescence versus market-mandated and warrantied product life-span.
"Gotta keep those enthusiasts buying new CPUs. Gotta pay for Conroe and Penryn R&D costs with sales of existing inventory, but pretty soon -- market-demand saturation, loss of revenue, no stock-dividends, etc. Well, maybe we can introduce Nehalem after the accountants tell us that all C2D sales from here on are almost pure gravy. OEMs are shipping to the DOD and those big defense contractors now. After a few more purchase orders like those trickling through the distribution chain back to Intel, we'll let rip with Nehalem and see how those enthusiasts jump on the ol' market-demand bandwagon!"
BonzaiDuck
Lifer
Not sure, but "Person1" came from another forum.
Of course, I was half-joking to say that I knew Person1, but his approach to things fit the profile of my friend who thinks his mother is poisoning him (he's 60 years old), and that the Vietnamese immigrants are gonna "tekk our jobs!! Tekk our jobs!!" [quoting from South Park on the episodes about those "eeemigrants from the fewww-tshurr . . . comin' through that time portal . . . ."]
My own real . . . ah . . . friend, whom I've described, worked in a big defense industry company before the Cold War's end in a QC department, testing electronic parts made for the Navy. He felt that the company wasn't being straight with the Navy -- selling parts that didn't measure up to DOD spec. So he threatened to sing like a canary to the government. In his story, he told of being hit by a car in the company parking-lot -- suggesting that "they" were trying to kill him. Maybe it's true -- I don't know. But . . . . that boy had problems -- real problems . . . :laugh:
One man's education is another man's propaganda. [Sorry, ladies -- I mean "person."]
But I'm occasionally annoyed by the lop-sided perceptions of people so specialized in one field that their understanding of other factors is either myopic, naive, or both.
Lemme get off my soap-box here. I've been humming the "Battle Hymn of the Republic" since the announcement Saturday . . . . I'm biden my time and waiting for the November showdown. [OFF the Soapbox!! OFF the Soapbox, you Idiot!! [Slap, slap.] [I slapped myself. Mea Culpa . . . Off topic and off forum . . . . Problems? What problems? Who said that? . . . . I-I-I don't talk to myself!!! No, Thur!! ]]
Overclocking is great fun and certainly can save you some bucks if you absolutely "need" the fastest CPU you can get...
On the flip side, if your livelihood depends on that system being perfect all the time, you wouldn't overclock it.
Obviously OCing usually does lead to a reduced lifespan on the CPU, but it doesn't always lead to it...besides, most overclockers sell or throw away their CPU well before that happens anyway. 🙂
JMHO
On the flip side, if your livelihood depends on that system being perfect all the time, you wouldn't overclock it.
Obviously OCing usually does lead to a reduced lifespan on the CPU, but it doesn't always lead to it...besides, most overclockers sell or throw away their CPU well before that happens anyway. 🙂
JMHO
ch33kym0use
Senior member
overclocking is something that I wouldn't do at work although at home it's ok
BonzaiDuck
Lifer
My impression was that Rubycon is hinting about people who have "too much time on their hands" . . . . . ah . . . . . who might I think that could be? 😉
imported_Champ
Golden Member
I have blown out 2 mobos in my life...before i started overclocking
if your comp explodes and your under the the max vcore then it would have a 99% chance of blowing up anyways
if your comp explodes and your under the the max vcore then it would have a 99% chance of blowing up anyways
BonzaiDuck
Lifer
Right -- and the culprits in the board-maker's quality-control division shouldn't have shipped it in the first place.
So far, the only mishaps I've had involved two motherboards of the same model -- the ASUS P4P800 and P4P800SE (a later revision).
I'd given the P4P800 to my brother, socketed with OCZ DDR-500 EL Gold RAM. It was over-clocked to DDR-480. Bro' didn't want to pay for a UPS battery-backup, and had a surge-protector plugged into a faulty household wiring system. When the washing machine kicked on, his system would hang. After we massaged our way through a couple of these crises and brought the system back to life, it lasted another six or eight months and just crapped out. That's how I gave up my precious "MOJO" system and started my C2D adventures -- Bro' got da MOJO -- a P4P800SE.
That SE board had replaced another SE board which I, also, had over-clocked to the full DDR-500 spec. It was, for about six months, a dream machine. Then suddenly the system would lock up -- or the screen would "gum up" and become unresponsive. That system had been equipped with the OCZ DDR-Booster, and I suspect the Booster had something to do with it even though I hadn't over-volted the RAM beyond warranty spec. It would continue to boot, and I ran MEMTEST86+ against it to discover what seemed like thousands of memory errors. The first kneejerk reaction was to RMA the RAM modules, and OCZ replaced the EL Golds with Platinum XTC DDR-500's. When I put those in the system and ran MEMTEST86+, I STILL got thousands of errors, and concluded that the memory-controller had crapped out on the board.
The generation of motherboards that hit the market in 2005 and 2006 included high-end units like my Striker Extreme and the ASUS P5W-- whatever it was, but Intel chipset and some PCI-X slots along with PCI and PCI_E. Those boards implemented improvement in component design, solid-state Mosfets or capacitors, and 8-cycle voltage regulation. I think the industry has "gotten wise," and is producing boards that are designed to run beyond mere processor FSB spec.
The P4P800's were good over-clockers, but we shortened their life, I think . . . . But I don't think I ever "lost" a processor to over-clocking.
Think massive tester farms combined with aggressive test conditions to force walking wounded chips to fatal outcomes in a hurry.
Kinda like how Intel's burn-test can find instability in your chip in a matter of minutes while Prime95 takes hours to uncover the same errror.
They also build-in a fair amount of self-diagnostic hardware onto those dimms that the testers are designed to access but our software programs do not.
BonzaiDuck
Lifer
Well, I haven't "shined-on" pushing my over-clock farther. I was at 3.7 Ghz with the E8400, and was able to get it to move from a 2-hour 40 minute failure under PRIME95 to a 6 hour 50 minute failure by bumping the VCORE one notch and the 1.2V_HT up from 1.3 to 1.35V (read by sensors as 1.36V).
This indicates to me, given my stubborn refusal to push CPU_VTT above 1.45V, that each successive increase may now require two notches up the VCORE ladder -- a hint of a "parabolic" or exponential slope. My personal "red-line" for VCORE was 1.35V, and it is set to 1.32125V now. It was rock-stable 15 hours PRIME no errs or warns at 3.645 GHZ, and I'm using those settings to drop it back to 3.60 and tighten my Corsair DHX latencies.
I had made some remarks about a possible -- even likely- scenario, where Intel cuts everything from the same cloth, modifying slightly to sell the differentiated product-line at a range of prices and thus capturing "more-than-normal" profits.
On the consumer end, we've discussed how many of us would rather pay $200 for a good over-clocker, than $1,000 for the flagship model. So here are my thoughts.
I remember a quote from some economist, who spoke of the two ineluctable factors in economic thinking: "Time, and Money." For many of us, Time IS Money. For some of us, we've stopped trying to value leisure-time as money lost-- having slipped over the boundary that separates the working-wounded, and the retrograde retired.
We spend days and weeks in time trying to get the $200 CPU to go faster than than the $1,000 flagship CPU.
If we think nothing of that fact, then we do this for the same reasons that people surf in sometimes risky waters. We get "juice" from it. We like it. It's a not-quite-cheap thrill.
So if the real Burt Munro as represented in the "World's Fastest Indian" movie could save up $2,000 of his pension money to just barely cover costs in his trip to Bonneville Salt Flats for running that souped-up motorcycle . . . . you geeks can do as you please.
And so will I. So will I.
If "Person1" chose to slam us for "da bidnis we're in" of over-clocking, perhaps he's frustrated by our teeny-bopper "junior" component, pockets filled with spare cash looking for new hardware thrills, who don't choose mature "red-line" voltages and speeds for their rigs.
But maybe that's just my frustration with the latest crop -- the generation-Z-'ers -- knowing full well that I felt as though I was the only one who knew anything when I, too, was between 15 and 20 years old.
You're born, you live, you get old, and you die. The old frown at the young, and the young snicker at the old.
It's the cycle of life on a generational front-side bus.
This indicates to me, given my stubborn refusal to push CPU_VTT above 1.45V, that each successive increase may now require two notches up the VCORE ladder -- a hint of a "parabolic" or exponential slope. My personal "red-line" for VCORE was 1.35V, and it is set to 1.32125V now. It was rock-stable 15 hours PRIME no errs or warns at 3.645 GHZ, and I'm using those settings to drop it back to 3.60 and tighten my Corsair DHX latencies.
I had made some remarks about a possible -- even likely- scenario, where Intel cuts everything from the same cloth, modifying slightly to sell the differentiated product-line at a range of prices and thus capturing "more-than-normal" profits.
On the consumer end, we've discussed how many of us would rather pay $200 for a good over-clocker, than $1,000 for the flagship model. So here are my thoughts.
I remember a quote from some economist, who spoke of the two ineluctable factors in economic thinking: "Time, and Money." For many of us, Time IS Money. For some of us, we've stopped trying to value leisure-time as money lost-- having slipped over the boundary that separates the working-wounded, and the retrograde retired.
We spend days and weeks in time trying to get the $200 CPU to go faster than than the $1,000 flagship CPU.
If we think nothing of that fact, then we do this for the same reasons that people surf in sometimes risky waters. We get "juice" from it. We like it. It's a not-quite-cheap thrill.
So if the real Burt Munro as represented in the "World's Fastest Indian" movie could save up $2,000 of his pension money to just barely cover costs in his trip to Bonneville Salt Flats for running that souped-up motorcycle . . . . you geeks can do as you please.
And so will I. So will I.
If "Person1" chose to slam us for "da bidnis we're in" of over-clocking, perhaps he's frustrated by our teeny-bopper "junior" component, pockets filled with spare cash looking for new hardware thrills, who don't choose mature "red-line" voltages and speeds for their rigs.
But maybe that's just my frustration with the latest crop -- the generation-Z-'ers -- knowing full well that I felt as though I was the only one who knew anything when I, too, was between 15 and 20 years old.
You're born, you live, you get old, and you die. The old frown at the young, and the young snicker at the old.
It's the cycle of life on a generational front-side bus.
Jschmuck2
Diamond Member
How to overclock your Toyota Camry XLE 3.5
Step 1: Purchase Toyota Camry XLE 3.5 (top of the line) for ~ 27k
Step 2: Acquire Lexus logo and front grill
Step 3: Remove Toyota front grill and trunk logo
Step 4: Attach Lexus front grill and trunk logo
Enjoy your new Lexus ES 350 for 8k cheaper.
Step 1: Purchase Toyota Camry XLE 3.5 (top of the line) for ~ 27k
Step 2: Acquire Lexus logo and front grill
Step 3: Remove Toyota front grill and trunk logo
Step 4: Attach Lexus front grill and trunk logo
Enjoy your new Lexus ES 350 for 8k cheaper.
ultra laser
Banned
Actually, guys, Intel and AMD want you to overclock. The reason being that if you break your chip, you'll have to buy a new one.
imported_Champ
Golden Member
drakkul is retarded...this made me laugh
hes hit the point where he knows that he has lost, now tenet you have to finish him off
ROFLMAO!!!!
You got it wrong.
Step1: Buy a Honda Civic DX
Step2: Get a Super charger and front race slicks.
Step3: Get NOS
Step4: You need the powered by Honda sticker. Remember Stickers increase overclocking.
Step5: Now you got an overclocked car.
- Status
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 24K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 24K
-
-
