Can you damage a Sandy Bridge by overclocking?

[BonzaiDuck]

I leave adaptive thermal monitor enabled. However my board has something like this. I can adjust my desired TDP limit.

It is an extremely good idea to keep a notebook with meticulously written, consistent entries when over-clocking. I over-clocked this particular bad-boy SB just short of a year ago, and I remember setting those features shown in your screenshot. But I think I set them after I'd pretty much obtained stable settings close to the ones I use today. Like you, I re-enabled the adaptive thermal monitoring after I was finished tweaking [Whoops! That's also methamphetamine slang.]

Since I didn't make note of that particular item, I have to depend on my memory (I apply for Medicare in July). Today, I had trouble remembering the name of George Harrison -- one of the Beatles. While that's intended to be funny, it's true, and it does indeed bother me. At least I didn't forget his face. . . . darn it.. . . Perhaps I need more Vitamin B12 for breakfast . . .

But apologizing for my needlessly verbose posts -- and getting to the point . . . I distinctly remember making those adjustments you mention only after watching the HWMonitor software's wattage reports (pertaining to TDP.)

I'm still not entirely sure which sources to believe. ShintaiDK has only given his best understanding of this -- although I detect a tone of impatience. But if the "Thermal Control Circuit" is not disabled, and this "Adaptive Thermal Monitoring" feature on the ASUS motherboards works in reference to TDP or processor wattages exceeding it . . . . well . . . . as I said . . . I'm not sure . . .

Is your post an indication of your consensus with me about this?

 

[cmdrdredd]

The thermal monitor as I understand it will enable throttling. Disable means no throttle. the only reason to disable it IMO is when you are on LN2 and are getting throttle at negative temps. The tdp limit is entirely separate as far as I can tell. It just Leta the board provide more power if the CPU asks for it. 140% is the default for the Asus ROG boards. It was set that way after a fresh bios flash and cmos reset.

 

[BonzaiDuck]

The thermal monitor as I understand it will enable throttling. Disable means no throttle. the only reason to disable it IMO is when you are on LN2 and are getting throttle at negative temps. The tdp limit is entirely separate as far as I can tell. It just Leta the board provide more power if the CPU asks for it. 140% is the default for the Asus ROG boards. It was set that way after a fresh bios flash and cmos reset.

The ASUS mobo "overclocking guide" says " . . doesn't disable TCC." I was under the impression that the TCC reacts to temperature, as opposed to TDP wattage. And the "guide" seems to conflict in meaning with the view of the short quote (likely from Intel docs and white papers) provided by ShintaiDK.

That mobo OC guide on the other hand says that the BIOS item "Adaptive Thermal Monitoring" reacts to TDP wattages.

This would seem to be a minor issue -- as they say "picking fly-s*** out of pepper" -- but in the 2009 forum link characterization I posted earlier, it's still part of the "haze of confusion."

On the "fly-s*** out of pepper" perspective, as everyone says here, unless you're pushing the processor to the TjMax limit, we probably shouldn't dwell on it, although I think it is unresolved unless we can prove that either the ASUS guide is inaccurate, or Shin's quote leaves something out.

But the ASUS guide clearly states to disable ATM while over-clocking. It doesn't say something like " . . . because you're liable to reach TjMax." And it speaks to throttling back or some sort of clock-cycle management based on the wattage and TDP -- not the temperatures.

I definitely remember noting that one core was reducing its frequency at ~73 to 75C with ATM ENABLED -- you could see the frequency read-out on my monitoring software during the stress-tests. That just happened to be the temperature: I could see that the wattages were exceeding the default TDP (in the features you mention and which I had yet to adjust). When I disabled the ATM feature, all cores were going full throttle under the same stress-tests. But this assertion may be received like the different witnesses in the Trayvon shooting. You can choose to believe what I say I saw, or say I was mistaken. And it would be silly for me to "go ballistic" in frustration about it.

 

[taltamir]

Short answer is sure. Long answer is unlikely, if you keep the voltage at auto and use good cpu cooling.

BIG WARNING!
Most mobos will horrendously over volt your CPU if you keep the voltage at auto while overclocking!

You MUST change it to manual, and manually set the voltage to your default (found via VID).

 

[cmdrdredd]

The ASUS mobo "overclocking guide" says " . . doesn't disable TCC." I was under the impression that the TCC reacts to temperature, as opposed to TDP wattage. And the "guide" seems to conflict in meaning with the view of the short quote (likely from Intel docs and white papers) provided by ShintaiDK.

That mobo OC guide on the other hand says that the BIOS item "Adaptive Thermal Monitoring" reacts to TDP wattages.

This would seem to be a minor issue -- as they say "picking fly-s*** out of pepper" -- but in the 2009 forum link characterization I posted earlier, it's still part of the "haze of confusion."

On the "fly-s*** out of pepper" perspective, as everyone says here, unless you're pushing the processor to the TjMax limit, we probably shouldn't dwell on it, although I think it is unresolved unless we can prove that either the ASUS guide is inaccurate, or Shin's quote leaves something out.

But the ASUS guide clearly states to disable ATM while over-clocking. It doesn't say something like " . . . because you're liable to reach TjMax." And it speaks to throttling back or some sort of clock-cycle management based on the wattage and TDP -- not the temperatures.

I definitely remember noting that one core was reducing its frequency at ~73 to 75C with ATM ENABLED -- you could see the frequency read-out on my monitoring software during the stress-tests. That just happened to be the temperature: I could see that the wattages were exceeding the default TDP (in the features you mention and which I had yet to adjust). When I disabled the ATM feature, all cores were going full throttle under the same stress-tests. But this assertion may be received like the different witnesses in the Trayvon shooting. You can choose to believe what I say I saw, or say I was mistaken. And it would be silly for me to "go ballistic" in frustration about it.

What did you use to monitor the individual core frequencies?

The problem is the manual for the motherboard is translated from Tiawanese or Chinese...something. Then the forums sometimes have users and reps from Asus who speak english as a second language. What they are trying to explain to us doesn't always come out as they intend.

 

[Don Karnage]

BIG WARNING!
Most mobos will horrendously over volt your CPU if you keep the voltage at auto while overclocking!

You MUST change it to manual, and manually set the voltage to your default (found via VID).

Every ROG board i've bought didn't overvolt by a lot when left on auto.

 

[cmdrdredd]

Every ROG board i've bought didn't overvolt by a lot when left on auto.

Good BIOS programming

 

[BonzaiDuck]

While I appreciate the advice on OC'ing, we created a long thread last year through end June.

There are three voltage adjustments on the Z68 P8Z68-V-[Pro/Deluxe/etc.] BIOS. I also looked at BIOS screens others had posted for the ROG Z68 boards -- they are hardly any different at all except for being "red" instead of "green."

First, the VCORE setting which also comes with a real-time report of VCORE. Second, the Voltage Offset. Finally, there is a third setting on a different BIOS page under "power management" for "additional voltage added under turbo-mode."

I did some considerable reading on SB over-clocking with the P67 and Z68 motherboards. I started by simply letting the board's "built-in" or "AI" over-clocking feature work to find an initial setting -- I think it gave me a 4.4 or 4.5 Ghz setting, and as one of our colleagues here noted, it seemed that the voltage was most certainly excessive. But I was already inclined to think that such is what it would do.

Then I did some testing to see how the board at default would raise voltage in turbo mode (going to 3.8), and collected some other data giving an indication of what "offset" would do.

All the prior advice on these boards and the SB, and the shared experience among those of us posting on that thread I mentioned -- suggested to leave the VCORE set to "Auto" while fixing the "Offset" -- then adjusting the offset either way until no BSOD or terminated Prime95 threads. Beyond that, however, I also discovered what the "Auto" setting on "Extra voltage on turbo" was giving, so I also fixed that to adjusted values.

So the board's own "AI" overclocking was pushing VCORE to 1.4V -- which I knew was excessive and I didn't like. The approach we all used allowed us to find the minimum settings for Offset -- and for me, the minimum for "Extra" in the third setting.

I got my 4.65Ghz setting to give a fully loaded value of somewhere around 1.30 to 1.32V. You can see what happens at the point where the stress-test removes the load, and the monitoring software shows it losing the VDroop and returning to about 1.36 to 1.37V before the speed drops back to the EIST value.

I could see where some people -- based on the old OC'ing wisdom to fix the VCORE at a particular value -- would do so with these boards. But for that one, single VCORE setting, we all agreed that leaving it on Auto and adjusting the other two items was all that was necessary.

I think there were some four (maybe five?) different LLC settings. To achieve absolute stability, I only moved it up to the first LLC notch.

I took about 20 pages of notes on legal copy/printer paper while I went through this exercise, and I could show them. But I doubt you'd want to review them.

Best computer I ever had, though. Not a glitch or a problem all year since I finished the tweaking. I'm using a 60GB Pyro drive to ISRT cache or "accelerate" a VelociRaptor; I've implemented the Lucid Virtu option. The worst that has happened is that there were glitches with the Lucid software -- and updating that software and cleaning up the registry with something like CCleaner was all that was ever needed. The SSD caching has been stellar. I could update the BIOS from the third or fourth revision I've got installed now; update the IRST Intel software -- but I don't want to mess up a good thing.

 

[BonzaiDuck]

What did you use to monitor the individual core frequencies?

The problem is the manual for the motherboard is translated from Tiawanese or Chinese...something. Then the forums sometimes have users and reps from Asus who speak english as a second language. What they are trying to explain to us doesn't always come out as they intend.

You might poo-pooh me on this, because the bundled "AI" software from ASUS has always been given a bad rap. or "bad rep." But it allows you to see all four cores simultaneously, as I recall. I'm pretty sure that's what I used. And if it doesn't allow seeing them simultaneously, you can flip through and view individually. Of course, one also has the graphic display with Task Manager "Performance," but I'm sure it was the ASUS monitoring. My memory is spotty on this, but I'm sure that's what I used.

I understand the "second-language" problem. I'm not sure we have fully resolved to an absolute certainty that business about Adaptive TM and the TCC. For me, it's not critical to anything because the TJunction sensor temperatures don't much go over 70+C even under extreme IBT loads.

 

[cmdrdredd]

Ahh...yeah i was gonna install the ai software for fan xpert then i decided to just set turbo in bios and leave it lol.

 

[BonzaiDuck]

Ahh...yeah i was gonna install the ai software for fan xpert then i decided to just set turbo in bios and leave it lol.

I raised the AI Suite II software to check the layout of the frequency monitor and refresh my memory. It shows all four cores as little bar graphs, but the readout at the top shows the numerical frequency for Core 0, and there is a button that switches to Core 1, Core 2, etc. That is apparently the way I was able to see the throttling with my mistake of leaving the Adaptive Thermal Monitoring feature Enabled during the over-clock stress-testing.

All in all, the monitor and recording features of that software seem pretty robust. You want to be careful with the overclocking features -- I simply chose not to use them. The thing will just keep going and going -- trying to get a higher setting until the system crashes and reboots. That's my memory of it.

 

[pantsaregood]

...as fot your 2500K, I don't believe a "safe" voltage has been established yet. Intel used to publish nominal VID values for their CPUs, but those based on the Sandy Bridge architecture just list every possible VID, which makes it fairly useless information.

I can't see 1.45v instantly blowing up a CPU, though. Ivy Bridge may be at a dangerous level there, but I doubt Sandy Bridge would be.

 

[cmdrdredd]

Just to test...I installed AI suite and monitored the cores last night. I ran IBT and never saw it go below 4599Mhz(accounting for variance). I leave thermal monitor on of course and temps were in the low to mid 80s and once saw 89c. Never throttled that I could see.

 

[BonzaiDuck]

...as fot your 2500K, I don't believe a "safe" voltage has been established yet. Intel used to publish nominal VID values for their CPUs, but those based on the Sandy Bridge architecture just list every possible VID, which makes it fairly useless information.

I can't see 1.45v instantly blowing up a CPU, though. Ivy Bridge may be at a dangerous level there, but I doubt Sandy Bridge would be.

We had made a fairly thorough discussion on this topic a year ago -- of course, they are old threads buried in the pile. Maybe even in this thread I'd remarked about a forum veteran named AigoMorla whose considerable experience with many machines (and many at once) -- together with his share of dying or dead processors -- allowed him to post a table of expected lifespans based on voltage overages.

We had made some conclusions about these SB CPU's based on the earlier generation I7/I5 models, since I think they were designed to a lithography spec of 45nm, while the socket-1155 Sandy Bridge is 32nm. It should be the case -- that the density of the core would be the most significant factor in an educated guess about a "safe maximum." "Upper safe limit" has variously been proposed at 1.3V to 1.35V, and I think (I THINK) AigoMorla felt sure it would be 1.35 -- with a 1 to 5% overage being insignificant.

But the point of it -- with "Turbo" overclocking, EIST and other features turned on, the processor will spend a lot more time at lower voltages than at voltages required for load conditions. Further, the highest voltages you will see will occur when the processor speed is still at the top of "Turbo" in Mhz, but the processor is unloaded. Therefore, amperage and wattage will be lower than under load. And those peak values will be fairly momentary.

Over-clocking has its risks. With less information coming from Intel, it may make those risks greater for not being able to follow their own assurances of "safe." "Safe" used to be precisely posted on their spec sheet, or "officially" defined as "safe" by Intel. But with Turbo Over-clocking those risks are also diminished somewhat. Probably more than somewhat.

I'm sure others may want to "throw in" about this, disagree or suggest corrections.

 

[BonzaiDuck]

Just to test...I installed AI suite and monitored the cores last night. I ran IBT and never saw it go below 4599Mhz(accounting for variance). I leave thermal monitor on of course and temps were in the low to mid 80s and once saw 89c. Never throttled that I could see.

Well -- it could be possible I was imagining what I saw as what I thought I saw. But I thought I saw one core drop back under the conditions I mentioned. I'm fairly sure of it. And again -- I just don't think we've resolved whether a feature responding to wattage or which involves clock-cycle management due to wattage above spec TDP (without adjustment of the limiting features you discussed) -- only occurs at TjMax.

Didn't you keep those wattage adjustments you posted in your BIOS screenies? What if you set them to default? You weren't clear about that factor in "what you did last night."

Like I said -- betwen "english second language," "Over-clocking guide versus Intel documentation," my observations of a year ago at only a year younger than my 64+ -- I don't know what to think. I'm pretty sure that with the wattage adjustments (like your screenies), my choice of a maximum loaded/unloaded voltage, etc. that what I "witnessed" was eliminated and I re-ENABLED ATM. So unless I start tweaking this year-old SB system again, it's not personally a pressing issue. It would still be nice to "know" though.

UPDATE: Oh, gee. you have the Ivy 3570K, socket 1155. I wondered when those would appear on the market. See -- I'm not keeping up. I had even wondered last year if I might be able to drop one of those into a Z68 board. Looking at yours, -- 4 cores, 3.4Ghz to "Turbo" and your clock speed, it wouldn't make sense. I better look and see if there's an I7 version of that CPU.

My excuse is that I'm getting old and [blank]ed up . . . . A great excuse. "Pardon me, sonny!"

 

[cmdrdredd]

Well -- it could be possible I was imagining what I saw as what I thought I saw. But I thought I saw one core drop back under the conditions I mentioned. I'm fairly sure of it. And again -- I just don't think we've resolved whether a feature responding to wattage or which involves clock-cycle management due to wattage above spec TDP (without adjustment of the limiting features you discussed) -- only occurs at TjMax.

Didn't you keep those wattage adjustments you posted in your BIOS screenies? What if you set them to default? You weren't clear about that factor in "what you did last night."

Like I said -- betwen "english second language," "Over-clocking guide versus Intel documentation," my observations of a year ago at only a year younger than my 64+ -- I don't know what to think. I'm pretty sure that with the wattage adjustments (like your screenies), my choice of a maximum loaded/unloaded voltage, etc. that what I "witnessed" was eliminated and I re-ENABLED ATM. So unless I start tweaking this year-old SB system again, it's not personally a pressing issue. It would still be nice to "know" though.

UPDATE: Oh, gee. you have the Ivy 3570K, socket 1155. I wondered when those would appear on the market. See -- I'm not keeping up. I had even wondered last year if I might be able to drop one of those into a Z68 board. Looking at yours, -- 4 cores, 3.4Ghz to "Turbo" and your clock speed, it wouldn't make sense. I better look and see if there's an I7 version of that CPU.

My excuse is that I'm getting old and [blank]ed up . . . . A great excuse. "Pardon me, sonny!"

Those settings are the defaults. I never touched anything except CPU PLL voltage set to 1.7125, +.13v offset, LLC high, memory timings 10-10-10-28 CR1 @ DDR3-2133 and 1.475v.

Everything else related to voltage or TDP allowance is how it was set from a fresh BIOS flash and CMOS reset.

Ivy will work in many Z68 boards with a new BIOS. The 3570k is 3.4Ghz with 3.8Ghz Turbo but I have it at 4.6Ghz. The i7 is the 3770k and from what I read online, it overclocks further as if Intel has selected the better performing chips for the i7 line. They run a little hotter but they can take more than a Sandy Bridge chip can. Generally speaking if you take an Ivy Bridge and clock it to 4.5Ghz it takes a Sandy Bridge CPU at 4.7Ghz or so to match it's performance so it's a little more efficient.

 

[BonzaiDuck]

Those settings are the defaults. I never touched anything except CPU PLL voltage set to 1.7125, +.13v offset, LLC high, memory timings 10-10-10-28 CR1 @ DDR3-2133 and 1.475v.

Everything else related to voltage or TDP allowance is how it was set from a fresh BIOS flash and CMOS reset.

Ivy will work in many Z68 boards with a new BIOS. The 3570k is 3.4Ghz with 3.8Ghz Turbo but I have it at 4.6Ghz. The i7 is the 3770k and from what I read online, it overclocks further as if Intel has selected the better performing chips for the i7 line. They run a little hotter but they can take more than a Sandy Bridge chip can. Generally speaking if you take an Ivy Bridge and clock it to 4.5Ghz it takes a Sandy Bridge CPU at 4.7Ghz or so to match it's performance so it's a little more efficient.

On the first topic, I can see that we're comparing a regular Navel Orange to a Cara-Cara Orange.

It would make sense -- the edge in efficiency. We had assumed last year that there would need to be a BIOS revision. However, this was the P8Z68-V-Pro board released in May 2011. They later released a P8Z68 "Gen 3" or something with that designation.

I'd think, if I wanted to do that at this point, I'd be inclined to replace the motherboard with a Z77 or whatever the successor chipset is. Looking at the over-clock possibilities, the benefits just don't outweigh the costs. Or in corollary, the uncertainty -- and the cost of the risk with the old board -- is too great.

I think, even so, that I could probably move the hard-disk setup and OS directly to such a successor motherboard, get it to successfully boot, and install the new chipset drivers. But that, too, is uncertain. I had confirmed with different nVidia chipsets some time ago -- and advised by their tech support how it would work -- that one could do that. But without further information, it is also a risk and probably too much trouble at the point of the hardware conversion if it doesn't work.

"I have to weigh the costs and the benefits." [a certain leader named O]

 

[taltamir]

Every ROG board i've bought didn't overvolt by a lot when left on auto.

I said MOST mobos not ALL mobos. ROG is the exception.

 

[cmdrdredd]

On the first topic, I can see that we're comparing a regular Navel Orange to a Cara-Cara Orange.

It would make sense -- the edge in efficiency. We had assumed last year that there would need to be a BIOS revision. However, this was the P8Z68-V-Pro board released in May 2011. They later released a P8Z68 "Gen 3" or something with that designation.

I'd think, if I wanted to do that at this point, I'd be inclined to replace the motherboard with a Z77 or whatever the successor chipset is. Looking at the over-clock possibilities, the benefits just don't outweigh the costs. Or in corollary, the uncertainty -- and the cost of the risk with the old board -- is too great.

I think, even so, that I could probably move the hard-disk setup and OS directly to such a successor motherboard, get it to successfully boot, and install the new chipset drivers. But that, too, is uncertain. I had confirmed with different nVidia chipsets some time ago -- and advised by their tech support how it would work -- that one could do that. But without further information, it is also a risk and probably too much trouble at the point of the hardware conversion if it doesn't work.

"I have to weigh the costs and the benefits." [a certain leader named O]

if I remember reading this right, but I am not 100%. The Gen3 designation was to differentiate the board from those that did not support PCIe 3.0 with the Ivy Bridge CPU.