Overclocking - Sandy Bridges and Voltages concerns????

[20mmrain]

I have been reading allot of posts lately on other forums stating concerns about overclocking SB and it's voltages. Most of what concerns me is the voltages on these chips and how easily they are supposed to be able to be killed.... For examples....

Max Memory voltages are not supposed to exceed 1.5 volts +5% which means max Intel voltage allowed is 1.58 volts. Which normally wouldn't worry me all that much but with seeing a few reports of this killing CPU's already I thought I would ask about it. Example do you think my XMS 3 running @ 1.65 for sustained amounts of time will be safe or should I buy something else in replacement to be safe?

Next been hearing reports of people killing off their SB chips with as low Vcore voltages of 1.28 volts and up. Has anyone here had anything like this happen to them? I thought the chip was supposed to handle up to 1.52 volts (some where in that area) safely (Per Intel)

I also see a general agreement lately over the net suggesting to overclockers not to run there SB chips over 1.38 volts for any extended time because of how these chips can be killed. Which is strange to me because I also see a tone of people running these chips way higher than that for extended periods of time with out problems. And also hitting clocks of 5.0 ghz or higher.

Also see reports of not exceeding VCCIO (VTT) voltages of 1.0815 which I always thought the rule was to keep this voltage with in .45 from DRAM. So if it is safe (Lets say) to run the DRAM voltage at 1.65 how would this be possible with that max voltage.

I am not trying to be a scary person about this but.... when reading this stuff I thought I would ask. Especially because while I do overclock.... I am also one of those overclockers that try to stay with in max allowed specs to get my overclocks. Are these rumors true or are they all fake.

please way in I would love to hear what people have been finding out about this???

Thanks

 

[brandonwh64]

This thread will help you out

http://www.techpowerup.com/forums/showthread.php?t=137670

 

[PaulieG]

I don't have time for a more complete answer right now, but here is the short version. I've had 4 2600k's. I've kept them all below 1.38v except for bench runs. None have seen any degrading. I gotta believe that running anywhere close to 1.52v on a 32nm chip will kill it at some point. Not a doubt in my mind. Intel has said that although it recommends 1.5v for power savings, that ram running at 1.65v is fine. I've run 3 different sets, all at 1.65v with no problems whatsoever. As far as absolute cpu/VICCIO voltages, I think it's really hard to tell what a relative safe max is right now, since the platform is so new.

 

[Frizz]

I wouldn't say 1.52v is the absolute max, but currently its hard to tell how long each chip will last so its not really worth the risk just yet.. My comfort zone for a 24/7 setting after reading most probably the same resources you've read online is

1.42Vcore MAX
1.650 Ram Voltage MAX
LLC either OFF or Level 1
VTT 1.2 MAX
Temps < 72 Degrees Celsius

What I would suggest to anyone with 1.65 rated RAM is to see if it can sustain its rated speeds at lower voltages, I am currently running off XMP Profile with 1.540 Ram Voltage and hasn't caused a problem for me in terms of stability.

 

[LifeOnMars]

20mmrain, where did you see reports of people killing off their chips at 1.28v? I don't believe that would be the reason they killed their chips but would be interested to see the source.

Also, I run my ram at 1866 which is above spec with less voltage (1.6v). Not one bsod, or strange issue at all. Take the time to read up on a few of the more noteworthy sites that don't tend to spread useless bullshit, they have really helped myself and alot of other people.

 

[cadaveca]

Max Memory voltages are not supposed to exceed 1.5 volts +5% which means max Intel voltage allowed is 1.58 volts. Which normally wouldn't worry me all that much but with seeing a few reports of this killing CPU's already I thought I would ask about it. Example do you think my XMS 3 running @ 1.65 for sustained amounts of time will be safe or should I buy something else in replacement to be safe?

Memory makers are releasing kits for SB @ 1.65v. I think it's fine, based on that. Going over 1.65v though...don't bother. SB really kills mem OC, so buy what speed you want, and don't mess around.

Next been hearing reports of people killing off their SB chips with as low Vcore voltages of 1.28 volts and up. Has anyone here had anything like this happen to them? I thought the chip was supposed to handle up to 1.52 volts (some where in that area) safely (Per Intel)

1.52v, as per intel, is safe, but not for full load, and not for long-term. It's labelled, as OK, but will shorten the lifespan of the cpu. Death of chips can be attributed to CURRENT, not voltage. Yes, my first build suffered catastrophic failure, and I got new CPUs.

I also see a general agreement lately over the net suggesting to overclockers not to run there SB chips over 1.38 volts for any extended time because of how these chips can be killed. Which is strange to me because I also see a tone of people running these chips way higher than that for extended periods of time with out problems. And also hitting clocks of 5.0 ghz or higher.

General consensus is to not go over 1.4v. Of course, there is a caveat...I try to not go over 1.35v, and most chips were buzzing along nicely @ 1.27v, 4.6ghz-4.9ghz. Dunno why, but 5.9ghz requires a shit-tonne of volts, and I only ever gave 1.575 to those high-multi chips. Not one died, or showed signs of degredation, but I didn't really push them far before selling.

Also see reports of not exceeding VCCIO (VTT) voltages of 1.0815 which I always thought the rule was to keep this voltage with in .45 from DRAM. So if it is safe (Lets say) to run the DRAM voltage at 1.65 how would this be possible with that max voltage.

I feel up to 1.2v is fine. I am running 1.15v for 2133mhz with 4 sticks. If something happens, you can be sure that I'll be letting everyone know. I'm running 4 sticks and "high volts" to specifically explore this.

I am not trying to be a scary person about this but.... when reading this stuff I thought I would ask. Especially because while I do overclock.... I am also one of those overclockers that try to stay with in max allowed specs to get my overclocks. Are these rumors true or are they all fake.

please way in I would love to hear what people have been finding out about this???

Thanks

Personally, i say rumour, and rumours started by those who aren't exactly 100% up to snuff on what's what. I mean, really....

I did this same thing with i7, C2D, Phenom1, 939, etc...platform gets launched, I buy many many chips, clock the crap outta them, and look for problems. Since i7....haven't really seen any deaths! I mean, it's possible to kill your chip...if you're an idiot only! Very few people killed C2D...I killed quite a few.

Intel has done a good job of breaking away the barriers of OC'ing, making it so there is no skill left in it. Heck, they even provide tools that do a pretty reliable job on OC'ing the chip for you, too. It won't max the chip out, but it'll push it far, and usually within a multi or two of the real max.

You could do 5GHZ, most chips do(I've yet to see one that didn't, but I'm sure they are out there, or stuck in the non-K bin), but even 4.5-4.8ghz is a piece of cake. I'm running either bone stock, or 4.5ghz, @ 1.24v.

 

[johnnyfiive]

In my opinion, 24/7 safe on air is all relevant to how cool you can keep the CPU. If you have the stock cooler, obviously you're progression will be limited due to temps and not voltage. During stress testing, an SB chip should stay under 80C. If you can't keep it under 80C, stop.

Speaking strictly voltage, I'd say the max period (air or even water), is about 1.45v. I personally won't run my 2500k 24/7 at anything above 1.4v, but that is my personal limit. Intel says the max is 1.52v, but that doesn't mean max "safe" voltage.

It's overclocking, it isn't warrantied, and nothing is guaranteed. Just keep that in mind and enjoy whatever result you're able to attain. Anything around 4.5GHz on these chips is insane enough already. People who literally complain about not being able to run 5GHz 24/7 seriously have a case of e-peen envy and they are dumb imo.

 

[LifeOnMars]

cadaveca - I would still love to know how you are getting those clocks with such little voltage.....care to share yet?

 

[cadaveca]

cadaveca - I would still love to know how you are getting those clocks with such little voltage.....care to share yet?

I did share...if ya missed it, you better go back to that thread.


I stand even more firmly behind my position on these chips now that I've bought, binned, and sold 10 more chips.

I OC using turbo, and merely increase current provided to the chip.


Don'y be confused though...I'm still on that same chip...becuase it reuired the least current for the clocks I wanted, and really low temps. This means I have done the opposite that I ahve in the past...I screened for LOW LEAKAGE, not HIGH LEAKAGE. Typically, higher-leakage chips scale better under cold, as the amout a chip leaks varies with temperatures, and coldbugs, in my opinion, occur when the chip is not capable of leaking enough current to remain stable. As I have no plans on going even close to 0c with this chip, I wanted the best for power consumption only, and on air.


You can really base the differences in how I am clocking my chips on board and bios only. There is no magic. NONE.

And to be honest, I pretty pissed that you guys even think that I'm doing something "special"...I am definitely NOT. I'm pissed enough to say that you guys are just doing it wrong, and also bought the wrong parts.

 

[LifeOnMars]

I did share...if ya missed it, you better go back to that thread.


I stand even more firmly behind my position on these chips now that I've bought, binned, and sold 10 more chips.

I OC using turbo, and merely increase current provided to the chip.


Don'y be confused though...I'm still on that same chip...becuase it reuired the least current for the clocks I wanted, and really low temps. This means I have done the opposite that I ahve in the past...I screened for LOW LEAKAGE, not HIGH LEAKAGE. Typically, higher-leakage chips scale better under cold, as the amout a chip leaks varies with temperatures, and coldbugs, in my opinion, occur when the chip is not capable of leaking enough current to remain stable. As I have no plans on going even close to 0c with this chip, I wanted the best for power consumption only, and on air.


You can really base the differences in how I am clocking my chips on board and bios only. There is no magic. NONE.

And to be honest, I pretty pissed that you guys even think that I'm doing something "special"...I am definitely NOT. I'm pissed enough to say that you guys are just doing it wrong, and also bought the wrong parts.

DAMMIT I shall have a looky. If I had the money to buy that many and then pick the best, I'm sure I could be as "magic" as you

 

[Zen_]

According to the Intel datasheet the maximum is 1.52v.

However, and I'm not claiming to be an electrical engineer, there have to be other factors.

For one a lot of mobos allow you to use a non-spec load line with the net result being more current at load than design.

You can also override the default limits for short term and long term TDP and core current.

The recall limited the supply of CPUs "in the field" for awhile, meaning less test cases for the effects of extreme overclocks.

Overclocking with the turbo multiplier only means less power draw at idle, which may skew the long term effects compared to regular OC method with a fixed clockrate. Just a guess.

At any rate nobody really knows what the long term safe voltage is. I'd love to see how long a SB CPU would last clocked @ 5.0 GHz, 1.52v crunching LinX 24/7.

 

[20mmrain]

In my opinion, 24/7 safe on air is all relevant to how cool you can keep the CPU. If you have the stock cooler, obviously you're progression will be limited due to temps and not voltage. During stress testing, an SB chip should stay under 80C. If you can't keep it under 80C, stop.

Speaking strictly voltage, I'd say the max period (air or even water), is about 1.45v. I personally won't run my 2500k 24/7 at anything above 1.4v, but that is my personal limit. Intel says the max is 1.52v, but that doesn't mean max "safe" voltage.

It's overclocking, it isn't warrantied, and nothing is guaranteed. Just keep that in mind and enjoy whatever result you're able to attain. Anything around 4.5GHz on these chips is insane enough already. People who literally complain about not being able to run 5GHz 24/7 seriously have a case of e-peen envy and they are dumb imo.

I agree and think you bring up a geood point. When taking into consideration theat we were braggin about hitting 4 ghz just last gen.... 4.5ghz is a hugh overclock and a very very fast chip. With that being said though... I have a few questions I am trying to get answered.... maybe I will have better luck in this thread then the other thread....

first one is....Is 1.33 volts too high for these chips on a 24.7 overclock at 4.5 Ghz.

Next question is....Is Vcore LLC okay to have enabled at 100% I have never had a board capable of differnt LLC %'s. SO I have 0%/25%/50%/75%/ and 100% which is okay to leave it at

C states on these chips seem to be a mistory leave them enabled or disabled???

 

[PaulieG]

I did share...if ya missed it, you better go back to that thread.


I stand even more firmly behind my position on these chips now that I've bought, binned, and sold 10 more chips.

I OC using turbo, and merely increase current provided to the chip.


Don'y be confused though...I'm still on that same chip...becuase it reuired the least current for the clocks I wanted, and really low temps. This means I have done the opposite that I ahve in the past...I screened for LOW LEAKAGE, not HIGH LEAKAGE. Typically, higher-leakage chips scale better under cold, as the amout a chip leaks varies with temperatures, and coldbugs, in my opinion, occur when the chip is not capable of leaking enough current to remain stable. As I have no plans on going even close to 0c with this chip, I wanted the best for power consumption only, and on air.


You can really base the differences in how I am clocking my chips on board and bios only. There is no magic. NONE.

And to be honest, I pretty pissed that you guys even think that I'm doing something "special"...I am definitely NOT. I'm pissed enough to say that you guys are just doing it wrong, and also bought the wrong parts.

Dave, I think people (including me) were confused, because not all boards have a current setting, or labeled as such. After switching out my Biostar for a Asus P8p67 Pro, I know exactly what you are talking about.

I agree and think you bring up a geood point. When taking into consideration theat we were braggin about hitting 4 ghz just last gen.... 4.5ghz is a hugh overclock and a very very fast chip. With that being said though... I have a few questions I am trying to get answered.... maybe I will have better luck in this thread then the other thread....

first one is....Is 1.33 volts too high for these chips on a 24.7 overclock at 4.5 Ghz.

Next question is....Is Vcore LLC okay to have enabled at 100% I have never had a board capable of differnt LLC %'s. SO I have 0%/25%/50%/75%/ and 100% which is okay to leave it at

C states on these chips seem to be a mistory leave them enabled or disabled???

I don't think 1.33v is too much for these chips. I think pushing up to 1.38v is too much, IMO. Vcore LLC is okay to have at 100%, as long as it isn't giving you way too much voltage under load. At first, I disabled C states as I would with with the x58 platform, thinking it would overclock better disabled. However, after testing enabled/disabled, I really found no difference with max OC. So, I now leave C states enabled.

 

[LAN_deRf_HA]

bought the wrong parts

What, did you finally say what board you were using? And even running my current up to max it's yet to make any impact on stability.

 

[cadaveca]

What, did you finally say what board you were using? And even running my current up to max it's yet to make any impact on stability.

Have you checked my reviews?



:shadedshu


If you've changed a setting pertaining to current, and it made no difference, then the setting you change didn't actually change the current. 100% fact...VoltsxAmps=Watts. Current=AMPs, and increasing that is either gonna give more heat, or allow for more clocks at lower voltage.


I said it before, the differences between how we are clocknig, is all down to the hardware.

 

[PaulieG]

Have you checked my reviews?



:shadedshu


If you've changed a setting pertaining to current, and it made no difference, then the setting you change didn't actually change the current. 100% fact...VoltsxAmps=Watts. Current=AMPs, and increasing that is either gonna give more heat, or allow for more clocks at lower voltage.


I said it before, the differences between how we are clocknig, is all down to the hardware.

The use of current adjustments certainly yields results. I was almost stable at 4.8ghz on 1.35v, passing 8 out of 10 cycles of IBT. Adjusting current to 110% with no change in vcore resulted in successfully completing all 10 cycles.

 

[cadaveca]

The use of current adjustments certainly yields results. I was almost stable at 4.8ghz on 1.35v, passing 8 out of 10 cycles of IBT. Adjusting current to 110% with no change in vcore resulted in successfully completing all 10 cycles.

Well of course, I know that already!

If it makes no difference, then either the bios is borked, the hardware doesn't allow for adjustments, or just plain-old sucks!

Until I've played with more boards, I can't make any judgements, but I do know what I will be looking for.

 

[PaulieG]

Well of course, I know that already!

If it makes no difference, then either the bios is borked, the hardware doesn't allow for adjustments, or just plain-old sucks!

Until I've played with more boards, I can't make any judgements, but I do know what I will be looking for.

Wiseass. I'd like to find out how other manufacturers are labeling this setting. I've rechecked the Biostar board, which is highly regarded, and can't find anything that relates to a current adjustment. I've got a Gigabyte UD4 on the way too, and I'm curious if they have a current setting. I've not seen any mention of it.

 

[LAN_deRf_HA]

Have you checked my reviews?



:shadedshu


If you've changed a setting pertaining to current, and it made no difference, then the setting you change didn't actually change the current. 100% fact...VoltsxAmps=Watts. Current=AMPs, and increasing that is either gonna give more heat, or allow for more clocks at lower voltage.


I said it before, the differences between how we are clocknig, is all down to the hardware.

What reviews? The last I see is the P7P55D-E.

 

[LifeOnMars]

On my P8P67 I have a primary plane adjustment which specifies CPU current (Ampage). It is done in incrememnts so I take it I should adjust in small increments and then run stability tests to see the difference at lower voltage?

What are the limits on this? I'm very interested but don't want to needlessly fry my little Sandy.