[Ixel]

Hi,
Well given no one replied yet to my previous question about undervolting PLL I went ahead and played around with that, the multiplier and the VCORE. Here's some of my findings.

My BIOS only allows certain specific voltages to be set, that aren't .005v apart, for PLL voltage, e.g. 1.715v, 1.753v, 1.80xv and so on, or auto.

Initially I tested my previous stable setup with a PLL voltage of 1.715v of the following:
- VCORE: 1.365v
- Multiplier: 45x
- BCLK: 100.2MHz

The above was stable, I didn't do an extensive stress test though. Most of my tests initially are 30-45 minutes, custom settings using up 3/4 of my memory, on Prime95.

I then tried the following with the same PLL voltage:
- VCORE: 1.365v
- Multiplier: 46x
- BCLK: 100.2MHz

Sadly the above crashed about 5 minutes into the test.

Since that I've managed to decrease my initial VCORE voltage from 1.365v to 1.350v and remain stable using the PLL voltage of 1.715v. Unfortunately a VCORE of 1.345v crashed about 7 minutes into the test. I will run an extensive test today and see what it turns up.

My temperatures are at most 73C on a Corsair H70 using AS5. They average around 65-70C after 15 minutes in the test.

Any suggestions or feedback are welcome.

 

[TwoCables]

Quote:

Originally Posted by Ixel;14530788
Hi,
Well given no one replied yet to my previous question about undervolting PLL I went ahead and played around with that, the multiplier and the VCORE. Here's some of my findings.

Before I begin, I just want to say that your other thread is only about 3 hours old. A thread here on OCN that is only 3 hours old is almost nothing: it can sometimes take 12-24 hours or longer to get a reply. For example: I saw it for the first time about 5 minutes ago because I browsed the Unanswered Threads for the first time 5 minutes ago.

Quote:

Originally Posted by Ixel;14530788
My BIOS only allows certain specific voltages to be set, that aren't .005v apart, for PLL voltage, e.g. 1.715v, 1.753v, 1.80xv and so on, or auto.

Initially I tested my previous stable setup with a PLL voltage of 1.715v of the following:
- VCORE: 1.365v
- Multiplier: 45x
- BCLK: 100.2MHz

The above was stable, I didn't do an extensive stress test though. Most of my tests initially are 30-45 minutes, custom settings using up 3/4 of my memory, on Prime95.

I then tried the following with the same PLL voltage:
- VCORE: 1.365v
- Multiplier: 46x
- BCLK: 100.2MHz

Sadly the above crashed about 5 minutes into the test.

Since that I've managed to decrease my initial VCORE voltage from 1.365v to 1.350v and remain stable using the PLL voltage of 1.715v. Unfortunately a VCORE of 1.345v crashed about 7 minutes into the test. I will run an extensive test today and see what it turns up.

My temperatures are at most 73C on a Corsair H70 using AS5. They average around 65-70C after 15 minutes in the test.

Any suggestions or feedback are welcome.

I don't have any suggestions, but I will say that this fascinates me because I've never considered manually adjusting my PLL voltage.

However, I am curious about one thing: why are you using 100.2 MHz instead of 100.0?

 

[Ixel]

Quote:

Originally Posted by TwoCables;14530901
Before I begin, I just want to say that your other thread is only about 3 hours old. A thread here on OCN that is only 3 hours old is almost nothing: it can sometimes take 12-24 hours or longer to get a reply. For example: I saw it for the first time about 5 minutes ago because I browsed the Unanswered Threads for the first time 5 minutes ago.

I don't have any suggestions, but I will say that this fascinates me because I've never considered manually adjusting my PLL voltage.

However, I am curious about one thing: why are you using 100.2 MHz instead of 100.0?

I see. I adjusted the BCLK by +.2MHz so it's basically an even 100MHz when in Windows, rather than 99.8MHz.

 

[TwoCables]

Quote:

Originally Posted by Ixel;14531200
I see. I adjusted the BCLK by +.2MHz so it's basically an even 100MHz when in Windows, rather than 99.8MHz.

Oh, I forgot that some boards do that.

 

[nezzarix]

I was also able to lower the voltage I needed by dropping PLL to 1.7v. It's a nice little miracle worker

Those temperatures are fine given your voltage and cooler. Still perfectly safe! Welcome to the forum

 

[TwoCables]

Were you guys at 1.8V before?

 

[nezzarix]

Quote:

Originally Posted by TwoCables;14531714
Were you guys at 1.8V before?

Yeap, that is the stock. I was at 1.75v PLL @ 4.5Ghz with 1.296v and managed to get to 4.6Ghz without a voltage increase. At first it was unstable but lowering PLL down to 1.7v fixed that problem.

 

[Ixel]

1.8v is roughly the default PLL voltage. Unfortunately unless it was just something I did, I had to increase vcore by .005v as it froze while I was doing something on the computer. I'm still less than my original vcore however.

What is the opinion on the reasonable/safer voltage to stay with for 24/7 use?

I use offset mode, which allows my system to dynamically decrease or increase the voltage based on the CPU usage. Loadline calibration is also at level 3 (originally level 5 before I started overclocking it). Currently offset I believe is set at +0.010v for the 45x multiplier, PLL voltage is still roughly 1.71v.

 

[nezzarix]

Quote:

Originally Posted by Ixel;14531753
1.8v is roughly the default PLL voltage. Unfortunately unless it was just something I did, I had to increase vcore by .005v as it froze while I was doing something on the computer. I'm still less than my original vcore however.

What is the opinion on the reasonable/safer voltage to stay with for 24/7 use?

I use offset mode, which allows my system to dynamically decrease or increase the voltage based on the CPU usage. Loadline calibration is also at level 3 (originally level 5 before I started overclocking it). Currently offset I believe is set at +0.010v for the 45x multiplier, PLL voltage is still roughly 1.71v.

Safe 24/7 voltage is a matter of opinion and varies from person to person. The Sandy Bridge spec sheet says 1.52v is the maximum but but 1.35-1.38v is recommended as the max to maximize the life of your CPU.

It all depends on your cooling, bravery, and current financial position. If 200-300 for a new CPU is a huge hit then I would keep the voltage <1.4v. 1.5v won't kill your CPU by any means but there are always extremely unlucky chips/conditions. People have been running 1.46+ for months without any problems.

 

[TwoCables]

Quote:

Originally Posted by nezzarix;14531742
Yeap, that is the stock. I was at 1.75v PLL @ 4.5Ghz with 1.296v and managed to get to 4.6Ghz without a voltage increase. At first it was unstable but lowering PLL down to 1.7v fixed that problem.

Whoa. When you say it was unstable, do you mean not stable enough to run Prime95 Blend for 12 hours, or what do you mean?

Quote:

Originally Posted by Ixel;14531753
1.8v is roughly the default PLL voltage. Unfortunately unless it was just something I did, I had to increase vcore by .005v as it froze while I was doing something on the computer. I'm still less than my original vcore however.

What is the opinion on the reasonable/safer voltage to stay with for 24/7 use?

The guy named Juan Jose ("J.J.") who is the Senior Technical Marketing Specialist at ASUS says that about 1.425V is about the maximum for 24/7 use, so I go by that.

Quote:

Originally Posted by Ixel;14531753
I use offset mode, which allows my system to dynamically decrease or increase the voltage based on the CPU usage. Loadline calibration is also at level 3 (originally level 5 before I started overclocking it). Currently offset I believe is set at +0.010v for the 45x multiplier, PLL voltage is still roughly 1.71v.

Well damn. This is intriguing. Although I'm using 48x, so I expect different results.

 

[nezzarix]

Quote:

Originally Posted by TwoCables;14531862
Whoa. When you say it was unstable, do you mean not stable enough to run Prime95 Blend for 12 hours, or what do you mean?

Yeap, nothing ever crashed but Prime95 would fail after an hour or two. Don't worry, the chip isn't leaking radioactive goo

 

[TwoCables]

Quote:

Originally Posted by nezzarix Yeap, nothing ever crashed but Prime95 would fail after an hour or two. Don't worry, the chip isn't leaking radioactive goo

Radioactive goo. hehe

Well, I just lowered the CPU PLL Voltage to 1.725V. I'll be able to begin a 12-hour Prime95 test in a few hours!

 

[Ixel]

Quote:

Originally Posted by TwoCables Radioactive goo. hehe Well, I just lowered the CPU PLL Voltage to 1.725V. I'll be able to begin a 12-hour Prime95 test in a few hours!

Radioactive goo, lol.

It will be interesting to see your results, I may try 1.40v for the vcore at the very upper limit to see what multiplier I can achieve at the lowest vcore setting after that. Given I'm using offset mode as opposed to fixed mode for the vcore I hope that as the system is idle a lot, or hardly using much CPU a lot of the time, it will help prolong the life of the CPU.

EDIT: I just saw the thread by Juan Jose, very interesting. He also says about decreasing the PLL voltage is possible while keeping system stability. He mentions some other voltages that may be decreasable just like PLL voltage, perhaps some experimentation is worth doing to see if I can cool the CPU slightly more than 73C max. First things first though, I will probably give up to 1.40v on LLC level 3 a try, see what the max multiplier I can get is, then gradually decrease the vcore until I start to see instability and just notch it up one or two from that. The only thing I may have to bear in mind is the max core temperature reached under a Prime95 for 15-30 minutes, as long as I keep it at least below 75-80C.

 

[CloudX]

Quote:

Originally Posted by Ixel Radioactive goo, lol. It will be interesting to see your results, I may try 1.40v for the vcore at the very upper limit to see what multiplier I can achieve at the lowest vcore setting after that. Given I'm using offset mode as opposed to fixed mode for the vcore I hope that as the system is idle a lot, or hardly using much CPU a lot of the time, it will help prolong the life of the CPU. EDIT: I just saw the thread by Juan Jose, very interesting. He also says about decreasing the PLL voltage is possible while keeping system stability. He mentions some other voltages that may be decreasable just like PLL voltage, perhaps some experimentation is worth doing to see if I can cool the CPU slightly more than 73C max. First things first though, I will probably give up to 1.40v on LLC level 3 a try, see what the max multiplier I can get is, then gradually decrease the vcore until I start to see instability and just notch it up one or two from that. The only thing I may have to bear in mind is the max core temperature reached under a Prime95 for 15-30 minutes, as long as I keep it at least below 75-80C.

Pretty sound idea, keep us posted on your trials. I've played with lowering the CPU PLL voltage since the very start of my SB overclocking a couple weeks ago. It seems like it keeps my voltage from going higher than my offset +0.004 adittional turbo voltage. Reason why I need the turbo voltage is simply because I cannot turn it off. On Auto it gets very generous with the v while under load after running for a while.

 

[Ixel]

Quote:

Originally Posted by CloudX Pretty sound idea, keep us posted on your trials. I've played with lowering the CPU PLL voltage since the very start of my SB overclocking a couple weeks ago. It seems like it keeps my voltage from going higher than my offset +0.004 adittional turbo voltage. Reason why I need the turbo voltage is simply because I cannot turn it off. On Auto it gets very generous with the v while under load after running for a while.

What PLL did you manage to achieve stable?

 

[CloudX]

PLL overvoltage is disabled btw. On my 4.7Ghz @ 1.408v sandy stable test it was at 1.75v. I'm at 48x and haven't done a full 12hr test yet. I will try, just had a baby so I have been using the PC when I have time, instead of messing with it. No issues at all just setting to 48x and 1.424v, 1.75pll. I've done a bunch of IBT (like tons) and Solidworks rendering and what not. It helped me bring my vcore down and keep it from jumping around without using the upper 2 LLC settings.

 

[Ixel]

An update on the progress.

I've reduced PLL voltage down to the lowest my BIOS allows me without any stability on my previous/current 4.5GHz clock that worked fine (see first post in the thread for more details). The lowest voltage option for PLL is roughly 1.52v.

There were no stability issues with that voltage on my previous/current 4.5GHz clock that worked fine. I am now trying (as I post this) the same PLL voltage with a drop in vcore by -0.010v, which previously crashed me. Also interestingly I don't know if it's me not noticing this before, but when watching the CPU clock I notice a slight .1-.5 variation every second or so in the MHz, I don't remember seeing that before reducing the PLL voltage considerably.

Anyone here on the thread have any interesting findings to add?

 

[nezzarix]

Quote:

Originally Posted by Ixel An update on the progress. I've reduced PLL voltage down to the lowest my BIOS allows me without any stability on my previous/current 4.5GHz clock that worked fine (see first post in the thread for more details). The lowest voltage option for PLL is roughly 1.52v. There were no stability issues with that voltage on my previous/current 4.5GHz clock that worked fine. I am now trying (as I post this) the same PLL voltage with a drop in vcore by -0.010v, which previously crashed me. Also interestingly I don't know if it's me not noticing this before, but when watching the CPU clock I notice a slight .1-.5 variation every second or so in the MHz, I don't remember seeing that before reducing the PLL voltage considerably. Anyone here on the thread have any interesting findings to add?

I've been looking for documents to prove/refute that such small PLL voltages is harmful for the CPU. It's well accepted that too much will kill a faster faster than anything, but I can't really find any info on too little... Intel's SB specifications state that 1.7v is the minimum but doesn't mention if it's detrimental to run a bit less.

1.7v has been working wonders for me but I want to know for sure before experimenting even further with it. PLL voltage is a dangerous thing to screw up.

 

[Ixel]

Quote:

Originally Posted by nezzarix I've been looking for documents to prove/refute that such small PLL voltages is harmful for the CPU. It's well accepted that too much will kill a faster faster than anything, but I can't really find any info on too little... Intel's SB specifications state that 1.7v is the minimum but doesn't mention if it's detrimental to run a bit less. 1.7v has been working wonders for me but I want to know for sure before experimenting even further with it. PLL voltage is a dangerous thing to screw up.

Well on the other thread I made about two hours before this, asking about PLL undervolting, a guy replied saying he had been using something like 1.3v, or w/e it was similar to that, for about 6 months upwards on his i7 9something, no problems at all. Search for threads made by me to find it, it's on this same category board.

EDIT: Additionally I've tried Prime95 for 10 minutes, OCCT for 15 minutes on medium data, and LinX using all memory 5 times, and no instability. I've knocked it down another -0.010v from the previous setting on the vcore, will test that for 30-45 minutes as errors usually occur in that timeframe.

 

[nezzarix]

Quote:

Originally Posted by Ixel Well on the other thread I made about two hours before this, asking about PLL undervolting, a guy replied saying he had been using something like 1.3v, or w/e it was similar to that, for about 6 months upwards on his i7 9something, no problems at all. Search for threads made by me to find it, it's on this same category board. EDIT: Additionally I've tried Prime95 for 10 minutes, OCCT for 15 minutes on medium data, and LinX using all memory 5 times, and no instability. I've knocked it down another -0.010v from the previous setting on the vcore, will test that for 30-45 minutes as errors usually occur in that timeframe.

It is most likely safe to use such little PLL voltage but I'd like to find a bit more support for it. I don't want to advice people to do something that may be detrimental to their CPU based on the experiences of a few people. The quest continues