OP, I believe your problem with shortening life span from your old C2D might be cooling. You need a better case like Haf X, put a nicer cooler like Hyper 212 + or evo, your components will last longer that way. In general use a 120mm cooler those 92mm ones are a bit weak for overclocked chips. Also another thing that can shorten component lifespan is the PSU, get a quality one will make your parts last longer by supplying more stable current.
Actually, my C2Duo was the longest lived system I owned. 4.5 years and was still rock solid when I built this latest machine. My A64 did eventually take a dump on me after two years, though.
My PSU is good..Corsair HX 520.
My PSU is good..Corsair HX 520.
Last edited:
UPDATE: Realized my system shutdown on reboot issues were not overclock or voltage caused, but rather the lovely Gigabyte Z68 boot loop issue (it would power cycle indefinitely after the failed restart). I updated to the latest BIOS that fixes the boot loop and have gone back down to -0.04V offset, and will likely try dropping it some more when I get home tonight. Right now I'm running P95 on it at that voltage, and temps have hit a max of 68C, but are sitting right now while churning at around 65-66C on the hottest core, 59-60C on the coolest. (been running for a while now...I'm monitoring with Splashtop from my work PC).
I also found out why my IBT scores were low...my machine hadn't yet downloaded Windows 7 SP1 until this morning, which enabled the AVX capability....now I get about 105 GFLOPs in IBT, roughly halving the run times. It also gets about 3C hotter when running that now.
A 50 minute video encode topped out at 61C.
What voltage do you think I would likely be able to get rock solid stable at 4GHz?
I also found out why my IBT scores were low...my machine hadn't yet downloaded Windows 7 SP1 until this morning, which enabled the AVX capability....now I get about 105 GFLOPs in IBT, roughly halving the run times. It also gets about 3C hotter when running that now.
A 50 minute video encode topped out at 61C.
What voltage do you think I would likely be able to get rock solid stable at 4GHz?
Last edited:
kalrith
Diamond Member
- Aug 22, 2005
- 6,628
- 7
- 81
What LLC are you using? What does cpuz say the voltage is at both idle and load?
Either way, each cpu is different, so your testing is the best thing to figure that out. I recommend running fixed voltage for ease of testing. As a starting point, set the voltage to 1.23v and LLC to 1. Then, monitor your voltage in cpuz while running Prime95. Run Prime95 for 10 minutes. Then reboot, reduce your voltage by .02v, run cpuz and Prime95 for 10 minutes. Rinse and repeat until it locks up or throws a Prime95 error. Then up the voltage by .01v from the lock-up voltage. If you don't get any errors, then let it run overnight. If it's still good in the morning, then reduce it by .005v (if it allows that fine of a setting) and test it again.
Another reason to test fixed voltage is because then you're not running into too low of an idle voltage. You might find that you have to run at LLC 3 in order to not lower your idle voltage too much.
Once you find the best fixed voltage, then you just have to find the offset and LLC to match that while still providing enough idle voltage to not crash.
Either way, each cpu is different, so your testing is the best thing to figure that out. I recommend running fixed voltage for ease of testing. As a starting point, set the voltage to 1.23v and LLC to 1. Then, monitor your voltage in cpuz while running Prime95. Run Prime95 for 10 minutes. Then reboot, reduce your voltage by .02v, run cpuz and Prime95 for 10 minutes. Rinse and repeat until it locks up or throws a Prime95 error. Then up the voltage by .01v from the lock-up voltage. If you don't get any errors, then let it run overnight. If it's still good in the morning, then reduce it by .005v (if it allows that fine of a setting) and test it again.
Another reason to test fixed voltage is because then you're not running into too low of an idle voltage. You might find that you have to run at LLC 3 in order to not lower your idle voltage too much.
Once you find the best fixed voltage, then you just have to find the offset and LLC to match that while still providing enough idle voltage to not crash.
I have not done anything with LLC...I frankly at the moment am not educated enough on what it does to feel comfortable messing with it.
Any good guides to what it does and how settings affect it? About the only thing I know is what it stands for and that it is supposed to do something with Vdroop.
Right now I've just set Vcore regulation to 'standard' and VID offset of -0.04.
My idle voltages are typically about 0.86V or so (and running at 1600MHz). Load voltages are what you see in the pic, namely 1.26V typical, but occasional fluctuations to 1.248V and 1.272V.
Any good guides to what it does and how settings affect it? About the only thing I know is what it stands for and that it is supposed to do something with Vdroop.
Right now I've just set Vcore regulation to 'standard' and VID offset of -0.04.
My idle voltages are typically about 0.86V or so (and running at 1600MHz). Load voltages are what you see in the pic, namely 1.26V typical, but occasional fluctuations to 1.248V and 1.272V.
Lazlo Panaflex
Platinum Member
- Jun 12, 2006
- 2,355
- 0
- 71
Guys, these are the 'bottom-feeder' Z68 boards by Gigabyte...meaning that you can only adjust vcore by fiddling with the offset. I know, because I have basically the same board (minus the "P"). Don't get me wrong, I think it's a nice board, just not very overclocking friendly 
Last edited:
kalrith
Diamond Member
- Aug 22, 2005
- 6,628
- 7
- 81
I think my voltage typically sits between 1.23 and 1.24v. That's at 43x, so I'm guessing you can drop your voltage quite a bit.
LLC resources:
http://www.overclock.net/t/657317/vdroop-or-not-llc-poll
http://www.anandtech.com/show/2404
Disclaimer: I'm not a Sandy Bridge overclocking expert, but here's how I understand it to work.
When you set a certain multiplier and leave everything else on auto, there's a set voltage that the motherboard thinks it needs to hit that multiplier. That number is often crazy high. Using just the offset will reduce that number by a certain amount (-.04v in your case). I believe the offset will affect both idle and load voltages.
With LLC enabled, the mobo will increase the voltage at load in order to counteract vdroop. At LLC disabled, the voltage will droop at load.
Here's the problem with determining the right LLC setting: Sandy Bridge has two set voltages, idle and load. With vdroop on a Core2 processor, you would see the voltage actually decrease at load, because there wasn't a different idle and load voltage (without 3rd party software). With vdroop on a Sandy Bridge processor, you'll see the voltage increase between idle and load, but it won't increase as much with LLC disabled.
All of this makes it hard to find the perfect combination of offset, LLC, idle, and load voltages all in one step. That's why I recommend running a fixed voltage and LLC 1 to determine your lowest stable voltage at 40x. Then, you can take that number and find an offset and LLC to match.
You might also find that you can overclock more at your present voltage without much, if any, impact on heat. I found that I could run my cpu at 43x at pretty low voltage (~1.23v), but I had to increase the voltage about 1.3v to hit 44x and 1.35v to hit 45x. Since I probably can't tell a real-world difference between 43x and 45x, I set my cpu to run at the much lower voltage but still with a 30% overclock.
Edit: I updated the above to reflect the fact that the OP's mobo only has enabled or disabled for LLC and not LLC of 1 – 5 like my mobo.
LLC resources:
http://www.overclock.net/t/657317/vdroop-or-not-llc-poll
http://www.anandtech.com/show/2404
Disclaimer: I'm not a Sandy Bridge overclocking expert, but here's how I understand it to work.
When you set a certain multiplier and leave everything else on auto, there's a set voltage that the motherboard thinks it needs to hit that multiplier. That number is often crazy high. Using just the offset will reduce that number by a certain amount (-.04v in your case). I believe the offset will affect both idle and load voltages.
With LLC enabled, the mobo will increase the voltage at load in order to counteract vdroop. At LLC disabled, the voltage will droop at load.
Here's the problem with determining the right LLC setting: Sandy Bridge has two set voltages, idle and load. With vdroop on a Core2 processor, you would see the voltage actually decrease at load, because there wasn't a different idle and load voltage (without 3rd party software). With vdroop on a Sandy Bridge processor, you'll see the voltage increase between idle and load, but it won't increase as much with LLC disabled.
All of this makes it hard to find the perfect combination of offset, LLC, idle, and load voltages all in one step. That's why I recommend running a fixed voltage and LLC 1 to determine your lowest stable voltage at 40x. Then, you can take that number and find an offset and LLC to match.
You might also find that you can overclock more at your present voltage without much, if any, impact on heat. I found that I could run my cpu at 43x at pretty low voltage (~1.23v), but I had to increase the voltage about 1.3v to hit 44x and 1.35v to hit 45x. Since I probably can't tell a real-world difference between 43x and 45x, I set my cpu to run at the much lower voltage but still with a 30% overclock.
Edit: I updated the above to reflect the fact that the OP's mobo only has enabled or disabled for LLC and not LLC of 1 – 5 like my mobo.
Lazlo Panaflex
Platinum Member
- Jun 12, 2006
- 2,355
- 0
- 71
Yeah, guess mine is the "low-low end" model, lol.
Good info there, Kalrith :thumbsup:
Good info there, Kalrith :thumbsup:
Last edited:
Puppies04
Diamond Member
- Apr 25, 2011
- 5,909
- 17
- 76
UPDATE: Realized my system shutdown on reboot issues were not overclock or voltage caused, but rather the lovely Gigabyte Z68 boot loop issue (it would power cycle indefinitely after the failed restart). I updated to the latest BIOS that fixes the boot loop and have gone back down to -0.04V offset, and will likely try dropping it some more when I get home tonight. Right now I'm running P95 on it at that voltage, and temps have hit a max of 68C, but are sitting right now while churning at around 65-66C on the hottest core, 59-60C on the coolest. (been running for a while now...I'm monitoring with Splashtop from my work PC).
I also found out why my IBT scores were low...my machine hadn't yet downloaded Windows 7 SP1 until this morning, which enabled the AVX capability....now I get about 105 GFLOPs in IBT, roughly halving the run times. It also gets about 3C hotter when running that now.
A 50 minute video encode topped out at 61C.
What voltage do you think I would likely be able to get rock solid stable at 4GHz?
That is looking a hell of a lot better than what you had before. You might squeeze the voltage a bit lower but if you had posted those numbers in the first place we most likely wouldn't be having this conversation.
Gz
Charlie98
Diamond Member
- Nov 6, 2011
- 6,292
- 62
- 91
I think I have the Lowest of low-low end... D2H-B3. I've looked everywhere and don't see any LLC adjustment, nor, I believe, voltage offset, just DVID. It works OK for me, but an OC monster it is not.
Ok, so now I need some help with where to go from here. I switched to direct setting of 1.235V, with LLC disabled. This gives idle voltages of around 1.235-1.26V, and the droop that occurs under load brings the voltage at full load to around 1.18V. Now, so far, it actually seems stable at that voltage. Passes IBT, and I'm running P95 now (though only 10 minutes in). This rather large drop in voltage meant my IBT temps dropped to 66C max (was 72) and running P95, I'm now at 62C on the hottest core after 10 minutes straight.
That part is really nice. This leads to some questions.
1. Assuming that my rig really is stable at 1.18V at full load and 4.0GHz, that would be an offset of -0.12V. I'm thinking that a drop of that much might make my system unstable at idle (it would result in idle voltages as low as about 0.70V). Thing is, the LLC settings are only available to me when setting voltage manually (otherwise they're grayed out). What's the deal with this, and can I really drop my offset that much?
2. I definitely want my CPU to drop voltage significantly under idle to both save power and reduce idle heat generation. If I'm understanding how this goes, it seems the LLC stuff only reduces Vdroop, and doesn't actually reduce the idle voltage...is that correct? If it is, how do I reduce load voltage without reducing idle voltage too much?
That part is really nice. This leads to some questions.
1. Assuming that my rig really is stable at 1.18V at full load and 4.0GHz, that would be an offset of -0.12V. I'm thinking that a drop of that much might make my system unstable at idle (it would result in idle voltages as low as about 0.70V). Thing is, the LLC settings are only available to me when setting voltage manually (otherwise they're grayed out). What's the deal with this, and can I really drop my offset that much?
2. I definitely want my CPU to drop voltage significantly under idle to both save power and reduce idle heat generation. If I'm understanding how this goes, it seems the LLC stuff only reduces Vdroop, and doesn't actually reduce the idle voltage...is that correct? If it is, how do I reduce load voltage without reducing idle voltage too much?
Puppies04
Diamond Member
- Apr 25, 2011
- 5,909
- 17
- 76
Forget it then, you are better running slightly higher at load and letting it drop down at idle then forcing it to constantly run at a slightly lower voltage than offset mode at full load.
Yeah...I tried going to 1.22 with LLC1, and I got a BSOD within a few seconds of IBT (it went to 1.15V). Went back to an offset of -0.07, which I think is where I'll probably keep it.
Interestingly enough, after the recovery of my BIOS, which recognized that I had a failed setting...my BCLK is now 0.5 MHz lower. I had it set to default, and it would always be at 100.3. After the BSOD, it's now at 99.8. I changed to manually specified, and now it's always 0.2MHz lower than what I set it. (I've got it set at 100.2 now, which is yielding exactly 100 in CPU-Z). So I'm running at exactly 4GHz now, instead of 4011.
I'm at 1.224V at full load now, and I'm going to run P95 for a while to make sure it's stable there...then I'll let it idle for a while and make sure it's stable at idle. If all goes well, I'll just leave it here.
Interestingly enough, after the recovery of my BIOS, which recognized that I had a failed setting...my BCLK is now 0.5 MHz lower. I had it set to default, and it would always be at 100.3. After the BSOD, it's now at 99.8. I changed to manually specified, and now it's always 0.2MHz lower than what I set it. (I've got it set at 100.2 now, which is yielding exactly 100 in CPU-Z). So I'm running at exactly 4GHz now, instead of 4011.
I'm at 1.224V at full load now, and I'm going to run P95 for a while to make sure it's stable there...then I'll let it idle for a while and make sure it's stable at idle. If all goes well, I'll just leave it here.
Charlie98
Diamond Member
- Nov 6, 2011
- 6,292
- 62
- 91
I reduced my DVID volts -.08v and it dropped my fully loaded LinX temps almost 10C... it does make a big difference.
Because I have no control over LLC, I think that's why my system crashed at any DVID adjustment over -.10v
I have my BCLK set at 100.2MHz so it works out to an even 100.MHz on the system... I like even numbers.
After one of my BSOD's I had to go back in and adjust EVERYTHING... it jacked with almost every setting. Also, if you use 'performance defaut' settings, it will reset the controller to IDE (instead of ACHI.) Just make sure you check that if you BSOD again and have a problem booting up.
Because I have no control over LLC, I think that's why my system crashed at any DVID adjustment over -.10v
I have my BCLK set at 100.2MHz so it works out to an even 100.MHz on the system... I like even numbers.
After one of my BSOD's I had to go back in and adjust EVERYTHING... it jacked with almost every setting. Also, if you use 'performance defaut' settings, it will reset the controller to IDE (instead of ACHI.) Just make sure you check that if you BSOD again and have a problem booting up.
Last edited:
So curiosity has gotten the better of me, so I just bumped it another 100MHz...running at 4.1 @ 1.236V, and ran Prim95 all night. I don't know whether it ran it perfectly, but at the very least it didn't crash the machine, as my computer had rebooted due to windows update (checked the system logs, and sure enough...system initiated restart for application of updates at 3:18AM). Still, that means that the computer at the very least stayed operational for 6 hours of P95. I'll run it during the day today to make sure it's stable at this voltage, but it passes IBT and was fine when I went to sleep last night after an hour and a half of P95, so I'm probably pretty good to go.
Temps are 64-66C on the hottest core during P95, so I'm happy with that. Thanks again for everyone's help. I am going to stop here...no need to really stress this thing now, and I'm running at the same voltage the system picks normally for 3.3GHz (I checked, it uses 1.24V). It's a -0.07 offset. I think I'm going to try and lower another .01V, but then I'm going to stop if that is 100% stable.
EDIT: Dropped to 1.224V and it has completed 20 runs of IBT fine, P95 seems fine, and PCMark finished just fine. Let it idle for an hour as well, and so far it looks solid. I'll see how it is in day to day life and may run an extended P95 test, but considering I was passing IBT at 4GHz and 1.17V, I think I'm going to be fine at 4.1 and 1.22. I might be able to go lower, as I'd imagine 1.20 would be just fine at 4.1, maybe even 1.19 or so, but the idle voltage at that point would be a little lower than where I'm comfortable...I don't want random silent data corruption, which I understand might be possible when running a little too undervolted.
Temps are 64-66C on the hottest core during P95, so I'm happy with that. Thanks again for everyone's help. I am going to stop here...no need to really stress this thing now, and I'm running at the same voltage the system picks normally for 3.3GHz (I checked, it uses 1.24V). It's a -0.07 offset. I think I'm going to try and lower another .01V, but then I'm going to stop if that is 100% stable.
EDIT: Dropped to 1.224V and it has completed 20 runs of IBT fine, P95 seems fine, and PCMark finished just fine. Let it idle for an hour as well, and so far it looks solid. I'll see how it is in day to day life and may run an extended P95 test, but considering I was passing IBT at 4GHz and 1.17V, I think I'm going to be fine at 4.1 and 1.22. I might be able to go lower, as I'd imagine 1.20 would be just fine at 4.1, maybe even 1.19 or so, but the idle voltage at that point would be a little lower than where I'm comfortable...I don't want random silent data corruption, which I understand might be possible when running a little too undervolted.
Last edited:
Lazlo Panaflex
Platinum Member
- Jun 12, 2006
- 2,355
- 0
- 71
Nice job, there...4.1 on stock volts is pretty sweet. I have mine undervolted to 1.044 (per CPU-Z, stock speeds).
Get yourself an SSD now
Get yourself an SSD now
Last edited:
I overclock my 2500k at 4.8 and 4.5, don't remember the specs they are clocked at off the top of my head, but I can check when I get home. I have a noctua NH-D14 and I rarely hit 55c at full load
Puppies04
Diamond Member
- Apr 25, 2011
- 5,909
- 17
- 76
Does that mean you rarely do anything stressfull with your CPU? What temps does it hit during 20 very high IBT runs?
I mean, my machine doesn't go much above 50 during most uses, even gaming (played MW3 for an hour last night and my CPU peaked at 51C). Hits 60C for video encoding on all four cores, and really doesn't get above that except when stress testing....but I'm doubting 4.8GHz and 55C on air running IBT.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 20K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 12K
-
Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)- Started by Vattila
- Replies: 13K
-
Facebook
Twitter