Overclock.net banner

1 - 18 of 18 Posts

·
Registered
Joined
·
8 Posts
Discussion Starter · #1 ·
Hey there overclockers. This is my first post and I've got a question that I can't seem to be able to determine the answer to. I've read as much as I could and followed as many guides as I could so if I perform a forum faux pas please just let me know. Anyways I recently installed a Kuhler 620 water cooler on my 2500k in preparation of my first ever overclock. I got the multiplier set to 45 and SVID auto adjusted my voltage to 1.355. I then proceeded to reduce voltage till errors/BSODs.
At 1.295v I got a BSOD at 15 minutes of prime95 blend.
At 1.305v I got a BSOD at the 51 minute mark of a prime95 blend.
At 1.315v the blend ran for a little under 12 hours without crashing however when I checked on it the following morning the second worker thread had failed at the 4 hour mark. It told me to check the stress.txt file

This is where confusion set in. I couldn't find a stress.txt file anywhere. I could restart the blend and the second core would start processing fine. I'm not too familiar with prime95 so I couldn't tell if this was a bug in the program, or a stability issue related to current clock speed and voltage. The only reason I thought it might be a bug of some sort is because it ran for 4 hours before worker failure and that seems pretty stable to me. Then again I don't have as much experience as all the intelligent, delightful, and helpful people on this forum
biggrin.gif
. I've included pictures below of my CMOS settings and screenshots of the prime95 worker failure as well.

Basically it boils down to this:
Is the worker failure a result of low voltage and I should up the voltage in the BIOS and start another 12 hour blend?
OR
Is my current multiplier and voltage stable and I should either ignore the worker failure or start another blend session at 1.315v?

This is the worker failure and error message
338

This is me wondering where "stress.txt" was
338

Advanced frequency settings
359

Advanced cpu core features
359

Advanced voltage settings
359

Intel i5 2500k @ 4.5GHz 1.315v
Gigabyte P67A-UD4-B3
16GB GSkill Ripjaw @ 1600MHz
Gigabyte GeForce GTX 560 Ti OC @ 900 MHz
Antec Kuhler H2O 620 Push/Pull setup
AZZA Dynamo 850W PSU

Sidenote: I read somewhere that with an overclocked sandy bridge the multiplier would actually drop when the CPU wasn't under load and only up the multiplier when load demanded it. I never saw the multiplier drop below 45x before or after running prime95. Do I have a setting wrong or is it just going to stay consistently at 4.5GHz?
 

·
Registered
Joined
·
329 Posts
Stress.txt is just their instruction/troubleshooting guide. It's in the folder with the prime.exe file. Your multiplier isn't dropping because you have C1E/C3/C6 disabled. They lower the multiplier at idle. You probably just need to up the voltage one more notch and go for a 24 hour run (that way prime can run through all the tests).
 

·
Registered
Joined
·
8 Posts
Discussion Starter · #3 ·
Thanks for the reply. I turned off the C1E/C3/C6 because I was following this guide for Sandy Bridge + Gigabyte P67 boards. If you scroll down he says to turn it off shortly after the third picture. Is it safe to turn those settings back on after my OC is stable? Also thanks for the voltage suggestion. Should I step it up once or twice? From 1.315v=>1.32v or 1.315v=>1.325v Was hoping I would have all this done in time to play Diablo 3 but I guess starting Saturday just wasn't enough time. The amount of time that I have to let my machine just sit there running prime is making me sad. Hopefully it will all be worth it in the end.
 

·
Registered
Joined
·
329 Posts
Quote:
Originally Posted by Helpless View Post

Thanks for the reply. I turned off the C1E/C3/C6 because I was following this guide for Sandy Bridge + Gigabyte P67 boards. If you scroll down he says to turn it off shortly after the third picture. Is it safe to turn those settings back on after my OC is stable? Also thanks for the voltage suggestion. Should I step it up once or twice? From 1.315v=>1.32v or 1.315v=>1.325v Was hoping I would have all this done in time to play Diablo 3 but I guess starting Saturday just wasn't enough time. The amount of time that I have to let my machine just sit there running prime is making me sad. Hopefully it will all be worth it in the end.
Just do one bump. If that's not stable, then do one more and that SHOULD be from where you're at. A lot of people turned C states off, but it's not a bad thing to leave them on. They help the CPU drop the multiplier and voltage at idle if you're using offset. With manual, turn them off cause they won't work anyway. Yeah, it took me a week of testing my computer for stability to finally get it there so I feel ya.
 

·
Registered
Joined
·
8 Posts
Discussion Starter · #5 ·
Thanks a lot Nihilo. You've been a tremendous help. Prime95 blend is running right now on 1.32v. After I get it stable I will turn C states back on and do ANOTHER 12-24 hour prime blend to make sure everything is honkay doray.
thumb.gif
 

·
Registered
Joined
·
203 Posts
Quote:
Originally Posted by Helpless View Post

Thanks for the reply. I turned off the C1E/C3/C6 because I was following this guide for Sandy Bridge + Gigabyte P67 boards. If you scroll down he says to turn it off shortly after the third picture. Is it safe to turn those settings back on after my OC is stable? Also thanks for the voltage suggestion. Should I step it up once or twice? From 1.315v=>1.32v or 1.315v=>1.325v Was hoping I would have all this done in time to play Diablo 3 but I guess starting Saturday just wasn't enough time. The amount of time that I have to let my machine just sit there running prime is making me sad. Hopefully it will all be worth it in the end.
Having one worker fail usually means you are almost but not quite stable. A minor bump to 1.320 might do it, but personally I would up the voltage from 1.315 to 1.325 as I don't like having 'just enough' voltage. I see you have LLC on Auto though, and in your screen shot CPU-Z is showing 1.284. What does CPU-Z show when under load? It might be that you just need to tighten up your LLC to get stable.

EIST and C states shouldn't really affect stability provided there is enough voltage to remain stable when downclocked / downvolted. C states definitely should be ON if you have an SSD, disabling them can cripple SSD performance.
 

·
Not new to overclock.net
Joined
·
1,588 Posts
In my experience, you may end up in the rut of adding cpu voltage until the cows come home, but youll still get a worker to fail at some point. I see you are running 16gb of ram, is that four sticks or two? Ive usually had single worker failure as a result of ram overclock/timing changes. You may want to look into adding VCCIO to 1.12 or thereabouts because you are running a substantial amount of ram., which may also require a ram voltage bump as well. I also see your doing some wierd stuff with the turbo core ratio. You may as well set it to all cores because you will get negligible to no difference doing what your trying to do there. I have fussed with scaling the turbo cores and it never worked for me. Although in about 10 minutes someone will probably jump in telling me Im wrong. Try settings turbo ratio to all cores, leaving voltage where you last tested moderatlely well, and raise your VCCIO up through 1.14 or so, and see where it works. Have you used intel burn test at all?
 

·
Registered
Joined
·
329 Posts
Quote:
Originally Posted by guitarmageddon88 View Post

In my experience, you may end up in the rut of adding cpu voltage until the cows come home, but youll still get a worker to fail at some point. I see you are running 16gb of ram, is that four sticks or two? Ive usually had single worker failure as a result of ram overclock/timing changes. You may want to look into adding VCCIO to 1.12 or thereabouts because you are running a substantial amount of ram., which may also require a ram voltage bump as well. I also see your doing some wierd stuff with the turbo core ratio. You may as well set it to all cores because you will get negligible to no difference doing what your trying to do there. I have fussed with scaling the turbo cores and it never worked for me. Although in about 10 minutes someone will probably jump in telling me Im wrong. Try settings turbo ratio to all cores, leaving voltage where you last tested moderatlely well, and raise your VCCIO up through 1.14 or so, and see where it works. Have you used intel burn test at all?
I'm with you on the core thing. Set them to scale together. Can you explain what VCCIO is since I'm not familiar with it at all. I'm also using 16GB of ram on two sticks.
 

·
Not new to overclock.net
Joined
·
1,588 Posts
Quote:
Originally Posted by Nihilo View Post

I'm with you on the core thing. Set them to scale together. Can you explain what VCCIO is since I'm not familiar with it at all. I'm also using 16GB of ram on two sticks.
VCCIO is the memory controller voltage. Whenever you overclock you ram (in our case, up to the next ram multi) many times youll need a VCCIO bump in addition to a ram voltage bump. I have low voltage memory, so Im at 1.25v 1600mhz stock. I upped my multi to 1866, had to increase voltage to 1.35v, loosen timings to 10-10-10-25, and add 1.13 of VCCIO to be stable. I got nowhere until VCCIO was added. I was trying for a while to get 2133, but no matter how far I loosened the timings, even to 13 (which probably has diminishing returns at that point) I couldnt get stable. 1866 at these specs is a good number to me. Granted, added VCCIO made add 1 or 2 degrees onto your load temps, but if you want stability with 16gb ram, you may need it. Leave everything as is, take out two sticks, and see where you test. That will tell you right away if it was ram causing the worker failures. Let me know
 
  • Rep+
Reactions: Nihilo

·
Not new to overclock.net
Joined
·
1,588 Posts
Quote:
Originally Posted by guitarmageddon88 View Post

VCCIO is the memory controller voltage. Whenever you overclock you ram (in our case, up to the next ram multi) many times youll need a VCCIO bump in addition to a ram voltage bump. I have low voltage memory, so Im at 1.25v 1600mhz stock. I upped my multi to 1866, had to increase voltage to 1.35v, loosen timings to 10-10-10-25, and add 1.13 of VCCIO to be stable. I got nowhere until VCCIO was added. I was trying for a while to get 2133, but no matter how far I loosened the timings, even to 13 (which probably has diminishing returns at that point) I couldnt get stable. 1866 at these specs is a good number to me. Granted, added VCCIO made add 1 or 2 degrees onto your load temps, but if you want stability with 16gb ram, you may need it. Leave everything as is, take out two sticks, and see where you test. That will tell you right away if it was ram causing the worker failures. Let me know
edit: I would check to see your memory is running at 2t instead of one as well. Youll probably want to just enable the XMP profile because it looks like its not

woops...didnt mean to quote myself
 

·
Registered
Joined
·
329 Posts
Quote:
Originally Posted by guitarmageddon88 View Post

VCCIO is the memory controller voltage. Whenever you overclock you ram (in our case, up to the next ram multi) many times youll need a VCCIO bump in addition to a ram voltage bump. I have low voltage memory, so Im at 1.25v 1600mhz stock. I upped my multi to 1866, had to increase voltage to 1.35v, loosen timings to 10-10-10-25, and add 1.13 of VCCIO to be stable. I got nowhere until VCCIO was added. I was trying for a while to get 2133, but no matter how far I loosened the timings, even to 13 (which probably has diminishing returns at that point) I couldnt get stable. 1866 at these specs is a good number to me. Granted, added VCCIO made add 1 or 2 degrees onto your load temps, but if you want stability with 16gb ram, you may need it. Leave everything as is, take out two sticks, and see where you test. That will tell you right away if it was ram causing the worker failures. Let me know
Ah nice. +rep for that explanation. I didn't have to up mine, but I'm running my sticks at stock, but this is great to remember for when I decide to try OCing them.
 

·
Not new to overclock.net
Joined
·
1,588 Posts
Quote:
Originally Posted by Nihilo View Post

Ah nice. +rep for that explanation. I didn't have to up mine, but I'm running my sticks at stock, but this is great to remember for when I decide to try OCing them.
Well what I meant was that by virtue of overloading the DIMMS that much you may actually need VCCIO at stock speeds. Try that. The more memory, the more stress on the other components, for probably, not too much gain. Unless of course you run servers and photoshop/CAD and such
 

·
Registered
Joined
·
8 Posts
Discussion Starter · #13 ·
Quote:
Originally Posted by iSeries View Post

I see you have LLC on Auto though, and in your screen shot CPU-Z is showing 1.284. What does CPU-Z show when under load? It might be that you just need to tighten up your LLC to get stable.
...
f you have an SSD, disabling them can cripple SSD performance.
How would I tighten LLC up? The only options for it is auto-enabled-disabled.
Don't have an SSD yet, one piece at a time.

Quote:
Originally Posted by guitarmageddon88 View Post

I see you are running 16gb of ram, is that four sticks or two? Ive usually had single worker failure as a result of ram overclock/timing changes. You may want to look into adding VCCIO to 1.12 or thereabouts because you are running a substantial amount of ram., which may also require a ram voltage bump as well.

I also see your doing some wierd stuff with the turbo core ratio. You may as well set it to all cores because you will get negligible to no difference doing what your trying to do there. I have fussed with scaling the turbo cores and it never worked for me. Although in about 10 minutes someone will probably jump in telling me Im wrong. Try settings turbo ratio to all cores, leaving voltage where you last tested moderatlely well, and raise your VCCIO up through 1.14 or so, and see where it works. Have you used intel burn test at all?
Ram is 4x4GB. I haven't changed any RAM timings. The only things I"ve touched in BIOS so far is multiplier and voltages.
As for the turbo core stuff, I haven't touched it. That's what it was like at default. I went and rechecked the guide by Sin0822 he actually recommended disabling it (which I missed) so I'll go back and change that.
 

·
Registered
Joined
·
203 Posts
Quote:
Originally Posted by Helpless View Post

How would I tighten LLC up? The only options for it is auto-enabled-disabled.
If on Auto you find your load voltage is considerably lower than your idle voltage then set it to enabled and check the difference again. That's not to say you will need LLC to be stable, you just need to make sure your load voltage is where you need it to be. LLC can help get a grip on vdroop though rather than just chucking more voltage at it.
 

·
Registered
Joined
·
8 Posts
Discussion Starter · #15 ·
So if I see idle voltage that is as high, or higher, than my load voltage then I should enable it?

edit: I just wanted to thank everyone who has chimed in on this thread. I've worked with computers for a while and I've always been intimidated by overclocking so this first time has been really daunting. Your input and suggestions have been put to good use and I'm just really glad that a community like this exists.
 

·
Registered
Joined
·
203 Posts
Quote:
Originally Posted by Helpless View Post

So if I see idle voltage that is as high, or higher, than my load voltage then I should enable it?
I see a certain amount of droop as a good thing (unlike the wife boom boom) but if Auto means no LLC and your idle voltage is considerably higher than your load voltage rather than chucking more and more voltage at it to get stable at load you may want to try and enable it. All LLC will do is lessen the gap between load voltage and idle voltage.
 

·
Not new to overclock.net
Joined
·
1,588 Posts
Well, you have a 2500k so I would not be running it at full speed all the time. Make use of the speedstep technology and let the thing downclock/use the offset voltage. Im sure that the feature of the chip staying downclocked during idle will significantly extend the life of these processors, unlike my 1090t which I burned out in less than a year.
 

·
Registered
Joined
·
8 Posts
Discussion Starter · #18 ·
Quote:
Originally Posted by guitarmageddon88 View Post

Well, you have a 2500k so I would not be running it at full speed all the time. Make use of the speedstep technology and let the thing downclock/use the offset voltage. Im sure that the feature of the chip staying downclocked during idle will significantly extend the life of these processors, unlike my 1090t which I burned out in less than a year.
I'm actually doing that right now. The system has been stable all day so I'm going to enable the C states and boot it back up. Then a quick 10 minute prime to make sure the multiplier is actually fluctuating like it should and there are no sudden voltage spikes. Then going on for the night. Wish me luck.
 
1 - 18 of 18 Posts
Top