Trouble getting my 2600k stable @4.5ghz

bigj1985

Distinguished
Mar 12, 2010
331
0
18,810
Ok so here it goes. I cannot for the life of me get my 2600k completely stable at 4.5ghz on a Gigabyte Z68XP-UD4 with the F5 Bios. I couldn't do it with F4 either. When I say stable I mean Prime stable. I am completely stable while gaming, browsing, benchmarks, etc. But I continue to recieve error codes in prime :cry:

For some reason the 2 most errors that I get are at the beginning of the test it will usually say (could not detect some of the hyperthreaded CPU's and therefore overide should be accurate bla bla bla). I get this error even at stock settings. The next one is (rounding was 0.5 and was expected 0.4) or something to that effect. What I am trying to figure out is whether it is my RAM or my cpu that is causing these errors?

At 4.5 ghz using 1.3Vcore the PC will boot into windows and be completely stable running everything I do except Prime. At 1.3V it will usually throw a rounding error at the 5-10 minute mark. The odd thing is adding more voltage does not seem to increase stability the way it should if it was the volts. What I mean is that it does not scale well at all wit the amount of voltage added vs the amount of time prime will run w/o a rounding error. For instance; jumping from 1.3Vcore under load to 1.34Vcore under load still seems to error in Prime after about 15-20 minutes. I even increased Vcore to 1.365V under load and will still get a rounding error at the 15-20 minute mark. Often times sooner. This has me wondering if it really is my Vcore that is the culprit. Also it seems that no matter what, even increasing the Vcore to 1.365V still throws out the(could not detect some of the hyperthreaded CPU's) error.

My RAM the way it sits is on XMP profile 1 with the (standard) performance enhance selected. From what I have researched online it seems that these Z68 Gigabyte boards need a little more tweaking than boards with the now standard UEFI bios. I figure if I can load into windows and start gaming and browsing and playing music etc. @4.5ghz with only 1.3Vcore there is no way in hell this chip needs a jump all the way to 1.37V just to be stable.

Please if anyone can help me because this is really really bugging the heck out of me not having a completely stable system. I bought a 2600k and dammit 4.5ghz should be no sweat. And the board has an excellent solid 18 phase VRM so there shouldn't be any issues. I thank all those willing to reply and have a great day:)
 
Solution
D
Change your RAM to 1.5v and loosen the timings to 9-9-9-24. You are overclocking the integrated memory controller on the CPU and Sandy Bridge CPUs can be sensitive to that. Technically running anything over 1.5v voids the CPU warranty. The cas 7 stuff gains you just about nothing over cas 9 anyway. Only way you will ever see a difference is in benchmarks and even there the difference is tiny.
D

Deleted member 217926

Guest
Change your RAM to 1.5v and loosen the timings to 9-9-9-24. You are overclocking the integrated memory controller on the CPU and Sandy Bridge CPUs can be sensitive to that. Technically running anything over 1.5v voids the CPU warranty. The cas 7 stuff gains you just about nothing over cas 9 anyway. Only way you will ever see a difference is in benchmarks and even there the difference is tiny.
 
Solution

bigj1985

Distinguished
Mar 12, 2010
331
0
18,810



Ok I will try that then. The only reason I left it at 1.6V is because I have read many conflicting reports saying that the 1.5V limit is all BS and that the chip can run 1.65V safely with no issue. This ram was also selected from my motherboards QVL and it just so happened to be CAS7 and on sale. That's why I bought it. But thank you I will attempt this later tonight and report back:)
 

bigj1985

Distinguished
Mar 12, 2010
331
0
18,810


Ya I'm using offset. I'm testing now. I clicked "load optimized defaults" in my bios and it set my RAM @11-11-28-1T @DDR3 1600 using 1.5V instead of the 1.6V the RAM calls for. I'm just leaving it there for now. 1.3V crashed 5 minutes into Prime. 1.32V crashed 20 minutes in. I'm now trying 1.33V and will report back
 

bigj1985

Distinguished
Mar 12, 2010
331
0
18,810
So far completely stable for 2 hours running @ 4.5 ghz and 1.34Vcore under load using offset. Turns out the rounding errors were caused by the RAM being at 1.6V. I was able to tighten the timings to 8-8-8-24 1600 @ 1.5V completely stable. I am just so thankful as I was going to be pissed if I got 1 of those dud of a 2600k that wouldn't do 4.5ghz. Seeing as that was my goal when I bought the system. Enough of a speed jump to notice a performance increase and fast enough to last another year or 2 w/o bottenecking any type of single card GFX setup. Yet conservative enough to not drastically shorten the life span of the chip. Thanks for all the help guys:)
 
D

Deleted member 217926

Guest
Some Sandy Bridge chips just hate having their memory controller overclocked. Did you leave the command rate at 1T? That can be very hard on most RAM and memory controllers. I would make sure it's set to 2T. So 8-8-8-24 2T.
 

bigj1985

Distinguished
Mar 12, 2010
331
0
18,810



If I can ask you one more thing. I have 2 ways to go. I can keep it at 8-8-8-24 @2T stable. If I want 1T I found I need to loosen the timings to 10-10-10-28 or face BSOD after 30 minutes of Prime. Which way do you think I will get more performance? I would think even with looser timings 1T would perform a little better but I could be wrong. Thank you for all of your help BTW>
 
D

Deleted member 217926

Guest
I would just set it to 8-8-8-24 2T and forget about it. Due to the integrated memory controller there is very little difference in performance once you reach DDR3 1600 cas 9.

I bought my system the week Sandy Bridge was released in January 2011 so besides getting caught up in the whole P67 chipset recall ( that was fun :( ) , I bought my memory before real testing had been done. If I had bought it later I would have just gone with DDR3 1600 cas 9.

Bit-Tech did an excellent article on it.

http://www.bit-tech.net/hardware/memory/2011/01/11/the-best-memory-for-sandy-bridge/


The whole article is worth reading but these are the important parts.

http://www.bit-tech.net/hardware/memory/2011/01/11/the-best-memory-for-sandy-bridge/4

http://www.bit-tech.net/hardware/memory/2011/01/11/the-best-memory-for-sandy-bridge/5

http://www.bit-tech.net/hardware/memory/2011/01/11/the-best-memory-for-sandy-bridge/6
 

bigj1985

Distinguished
Mar 12, 2010
331
0
18,810



Thanks for everything bro. Extremely informative. I'm actually happy to report that I was able to get my memory stable as well at 1600 7-8-7-24 2T using 1.6V. I found that my board was setting my ram at 1T all by itself and was causing instability. It was the bios I was using. F5. Once I went to F6 it stopped doing that