Sign in with
Sign up | Sign in
Your question

Fatal Error: Rounding was 0.5 2600k on P8Z68 Deluxe/Gen3 after 8 hour

Last response: in Overclocking
Share
December 31, 2011 8:25:27 PM

Quick rundown.

New build with Asus p8z68 Deluxe/gen3 MOBO. 2600k and Corsair Vengeance 9-9-9-24 DDR3 1600 ram.

I overclocked the processor to 4.5 GHZ and running prime95 blend was going well until after 8 hours I got errors on 2 threads within second of each other.

I'm really not sure what adjust. It's close to stable. Right now the CPU is running at 1.3v .. ram is at suggested 1.5v.

I'm not really very experienced overclocking. I've done 3 other systems and made them stable but this one i'm not sure what to do next.

Does this sound more like a RAM problem or CPU issue? Where should I start adjusting from here?

I'm guessing it's a RAM issue because I ran the small FFT test for 12 hours with no error. And after those 2 threads received fatal errors my motherboard temp sensor stopped working and showed my Mobo at -1C .. which is not correct hehe.

Any help is appreciated. This system is a new workstation so it's critical to keep this thing stable. I'm happy to provide any more info if it's necessary.

Thanks
January 1, 2012 11:40:51 PM

I may have posted this in the wrong section. But I wanted to give an update.

I've been running in offset mode so the cpu voltage lessens when there is no load. I like this feature and want to take advantage of it. Not really understanding this feature, and other new features with these intel chips, is probably where my trouble had been.

After reading up on these new features i've determined that I need to raise the value of "additional turbo voltage" found in the "cpu power management" section of the Bios.

I nudged it up by .012 volts and running a p95 blend test.. going ok for past 2 hours.

If I still receive errors i'm going to turn back the speed of my ram from 1600 to 1333. I should have done this before I started trying to OC the processor but it's too late now.

I'll let ya'll know how this test finishes. I'm giving another 22 hours if there are no errors. We'll see.


m
0
l
January 2, 2012 12:16:15 AM

If you are getting Prime errors your overclock is not stable. You should either reduce the overclock or increase the voltage. Don't forget that increasing the voltage will reduce the life of your processor. I think it may be better to reduce the overclock slightly., a few percent difference in performance will not be noticed in everyday usage.
m
0
l
Related resources
a b K Overclocking
January 2, 2012 1:26:32 AM

pjmelect said:
If you are getting Prime errors your overclock is not stable. You should either reduce the overclock or increase the voltage. Don't forget that increasing the voltage will reduce the life of your processor. I think it may be better to reduce the overclock slightly., a few percent difference in performance will not be noticed in everyday usage.

His voltage is very very safe as long as he has good temps. Just don't go too far past 1.4v. 1.45v if you have very good cooling. As for the issue, I had the same problem on my 2500k. No need to mess with the RAM. I run mine at 1600mhz and didn't have to touch it. Although, running the memory @ 1.55v may help a bit. Anyway, I ran my voltage at auto with seemed to yield best voltage results and did not give me any errors. I had errors running a fixed voltage even when it was higher than my maximum auto voltage. Try running the voltage on auto and make sure it doesn't pass 1.35 - 1.36v maximum while stressing it. That really wouldn't be a dangerous voltage, but sometimes auto overvolts and in that situation it would be overvolting by a marginal amount.

What cooler do you have?
m
0
l
January 2, 2012 1:44:55 AM

HostileDonut said:
His voltage is very very safe as long as he has good temps. Just don't go too far past 1.4v. 1.45v if you have very good cooling. As for the issue, I had the same problem on my 2500k. No need to mess with the RAM. I run mine at 1600mhz and didn't have to touch it. Although, running the memory @ 1.55v may help a bit. Anyway, I ran my voltage at auto with seemed to yield best voltage results and did not give me any errors. I had errors running a fixed voltage even when it was higher than my maximum auto voltage. Try running the voltage on auto and make sure it doesn't pass 1.35 - 1.36v maximum while stressing it. That really wouldn't be a dangerous voltage, but sometimes auto overvolts and in that situation it would be overvolting by a marginal amount.

What cooler do you have?



I have a Zalman cooler. I've had good experiences with them in the past. Quiet and efficient. Temps are peaking at about 67 C under full load at 1.320 volts peak.

I'm trying to find the balance between stability and efficiency. The build is for a new audio workstation so the demand on the processor varies. That's the reason I want to keep the offset engaged to reduce voltage during times when the processor isn't being asked to do much.

The ram I don't think needs to go past stock. But just for the sake of ruling the RAM out as a cause for the errors in P95 I figured i'd turn em back to 1333 if this current test in P95 fails before 24 hours.

The RAM voltage is currently set to 1.5.. I had it on auto and set it to manual just to see if it made a difference. It didn't.

I'm about 3 hours in to this p95 run and no errors. The workers seem to all be close to in sync. Before I had several workers falling pretty far behind with no extra programs running. So that's good I would think lol.

All these new options and improved auto clocking features have really thrown a curve ball at me but I think it's still going well. I'll keep you guys posted as to how the rest of this tune goes. Hopefully I can wrap this up over the next 48 hours and then come the joys of installing software for 2 days. :pt1cable: 
m
0
l
a b K Overclocking
January 2, 2012 2:16:18 AM

Okay, 67C isn't bad. Just don't exceed 70C. Also, auto voltage will tune down your voltage like offset when it is not needed.
m
0
l
January 2, 2012 2:22:38 AM

HostileDonut said:
Okay, 67C isn't bad. Just don't exceed 70C. Also, auto voltage will tune down your voltage like offset when it is not needed.


OK. That's good to know. I'll keep it in mind if this p95 run errors. May be worth a shot to see if it makes a difference. Thanks!
m
0
l
January 2, 2012 2:29:57 AM

I personally don't like to increase my voltages and I would not be happy with temperatures above 60C at full load, but then I am a conservative overclocker and my processors are precious to me. :sweat: 

You also have to consider the ambient temperature, It is winter where I am from and my room temperature is only 18C, in the summer the temperature can go to over 30C. Your overclock may not be stable with higher ambient temperatures and your processor temperature will get much to hot.
m
0
l
January 2, 2012 3:06:21 AM

pjmelect said:
I personally don't like to increase my voltages and I would not be happy with temperatures above 60C at full load, but then I am a conservative overclocker and my processors are precious to me. :sweat: 

You also have to consider the ambient temperature, It is winter where I am from and my room temperature is only 18C, in the summer the temperature can go to over 30C. Your overclock may not be stable with higher ambient temperatures and your processor temperature will get much to hot.



From what other people have experienced with this processor my current temps are right where they should be during a p95 blend test with air cooling. I've seen people with higher temps than this with stock voltage. I've also seen several tests with the stock HSF running a p95 large FFT test with temps in the 80's.

I think that if Intel ships a processor and HSF the HSF would be sufficient to cool the processor. And with temps in the 80's with the stock HSF i'm pretty comfortable at 67C.

I know that Intel tests the 2500 and 2600 up to 1.5+ volts. Running .020 above stock isn't going to damage anything.

As far as temps go. Right now my fans are running at about have speed for both the case and HSF. I have the fans set to ramp up to 3/4 speed when CPU temp hits 70. If it were warmer in here, which it won't ever be because this is a climate controlled studio, the fans would spin faster.

As far as temps go i'm very comfortable with them where they are. And as the paste sets the temps will be further reduced.
m
0
l
a b K Overclocking
January 2, 2012 3:19:34 AM

Push said:
From what other people have experienced with this processor my current temps are right where they should be during a p95 blend test with air cooling. I've seen people with higher temps than this with stock voltage. I've also seen several tests with the stock HSF running a p95 large FFT test with temps in the 80's.

I think that if Intel ships a processor and HSF the HSF would be sufficient to cool the processor. And with temps in the 80's with the stock HSF i'm pretty comfortable at 67C.

I know that Intel tests the 2500 and 2600 up to 1.5+ volts. Running .020 above stock isn't going to damage anything.

As far as temps go. Right now my fans are running at about have speed for both the case and HSF. I have the fans set to ramp up to 3/4 speed when CPU temp hits 70. If it were warmer in here, which it won't ever be because this is a climate controlled studio, the fans would spin faster.

As far as temps go i'm very comfortable with them where they are. And as the paste sets the temps will be further reduced.

You are right on the money! In the low 60C is good for 24/7 use, and Prime95 generates more heat than normal applications, meaning your CPU should be in the low 60s. Sounds good for 24/7 usage.
m
0
l
January 2, 2012 3:35:21 PM

Happy to say that the new build has successfully completed 16hours of P95 blend testing. Temps overnight peaked at 62C briefly. It seems that the thermal paste has set.

The only change I made in the bios was upping the "additional turbo voltage" to .008. Previously I had said that I upped it by .012 but I was wrong.

So today I dialed that setting back one notch to .004 and i'm running another blend test in P95. We'll see how that goes.

Easy overclock. I got a little nervous the first time around because others had such a consistently easy time with this processor I was nervous that my ram had issues when p95 got an error in it's first test.

Once this run of p95 finishes i'll start on the memory. See if we can wiggle a little something out of it.
m
0
l
a b K Overclocking
January 2, 2012 3:49:31 PM

Push said:
Happy to say that the new build has successfully completed 16hours of P95 blend testing. Temps overnight peaked at 62C briefly. It seems that the thermal paste has set.

The only change I made in the bios was upping the "additional turbo voltage" to .008. Previously I had said that I upped it by .012 but I was wrong.

So today I dialed that setting back one notch to .004 and i'm running another blend test in P95. We'll see how that goes.

Easy overclock. I got a little nervous the first time around because others had such a consistently easy time with this processor I was nervous that my ram had issues when p95 got an error in it's first test.

Once this run of p95 finishes i'll start on the memory. See if we can wiggle a little something out of it.

I wouldn't even try the RAM. You will gain around 1 FPS on games with faster RAM, and if the RAM has an error when OCing it, you can make your WHOLE BIOS bad, resulting in a broken computer.
m
0
l
January 2, 2012 10:55:15 PM

HostileDonut said:
I wouldn't even try the RAM. You will gain around 1 FPS on games with faster RAM, and if the RAM has an error when OCing it, you can make your WHOLE BIOS bad, resulting in a broken computer.


Yep, you are correct in that the gains for a gaming PC are insignificant. But for encoding and activities such as that the gains are significant. I've seen plenty of benchmarks where an OC from 1600 to 1833 gave only a 1 FPS increase in games, which i'd imagine aren't even noticeable.

But the kind of work i'm doing the memory can hold me back. It's both memory and CPU intensive. In these benchmarks there is a significant increase in performance.

I've overclocked ram before. Never had any issues with the losing or corrupting the BIOS. Never have I heard of such a thing either. Not saying that it couldn't happen because i'm no technical genius. Just saying that for me it's worth what little risk is associated with trying to get an extra 200 mhz out of my ram.

We'll see how it goes. I've put together 3 other machines for the same purpose with success. I don't think it should be a problem.

I'm running another stress test because I found some other *** in this BIOS that I didn't understand so i've tuned back the voltages. We'll see. So far so good. Thanks for you help m8
m
0
l
a b K Overclocking
January 2, 2012 10:59:13 PM

Sounds good! :) 
m
0
l
January 5, 2012 7:55:22 PM

Just an update.

I don't like all these new OC features on Asus boards. Some of the biggest features seem redundant and totally unnecessary. Turbo? Additional turbo voltage, offset mode, XMP, by all cores and by per core. All of these things seem to be either unnecessary or counter act another process.

Per cores and all cores makes absolutely no difference in my tests. And it seems to not make any difference to anyone else either. Turbo voltage can be ramped up but offset/vdroop counteract that adjustment. Same could be sed for adjusting offset.

OC tuner and AI overclock tuner do the same thing. Except one actually works and the other does not.

All in all, I would say again that this is a good board. But there are loads of settings and options that are completely unnecessary and don't seem to serve the user in any beneficial way.

I ended up finishing this OC with a tune of 4.2 running at less than stock voltage. Essentially everything is disabled Turbo, offset, xmp, eist/speedstep, llc, vrm spread spectrum... all a waste of my time. Completely and totally. Mostly crap for kids.
m
0
l
a b K Overclocking
January 5, 2012 7:59:24 PM

All of those things do something. Obviously you don't get it....
m
0
l
January 5, 2012 8:06:45 PM

HostileDonut said:
All of those things do something. Obviously you don't get it....



Oh no, I get it completely. When LLC is counter acting the additional voltage set through turbo I understand that. When XMP is enabled i'm ensured my memory is running at the advertised speed.. but I can manually set it. Per cores and by all cores doesn't make any difference what-so-ever.. my tests and many other benchmarks prove that.

I do get it. It's completely and entirely unnecessary. I do get it though. I'm a pretty bright guy and have wasted much of the last week make sure that I did understand these things. And they are unnecessary and complicate the process of overclocking.
m
0
l
a b K Overclocking
January 5, 2012 8:30:40 PM

Push said:
Oh no, I get it completely. When LLC is counter acting the additional voltage set through turbo I understand that. When XMP is enabled i'm ensured my memory is running at the advertised speed.. but I can manually set it. Per cores and by all cores doesn't make any difference what-so-ever.. my tests and many other benchmarks prove that.

I do get it. It's completely and entirely unnecessary. I do get it though. I'm a pretty bright guy and have wasted much of the last week make sure that I did understand these things. And they are unnecessary and complicate the process of overclocking.

By disabling cores, you save power too. 4 cores vs. 1 core, the 1 core will use less power. XMP makes it so you don't have to manually set more memory. It is a luxury, it doesn't overclock... Of course you could manually set things! Why would they make it so that you could not manually set things with XMP?! And yes, your cores should all be pretty similar so that they all give at the same time, although, my core #2 is weaker than the rest of my cores. So I may be able to OC to 5ghz on cores 0,1, and 3, but only 4.8Ghz on cores 0,1,2, and 3. I don't know too much about LLC, so I can't say much on that, but they don't just through crap in to make people mad.... They put in extra stuff so people who really want to tweak a lot can do it as they please. Turbo voltage helps when turboing. It adds voltage, but what if you don't need 1.3v? You may only need 1.25v at some point, don't you want that voltage to be tuned down to 1.25v while it doesn't need the extra .05v?
m
0
l
January 5, 2012 9:44:13 PM

Quote:
I ended up finishing this OC with a tune of 4.2 running at less than stock voltage. Essentially everything is disabled Turbo, offset, xmp, eist/speedstep, llc, vrm spread spectrum... all a waste of my time. Completely and totally. Mostly crap for kids.


I am glad to see that you took my advice, of reducing the overclock slightly, It should make for a more stable system and better temperatures and give the processor a longer life. I think that 4.5 GIG on your processor is achievable, but not worth it in the long term as regards processor life and stability.
m
0
l
!