A few questions about installed water cooler.

[Archer27]

So tonight I installed my water cooler(Lepa Aquachanger 240) and for my first liquid cooling install I think I did a decent job.

I removed the CPU from it's socket to clean the old paste off more effectively and am powering the pump directly from the power supply via Sata/Peripirhal to Molex .

Upon booting up after the change I got a few messages:

"Overclock failed"
"Asus anti-surge was triggered to protect system from unstable power supply"
"Keyboard"
"New Processor Installed"

After a reboot, these messages went away and haven't come back. Is it because I reseated the CPU and added the pump and 2 radiator fans to the electrical load? Is it normal? Should I be concerned?

Power supply is stable, voltages read what they are supposed to. VCCIN has actually increased from 1690 to 1740 at idle however actual core voltage remains roughly the same as before.

Ran Short FFT in Prime 95 for 10 mins, stable with core temps maxing out at 58c...much better than before.

 

[Darkbreeze]

Is your system plugged into a power strip/surge protector, or UPS? If so, unplug it, plug directly into the wall socket and see if it makes any difference. Fluctuations in power delivery and quality/noise/ripple/regulation can affect GPU card performance and everything else really. There have been a few occasions when I've seen low quality power strips or incompatible wave form UPSs cause some crazy little problems too.

 

[Archer27]

Is your system plugged into a power strip/surge protector, or UPS? If so, unplug it, plug directly into the wall socket and see if it makes any difference. Fluctuations in power delivery and quality/noise/ripple/regulation can affect GPU card performance and everything else really. There have been a few occasions when I've seen low quality power strips or incompatible wave form UPSs cause some crazy little problems too.

I'm plugged into a Belkin surge protector that has 5 outlets. It's actually a pretty old surge protector I got 5 years ago. I will try what you suggested.

BTW did you get my system specs?

 

[Archer27]

Alright, so tonight I tryed your suggestion(plugging direct into the wall) unfortunately there was no change at all.

But, I have absolutely confirmed without a doubt that the CPU is not throttling nor the GPU.

I did a world boss in GW2 tonight, which is incredibly CPU intensive as it has to render so many things in a ridiculously low amount of time and the fight lasted 10 minutes.

CPU usage didn't go above 75% and GPU usage didn't go above 65%

Highest individual core usage was 88%, 77%, 65%, 55% approximately.

FPS on the fight was 17-44fps and again, this is on Ultra 1080p

Between environment rendering, over 50 players on the screen, rendering everyone's attacks, player pets, the boss and its attacks, the portals and NPC's that spawn and the fact this game is horrible utilizing any more than 3 cores I think the adjustment I made in the BIOS earlier helped.

I disabled a setting that changes voltage for power conservation, I forget exactly what it was called.

Overall, I noticed a slight improvement in other games as well after the change.

Still, the fps stutter remains however...not as bad.

 

[Darkbreeze]

Is your control panel power profile set to performance?

CPU/GPU "usage" and "throttling" are not the same. Throttling is when the system reduces the voltage and multiplier because of thermal considerations, which may be based on sensors you cannot monitor but the system can and does. Not in every case, but sometimes.

Usage is simply how much of the estimated resources are being utilized at the time. Still, generally you'll see a reduction in max frequency when this happens, so the CPU would refuse to go above 2.5Ghz, for example, on a 3.5Ghz stock setting, which will of course reduce the overall performance. If you're not seeing a reduction in frequency/voltage during these dips, then it's not being throttled.

It might be a long shot, but sometimes FPS drops in online gameplay is network adapter related. What is the network adapter name in your device manager properties for the ethernet adapter, or are you using Wi-Fi?

 

[Archer27]

Is your control panel power profile set to performance?

CPU/GPU "usage" and "throttling" are not the same. Throttling is when the system reduces the voltage and multiplier because of thermal considerations, which may be based on sensors you cannot monitor but the system can and does. Not in every case, but sometimes.

Usage is simply how much of the estimated resources are being utilized at the time. Still, generally you'll see a reduction in max frequency when this happens, so the CPU would refuse to go above 2.5Ghz, for example, on a 3.5Ghz stock setting, which will of course reduce the overall performance. If you're not seeing a reduction in frequency/voltage during these dips, then it's not being throttled.

It might be a long shot, but sometimes FPS drops in online gameplay is network adapter related. What is the network adapter name in your device manager properties for the ethernet adapter, or are you using Wi-Fi?

1. Yes it is set to High Performance

2. I understand that, I have confirmed that voltages are great and frequency is constant.

3. With respect, I have mentioned at least 4 times in this thread that I'm on a wireless connection.
In the spec list you asked for, is the model of my wireless card.

 

[Darkbreeze]

Lots of gamers have wireless adapters, and don't use them when the system is at home. Just because it has wifi doesn't mean that's the primary connection type. My appologies for not absorbing the fact that you're on a wireless connection. I'm not certain that it's relevant anyhow, but again, sometimes network issues cause drops in frame rates. If the system doesn't get the data it needs, it can't very well process that data and therefore rates will drop correspondingly.

Is there any way you can try connecting via ethernet to see if the problem goes away, or is that not an option? You might try seeing if there is an updated driver available, however doubtful that might be. Given the fact that the ASUS website has been acting like crap for the last few days, you may or may not even be able to find that information. I think they are restructuring the support aspect of the product pages as I haven't been seeing normal behavior or support links on any motherboard or other product pages for the last couple of days regardless of what browser I use. Maybe you'll have more success, or it may not even be an issue if you already know you've manually installed the latest driver.

 

[Archer27]

Lots of gamers have wireless adapters, and don't use them when the system is at home. Just because it has wifi doesn't mean that's the primary connection type. My appologies for not absorbing the fact that you're on a wireless connection. I'm not certain that it's relevant anyhow, but again, sometimes network issues cause drops in frame rates. If the system doesn't get the data it needs, it can't very well process that data and therefore rates will drop correspondingly.

Is there any way you can try connecting via ethernet to see if the problem goes away, or is that not an option? You might try seeing if there is an updated driver available, however doubtful that might be. Given the fact that the ASUS website has been acting like crap for the last few days, you may or may not even be able to find that information. I think they are restructuring the support aspect of the product pages as I haven't been seeing normal behavior or support links on any motherboard or other product pages for the last couple of days regardless of what browser I use. Maybe you'll have more success, or it may not even be an issue if you already know you've manually installed the latest driver.

Unfortunately without adding an Ethernet outlet upstairs here, there is no way I can connect hardwire.

Ya I am having trouble with the ASUS site at the moment. The driver I am using is the same one that came with the card, so it may be out of date.

I've made a decision however, I am going to upgrade my CPU to a K model because I really want to overclock(hence why I got the Z97 board) I was planning such an upgrade later this year, but due to this issue, I feel it may be worth it.

It will accomplish more goals than one. I will get to pull the cooler and inspect the paste and check the torque on the block bolts as well.

Getting an I5 4690k. Will update when installed.

 

[Archer27]

Vouched for something a little more extreme. Wish me luck!

 

[Darkbreeze]

That should offer somewhat better performance, but I'm not sure it's the solution. Your current chip had very similar single core performance and it would take an extremely optimized title, optimized for a high number of threads/cores, to offer any advantage due to the additional hyperthreads on the 4790k.

The faster base clock on the 4790k might be the difference maker though. I'm skeptical that this is the solution, but maybe, and it certainly can't hurt.

 

[Archer27]

That should offer somewhat better performance, but I'm not sure it's the solution. Your current chip had very similar single core performance and it would take an extremely optimized title, optimized for a high number of threads/cores, to offer any advantage due to the additional hyperthreads on the 4790k.

The faster base clock on the 4790k might be the difference maker though. I'm skeptical that this is the solution, but maybe, and it certainly can't hurt.

Installed! I have also overclocked it already to 4.6ghz stable(I have run multiple stress tests)

This runs hotter than my 4460 but I assume it's due to the higher frequency.

My Vcore is at 1.197. I assume this is a good thing?

Will test in a few mins on some games!

 

[Darkbreeze]

How long did you stress test?

Prime95 version 26.6 should be run for 15 minutes to determine thermal compliance during each step of the process, and stability should be tested by running for no less than 15 hrs. Both should be done using Small FFT on Prime95 version 26.6. No other version should be used due to the use of AVX instructions on newer versions, which creates unrealistic thermal conditions.

http://windows-downloads-center.blogspot.com/2011/04/prime95-266.html


I'd go out on a limb here and say I could almost guarantee your voltage is too low to be stable. Most Devil's canyon chips need between 1.2 and 1.3v to be stable at 4.5-4.6Ghz, and 1.25 to 1.29 is the most common range.

Some require more. The need for less would be rare in the extreme. Sure, it will run, but running and running stable are entirely different things. And running Intel burn test or some other quick utility is not going to tell you the system is stable. The best method is to gradually, incrementally determine the overclock, because every chip is different and has different characteristics.

This covers most of the basic tenets of overclocking:

http://www.overclock.net/t/91/ultimate-overclocking-guide


And this is specific to Haswell cpus.:

http://www.overclock.net/t/1411077/haswell-overclocking-guide-with-statistics


Here are some common results. Scroll down to the bottom of the spreadsheet for Devils canyon specific results:

https://docs.google.com/spreadsheets/d/1JfvjG0xKvQxUY4z7CNr_pCMhzH0Y3YaiYLWWJJY9Opw/edit#gid=0

 

[Archer27]

How long did you stress test?

Prime95 version 26.6 should be run for 15 minutes to determine thermal compliance during each step of the process, and stability should be tested by running for no less than 15 hrs. Both should be done using Small FFT on Prime95 version 26.6. No other version should be used due to the use of AVX instructions on newer versions, which creates unrealistic thermal conditions.

http://windows-downloads-center.blogspot.com/2011/04/prime95-266.html


I'd go out on a limb here and say I could almost guarantee your voltage is too low to be stable. Most Devil's canyon chips need between 1.2 and 1.3v to be stable at 4.5-4.6Ghz, and 1.25 to 1.29 is the most common range.

Some require more. The need for less would be rare in the extreme. Sure, it will run, but running and running stable are entirely different things. And running Intel burn test or some other quick utility is not going to tell you the system is stable. The best method is to gradually, incrementally determine the overclock, because every chip is different and has different characteristics.

This covers most of the basic tenets of overclocking:

http://www.overclock.net/t/91/ultimate-overclocking-guide


And this is specific to Haswell cpus.:

http://www.overclock.net/t/1411077/haswell-overclocking-guide-with-statistics


Here are some common results. Scroll down to the bottom of the spreadsheet for Devils canyon specific results:

https://docs.google.com/spreadsheets/d/1JfvjG0xKvQxUY4z7CNr_pCMhzH0Y3YaiYLWWJJY9Opw/edit#gid=0

I should have asked you before I started the world boss....I BSOD.
I rejoined the fight after increasing the voltage to 1.25 and was able to finish it.
I should have tested with Prime 95 longer than 10 mins...
Running Prime 95 again Short FFT on version 26.6 just like last time at the new voltage.
While the test is running, I will read those articles.

 

[Darkbreeze]

Everything you need to know is in there. Pretty much. It's no substitute for experience, but you have to start somewhere and those are a very good place to start.

 

[Archer27]

Everything you need to know is in there. Pretty much. It's no substitute for experience, but you have to start somewhere and those are a very good place to start.

Very informative indeed.
Approaching 1 hour stable.

Is it really necessary to run for 15 hours? I won't be able to sleep with the fans blazing away lol.

 

[Darkbreeze]

24hrs is what's actually recommended. I've seen dozens of overclocked systems error out at between 15-24hrs. I've pretty much never seen one error out after 24hrs. 15hrs is enough to pretty well guarantee that any instability there might potentially be is so minimal though, that you'd probably never see it occur in real world use, only when running prime numbers or a similar high difficulty calculation cycle.

Systems that fold for example, can't afford to have ANY kind of instability. Instabilities introduce micro-errors into the data that is handled, and while that might not be noticeable or apparent at first, over time it can corrupt your OS or other data to the point where it becomes unsalvageable. That would take some time, but it can happen. The more serious the problem, the quicker it can occur. Mostly stability will affect actual performance though, with errors causing restarts and blue screens, but that's only when something is clearly not configured correctly.

You don't necessarily have to do the full run right now. If it's still stable by bedtime, call it a day and do the longer run later when you can let it go and come back to check it later.

 

[Archer27]

24hrs is what's actually recommended. I've seen dozens of overclocked systems error out at between 15-24hrs. I've pretty much never seen one error out after 24hrs. 15hrs is enough to pretty well guarantee that any instability there might potentially be is so minimal though, that you'd probably never see it occur in real world use, only when running prime numbers or a similar high difficulty calculation cycle.

Systems that fold for example, can't afford to have ANY kind of instability. Instabilities introduce micro-errors into the data that is handled, and while that might not be noticeable or apparent at first, over time it can corrupt your OS or other data to the point where it becomes unsalvageable. That would take some time, but it can happen. The more serious the problem, the quicker it can occur. Mostly stability will affect actual performance though, with errors causing restarts and blue screens, but that's only when something is clearly not configured correctly.

You don't necessarily have to do the full run right now. If it's still stable by bedtime, call it a day and do the longer run later when you can let it go and come back to check it later.

Ah I see. Wow I'm learning alot from all this and it all started with a watercooler installation lol.
We are at 3 hours stable now.
Do you think the BSOD while gaming was due to the low voltage OC? The first Prime(10min run) and Intel Burn ran fine without BSOD at the voltage.

I'll run it for the full 15 hours tomorrow. Thanks for your patience with me Darkbreeze!(btw the Asus site is still being a pain)

 

[techgeek]

The true stability test is to use a wide range of loads / load types. I've seen guys run Prime 95 for days stable, then they run BF4 or some other game and it crashes / BSOD's in 10 minutes. Or just browsing in Chrome and watching something in Flash, and BSOD. These are usually the same guys that will argue with you up and down that their system is stable and refuse to return their system back to stock to even try and see if the game crashes continue.

Prime 95 is a great stress test for thermal limits (ie it gives you a good idea if your cooling is sufficient), but it may not tell you if you are 24 / 7 stable under average loads. Realbench by ASUS is a good stress tester, variable workloads. If you are using XMP memory (memory overclocking), Hyper / Super Pi is a good stress test for RAM. CineBench is another good stress test for both CPU and GPU (moreso for CPU though). Of course gaming is a good stress test and if your systems primary use is gaming, then playing games is great to test stability.

Really the best is a variety of workloads. If you stick to just one stress test, then you won't have the entire picture. I'm not saying that you can't use one to dial in a max OC, but don't leave off with just one stress test when you've reached your goal OC.

 

[Darkbreeze]

True enough. But for the primary configuration to determine whether the system has probable stability, the prime number method is almost unimpeachable. There is no system that won't crack under some kind of pressure. And every CPU known has some form of weakness. Many of them have known faults that have never been given microcode patches, or none exists to give them due to the problem being too central to the core architecture.

So once you've determined the system is primarily stable, via Prime, then other stress tests are of course worth trying out to see if there are any lingering general concerns. Trying to worry about ABSOLUTE stability is a snipe hunt though because there is no such thing. I usually say take your overclock to the highest you can stabilize it at and still remain below thermal limits, primarily, verify stability using Prime and whatever else you wish to use, and then back it off by 100mhz via multiplier if it's going to be a daily driver that you want to last a good long while.

If you can get it primarily stable at 4.6Ghz at whatever voltage, and it passes a 15hr Prime run and whatever else you want to throw at it, then back it off to 4.5Ghz and it should be essentially rock solid so long as you are thermally comfortable.

Everybody has a preferred methodology though. To each their own.

 

[techgeek]

True enough. But for the primary configuration to determine whether the system has probable stability, the prime number method is almost unimpeachable. There is no system that won't crack under some kind of pressure. And every CPU known has some form of weakness. Many of them have known faults that have never been given microcode patches, or none exists to give them due to the problem being too central to the core architecture.

So once you've determined the system is primarily stable, via Prime, then other stress tests are of course worth trying out to see if there are any lingering general concerns. Trying to worry about ABSOLUTE stability is a snipe hunt though because there is no such thing. I usually say take your overclock to the highest you can stabilize it at and still remain below thermal limits, primarily, verify stability using Prime and whatever else you wish to use, and then back it off by 100mhz via multiplier if it's going to be a daily driver that you want to last a good long while.

If you can get it primarily stable at 4.6Ghz at whatever voltage, and it passes a 15hr Prime run and whatever else you want to throw at it, then back it off to 4.5Ghz and it should be essentially rock solid so long as you are thermally comfortable.

Everybody has a preferred methodology though. To each their own.

Don't get me wrong, I wasn't casting doubt on Prime 95's usefulness. I was just pointing out that a lot of people think that stability testing starts and ends with Prime 95 and this approach is completely wrong. Prime 95 is great at checking thermal headroom and a singular type of workload. Many people mistakenly think that if they are Prime stable, then they are stable, and that's simply not always the case. Like I said at the end of my post, it's good for dialling in a final OC, but once you've reached your proposed final overclock, then more thorough stability testing starts.

 

[Archer27]

Just got home, Prime 95 is still running on all 8 threads 7 hours strong. Maximum temps 68-77
Looking good so far guys!

 

[Darkbreeze]

You're probably in the clear then, but I'd let it finish up the full 15hrs and then proceed from there. Probably you're on the right track though.

 

[Archer27]

You're probably in the clear then, but I'd let it finish up the full 15hrs and then proceed from there. Probably you're on the right track though.

Glad to hear it. At 12 hours now, still running strong. It would seem some of the more extreme tests have pushed the maximum temps recorded up slightly. Max temps are 77, 76, 78, and 70. Assuming this is normal

 

[Darkbreeze]

While running full 100% steady state stress tests anything below 80°C is desirable. Anything below 85°C is acceptable. Anything below 90°C PROBABLY won't damage anything. Anything over 90°C is bad. Anything over 100°C is probably causing damage or throttling the system down to where it can't cause damage.


Everything you want to know about it is here:

http://www.tomshardware.com/forum/id-1808604/intel-temperature-guide.html

 

[Archer27]

While running full 100% steady state stress tests anything below 80°C is desirable. Anything below 85°C is acceptable. Anything below 90°C PROBABLY won't damage anything. Anything over 90°C is bad. Anything over 100°C is probably causing damage or throttling the system down to where it can't cause damage.


Everything you want to know about it is here:

http://www.tomshardware.com/forum/id-1808604/intel-temperature-guide.html

Thanks for the article, again very informative.
So it would seem I'm in the desirable range then. Nearing the 14 hour mark stable. I guess the 1.20 volts I was running before when the game BSOD on me was too little for 4.6 as you said. Upping it to the 1.25 made all the difference it seems.

 

[Archer27]

I have noticed something odd.

Core temps are in the 70s and the "CPU temp" is 43-45c.(Motherboard temp is 19-23c)

Why is there such a big difference?

(btw I'm using realtemp for core temps and AI suite 3 for mobo and CPU temp)