mersenneforum.org  

Go Back   mersenneforum.org > New To GIMPS? Start Here! > Information & Answers

Reply
 
Thread Tools
Old 2014-07-03, 16:25   #1
Pleco
 
Pleco's Avatar
 
Jul 2014

108 Posts
Default Prime95 and Haswell

Hi,

I have just built a new system with an Intel i7-4790K processor and Gigabyte Z97X-UD5H-BK motherboard. I was planning to run stability tests with Prime95, however I read on some postings on various forums saying that Prime95 does not go well with Haswell CPUs. Some said Prime95 will overvolt the CPU during the test and cause potential damage. Others recommended changing the vcore to manual in the BIOS before using Prime95. I have used Prime95 in the past without any issues and wondering if what they are saying is true? Can someone help shed some light on this please?

Thanks
Pleco is offline   Reply With Quote
Old 2014-07-03, 17:12   #2
henryzz
Just call me Henry
 
henryzz's Avatar
 
"David"
Sep 2007
Cambridge (GMT/BST)

133708 Posts
Default

When AVX instructions are used I am pretty certain a voltage boost(0.1v or 0.2v can't remember) is added automatically(for any program that uses those instructions not just prime95). If the voltage has been fiddled with already or the bios sets it high by default then this could mean that the voltage is too high. I would also expect increased heat due to the avx instructions.
henryzz is offline   Reply With Quote
Old 2014-07-03, 17:21   #3
TheMawn
 
TheMawn's Avatar
 
May 2013
East. Always East.

11·157 Posts
Default

I have heard of this, too. The CPU automatically volts itself up whenever AVX instructions are used. This is completely Intel specification; I stress that Prime95 has nothing to do with it.

If Prime95 burns out a stock settings CPU then Intel is to blame and this should be under warranty. If you've overclocked and increased the voltage already, then do be careful. However, I have heard of certain boards that will give you manual control over that overvoltage feature, to either lower or eliminate the offset.
TheMawn is offline   Reply With Quote
Old 2014-07-03, 20:17   #4
Pleco
 
Pleco's Avatar
 
Jul 2014

23 Posts
Default

I'm currently running on stock settings with an aftermarket heatsink (Cooler Master Hyper 212 EVO) and have no plans to overclock. I left the voltage settings on auto in the BIOS, Prime95 or any other AVX intense tasks shouldn't be able to push the voltage to dangerous levels right?

I'm also want to run some stress testing with Intel Extreme Tuning Utility, but I am not sure how long I should run it for optimum results. Is it the same as Prime95, minimum of 24 hours?

Thanks
Pleco is offline   Reply With Quote
Old 2014-07-03, 20:46   #5
henryzz
Just call me Henry
 
henryzz's Avatar
 
"David"
Sep 2007
Cambridge (GMT/BST)

23·3·5·72 Posts
Default

Quote:
Originally Posted by Pleco View Post
I'm currently running on stock settings with an aftermarket heatsink (Cooler Master Hyper 212 EVO) and have no plans to overclock. I left the voltage settings on auto in the BIOS, Prime95 or any other AVX intense tasks shouldn't be able to push the voltage to dangerous levels right?

I'm also want to run some stress testing with Intel Extreme Tuning Utility, but I am not sure how long I should run it for optimum results. Is it the same as Prime95, minimum of 24 hours?

Thanks
The length depends on how stable you want it. 24 hours would probably be fine for most people. If you are planning to be running prime95 further I would probably do 48 or more.
henryzz is offline   Reply With Quote
Old 2014-07-04, 01:10   #6
TheMawn
 
TheMawn's Avatar
 
May 2013
East. Always East.

11×157 Posts
Default

No, you should be fine. Remember, AVX instructions are not some mod hacked into the CPU by any third-part software. They are the most basic, low-level bits of mathematics that the CPU does, and they have been built into the CPU by Intel, the manufacturer.

The manufacturer set the CPU up to increase voltage whenever the AVX instructions are being used. It's not Prime95 saying "Yo, CPU, jack up da voltage," it's the CPU and Motherboard asking for more because the AVX instructions are being used.


Admittedly, Intel probably wasn't "expecting" that you would be using AVX instructions 24/7 but they would (or should) never allow a CPU to leave the factory if it couldn't survive 24/7 AVX instructions.

The people raising concerns are already overclocking and overvolting, and they are correct to be concerned. My CPU is at a nice 1.200V right now, doing 4.5 GHz. If I was a Haswell user and I didn't know any better, my CPU might be at 1.400V instead, and that's getting to be a bit high. 1.200V is not by any stretch a dangerous voltage either.

Back when I was doing 4.6 GHz, I was pushing 1.350V before I figured out I could manage it with 1.300V instead. 1.55V would be BAD for 24/7 usage, regardless of my cooling setup. That's where the concerns are. However, I can pretty much guarantee that at stock / AUTO settings, the extra you get from the AVX instructions will not be harmful.

Well, it won't be if your cooling is up to the task. The idea of Prime95 stress testing is to put your CPU under 100% load with the bonus of the CPU being hot, because it will be less stable at high heat. In a sense, the fact that the AVX instructions call in an extra 0.200V makes the CPU even hotter and more likely to de-stabilize.
TheMawn is offline   Reply With Quote
Old 2014-07-04, 02:01   #7
kracker
 
kracker's Avatar
 
"Mr. Meeseeks"
Jan 2012
California, USA

23·271 Posts
Default

BTW: AVX does not jack up the voltage in Haswell under adaptive. AVX2 and FMA3(what Prime95 has/uses) does.
kracker is offline   Reply With Quote
Old 2014-07-04, 12:26   #8
Pleco
 
Pleco's Avatar
 
Jul 2014

23 Posts
Default

I ran Prime95 and tried two different stress tests, Blend and FFT.

Blend test was normal, CPU core voltage was at 1.170V and core temperatures 62-67C.

FFT test on the other hand is what worries me, the second I run the test, the core temperatures instantly jump to 90-100C though the voltage was about the same as the Blend test. I stopped the test as fast as I could, not sure if the excessive temperatures caused any damage to my CPU.

Last fiddled with by Pleco on 2014-07-04 at 12:33
Pleco is offline   Reply With Quote
Old 2014-07-04, 12:40   #9
retina
Undefined
 
retina's Avatar
 
"The unspeakable one"
Jun 2006
My evil lair

22·1,549 Posts
Default

Quote:
Originally Posted by Pleco View Post
... not sure if the excessive temperatures caused any damage to my CPU.
I doubt it. You would need a lot more than that to cause permanent damage. The most likely outcome of overtemp will be a CPU shutdown. The CPUs have internal circuitry to shut it down when it gets too hot. The result is that you have to wait a few minutes for things to cool down before turning it on again.

Last fiddled with by retina on 2014-07-04 at 12:41
retina is offline   Reply With Quote
Old 2014-07-04, 16:30   #10
TheMawn
 
TheMawn's Avatar
 
May 2013
East. Always East.

11·157 Posts
Default

Quote:
Originally Posted by Pleco View Post
FFT test on the other hand is what worries me, the second I run the test, the core temperatures instantly jump to 90-100C though the voltage was about the same as the Blend test. I stopped the test as fast as I could, not sure if the excessive temperatures caused any damage to my CPU.
Now that is very surprising.

What temperature monitoring software do you use? Certain ones like the one I have also have a power draw monitor, which could be of interest to this discussion. If the CPU is drawing 120W or some crazy business then something is very strange.

Don't worry about CPU damage. Like Retina said, they're very good at protecting themselves. If it ran for a few months at 100C, that would be a different story. A couple of minutes, though, no worries.
TheMawn is offline   Reply With Quote
Old 2014-07-04, 16:59   #11
Pleco
 
Pleco's Avatar
 
Jul 2014

10002 Posts
Default

I was using three different softwares with temperature monitoring. All displayed similar results.

1)Intel Extreme Tuning Utility
2)Real Temp
3)HWMonitor

Unfortunately I didn't take note of the wattage the CPU was drawing. Once I get home I'll check the logs of XTU, hopefully it was recorded.
Pleco is offline   Reply With Quote
Reply



Similar Threads
Thread Thread Starter Forum Replies Last Post
18 core Haswell/P-1 CPU load nucleon Hardware 10 2015-04-26 14:48
Haswell-E Prelim. Benchmark sdbardwick Hardware 37 2015-02-10 18:49
Haswell Rig Mini-Geek Hardware 64 2014-05-27 13:22
LLR benchmarks for Haswell Unregistered Information & Answers 0 2013-06-16 21:48
Haswell New Instructions / AVX2 ixfd64 Hardware 72 2013-03-20 00:00

All times are UTC. The time now is 05:50.


Sat Jul 17 05:51:00 UTC 2021 up 50 days, 3:38, 1 user, load averages: 1.31, 1.55, 1.79

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.