mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Hardware

Reply
 
Thread Tools
Old 2018-01-26, 04:48   #1
stalag
 
Jan 2018

910 Posts
Default settings for older processors

Hi all,


I have an Intel 4790K Haswell based processor. The latest version of Prime 95 I am using is Windows 65, Prime95,v.29.4.build 7. This is on a windows 7. It is a known issue that the Haswell processors run very hot, mostly due to the AVX instruction sets for the processors. Apparently to what I have read, the AVX and variant instruction sets were not meant in these processors to run flat out. Use of these is not recommended for extended periods of time. I applied the following CpuSupportsAVX2=0 line in the local.txt file. The documentation is so slim for adjusting Prime 95 in situations where the processor is cooking itself, that I am asking for someone who knows the program to write a proper primer as to what to disable in the instruction set to set the processor to run cooler. Later processors are more tolerant of use of the AVX instruction set due in part to better contact of the processor core to the lids over the processor core. I am running a H75 corsair AIO solution for my processor cooling, and near idle the processor runs about 35C. Under use of Prime 95 without the CPU supports instruction the processor runs at 88C. Use of the CPU supports instruction drops the temperature to a much safer level of around 68C. There needs to be a switch to use prime 95 to not exercise the CPU to such a degree that one may risk burning up the processor. Please consider this as this is an issue that needs addressed.
stalag is offline   Reply With Quote
Old 2018-01-26, 05:47   #2
Mysticial
 
Mysticial's Avatar
 
Sep 2016

7·47 Posts
Default

If your goal of running Prime95 is to join GIMPs and help search for primes, then my advice is to lower your overclock to that it can handle the AVX.

The throughput will be better with AVX at a lower clock than without it at a higher clock.

If your goal is to stress-test the system, then you're not stable unless it can handle everything that's thrown at it - including Prime95 AVX. Avoiding Prime95 AVX because it runs too hot is the same as lying to yourself. Prime95 isn't the only thing in the world that uses AVX.

A better way to avoid AVX is to disable it globally in the OS via bcdedit. That way no programs are allowed to use AVX and therefore the AVX cannot overheat your system.

Alternatively, if your BIOS supports it, you can set Tj.Max to something you're comfortable with. Then if a particular load pushes the temps up to that level, the processor throttles. That way you get the best of both worlds. This is what I do on my systems. And they will throttle under Prime95.

Last fiddled with by Mysticial on 2018-01-26 at 06:03
Mysticial is offline   Reply With Quote
Old 2018-01-26, 08:21   #3
axn
 
axn's Avatar
 
Jun 2003

7·23·29 Posts
Default

Quote:
Originally Posted by stalag View Post
Apparently to what I have read, the AVX and variant instruction sets were not meant in these processors to run flat out.
Then you have been lied to.
axn is online now   Reply With Quote
Old 2018-01-26, 09:53   #4
S485122
 
S485122's Avatar
 
Sep 2006
Brussels, Belgium

155010 Posts
Default

If your problem is not overclocking it might just be a cooling problme : it could be dust build up in a good cooling system or a bad or "undersized" cooler or bad fans or a bad conception of the computer case.)

I run a i7-4771 flat out since a few years and have no problems, the temperature of the different cores is around 50°C.

Jacob
S485122 is offline   Reply With Quote
Old 2018-01-26, 10:36   #5
LaurV
Romulan Interpreter
 
LaurV's Avatar
 
Jun 2011
Thailand

5×11×157 Posts
Default

Quote:
Originally Posted by stalag View Post
The documentation is so slim for adjusting Prime 95 in situations where the processor is cooking itself,
<snip>
There needs to be a switch to use prime 95 to not exercise the CPU to such a degree that one may risk burning up the processor. Please consider this as this is an issue that needs addressed.
There is no issue. Additional to what other people said above, the documentation for P95 (and any other "stress tester" program) is very clear about it: if your system is cooking itself, there is only one thing to do: solve your thermal issues. This can mean reapply the thermal paste, or pads, remove clogs of dust from the fans and heat sinks, de-lid (this are "extreme measures", not needed for your type of the CPU, that CPU has quite good internal cooling actually), use a cooler with higher dissipation (not always the most expensive and funny looking, with colored lights, etc. is the best), and/or reduce the clocks to more "normal" values. If your CPU consumes 100 watts, the only way to keep it cool is that the cooler you use dissipates 100 watts. If it dissipates only 80, then your processors feels like having a 20 watts fire under his ass...

On the other hand (related to the second paragraph) the CPU can not "burn up", they are very clever devices and will throttle down the clock to cool themselves when the temperatures get to the dangerous zone. In fact, I remember that Asus offered warranty for their systems, if you can fry it (by overclocking, or whatever, of course, without using a hammer, or changing the bios, etc, which causes you to lose the warranty) then they will change it for free. You should be worried more about the performance of your system getting down because of insufficient heat dissipation (i.e. getting hot), than about the fact that your CPU will burn.

And to add salt, this is from P95 readme file (stress.txt, last Q&A), this fragment I really like to post here and there... it reminds me years ago, when I was like you, believing my system is the best, and if some programs can crash it, then the programs were wrong...

Quote:
Q) A forum member said "Don't bother with prime95, it always pukes on me,
and my system is stable!. What do you make of that?"

or

"We had a server at work that ran for 2 MONTHS straight, without a reboot
I installed Prime95 on it and ran it - a couple minutes later I get an error.
You are going to tell me that the server wasn't stable?"

A) These users obviously do not subscribe to the 100% rock solid
school of thought. THEIR MACHINES DO HAVE HARDWARE PROBLEMS.
But since they are not presently running any programs that reveal
the hardware problem, the machines are quite stable. As long as
these machines never run a program that uncovers the hardware problem,
then the machines will continue to be stable.
LaurV is offline   Reply With Quote
Old 2018-01-26, 10:55   #6
heliosh
 
Oct 2017
++41

7D16 Posts
Default

Quote:
Originally Posted by stalag View Post
exercise the CPU to such a degree that one may risk burning up the processor.
There's no risk for that. Since many years CPUs throttle their clock and voltage when they run at their thermal limit, usually around 100°C. Running at this temperature doesn't harm the CPU.
heliosh is offline   Reply With Quote
Old 2018-01-26, 12:50   #7
science_man_88
 
science_man_88's Avatar
 
"Forget I exist"
Jul 2009
Dumbassville

26·131 Posts
Default

There was a THROTTLE= x to only use x% of processors i think. Not sure of the capitalization.
science_man_88 is offline   Reply With Quote
Old 2018-01-26, 18:12   #8
VictordeHolland
 
VictordeHolland's Avatar
 
"Victor de Hollander"
Aug 2011
the Netherlands

100100110002 Posts
Default

Reading the thread title, I thought he had something like a Pentium 4 or Core 2 Duo . By his standards my Sandy- and IvyBridge CPUs are ancient.

Modern CPUs are almost unbreakable, they'll throttle if they get too hot. You should check your cooling solution (thermal paste, dust, etc.). If that doesn't take care of the problem, you could lower your (OC/turbo) clocks. For instance running 3.7GHz with AVX2 enabled gives better throughput than 4.0GHz with AVX disabled.
VictordeHolland is offline   Reply With Quote
Old 2018-01-26, 22:11   #9
Mark Rose
 
Mark Rose's Avatar
 
"/X\(‘-‘)/X\"
Jan 2013
Ͳօɾօղէօ

5×563 Posts
Default

I still have a Sandy Bridge machine doing reliable DCLL work.
Mark Rose is offline   Reply With Quote
Old 2018-01-27, 04:17   #10
stalag
 
Jan 2018

32 Posts
Default

Quote:
Originally Posted by science_man_88 View Post
There was a THROTTLE= x to only use x% of processors i think. Not sure of the capitalization.
It is this switch I would like to use. I have already backed down some of my overclock. Also the H75 Corsair AIO cooler is a new in place cooler. I have also made sure enough cooling hits the motherboard to lower the temperatures. Without changing cases, I would love to find out how to lower the processor voltage by about 10% from the peak 1.38V to run at a consistent 1.32V. To do this I need some directions. My mother board is a Sabertooth Z97 Mark 2. The other solution was suggested by use of the throttle command from 100% usage of the 4 cores used to around 80%. I want to run prime 95, but not flat out on my hardware but yet distributed. Under Linux I know I could use the nice command, but I don't know the windows equivalent. Another option would be to use the Throttle command above but current documentation does not seem to exist. I would be more comfortable with temperatures in the 60C-70C range than the current 80C-90C range. No other program has taxed my system and forced me to implement better cooling solutions. If I knew how to tweak the settings I do have in bios successfully, I might be able to get the best solution to use the AVX code without driving temperatures into the 90C-100C range. Realize that this is far better point than what I am at before I started trying to tweak my system after the Microsoft Jan 4 2018 patch.
stalag is offline   Reply With Quote
Old 2018-01-29, 00:01   #11
kladner
 
kladner's Avatar
 
"Kieren"
Jul 2011
In My Own Galaxy!

32×1,103 Posts
Default

Quote:
I would love to find out how to lower the processor voltage by about 10% from the peak 1.38V to run at a consistent 1.32V. To do this I need some directions. My mother board is a Sabertooth Z97 Mark 2.
On a Sabertooth Z-170 Mk1 (like mine), you would go to the AI Tweaker tab in the BIOS. On this board, you scroll down to "CPU Core/Cache Voltage." Set the mode to Manual. This should show something like "CPU Voltage Override". Set the voltage down a couple of ticks, (right side + - keys). Save and restart. If you have problems restarting, hold down the power button until the machine shuts down. On my system, when you turn it back on, it will start in BIOS "safe mode" so you can change things.

It is hard to cover all possible details, and your board probably has differences from mine. Post back with what you encounter, and we can work from there.

EDIT: Don't expect the voltage you set in BIOS to be what you get under load in Windows. It will likely be higher, due to Load Line compensation. This takes a few reboots to learn the approximate relationship between the BIOS setting, and the "under load" voltage.

Last fiddled with by kladner on 2018-01-29 at 00:09
kladner is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Using older version of Prime95 ZFR PrimeNet 9 2015-04-25 07:03
new idea on an older one just can't find thread science_man_88 Miscellaneous Math 2 2011-01-01 18:18
Whither Older CPUs? Rodrigo Operation Billion Digits 4 2010-10-20 15:12
Older Linux Versions of mprime? jinydu Software 2 2009-08-15 16:56
older files archived? tom11784 Lone Mersenne Hunters 0 2004-05-24 12:02

All times are UTC. The time now is 11:49.

Mon Aug 10 11:49:28 UTC 2020 up 24 days, 7:36, 3 users, load averages: 2.28, 1.91, 1.65

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, 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.