![]() |
![]() |
#1 |
Mar 2014
3416 Posts |
![]()
New I7-7700K 4.2GHz system running Windows 10 arrived yesterday.
I set it to work on a doublecheck, and was seeing between 3 and 3.5 ms per iteration. (And was happy. Benchmark confirmed it was something like 30% faster than my previous fastest machine had been.) This morning it was slower than 5 ms per iteration. Various Windows updates, etc, had been applied, but no processes tying up the cpu showing in task manager or anything. I thought perhaps it had an affinity issue (running 4 threads on only 2 of the CPUs rather than spreading across all four.) Upgrading to version 29.1 only resulted in a very modest improvement, to about 4.8ms per iteration. If I manually adjust the number of threads, I get 8ms/iter with one, 5ms/iter with two, very little additional benefit from 3 or 4. I can't think of anything else that could have changed overnight. Any ideas for me? |
![]() |
![]() |
![]() |
#2 |
"/X\(‘-‘)/X\"
Jan 2013
47·67 Posts |
![]()
What is your RAM speed? What is your CPU temperature?
|
![]() |
![]() |
![]() |
#3 |
Mar 2014
22×13 Posts |
![]()
RAM is 16GB of DDR4 at 2400MHz.
CPUs are reporting 70°C. The CPU status report in CAM is reporting "core clock: 3990 MHz". Does this mean it is self-throttling rather than maintaining 4.2GHz? CPU fan speed is not reported, GPU fan speed is. Hm, wonder if that means there is actually a problem with the CPU fan, or if it's just not reported. |
![]() |
![]() |
![]() |
#4 | |
"/X\(‘-‘)/X\"
Jan 2013
47×67 Posts |
![]()
Have you confirmed it's running at that speed?
Quote:
|
|
![]() |
![]() |
![]() |
#5 |
P90 years forever!
Aug 2002
Yeehaw, FL
200616 Posts |
![]()
Is there some kind of Windows power save feature enabled?
|
![]() |
![]() |
![]() |
#6 |
Mar 2014
22·13 Posts |
![]()
The Windows power options *say* it is using 'active' management (speeding up the fan before slowing down the processor.)
An afternoon running the BIOS diagnostics (F12 at bootup) and then CAM and OpenHardwareMonitor was interesting. The BIOS says there is a "motherboard fan" and a "processor fan", which both pass their tests. But when the F12 diagnostics get to testing the processor itself, the processor fan doesn't speed up; the processor temperature levels out at 70C; and the processor test takes longer than expected. Monitoring in CAM, the GPU fan appears to really be a GPU fan (it responds to changing GPU load when I start and stop mfaktc, and is able to keep the GPU temp stable at 80C.) The processors... as far as I can tell are being regulated by changing the clock speed rather than anything else. If I run 1 or 2 threads in prime95, 1 or 2 processors rise to about 75C while the others remain cool, and the core clock stays at 4.2GHz. If I run 3 or 4 threads, the clock quickly pulls back to 3.9GHz. At this point it DOESNT look like a prime95 threading problem, but a cooling problem - which I (and my systems guru) am somewhat out of my depth to address. Edited to add for Mark: yes, CAM reports the memory is at 2395GHz. Last fiddled with by Siegmund on 2017-06-02 at 22:51 |
![]() |
![]() |
![]() |
#7 |
"Kieren"
Jul 2011
In My Own Galaxy!
2·3·1,693 Posts |
![]()
What does Task Manager show on the Performance tab when P95 has kind of 'settled in'. If hyperthreading is enabled, there should be 4 cores maxed out, with the other 4 relatively quiet if nothing else major is running. Those 4 cores should be either 0 or 1, 2 or 3, 4 or 5, 6 or 7.
I mostly have not had problems running 29.1 with all four threads doing one DC. The affinities seemed to stabilize fairly quickly on individual cores. When I tried running one DC on three cores, and P-1 on the other, there was some tendency for at least one thread to divide itself between the two hyperthreads. The fix for this is to put something like this in your local.txt: [Worker #1]I am currently running all for cores on one DC. The commented lines are things I have messed around with. |
![]() |
![]() |
![]() |
Thread Tools | |
![]() |
||||
Thread | Thread Starter | Forum | Replies | Last Post |
[Patch] CPU affinity prompt problem in mprime Linux / OS X build | Explorer09 | Software | 1 | 2017-03-01 02:34 |
WTF ??? P-1 slowdown | PageFault | Software | 2 | 2012-03-01 21:53 |
XP 32-bit to Win7 64-bit slowdown | willmore | Software | 6 | 2009-11-07 05:33 |
WinXP P95 slowdown | sgrupp | Software | 21 | 2003-12-28 18:48 |
laptop slowdown | nomadicus | Hardware | 7 | 2003-07-26 03:43 |