![]() |
|
|
#298 | |
|
Jan 2003
3138 Posts |
Quote:
If we use faster memory to relieve the memory bottleneck, the CPU may also be stressed more and in turn that may affect stability again. |
|
|
|
|
|
|
#299 |
|
May 2013
East. Always East.
11·157 Posts |
If you're having problems with the memory, look into increasing the memory controller's juice. For Ivy Bridge, anyway, that's the VCCSA, mostly, and also the VCCIO. Look them up for your particular hardware, first, as it's been known that there is a fine line.
|
|
|
|
|
|
#300 |
|
"Serge"
Mar 2008
Phi(4,2^7658614+1)/2
251916 Posts |
I think I finally got my system stable with a (not so recommended) VCCSA=1.22v, and all 2400MHz XMP specs intact (10-12-12-31-2T), + CPU Freq 4200 + air cooling (CMaster TPC 612). But this is probably still not over.
|
|
|
|
|
|
#302 |
|
"/X\(‘-‘)/X\"
Jan 2013
2·5·293 Posts |
I have a Haswell (i7-4770 CPU @ 3.40GHz, hyperthreading enabled), but only 1600 MHz memory. I am considering running 28.1.
I realize I haven't got the memory bandwidth to even fully utilize the 4 cores. I've tried setting NumCPUs=3 in prime.txt, but mprime always queues work and runs 4 processes. Is there a proper way to get mprime to only use 3 cores? If I can keep a core free I'll turn off hyperthreading. |
|
|
|
|
|
#304 | |
|
P90 years forever!
Aug 2002
Yeehaw, FL
19×397 Posts |
Quote:
|
|
|
|
|
|
|
#305 | |
|
"Oliver"
Mar 2005
Germany
11×101 Posts |
Yes, so far no issues. The difference is too big for just 100 or 200MHz difference when you compare 27.9 to 28.1. This system isn't doing some real work, just testing, etc. so far. But my feeling (and the fact, that Linpack still crashes much faster than v28.1 on higher clocks) tells me, that 4000MHz is OK for my system.
Quote:
Oliver |
|
|
|
|
|
|
#306 |
|
"/X\(‘-‘)/X\"
Jan 2013
2×5×293 Posts |
Putting NumCPUs in local.txt worked perfectly. Thanks!
I noticed that I only loose 15% performance per core when I go from 3 to 4 cores. So I still get a 13% improvement running four cores with 1600 MHz memory using 27. I'll test with 28 later. Maybe there will be no benefit to the 4th core then. Last fiddled with by Mark Rose on 2013-09-13 at 21:59 |
|
|
|
|
|
#307 |
|
"/X\(‘-‘)/X\"
Jan 2013
2·5·293 Posts |
Okay, so I got a noticeable improvement from 28.1. And I've overclocked my RAM to 2000 MHz (and increased the timings from 9's to 10's) and I seem to be getting near the max output from all four cores now. Awesome.
|
|
|
|
|
|
#308 |
|
P90 years forever!
Aug 2002
Yeehaw, FL
19×397 Posts |
I'm upgrading my i7-860 computer by replacing the motherboard, CPU, and memory. The CPU is the Haswell 4670K, memory is DDR3-2400.
I recycled the i7-860 CPU cooler: http://www.newegg.com/Product/Produc...82E16835103064 Right now, with CPU at 3.6GHz, voltage at 1.08 (I can't turn off adaptive voltage, so this should mean 1.18V running AVX instructions), and memory at DDR3-2400, I'm getting temps of 90C running prime95 (up to 100C running the small torture test). Does this sound right? Tomorrow, I'll try either 1) reapplying the thermal compound, or 2) using the stock CPU cooler, or both to see if temperatures get any better. I had hoped to get this running at least somewhat close to my first Haswell box running at 4.1GHz, 1.2V (adaptive voltage off), DDR3-2400 and temps of around 80C. |
|
|
|
![]() |
| Thread Tools | |
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| Haswell-E Prelim. Benchmark | sdbardwick | Hardware | 37 | 2015-02-10 18:49 |
| Prime95 and Haswell | Pleco | Information & Answers | 22 | 2014-07-13 16:03 |
| Haswell Rig | Mini-Geek | Hardware | 64 | 2014-05-27 13:22 |
| Prime95 version 27.1 early preview, not-even-close-to-beta release | Prime95 | Software | 126 | 2012-02-09 16:17 |
| Missing mouse-over preview text | retina | Forum Feedback | 1 | 2011-09-12 15:32 |