| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: OpenPandemics - COVID-19 Project
Thread: Power Consumption |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 14
|
|
| Author |
|
|
highwind
Advanced Cruncher Joined: Aug 22, 2009 Post Count: 56 Status: Offline Project Badges: 
|
I am a big crunch for MCM and have recently switched to COVID-19 just to help out the pandemic.
I am having a few machines running: 1. i7-7700HQ 2. i7-8565U 3. Ryzen 9 3900X 4. ThreadRipper 3960X 5. Ryzen 7 2700X I am always running all CPU 24/7 100%. I notice that COVID-19 project draws about 10% less power that MCM and the heat generated from COVID-19 is about 4 degree less at full load. Anyone else see this happens as well? It looks like MCM is quite powerful hungry or COVID-19 is more energy efficient. |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1684 Status: Offline Project Badges: 
|
I notice that COVID-19 project draws about 10% less power that MCM and the heat generated from COVID-19 is about 4 degree less at full load. And OPN1 is bringing at least 10% less credit than MCM on Linux (Windows seems to be less impacted). I discovered the situation last night and I planed to make a deeper analysis prior to share my observations with the Community. I do not measure the effective electricity power consumption and CPU temperature is partially related to the room temperature, and it became fresher during the last days by us. But your observations fit very well with mine. Cheers, Yves --- PS: Maybe adriverhoef can confirm with his extensive data collection. |
||
|
|
hchc
Veteran Cruncher USA Joined: Aug 15, 2006 Post Count: 865 Status: Offline Project Badges: 
|
Could've sworn that MCM's application takes advantage of Advanced Vector Extensions AVX instruction set. Not sure if it's AVX2 or just AVX. This typically uses more current and generates more heat.
----------------------------------------
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Doubt it since there's just single 32/64 bit compiles that need to work for the lowest common denominator. I.e. when 2010 CPU meets 2020 CPU they still should be able to have a bit for bit matching comparison.
|
||
|
|
William Albert
Cruncher Joined: Apr 5, 2020 Post Count: 41 Status: Offline Project Badges:
|
Doubt it since there's just single 32/64 bit compiles that need to work for the lowest common denominator. I.e. when 2010 CPU meets 2020 CPU they still should be able to have a bit for bit matching comparison. An executable needs to be explicitly compiled for a 32- or 64-bit architecture. For architectural extensions like AVX, the program can test at runtime if a set of extensions is present, and can use separate code paths (e.g., AVX2, AVX, SSE4, SSE3) depending on what extension is available. |
||
|
|
Falconet
Master Cruncher Portugal Joined: Mar 9, 2009 Post Count: 3315 Status: Offline Project Badges:
|
Yes, I noticed lower power consumption too.
----------------------------------------I posted it on another thread last week but I think my Ryzen 1400 dropped about 6 watts.  - AMD Ryzen 5 1600AF 6C/12T 3.2 GHz - 85W - AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W - AMD Ryzen 7 7730U 8C/16T 3.0 GHz |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Doubt it since there's just single 32/64 bit compiles that need to work for the lowest common denominator. I.e. when 2010 CPU meets 2020 CPU they still should be able to have a bit for bit matching comparison. An executable needs to be explicitly compiled for a 32- or 64-bit architecture. For architectural extensions like AVX, the program can test at runtime if a set of extensions is present, and can use separate code paths (e.g., AVX2, AVX, SSE4, SSE3) depending on what extension is available. All well and nice, but in a quorum 2 you cant have these extensions used by 1 and not the other. Homogeneity of the whole database of results is critical. More here: https://boinc.berkeley.edu/trac/wiki/HomogeneousRedundancy Don't think WCG uses HR beyond the general platforms they support, Windows, Linux, OSX, Android Arm, Linux Arm. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I notice that COVID-19 project draws about 10% less power that MCM and the heat generated from COVID-19 is about 4 degree less at full load. And OPN1 is bringing at least 10% less credit than MCM on Linux (Windows seems to be less impacted). I discovered the situation last night and I planed to make a deeper analysis prior to share my observations with the Community. I do not measure the effective electricity power consumption and CPU temperature is partially related to the room temperature, and it became fresher during the last days by us. But your observations fit very well with mine. Cheers, Yves --- PS: Maybe adriverhoef can confirm with his extensive data collection. And MIP creates even less heat. |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1684 Status: Offline Project Badges:
|
And MIP creates even less heat. We know the reason: > By MIP1, the CPU waits for the cache. It is not the case for OPN1. Additionally on my side, the Linux machines are mostly impacted; for example: > i7 4770K with Windows 7 Pro SP1 x64: 33.47 points /core /hour > i7 6700K with Linux Mint 18.3 x64 (Ubuntu 16.04): 34.22 points /core /hour Usually the i7 6700K brings about 30% more than the i7 4770K. For the AMD machines (Phenom II x6, Ryzen 7 2700) - the 4 machines run Ubuntu or Linux Mint - all machines experience about 20% point decrease with OPN1 (not considering SCC1). At this time, I don't have any idea if it is only a question of crediting or if the machines are really performing less good with OPN1 on Linux. @WCG-Tech: do you notice a similar figure on your own machines? Cheers, Yves ---------------------------------------- [Edit 1 times, last edit by KerSamson at May 25, 2020 2:28:09 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
And MIP creates even less heat. We know the reason: > By MIP1, the CPU waits for the cache. It is not the case for OPN1. Additionally on my side, the Linux machines are mostly impacted; for example: > i7 4770K with Windows 7 Pro SP1 x64: 33.47 points /core /hour > i7 6700K with Linux Mint 18.3 x64 (Ubuntu 16.04): 34.22 points /core /hour Usually the i7 6700K brings about 30% more than the i7 4770K. For the AMD machines (Phenom II x6, Ryzen 7 2700) - the 4 machines run Ubuntu or Linux Mint - all machines experience about 20% point decrease with OPN1 (not considering SCC1). At this time, I don't have any idea if it is only a question of crediting or if the machines are really performing less good with OPN1 on Linux. @WCG-Tech: do you notice a similar figure on your own machines? Cheers, Yves I have tried OPN on an i5-3210M and a FX-8320, both with Win10 and Linux Mint 19.3 (based on Ubuntu 18.04) kernel 5.3 There is very little difference in run time and points for the FX-8320 between Win10 and Mint, yet the i5 mobile CPU does run better under Mint, slightly shorter processing times and better points. However, Linux isn't all equal, have a look here: https://www.phoronix.com/scan.php?page=articl...ng-2020-distros&num=8 The benchmarks of different Linux distributions shows that there are differences in performance between different Distros. The guys at phoronix.com have done several such benchmark comparisons between different distros over the years, maybe Ubuntu 16.04 isn't the right base anymore? |
||
|
|
|