| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: Mapping Cancer Markers Forum
Thread: i5 faster than i7 ? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 11
|
|
| Author |
|
|
EddyT
Cruncher Germany Joined: Oct 17, 2013 Post Count: 9 Status: Offline Project Badges: 
|
Hey there Crunchers,
----------------------------------------my problem is that my i7 CPU takes about ~5-7hrs for one WU to finish while the i5 just needs ~3-4hrs. i5 is clocked at 3,6GHz and the i7 at 3,2GHz. This should not do much of a difference.. i7 is an i7-870 and the i5 an i5-4670. Task-Manager shows me for the i7 that every WU is running at 12% and on the i5 25% since the i7 is an 4 Core 8 Thread and the i5 a 4 C 4 T. What can I do to improve my i7? Thanks for answers Cheers Eddy  |
|||
|
||||
|
deltavee
Ace Cruncher Texas Hill Country Joined: Nov 17, 2004 Post Count: 4894 Status: Offline Project Badges: 
|
What can I do to improve my i7? What's to improve? The faster newer processor is outperforming a slower one with hyperthreading. If what you want to improve is the "hrs for one WU to finish" then turn off the i7 hyperthreading and it will complete WUs much faster. If you want maximum points or runtime then you should probably leave it the way it is. Try it and see for yourself. |
||
|
|
Dayle Diamond
Senior Cruncher Joined: Jan 31, 2013 Post Count: 452 Status: Offline Project Badges: 
|
Deltavee is right, hyperthreading helps processors do much more work. It's true that your i7 from seven years ago can do eight tasks at once. That said, processor speeds (in GHz) should only be compared within a manufacturing generation, and your i5 is three years old.
A lot has happened since then an a upgrade would allow you to double your contribution for the same amount of electricity. |
||
|
|
AMuthig
Advanced Cruncher USA Joined: Nov 30, 2013 Post Count: 59 Status: Offline Project Badges: 
|
You can use this site as a rough guide to determine CPU performance. In my experience, the benchmark measurements closely correspond to the amount of WCG work you will complete. http://www.cpubenchmark.net/
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
This thread should be moved to the hardware forum...
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Actually, clock speed is an ok way to measure how quickly a single task will complete. I have a E6600 overclocked to 3.2Ghz and it does MCM tasks in 3-4hrs.
Modern processors will do things in a more energy efficient manner and add new features, but a 3.2ghz processor is a still a 3.2ghz processor. Have you got the same bonic profile applied to each machine? |
||
|
|
kva.pl
Cruncher Joined: Aug 2, 2014 Post Count: 38 Status: Offline Project Badges:
|
Actually, clock speed is an ok way to measure how quickly a single task will complete. I have a E6600 overclocked to 3.2Ghz and it does MCM tasks in 3-4hrs. Modern processors will do things in a more energy efficient manner and add new features, but a 3.2ghz processor is a still a 3.2ghz processor. What? Pentium G3258 with 3,2Ghz needs less than 2,5h for the same job. So no, mhz is relevant only with the same architekture. With different CPU cores it means almost nothing. Another thing is that modern CPU will be not only faster but also have much lower power draw. Crunching with old CPU is simply inefficient. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
kva, your example is not representative of the MCM workload.
Look here: https://secure.worldcommunitygrid.org/forums/...d_thread,36119_offset,740 The average time taken to complete a MCM unit is 3.9 hrs this week. It has been extremely stable for the last few months. Also, look at your results status page - you'll see that your wingman almost always completes tasks in 3.5-4.5 hrs. There would be a much wider range if processor architecture made a significant difference for task completion time. My ancient processor (E6600) is completing tasks in about the average time. Generally, you're correct; you can't use Ghz to compare processors of different architectures. Yes, you're completely correct about the power efficiency. But please note that I'm talking about the very specific example of completing a single MCM task. By definition, the clock speed is how long it takes to execute an instruction and as far as I understand, the tasks don't take advantage of modern instruction sets in order to maintain backwards compatibility. So, in this specific case, Ghz is a OK approximation. |
||
|
|
kva.pl
Cruncher Joined: Aug 2, 2014 Post Count: 38 Status: Offline Project Badges:
|
Pentium Haswell @3,2Ghz - less than 2,5h per MCM WU.
Pentium E6600 @3,2Ghz - 3-4h per MCM WU. Q.E.D. I have no idea what avg time from desktops, laptops, servers, small factor PCs etc could prove here. "By definition, the clock speed is how long it takes to execute an instruction" It is almost true. But there is also IPC, branch prediction, cache and other factors playing their roles. "as far as I understand, the tasks don't take advantage of modern instruction sets in order to maintain backwards compatibility" No, there are different apps optimized for different sets of extensions. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Read carefully: Try naming 3 sciences over the history of WCG that were optimized in such a way that it led to a subset of x86 CPUs by design being excluded from participating. You'll be challenged to name 2, as there is ever only been mentioned 1, and that one completes right now.
QED? Are you comparing CPUs with and without hyperthreading? E6600 has not, the Haswell has. Comparing same OS platform at that? https://www.pugetsystems.com/labs/hpc/Hyper-T...Parallel-Performance-578/ |
||
|
|
|