Well productivity wise whatever beautiful theories you want to say the results/numbers beating all your words. I am showing the solid results but you and few other religious hard liners are sticking to theories rather than facts. So you'd better stick with your theories and I'd better stick with my results. Since E@H cannot provide fine benchmark list to cpu's.
Well productivity wise whatever beautiful theories you want to say the results/numbers beating all your words. I am showing the solid results but you and few other religious hard liners are sticking to theories rather than facts. So you'd better stick with your theories and I'd better stick with my results. Since E@H cannot provide fine benchmark list to cpu's.
'religious hardliners' ?? :o
Um, Orgil ..... archae86 is a retired Intel electronic engineer ( chip design, fabrication ... ) a midwife so to speak from many of the earlier x86 designs onwards ( see the gag in his sig? ). He's probably forgotten more than we'll ever collectively know. His CV is truly amazing and we are pleased to have his insights expressed here.
[ We can tell that you've totally mis-understood the HT issue, and can't bring yourself to admit that. But maybe, just maybe, you might want to review your response to one of the world's leading chip designers? ]
Your post's wording is close to objectionable flaming and ad-hominem ( ie. deletion ), so can we stay civil on this one? :-)
Cheers, Mike.
( edit ) For instance see here where Peter Stoll is mentioned ( note 12 ) as a designer. Or here. Or here. Or .....
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
To drift off-thread for one moment, Orgil, noting your listed country and your team name, it might interest you to know I attended a few of the events for this year's Naadam. Opening ceremony, some wrestling, one horse race, archery, and a little ankle-bone shooting. Some of the pictures I took are here, though the Naadam pictures are in subdirectories under it and the maps and overhead imagery are not my photos, though they do display track log data from the GPSr I carried.
Yes moderator, 'tis off-topic, and I'll cheerfully submit to deletion.
Whole point is I simply offered result numbers but you all speaking some super theory that ignoring my numbers and according to your theories I must be wrong and you must right something. So out of some wondering/desperation I am concluding that your points are not much different than how religious hardliners behave. The whole point is in this case traditionally believed thing of mobile per core productivity is slightly higher than desktop cpu core and even server cpu core. If non oc'd mobile cpu core is outdoing non oc'd server cpu core what is not so much acceptable for computing culture or religion...
In non oc'd original case most xeons doing 29k or 30k's versus mostly mobile core i5's doing 22'k or less that these numbers represent clear productivity difference.
ps. thanks for your naadam things that looks cool but well that is off topic.
Your figures may be hard numbers, but they are resting on quicksand. I don't need a complicated theory to see that. At this moment I don't trust them to support my weight.
"For each processor core that is physically present, the operating system addresses two virtual processors, and shares the workload between them when possible."
Hyper-threading means that every core is doing 2 tasks simultaneously
which makes 1 task to complete for longer time
(but usually the sum of the complete times for 2 tasks is better than if 2 tasks are done one after another on the same core with NO Hyper-threading)
Analogy: You want to cook 2 dishes for dinner:
NO Hyper-threading (on 1 core CPU = 1 cook):
You cook the first dish and finish it in 1 hour, then start to do the second which also takes 1 hour and you have the dinner ready after 2 hours
With Hyper-threading (on 1 core CPU = 1 cook (but which moves so fast that it appears to the observer that there are 2 cooks in the kitchen)):
You start to cook 2 dishes simultaneously (you split and optimize your time between the 2 pots) and finish both (dinner is ready) after 1.5 hours (FASTER total time)
but making of every dish will also take 1.5 hours (Slower time for a "task") (you can't eat any of them before 1.5 hours "mark")
NO Hyper-threading (on 2 core CPU = 2 real cooks in the kitchen):
2 cooks start to make 2 dishes at the same time, both finish after 1 hour
- every dish is ready after 1 hour
- dinner (of 2 dishes) is ready after 1 hour
Before SETI sort of came back up, I was crunching Einstein full time. At first I had hyperthreading turned off, as my rig was configured as a GPU cruncher. At Bikeman's suggestion I turned on hyperthreading which showed up as 12 processors instead of 6. The result was that I added about 1 hour and 20 minutes to each crunch time, but because of all 12 threads producing at once, my total output increased dramatically. I actually made it to #3 in the top computers. I am heavilly overclocked as well. It really does make a difference if hyperthreading is turned on or off. As BilBg pointed out, the times do increase for each work unit, on the same processor, at the same clock speed. There are several things that decide how many seconds a WU takes to complete.
Is hyperthreading turned on or off.
What is the clock speed of the processor.
What is the clock speed of the memory.
How big is the cache level for the CPU, and how fast can that cache be fed to the processor.
There are other factors as well, so in order to get a straight comparison, you would need to know all that information about each computer being tested.
Whole point is I simply offered result numbers but you all speaking some super theory that ignoring my numbers and according to your theories I must be wrong and you must right something.
Take a look back, noone questioned your numbers. The problem is just "what do they mean?" And the folowing is actually not complicated at all - just tell us if you're running 2 or 4 WUs in parallel :)
Last night I tried e@h on
)
Last night I tried e@h on core i5 460M and it is doing 19k.
i5 460 is around 20% faster cpu than my i5 430.
Well productivity wise
)
Well productivity wise whatever beautiful theories you want to say the results/numbers beating all your words. I am showing the solid results but you and few other religious hard liners are sticking to theories rather than facts. So you'd better stick with your theories and I'd better stick with my results. Since E@H cannot provide fine benchmark list to cpu's.
Hi! Let me try to bring
)
Hi!
Let me try to bring some light into this matter:
I have a i5 430 M notebook myself (I love it!!). A typical S5GC1HF task takes 28k sec (that's the stock Linux version of the app).
It's not only configured to enable hyperthreading, it's actually DOING 4 tasks in parallel.
It's easy to check. Go to the BOINC manager advanced view and look at the "Tasks" tab.
Here's an example picture I just took:
See? There are 4 tasks marked "active" (well in German, but you get the idea :-) ).
Is your screen the same? 4 E@H tasks marked Active, or just 2?? And if 4, are those 4 E@H tasks, or a mix of E@H and other projects' tasks?
Happy Crunching
HBE
RE: Well productivity wise
)
'religious hardliners' ?? :o
Um, Orgil ..... archae86 is a retired Intel electronic engineer ( chip design, fabrication ... ) a midwife so to speak from many of the earlier x86 designs onwards ( see the gag in his sig? ). He's probably forgotten more than we'll ever collectively know. His CV is truly amazing and we are pleased to have his insights expressed here.
[ We can tell that you've totally mis-understood the HT issue, and can't bring yourself to admit that. But maybe, just maybe, you might want to review your response to one of the world's leading chip designers? ]
Your post's wording is close to objectionable flaming and ad-hominem ( ie. deletion ), so can we stay civil on this one? :-)
Cheers, Mike.
( edit ) For instance see here where Peter Stoll is mentioned ( note 12 ) as a designer. Or here. Or here. Or .....
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
To drift off-thread for one
)
To drift off-thread for one moment, Orgil, noting your listed country and your team name, it might interest you to know I attended a few of the events for this year's Naadam. Opening ceremony, some wrestling, one horse race, archery, and a little ankle-bone shooting. Some of the pictures I took are here, though the Naadam pictures are in subdirectories under it and the maps and overhead imagery are not my photos, though they do display track log data from the GPSr I carried.
Yes moderator, 'tis off-topic, and I'll cheerfully submit to deletion.
Whole point is I simply
)
Whole point is I simply offered result numbers but you all speaking some super theory that ignoring my numbers and according to your theories I must be wrong and you must right something. So out of some wondering/desperation I am concluding that your points are not much different than how religious hardliners behave. The whole point is in this case traditionally believed thing of mobile per core productivity is slightly higher than desktop cpu core and even server cpu core. If non oc'd mobile cpu core is outdoing non oc'd server cpu core what is not so much acceptable for computing culture or religion...
In non oc'd original case most xeons doing 29k or 30k's versus mostly mobile core i5's doing 22'k or less that these numbers represent clear productivity difference.
ps. thanks for your naadam things that looks cool but well that is off topic.
Your figures may be hard
)
Your figures may be hard numbers, but they are resting on quicksand. I don't need a complicated theory to see that. At this moment I don't trust them to support my weight.
Everybody is telling you to
)
Everybody is telling you to mind Hyper-threading but you refuse to.
http://en.wikipedia.org/wiki/Hyper-threading
"For each processor core that is physically present, the operating system addresses two virtual processors, and shares the workload between them when possible."
Hyper-threading means that every core is doing 2 tasks simultaneously
which makes 1 task to complete for longer time
(but usually the sum of the complete times for 2 tasks is better than if 2 tasks are done one after another on the same core with NO Hyper-threading)
Analogy: You want to cook 2 dishes for dinner:
NO Hyper-threading (on 1 core CPU = 1 cook):
You cook the first dish and finish it in 1 hour, then start to do the second which also takes 1 hour and you have the dinner ready after 2 hours
With Hyper-threading (on 1 core CPU = 1 cook (but which moves so fast that it appears to the observer that there are 2 cooks in the kitchen)):
You start to cook 2 dishes simultaneously (you split and optimize your time between the 2 pots) and finish both (dinner is ready) after 1.5 hours (FASTER total time)
but making of every dish will also take 1.5 hours (Slower time for a "task") (you can't eat any of them before 1.5 hours "mark")
NO Hyper-threading (on 2 core CPU = 2 real cooks in the kitchen):
2 cooks start to make 2 dishes at the same time, both finish after 1 hour
- every dish is ready after 1 hour
- dinner (of 2 dishes) is ready after 1 hour
Get this program to find/know many things about your CPU/computer:
System Information Viewer
http://rh-software.com/
http://rh-software.com/downloads/siv.zip
Examples:
NO Hyper-threading - 4 cores acting as 4 CPUs:
With Hyper-threading - 6 cores acting as 12 CPUs:
[pre] [/pre]
Before SETI sort of came back
)
Before SETI sort of came back up, I was crunching Einstein full time. At first I had hyperthreading turned off, as my rig was configured as a GPU cruncher. At Bikeman's suggestion I turned on hyperthreading which showed up as 12 processors instead of 6. The result was that I added about 1 hour and 20 minutes to each crunch time, but because of all 12 threads producing at once, my total output increased dramatically. I actually made it to #3 in the top computers. I am heavilly overclocked as well. It really does make a difference if hyperthreading is turned on or off. As BilBg pointed out, the times do increase for each work unit, on the same processor, at the same clock speed. There are several things that decide how many seconds a WU takes to complete.
Is hyperthreading turned on or off.
What is the clock speed of the processor.
What is the clock speed of the memory.
How big is the cache level for the CPU, and how fast can that cache be fed to the processor.
There are other factors as well, so in order to get a straight comparison, you would need to know all that information about each computer being tested.
Steve
Crunching as member of The GPU Users Group team.
RE: Whole point is I simply
)
Take a look back, noone questioned your numbers. The problem is just "what do they mean?" And the folowing is actually not complicated at all - just tell us if you're running 2 or 4 WUs in parallel :)
MrS
Scanning for our furry friends since Jan 2002