PrimeGrid
Please visit donation page to help the project cover running costs for this month

Advanced search

Message boards : Problems and Help : If X =…

Author Message
Profile The Sarge
Send message
Joined: 27 Dec 17
Posts: 1
ID: 964240
Credit: 347,694,149
RAC: 0
PPS LLR Bronze: Earned 10,000 credits (13,053)PSP LLR Bronze: Earned 10,000 credits (23,712)SR5 LLR Bronze: Earned 10,000 credits (23,752)Generalized Cullen/Woodall Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,040,379)PPS Sieve Double Silver: Earned 200,000,000 credits (344,458,893)AP 26/27 Bronze: Earned 10,000 credits (12,129)GFN Ruby: Earned 2,000,000 credits (2,121,712)
Message 113821 - Posted: 27 Jan 2018 | 8:24:39 UTC

Greetings all. My question is related to hyper threading, and what, if anything it benefits the project and/or a contributor.

Admittedly, I am having trouble understanding the math of hyperthreading, but at the basic level, this is what I think I understand.

If, a host, for example, an 8 core CPU/system commits to running PrimeGrid LLR tasks (sans GPU), say PPS, what is the advantage of creating an app_config file to invoke hyperthnreading?

If I understand what I’ve read, is that instead of having 8 separate tasks, running discretely on 8 cores, the xml config file can be crafted to run a single task across 8 cores, and instead of having 8 tasks running simultaneously, there will be one task running on 8 cores. I hope I’ve described that accurately. I think Amicable Numbers does this in a similar way in that you can set the limit of how many CPUs are dedicated to a task.
My question is what’s the advantage, of running a single LLR task across 8 cores, versus running 8 LLR across 8 cores? If I understand the math of my question, are they not, quantitatively, the same? Is it 8x1=8 or 1x8=8, or ??? Like I said, I don’t understand the underlying math, but I want to.

If there is a distinct advantage to creating a custom config file, then I’m interested, but if the mathematical equation, and credit is the same after the same time, then what’s the point?

Thank you in advance for your patience.



Profile composite
Volunteer tester
Send message
Joined: 16 Feb 10
Posts: 838
ID: 55391
Credit: 763,326,750
RAC: 403,264
Discovered 2 mega primesFound 1 prime in the 2018 Tour de Primes321 LLR Turquoise: Earned 5,000,000 credits (5,477,467)Cullen LLR Gold: Earned 500,000 credits (776,297)ESP LLR Ruby: Earned 2,000,000 credits (3,433,680)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,093,491)PPS LLR Sapphire: Earned 20,000,000 credits (26,886,458)PSP LLR Turquoise: Earned 5,000,000 credits (6,104,304)SoB LLR Sapphire: Earned 20,000,000 credits (31,633,596)SR5 LLR Turquoise: Earned 5,000,000 credits (6,110,877)SGS LLR Ruby: Earned 2,000,000 credits (3,486,285)TRP LLR Turquoise: Earned 5,000,000 credits (7,070,795)Woodall LLR Amethyst: Earned 1,000,000 credits (1,693,614)321 Sieve (suspended) Emerald: Earned 50,000,000 credits (50,256,050)Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,571,178)Generalized Cullen/Woodall Sieve (suspended) Emerald: Earned 50,000,000 credits (50,009,610)PPS Sieve Double Silver: Earned 200,000,000 credits (368,403,727)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Jade: Earned 10,000,000 credits (10,165,888)TRP Sieve (suspended) Sapphire: Earned 20,000,000 credits (20,071,454)AP 26/27 Turquoise: Earned 5,000,000 credits (6,616,128)WW Gold: Earned 500,000 credits (932,000)GFN Emerald: Earned 50,000,000 credits (53,771,465)PSA Double Bronze: Earned 100,000,000 credits (102,762,384)
Message 113822 - Posted: 27 Jan 2018 | 8:38:24 UTC
Last modified: 27 Jan 2018 | 9:10:42 UTC

There are 2 advantages to multithreading.

1) A multithreaded task runs faster than a single-threaded task. If a workunit results in a prime number, the first valid reporter is credited as the discoverer, while the other reporter is the double-checker. A double-check has half the prime-score of a discovery. And for the TdP competition, only the discovery is credited. If you do 8 tasks in parallel, you might get the same throughput, but every one of those tasks is likely to be eclipsed in turnaround time by a wingman using multithreading.

2) Some CPU chips have higher throughput by multithreading because they make more effective use of the CPU's memory cache.

EDIT:
3) Some of the longer-running tasks would not complete before the deadline on a slower CPU without multithreading.

NB. Multithreading is the method of using multiple threads for one task. Hyperthreading is a hardware technology which creates an extra logical CPU in each core by sharing most of the core between 2 logical CPUs. Its benefit comes from increasing utilization of the core, by running the second thread on it while the first thread is waiting for a memory access to complete.

The downside of more core utilization is more heat and higher temperature which could damage the chip. If the CPU's cooler can't handle the greater heat dissipation, some CPUs reduce the frequency for a while to cool the chip, so you've lost the benefit of higher throughput. LLR is highly optimized to use the hardware, so LLR tends to put more heat stress on CPUs than most other software. Borderline cooling solutions may fail to keep the CPU chip cool enough to prevent computation errors.

EDIT EDIT:
The marginal speed gain from multithreading is sub-linear. Usually, doubling the number of threads working on a task does not reduce the runtime by half, unless the cache was overloaded in the first place and the CPU was running inefficiently.

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Avatar
Send message
Joined: 21 Jan 10
Posts: 13634
ID: 53948
Credit: 281,255,136
RAC: 21,117
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de PrimesFound 1 prime in the 2020 Tour de PrimesFound 2 primes in the 2021 Tour de Primes321 LLR Turquoise: Earned 5,000,000 credits (5,132,712)Cullen LLR Turquoise: Earned 5,000,000 credits (5,038,114)ESP LLR Turquoise: Earned 5,000,000 credits (6,177,890)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Sapphire: Earned 20,000,000 credits (20,751,038)PSP LLR Turquoise: Earned 5,000,000 credits (5,197,957)SoB LLR Sapphire: Earned 20,000,000 credits (36,067,618)SR5 LLR Jade: Earned 10,000,000 credits (10,007,110)SGS LLR Ruby: Earned 2,000,000 credits (3,718,606)TRP LLR Turquoise: Earned 5,000,000 credits (5,084,329)Woodall LLR Ruby: Earned 2,000,000 credits (3,872,218)321 Sieve (suspended) Jade: Earned 10,000,000 credits (10,061,196)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (22,885,121)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Jade: Earned 10,000,000 credits (10,118,303)WW Sapphire: Earned 20,000,000 credits (32,204,000)GFN Emerald: Earned 50,000,000 credits (78,031,938)PSA Jade: Earned 10,000,000 credits (12,445,029)
Message 113827 - Posted: 27 Jan 2018 | 9:52:43 UTC

You appear to be confusing Hyper-threading with multithreading.

You used the word "Hyper-threading", but you were actually asking about multithreading. The two are completely different.

Multithreading is a software technique whereby a program uses multiple, independently running threads to do multiple things simultaneously, such as communicating on a network while also drawing something on the screen. In the case of LLR, this technique is used to allow LLR to perform calculations simultaneously on multiple CPU cores, thus speeding up the calculation.

Hyper-threading is a proprietary Intel hardware design that allows a single CPU core to run two hardware threads simultaneously. In essence, some parts of the CPU core are duplicated for the second thread while others are not. It's a way of boosting performance. It's not particularly well suited for use with LLR or the CPU version of Genefer, but it is beneficial for sieving.

(I suggest reading the Wikipedia links above if you want a more detailed explanation.)

Getting back to your question about multithreading, it's possible to run LLR in multithreaded mode, using all the CPU cores together to perform the calculation faster. For large LLR tasks (e.g., PPS-MEGA or larger), not only is it faster, it's also more efficient. Running a single SoB task on all 4 cores is likely to run more than 4 times as fast as running 4 separate SoB tasks on a single core, so over time, you'll get more SoB tasks completed using multithreading.

Whether or not you're using multithreading, Hyper-threading shouldn't be used with LLR. If you have a 4-core/8-thread CPU, you should only run LLR on 4 of the cores.
____________
My lucky number is 75898524288+1

XAVER
Avatar
Send message
Joined: 21 Jun 16
Posts: 17
ID: 450130
Credit: 90,275,237
RAC: 0
Discovered the World's First base 79 Generalized Cullen prime!!!Discovered 1 mega prime321 LLR Ruby: Earned 2,000,000 credits (2,013,360)Cullen LLR Amethyst: Earned 1,000,000 credits (1,765,277)ESP LLR Amethyst: Earned 1,000,000 credits (1,212,056)Generalized Cullen/Woodall LLR Amethyst: Earned 1,000,000 credits (1,226,840)PPS LLR Ruby: Earned 2,000,000 credits (2,474,233)PSP LLR Amethyst: Earned 1,000,000 credits (1,105,896)SoB LLR Ruby: Earned 2,000,000 credits (4,085,556)SR5 LLR Amethyst: Earned 1,000,000 credits (1,485,698)SGS LLR Amethyst: Earned 1,000,000 credits (1,001,949)TRP LLR Ruby: Earned 2,000,000 credits (2,264,223)Woodall LLR Amethyst: Earned 1,000,000 credits (1,711,183)321 Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,000,211)Generalized Cullen/Woodall Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,002,733)PPS Sieve Sapphire: Earned 20,000,000 credits (20,057,450)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,065,490)TRP Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,000,543)AP 26/27 Turquoise: Earned 5,000,000 credits (7,289,529)WW Bronze: Earned 10,000 credits (12,000)GFN Sapphire: Earned 20,000,000 credits (25,042,405)PSA Jade: Earned 10,000,000 credits (13,458,605)
Message 113841 - Posted: 27 Jan 2018 | 17:28:28 UTC

My i7-5960X 8/16 cores 32Gb memory runs LLR Wus faster with hyperthreading and multithreading on (12 c per wu).

e.g.: llrESP 12c (HT and MT on) around 2h 30m per wu
llrESP 6c (HT off MT on) around 3h+ per wu
this also goes for larger llr wus

so this seems to indicate that some cpus are better with both HT and MT set to on, or is there an error on my side?

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Avatar
Send message
Joined: 21 Jan 10
Posts: 13634
ID: 53948
Credit: 281,255,136
RAC: 21,117
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de PrimesFound 1 prime in the 2020 Tour de PrimesFound 2 primes in the 2021 Tour de Primes321 LLR Turquoise: Earned 5,000,000 credits (5,132,712)Cullen LLR Turquoise: Earned 5,000,000 credits (5,038,114)ESP LLR Turquoise: Earned 5,000,000 credits (6,177,890)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Sapphire: Earned 20,000,000 credits (20,751,038)PSP LLR Turquoise: Earned 5,000,000 credits (5,197,957)SoB LLR Sapphire: Earned 20,000,000 credits (36,067,618)SR5 LLR Jade: Earned 10,000,000 credits (10,007,110)SGS LLR Ruby: Earned 2,000,000 credits (3,718,606)TRP LLR Turquoise: Earned 5,000,000 credits (5,084,329)Woodall LLR Ruby: Earned 2,000,000 credits (3,872,218)321 Sieve (suspended) Jade: Earned 10,000,000 credits (10,061,196)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (22,885,121)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Jade: Earned 10,000,000 credits (10,118,303)WW Sapphire: Earned 20,000,000 credits (32,204,000)GFN Emerald: Earned 50,000,000 credits (78,031,938)PSA Jade: Earned 10,000,000 credits (12,445,029)
Message 113843 - Posted: 27 Jan 2018 | 17:40:43 UTC - in response to Message 113841.

My i7-5960X 8/16 cores 32Gb memory runs LLR Wus faster with hyperthreading and multithreading on (12 c per wu).

e.g.: llrESP 12c (HT and MT on) around 2h 30m per wu
llrESP 6c (HT off MT on) around 3h+ per wu
this also goes for larger llr wus

so this seems to indicate that some cpus are better with both HT and MT set to on, or is there an error on my side?


There appears to be an error in your test strategy.

Unless there's some typos there, or some details you left out, that's not the conclusion I'd get from those results.

It's an 8 core machine. If you tell it to use 12 "cores", it's actually using the 8 real cores plus 4 hyperthreads. The 4 hyperthreads probably don't add much, so you're simply using all 8 cores on the CPU.

In the second test, you have it set to use 6 cores, so it's using 6 of the 8 cores.

What that test shows is not a comparison between HT on and HT off, but a comparison between using 6 cores and using 8 cores. Of course the test where you have all 8 cores running is going to be faster than when you're using only 6 cores. Any other result would be very surprising.

____________
My lucky number is 75898524288+1

Profile composite
Volunteer tester
Send message
Joined: 16 Feb 10
Posts: 838
ID: 55391
Credit: 763,326,750
RAC: 403,264
Discovered 2 mega primesFound 1 prime in the 2018 Tour de Primes321 LLR Turquoise: Earned 5,000,000 credits (5,477,467)Cullen LLR Gold: Earned 500,000 credits (776,297)ESP LLR Ruby: Earned 2,000,000 credits (3,433,680)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,093,491)PPS LLR Sapphire: Earned 20,000,000 credits (26,886,458)PSP LLR Turquoise: Earned 5,000,000 credits (6,104,304)SoB LLR Sapphire: Earned 20,000,000 credits (31,633,596)SR5 LLR Turquoise: Earned 5,000,000 credits (6,110,877)SGS LLR Ruby: Earned 2,000,000 credits (3,486,285)TRP LLR Turquoise: Earned 5,000,000 credits (7,070,795)Woodall LLR Amethyst: Earned 1,000,000 credits (1,693,614)321 Sieve (suspended) Emerald: Earned 50,000,000 credits (50,256,050)Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,571,178)Generalized Cullen/Woodall Sieve (suspended) Emerald: Earned 50,000,000 credits (50,009,610)PPS Sieve Double Silver: Earned 200,000,000 credits (368,403,727)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Jade: Earned 10,000,000 credits (10,165,888)TRP Sieve (suspended) Sapphire: Earned 20,000,000 credits (20,071,454)AP 26/27 Turquoise: Earned 5,000,000 credits (6,616,128)WW Gold: Earned 500,000 credits (932,000)GFN Emerald: Earned 50,000,000 credits (53,771,465)PSA Double Bronze: Earned 100,000,000 credits (102,762,384)
Message 113844 - Posted: 27 Jan 2018 | 17:49:10 UTC - in response to Message 113841.

I concur with Michael, "12 cores" doesn't make sense for that CPU. Maybe you mean 12 threads? And you don't say which subproject. The FFT size makes a difference on optimal cache usage.

I say, use whatever configuration results in the most throughput for your machine. You might stumble onto a non-obvious configuration that is best for your machine. You have to test them all.

XAVER
Avatar
Send message
Joined: 21 Jun 16
Posts: 17
ID: 450130
Credit: 90,275,237
RAC: 0
Discovered the World's First base 79 Generalized Cullen prime!!!Discovered 1 mega prime321 LLR Ruby: Earned 2,000,000 credits (2,013,360)Cullen LLR Amethyst: Earned 1,000,000 credits (1,765,277)ESP LLR Amethyst: Earned 1,000,000 credits (1,212,056)Generalized Cullen/Woodall LLR Amethyst: Earned 1,000,000 credits (1,226,840)PPS LLR Ruby: Earned 2,000,000 credits (2,474,233)PSP LLR Amethyst: Earned 1,000,000 credits (1,105,896)SoB LLR Ruby: Earned 2,000,000 credits (4,085,556)SR5 LLR Amethyst: Earned 1,000,000 credits (1,485,698)SGS LLR Amethyst: Earned 1,000,000 credits (1,001,949)TRP LLR Ruby: Earned 2,000,000 credits (2,264,223)Woodall LLR Amethyst: Earned 1,000,000 credits (1,711,183)321 Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,000,211)Generalized Cullen/Woodall Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,002,733)PPS Sieve Sapphire: Earned 20,000,000 credits (20,057,450)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,065,490)TRP Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,000,543)AP 26/27 Turquoise: Earned 5,000,000 credits (7,289,529)WW Bronze: Earned 10,000 credits (12,000)GFN Sapphire: Earned 20,000,000 credits (25,042,405)PSA Jade: Earned 10,000,000 credits (13,458,605)
Message 113847 - Posted: 27 Jan 2018 | 18:43:49 UTC

You are right, with 12 (out of 16) cores i mean threads (in boinctasks they are abbreviated as "C").

I use 14 threads (HT on) for Boinc (87.5%). 12 for one llr wu, 2 for genefer.
With HT off 6 (full cores?) for one llr wu and 1 for genefer (7 cores for Boinc/87.5%).

I tried this two settings on every llr subprojects (except SGS and PPSE) and found that llr runs with HT on on my system definitely faster than with HT off.

Hope this clarifies the confusion. Thanks also for your support

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Avatar
Send message
Joined: 21 Jan 10
Posts: 13634
ID: 53948
Credit: 281,255,136
RAC: 21,117
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de PrimesFound 1 prime in the 2020 Tour de PrimesFound 2 primes in the 2021 Tour de Primes321 LLR Turquoise: Earned 5,000,000 credits (5,132,712)Cullen LLR Turquoise: Earned 5,000,000 credits (5,038,114)ESP LLR Turquoise: Earned 5,000,000 credits (6,177,890)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Sapphire: Earned 20,000,000 credits (20,751,038)PSP LLR Turquoise: Earned 5,000,000 credits (5,197,957)SoB LLR Sapphire: Earned 20,000,000 credits (36,067,618)SR5 LLR Jade: Earned 10,000,000 credits (10,007,110)SGS LLR Ruby: Earned 2,000,000 credits (3,718,606)TRP LLR Turquoise: Earned 5,000,000 credits (5,084,329)Woodall LLR Ruby: Earned 2,000,000 credits (3,872,218)321 Sieve (suspended) Jade: Earned 10,000,000 credits (10,061,196)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (22,885,121)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Jade: Earned 10,000,000 credits (10,118,303)WW Sapphire: Earned 20,000,000 credits (32,204,000)GFN Emerald: Earned 50,000,000 credits (78,031,938)PSA Jade: Earned 10,000,000 credits (12,445,029)
Message 113848 - Posted: 27 Jan 2018 | 19:05:35 UTC - in response to Message 113847.

I tried this two settings on every llr subprojects (except SGS and PPSE) and found that llr runs with HT on on my system definitely faster than with HT off.


Please reread what I wrote. Your testing methodology is flawed and you're not actually testing the difference between running with HT-on and HT-off. Your test shows the difference between using all CPU cores and using 6 of 8 CPU cores.
____________
My lucky number is 75898524288+1

XAVER
Avatar
Send message
Joined: 21 Jun 16
Posts: 17
ID: 450130
Credit: 90,275,237
RAC: 0
Discovered the World's First base 79 Generalized Cullen prime!!!Discovered 1 mega prime321 LLR Ruby: Earned 2,000,000 credits (2,013,360)Cullen LLR Amethyst: Earned 1,000,000 credits (1,765,277)ESP LLR Amethyst: Earned 1,000,000 credits (1,212,056)Generalized Cullen/Woodall LLR Amethyst: Earned 1,000,000 credits (1,226,840)PPS LLR Ruby: Earned 2,000,000 credits (2,474,233)PSP LLR Amethyst: Earned 1,000,000 credits (1,105,896)SoB LLR Ruby: Earned 2,000,000 credits (4,085,556)SR5 LLR Amethyst: Earned 1,000,000 credits (1,485,698)SGS LLR Amethyst: Earned 1,000,000 credits (1,001,949)TRP LLR Ruby: Earned 2,000,000 credits (2,264,223)Woodall LLR Amethyst: Earned 1,000,000 credits (1,711,183)321 Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,000,211)Generalized Cullen/Woodall Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,002,733)PPS Sieve Sapphire: Earned 20,000,000 credits (20,057,450)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,065,490)TRP Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,000,543)AP 26/27 Turquoise: Earned 5,000,000 credits (7,289,529)WW Bronze: Earned 10,000 credits (12,000)GFN Sapphire: Earned 20,000,000 credits (25,042,405)PSA Jade: Earned 10,000,000 credits (13,458,605)
Message 113851 - Posted: 27 Jan 2018 | 19:42:10 UTC

So you mean with HT on Boinc is still using all 8 cores (16 threads) even when set to use 87.5%) for running llr. I thought it would only use 7 cores (14 threads).

So how can i test the real difference between HT on and HT off? (cores vs. threads)

Thanks for clarification of my rather stupid assumptions.

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Avatar
Send message
Joined: 21 Jan 10
Posts: 13634
ID: 53948
Credit: 281,255,136
RAC: 21,117
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de PrimesFound 1 prime in the 2020 Tour de PrimesFound 2 primes in the 2021 Tour de Primes321 LLR Turquoise: Earned 5,000,000 credits (5,132,712)Cullen LLR Turquoise: Earned 5,000,000 credits (5,038,114)ESP LLR Turquoise: Earned 5,000,000 credits (6,177,890)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Sapphire: Earned 20,000,000 credits (20,751,038)PSP LLR Turquoise: Earned 5,000,000 credits (5,197,957)SoB LLR Sapphire: Earned 20,000,000 credits (36,067,618)SR5 LLR Jade: Earned 10,000,000 credits (10,007,110)SGS LLR Ruby: Earned 2,000,000 credits (3,718,606)TRP LLR Turquoise: Earned 5,000,000 credits (5,084,329)Woodall LLR Ruby: Earned 2,000,000 credits (3,872,218)321 Sieve (suspended) Jade: Earned 10,000,000 credits (10,061,196)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (22,885,121)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Jade: Earned 10,000,000 credits (10,118,303)WW Sapphire: Earned 20,000,000 credits (32,204,000)GFN Emerald: Earned 50,000,000 credits (78,031,938)PSA Jade: Earned 10,000,000 credits (12,445,029)
Message 113858 - Posted: 27 Jan 2018 | 20:21:12 UTC - in response to Message 113851.
Last modified: 27 Jan 2018 | 20:29:54 UTC

So you mean with HT on Boinc is still using all 8 cores (16 threads) even when set to use 87.5%) for running llr. I thought it would only use 7 cores (14 threads).


Yes and no, although technically it's Windows or Linux that's assigning the cores, not BOINC. BOINC has no control over what cores are used.

All modern operating systems are aware of Hyperthreading and should always use the real cores first before using the Hyper-threads.

If you set BOINC to 87.5%, it will only use 14 threads -- but it has no control over WHICH 14 threads it's using. That's determined by Windows. So, while you might intuitively think if it's using 14 threads, that means it's using 7 cores, you're wrong. Windows will use all 8 real cores first, so anything over 7 threads (i.e., 8 to 16 threads) will always use all 8 real cores on the CPU.


So how can i test the real difference between HT on and HT off? (cores vs. threads


The only way to do that while running less than all the cores is to manually set core affinity for the task. However, if you do that, you WILL affect the way the task is running, so the results won't really be indicative of how the task would normally run.

So the only way to do this test on an 8 core machine is to run the test with 8 cores (i.e., the 8 real cores) vs. 16 cores (the 8 real cores plus the 8 Hyper-threads). You need to have all the cores running. You really can't test 6 cores vs 12 cores -- unless you want to physically replace the 8 core CPU with a 6 core CPU. :)
____________
My lucky number is 75898524288+1

Message boards : Problems and Help : If X =…

[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2021 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 1.72, 1.80, 1.56
Generated 4 Aug 2021 | 19:17:45 UTC