What application?? I assume you mean N-body since that is the only one left and you only mention a cpu.

You probably can find one somewhere in the Top 100 lists.

Yes Keith apologies I didn't say the application yes "N body" I had a look through the "top computer" list but couldn't see 1 at a quick glance

I had to use BoincStats to find a 7900X WAY down the lists. ~2300 or so before one showed up. But not active recently enough to show what its N-body times were.

---

Thanks Keith, have sent you a pm

I see what looks like CPU thrashing for an i9-13950HX.
Is there any way to smooth out the loading on CPU so that we don't lose approximately 5% performance represented by all those dips? The particular tasks include de_nbody_11_02_2023_v183_pal5_data__2_1705435140_464910_0 as an example.

I don't know why you're asking in this thread, because as per the title it's about benchmarking times.

I suggest you look for a thread with a similar subject to yours, or start a new one.

---

Team AnandTech - SETI@H, DPAD, F@H, MW@H, A@H, LHC, POGS, R@H, Einstein@H, DHEP, WCG

Main rig - Ryzen 5 3600, MSI B450 G.Pro C. AC, RTX 3060Ti 8GB, 32GB DDR4 3200, Win 10 64bit
2nd rig - i7 4930k @4.1 GHz, HD 7870 XT 3GB(DS), 16GB DDR3 1866, Win7

It was probably my MISTAKE.

i have a 4080 super, 3090, 1080, and 1060 6gb i would like to contribute to the benchmark system. how can i help?

i have a 4080 super, 3090, 1080, and 1060 6gb i would like to contribute to the benchmark system. how can i help?

You can't. There are no gpu applications here.

---

You can't. There are no gpu applications here.

And every work unit requires different amount of computation, so it's not even possible to submit any CPU times. Well, it's possible, but doesn't make much sense. ;-)

Perhaps time to unpin this thread?

---

I believe only the originator of the thread can unpin it. That looks to be user https://milkyway.cs.rpi.edu/milkyway/show_user.php?userid=146600

---

It was probably my MISTAKE.

Try reading the thread title and op at least then :p

You can't. There are no gpu applications here.

...... anymore.

You can't. There are no gpu applications here.

And every work unit requires different amount of computation, so it's not even possible to submit any CPU times. Well, it's possible, but doesn't make much sense. ;-)

Perhaps time to unpin this thread?

Their was a time (upto a couple of years ago or so) that the majority of WUs had pretty much a set amount of computation, and hence very nearly identical credit for each WU(result). Therefore you could benchmark GPUs and CPUs with MW with a common credit WU.

I believe only the originator of the thread can unpin it. That looks to be user https://milkyway.cs.rpi.edu/milkyway/show_user.php?userid=146600

Nope, that's down to the mods, I never pinned it in the first place (although I think I did ask), and I'm pretty sure we users can't pin threads.
Actually, I think I asked the mods to unpin this thread a few years ago, but it seems the mods are non-existent here.

---

Team AnandTech - SETI@H, DPAD, F@H, MW@H, A@H, LHC, POGS, R@H, Einstein@H, DHEP, WCG

Main rig - Ryzen 5 3600, MSI B450 G.Pro C. AC, RTX 3060Ti 8GB, 32GB DDR4 3200, Win 10 64bit
2nd rig - i7 4930k @4.1 GHz, HD 7870 XT 3GB(DS), 16GB DDR3 1866, Win7

i have a 4080 super, 3090, 1080, and 1060 6gb i would like to contribute to the benchmark system. how can i help?

Thanks for the offer, but unfortunately MW results credits vary widely now, and their are no longer any GPU apps for MW I believe.
Also, this stupid forum platform doesn't allowing editing after about 15 minutes, so I can't update the op.

---

Team AnandTech - SETI@H, DPAD, F@H, MW@H, A@H, LHC, POGS, R@H, Einstein@H, DHEP, WCG

Main rig - Ryzen 5 3600, MSI B450 G.Pro C. AC, RTX 3060Ti 8GB, 32GB DDR4 3200, Win 10 64bit
2nd rig - i7 4930k @4.1 GHz, HD 7870 XT 3GB(DS), 16GB DDR3 1866, Win7

Hi, I just moved to a Windows 11 computer, Intel(R) Core(TM) i7-14700, 32 GB of RAM. I have been very happy with the reported results. However, I decided to take a look at Task Manager this afternoon and was surprised to discover that it shows the current MilkyWay WU using only 20-50% of the CPU - BOINC Manager shows it running on "16 CPUs." My Computing Preferences are set at 100%. Intel tells me that it has:
Total Cores 20
# of Performance-cores 8
# of Efficient-cores 12
Total Threads 28

My (old) Windows 10 computer ran at nearly 100% all the time.
Would someone please explain this to me?
How do I get it to use more of its capability?

Thanks.

# of Performance-cores 8
# of Efficient-cores 12
Total Threads 28

Set "Max # of threads for each MilkyWay@home task" in your MilkyWay@home preferences to 14 or 7, than your computer will either run 2 WUs with 14 threads each or 4 WUs with 7 threads each. Can't predict which is going to be faster on your CPU with those P- and E-cores, but at least all cores should be in use. If I should guess, 7 threads per WU should be fastest since your CPU has lots of cache, but YMMV.

EDIT: you might also need to increase the priority at which the tasks run, like this user, otherwise they might run only on the E-cores.

---

Thanks. I changed Max # of threads to 7. Updated and got new tasks. No change. I changed the priority to high in Task Manager for the current WU. Still no change. Please advise.

Also, if I change the priority in Task Manager, that applies only to the current WU, not the next one. Where do I change priority so that it applies to all WUs?

[/code]

More information: right now I have one MilkyWay WU showing 16 CPUs, and 12 WCG work units, all running simultaneously. Task Manager shows up to 70% use.

Thanks. I changed Max # of threads to 7. Updated and got new tasks. No change.

Did you save the settings? It's not enough to select 7, you also need to click on "save". Check again, this is a common mistake.
The display for old WUs in BOINC Manager won't change for WUs you already had in your cache and the running ones will also continue running on 16 threads, but those, which start after the change, will run on 7 threads (even if they will still show 16). New WUs should however show the number of threads correctly. So check your settings on the website and if there's a 7, check your client_state.xml, app_version should look like this (the flops value might be different of course):

<app_version>
    <app_name>milkyway_nbody</app_name>
    <version_num>194</version_num>
    <platform>windows_x86_64</platform>
    <avg_ncpus>7.000000</avg_ncpus>
    <flops>37638326516.094742</flops>
    <plan_class>mt</plan_class>
    <api_version>7.6.33</api_version>
    <cmdline>--nthreads 7</cmdline>
    <file_ref>
        <file_name>milkyway_nbody_1.94_windows_x86_64__mt.exe</file_name>
        <main_program/>
    </file_ref>
</app_version>

Also, if I change the priority in Task Manager, that applies only to the current WU, not the next one. Where do I change priority so that it applies to all WUs?

As descibed in this post.

---

First - every computer type is different, so what works well on one may be totally wrong for any other computer. for example Link's insistence that 7-cores is the best for all is absolutely WRONG and should be ignored.

Second his example config file contains out of date and inaccurate information, with NO explanation what the sections actually do.; DO NOT use it.

Third, don't waste your time thinking that changing the run priority of a task will have an significant effect on their run time. It may improve by a small (1 or 2 percent at best), or it may render your computer unusable.


Now, considering the app_config.xml file (for a standard windows 10/11 installation is to be found in the folder C:\ProgramData\BOINC\projects\milkyway.cs.rpi.edu_milkyway )

<app_version>
<app_name>milkyway_nbody</app_name>

Required header section, using generic name, milkyway_nbody_orbit_fitting is an alternative name

<version_num>194</version_num>
<platform>windows_x86_64</platform>

These lines are not needed

<avg_ncpus>7.000000</avg_ncpus>

The average number of cpu threads to be used.

<flops>37638326516.094742</flops>

Only used by BOINC server to calculate the number of tasks to be sent, has no impact on the rate of calculation.

<plan_class>mt</plan_class>

Tells BOINC, and thus the applcaiton that tasks are to be considered to be multi-thread

<api_version>7.6.33</api_version>

Old api, should rally be using 8.0.2 - may have no detrimental impact just now, but could do in the future

<cmdline>--nthreads 7</cmdline>

The actual maximum number of threads to be used. While avg_ncpus and nthreads can be different doing so can cause all sorts of unexpected behaviour.

<file_ref>
<file_name>milkyway_nbody_1.94_windows_x86_64__mt.exe</file_name>
<main_program/>
</file_ref>

Ties you to one specific version of MilkyWays's application, and is not neede, and so can be removed.

</app_version>

Given your processor has two distinct type of core, it is almost certain that you need to do some tests to make sure you are running the best possible number of threads. Remember that Hyper-threading (or whatever Intel call it) is only available on your "performance" cores, and there appears to be a very big performance difference between the "performance" and economy cores, so you may end up with a strange compromise.

One option not mentioned is "project_max_concurrent" - this does what its name suggests (for a change!!)

A simpler version of app_config.xml might be:

<app_config>
   <app_version>
      <app_name>milkyway_nbody_orbit_fitting</app_name>
      <plan_class>mt</plan_class>
         <avg_ncpus>x</avg_ncpus>
         <cmdline>--nthreads x</cmdline>
   </app_version>
   <project_max_concurrent>y</project_max_concurrent>
</app_config>

where x is the number of threads (cores) and y is the number of concurrent MW tasks.

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

First - every computer type is different, so what works well on one may be totally wrong for any other computer. for example Link's insistence that 7-cores is the best for all is absolutely WRONG and should be ignored.

What do you suggest for a CPU with 28 threads? There are not many sensible possibilities: 28 is definitely too many, so there are only 14 or 7 left. 2 is likely going to be slower.

Second his example config file contains out of date and inaccurate information, with NO explanation what the sections actually do.; DO NOT use it.

Perhaps you should read AND understand what it posted before telling people what to do with it or what not to do. I didn't post an app_config.xml, I told him to check how the app_version section looks like in his client_state.xml. That's all.

Ties you to one specific version of MilkyWays's application, and is not neede, and so can be removed.

No, nothing should be removed from the part of client_state.xml I posted, removing it will break it.

---