BOINC Pentathlon 2021

[LAR_Systems]

Just now, GOTSpectrum said:

Any chance you're willing to swap your GPUs from folding to primegrid once that event starts? 

Afraid not... means I can't do development on the folding DB / Web client, need my systems running for testing and some of the uptime tests I have setup.
Have the python terminal based client prototype running, but needs work.

[GOTSpectrum]

3 minutes ago, LAR_Systems said:

Afraid not... means I can't do development on the folding DB / Web client, need my systems running for testing and some of the uptime tests I have setup.
Have the python terminal based client prototype running, but needs work.

Fair enough, We all appreciate what you're doing with that sweet sweet database 

[leadeater]

2 hours ago, GOTSpectrum said:

 

yep we did... 

 

2 hours ago, J-from-Nucleon said:

Yea, sorry. Didn't read through everything

 

Really not that big of a deal, it's not like we bunkered much or anything (like only due to validation pending literally) and I've still got a huge amount pending even as it is now. Won't know what the actual RAC is for teams for another like 2 days, ours however is in general correct and will increase as pending's get processed, other teams on the other hand might slightly decrease. Who really knows.

 

Regardless 3rd shouldn't be out of the realm of achievable at all.

 

I've still got around 10 2x 2690v3 (24c/48t, 240c,480t) and 7 2x 6242R (40c,80t, 280c/560t) servers I can throw in for some amount of time, maybe not the whole event but at least some of it and an increase of 1040 threads even if only for a few days is a damn good amount.

 

Still got some other additional servers I can chuck in longer term too, but not too many, only 2-4 2690v1 (16c/32t each).

[Windows7ge]

5 minutes ago, leadeater said:

I've still got around 10 2x 2690v3 (24c/48t, 240c,480t) and 7 2x 6242R (40c,80t, 280c/560t) servers I can throw in for some amount of time, maybe not the whole event but at least some of it and an increase of 1040 threads even if only for a few days is a damn good amount.

 

Still got some other additional servers I can chuck in longer term too, but not too many, only 2-4 2690v1 (16c/32t each).

Tells us he won't have as much power this year as he did last year.

 

Keeps pulling exorbitant amounts of CPU out of nowhere to add in like he did last year.

[leadeater]

Just now, Windows7ge said:

Tells us he won't have as much power this year as he did last year.

 

Keeps pulling exorbitant amounts of CPU out of nowhere to add in like he did last year.

Well 5 of those newer servers literally turned up yesterday and I can delay putting them through to production, for how long I don't know lol

[LAR_Systems]

1 minute ago, leadeater said:

Well 5 of those newer servers literally turned up yesterday and I can delay putting them through to production, for how long I don't know lol

The real reason new hires wait weeks to get their email and shared network drive

[Windows7ge]

Just now, leadeater said:

Well 5 of those newer servers literally turned up yesterday and I can delay putting them through to production, for how long I don't know lol

Just tell your boss they need firmware updates but you can't get to it for two weeks.

 

Unless this idea is just gonna get me the grumpy old man gif again...

[leadeater]

1 minute ago, Windows7ge said:

Unless this idea is just gonna get me the grumpy old man gif again...

[Windows7ge]

4 minutes ago, leadeater said:

[Gorgon]

7 hours ago, Den-Fi said:

Aiight. I woke up an hour early at 4AM to get the 3950X up and running.

These long days suuuuuuuuuuuck, but it's marathoning now.

Let those Threads Rip! (Sorry, I couldn't resist)

6 hours ago, Den-Fi said:

... Most of my rigs were moved to my warehouse which we just moved this week. There's no networking to that area yet. ...

Ah, that explains it. I noticed your output on F@H was uncharacteristically low.

[Gorgon]

19 minutes ago, leadeater said:

Well 5 of those newer servers literally turned up yesterday and I can delay putting them through to production, for how long I don't know lol

I smell a burn-in test. I think that might take at least 12 days.

[Gorgon]

21 minutes ago, Windows7ge said:

Just tell your boss they need firmware updates but you can't get to it for two weeks.

 

Unless this idea is just gonna get me the grumpy old man gif again...

Umm - I claim prior precedence to being the Grumpy one. There is a reason for my Avatar after all.

 

Isn't there a new variant of Spectre that needs addressing as well? It would seem prudent to me to apply those mitigations and do some extensive testing under a high sustained development load for a while.

[Guest]

9 minutes ago, Gorgon said:

Let those Threads Rip! (Sorry, I couldn't resist)

Ah, that explains it. I noticed your output on F@H was uncharacteristically low.

Yeah, I have not had time for a single thing. A lot of projects that COVID halted all started back up at the exact same time.

[Windows7ge]

15 minutes ago, Gorgon said:

Umm - I claim prior precedence to being the Grumpy one. There is a reason for my Avatar after all.

 

Isn't there a new variant of Spectre that needs addressing as well? It would seem prudent to me to apply those mitigations and do some extensive testing under a high sustained development load for a while.

I didn't grow up watching the shows or owning any of the bears so...

 

LMG did recently release a video about that. Dataceters might implement a mitigation for it but that probably depends on how bad the performance hit is. It kind of sucks that the only real vulnerabilities anybody's finding as of late in CPU's are ones where patching it makes the system perform worse. I guess the devs either didn't know or didn't care all for the sake of performance.

 

Which now that I think about it has kind of been the end game between AMD & Intel. Who can build the faster processor. When your bosses tell you to make it go faster I guess security goes out the window if you want to keep your job.

[Gorgon]

17 minutes ago, Windows7ge said:

I didn't grow up watching the shows or owning any of the bears so...

I'm way too old for them and our kids way too young but our eldest got hooked on the videos as a toddler and Santa went a little nuts on eBay one year and the rest, as they say, is history.

 

17 minutes ago, Windows7ge said:

LMG did recently release a video about that. Dataceters might implement a mitigation for it but that probably depends on how bad the performance hit is. It kind of sucks that the only real vulnerabilities anybody's finding as of late in CPU's are ones where patching it makes the system perform worse. I guess the devs either didn't know or didn't care all for the sake of performance.

 

Which now that I think about it has kind of been the end game between AMD & Intel. Who can build the faster processor. When your bosses tell you to make it go faster I guess security goes out the window if you want to keep your job.

The recently announced exploits are esoteric and likely only of interest to researchers as they require an already running process to exploit them but if the microcode patches make it into Windows Updates or Linux Kernel updates then likely most systems will have to take the hit unless the admins do some work to specifically deny them. Likely only worth the risk on SQL servers and other systems where significant tuning is required.

[Arrogath]

I tossed in the contribution of my 3950x to the marathon run, I'll try to drop a gpu or two on other projects

[GOTSpectrum]

4 minutes ago, Arrogath said:

I tossed in the contribution of my 3950x to the marathon run, I'll try to drop a gpu or two on other projects

Happy to see you over here, thanks for responding to my message beating of the drums of war

[GOTSpectrum]

That's our fourth update over 50k points, GG people

 

 

[MetaDrow]

Reporting from F@H division

Just got to work out some small issues cause windows being windows...

[rkv_2401]

I'm going to try and put another 12-16 threads on SiDock@home, at least once this Javelin throw is complete. After that I'm going to just run SiDock all the way till the end.

[LazyDev]

I'm very busy at this time, but have setup Boinc to use all of my Ryzen 7 5800X threads

 

-Edit

 

Now have my Ryzen 3 1200 on my test system with 4 threads running. Do have a Ryzen 5 1600x cpu, that is untested. Just need to fix one of the many motherboards that I have.

[Scarfy]

I've missed the start of the marathon , but i'm all set up for the javelin, 32 cores crunching away.

[leadeater]

I think we may need to actually push hard on the next Javelin throw, Sprint is going to be a CPU project (I assume) and I predict a likely overlap between Sprint and one of the Javelin throws so it might be easier to do the next up coming one and skip the one during the sprint.

 

Thoughts?

[Favebook]

Is there no easy way to bunker some projects while still uploading others? It's been long since I've done this. I guess it is not possible with single instance of BOINC manager, am I right?

[leadeater]

6 hours ago, Favebook said:

Is there no easy way to bunker some projects while still uploading others? It's been long since I've done this. I guess it is not possible with single instance of BOINC manager, am I right?

Well there is if you planned ahead. You can run multiple instances of BOINC on a computer and attach each one to a different project then suspend projects on all the other instances than the one you want to actually have doing CPU time. For each instance you can turn on or off network so that means you can finish up everything you have gotten, leave network off and then let another instance run and get tasks or w/e.

 

This means that you can have multiple instances doing the same project i.e. you've finished everything so run up another instance to get more.

 

There's a lot you can do, but that's all down to how much effort you want to put in and time you have.

 

Failing the above just edit your host file and change the project URL to point to 127.0.0.1, or if you have a firewall block it there.