Posts by Jim1348

1) Message boards : Rosetta@home Science : IPD and AmgenScience (Message 91317)
Posted 11 days ago by Jim1348
Post:
Very interesting, but I am not sure how much of that applies to us. Will they be sending that type of design work here?
We don't really have an overview of how much work is in the pipeline, or where it comes from.

Should we be building more machines? If they are not needed, it just adds to the power load.
2) Message boards : Number crunching : Good bye Rosetta !!! (Message 91191)
Posted 1 Oct 2019 by Jim1348
Post:
The credit system here makes 0 sense to me. It seems like if a workunit of the same type is completed by quite a few people, the credit given is drasttttttically reduced, regardless of the machine you're on, which makes it hard to benchmark and very frustrating especially since it makes one wonder if their CPU cycles are actually contributing to anything or if you're just in limbo.

Me too. But I originally got about 200 points per WU on the first set of work units on my i7-9700 (Ubuntu 18.04) when running the 24-hour work.
I just tried the 12-hour work units, and got 977.04 points. I like that better.
https://boinc.bakerlab.org/rosetta/results.php?hostid=3735813

It is not the credit per se, but I have no idea what is actually going on. I would think a big (BIG) project like Rosetta could do better.
3) Message boards : Number crunching : Good bye Rosetta !!! (Message 91185)
Posted 1 Oct 2019 by Jim1348
Post:
Jim, I see you got a Ryzen 3700x. I'm very curious to see how that pans out for you.

I had high hopes for it. It started out very well, around 1200 points for the 24-hour work units. Then after a few days it dropped down to around 200 points, so I stopped.

Ugh. I'm looking at this for my next system. I get more than that now using an FX8370

Sometimes the shorter (12 or 18 hour) work units do better. I think the scoring system is off track. The Ryzen 3700x is a nice chip, and there are plenty of other uses for it if it doesn't work out for Rosetta.
4) Message boards : Number crunching : Good bye Rosetta !!! (Message 91151)
Posted 22 Sep 2019 by Jim1348
Post:
Jim, I see you got a Ryzen 3700x. I'm very curious to see how that pans out for you.

I had high hopes for it. It started out very well, around 1200 points for the 24-hour work units. Then after a few days it dropped down to around 200 points, so I stopped.

I have seen this behavior before. In fact, it is the rule rather than the exception, as I have posted on multiple times elsewhere. My best (most consistent) results are with an i7-3700 or Ryzen 2600 running Ubuntu. I am presently trying a Ryzen 2700 (my Ryzen 1700 produces segmentation errors, even though it is one of the "fixed" ones).

It is possible that the shorter work units (maybe 12 or 18 hours) would be more consistent. I think it has something to do with their scoring function, but it is not possible for me to troubleshoot it.
Give it a try; it might work for you. The Ryzen 3700x is a very nice chip; it is doing very well on WCG now.

NOTE: My i7-9700 was equally bad if not worse; it was always 200 points. So it is not Intel versus AMD.
5) Message boards : Number crunching : Good bye Rosetta !!! (Message 91145)
Posted 22 Sep 2019 by Jim1348
Post:
An interesting thing I found working with different projects, and I'm not sure if this is due to excessive disk swapping, but throwing a project like Asteroids at this Ryzen 1800x spikes temps up past 65 c and the package power is right around 120 W. Throwing Rosetta at it, temperatures plummet down to around 55 steady and the package power slides down to about 90 w. The fans are at a lower RPM too. Does this mean I'm losing efficiency somewhere along the line? I have tried setting Rosetta to only use 12 threads, but that didn't seem to bring out much of a change in power or temperatures. I'm not sure if I should be looking at the number of decoys found in each workunit or credit granted or something else.

I am running a Ryzen 1700 (all 16 cores) on Rosetta (Ubuntu 18.04), and Psensor/CPUTIN shows 48C, so I guess it is normal enough. As for "efficiency", that is a slippery concept. All I know is that BoincTasks shows the CPU% at 99.5% on all the cores, so that looks good to me. Some programs just exercise the chip more than others. For example, those that use AVX2 usually run hot. I don't think there is anything you can do about it. That is for the program developers. Just let it do its thing.
6) Message boards : Number crunching : Rosetta 4.0+ (Message 91141)
Posted 21 Sep 2019 by Jim1348
Post:
Reason: Out Of Memory(C++ Exception) (0xe06d7363) at address 0x761BC5AF

You have 4 GB of memory. Windows will need 1 GB, and each work unit can take 1 GB (sometimes more).


No. It's a 8gb machine (Intel(R) Core(TM) i7-6700 CPU, memory 8088.59 MB)
So, i have to crunch 3 wus with 8 gb of ram?
I think it's better if they work on memory optimization/pointers/etc

I was referring to the post by James W. His machine is 4 GB.
https://boinc.bakerlab.org/rosetta/show_host_detail.php?hostid=3710630
7) Message boards : Number crunching : Rosetta 4.0+ (Message 91112)
Posted 12 Sep 2019 by Jim1348
Post:
Reason: Out Of Memory(C++ Exception) (0xe06d7363) at address 0x761BC5AF

You have 4 GB of memory. Windows will need 1 GB, and each work unit can take 1 GB (sometimes more).
So don't run more than three at a time.
8) Message boards : Number crunching : Good bye Rosetta !!! (Message 91094)
Posted 7 Sep 2019 by Jim1348
Post:
I don't think you need to worry about the virtual memory. That is just the disk space it might require, and I don't know if it actually uses it anyway. As for working memory, 1 GB per work unit is enough. Occasionally you will see a work unit (probably in development) that uses more, but never all of them at once that I have seen. I think even if it did occur, the worst that would happen is that some work units would not run until more memory became available.

However, I like lots of memory too, and it is cheap, so why not? It might be useful later.
9) Message boards : Number crunching : Good bye Rosetta !!! (Message 91092)
Posted 6 Sep 2019 by Jim1348
Post:
Now that's very odd and unfortunate. You'd think Intel's newest would be walking all over the place.
Assuming all of these machines have enough ram to get the job done? I have a ryzen 1800x here but only has 16 gb of ram so I think I'll be reserving 10 or so jobs at once instead of 16.

They all have at least 16 GB, and the i7-9700 has 32 GB. It was sized for the new CPDN work units, but I thought I would try it on Rosetta until they become available. But no luck; it is now on LHC. It may be an oddity of the Rosetta scoring system.
10) Message boards : Number crunching : Good bye Rosetta !!! (Message 91090)
Posted 6 Sep 2019 by Jim1348
Post:
I have various degrees of success with different machines, and can't always predict which will work the "best". For me, that means the highest and most consistent output on the 24-hour work units.

But for the past few months, my Ryzen 2600 (Ubuntu 18.04) has been doing very well; the best I have ever seen.
https://boinc.bakerlab.org/rosetta/results.php?hostid=3718408

My i7-3770 does very well too, but for some reason my i7-4770 and 4790 are falling down on the job. It may just be a scoring abnormality by Rosetta. And my i7-8700 started out well, but then fell down. Most surprisingly, my i7-9700 (8 full cores), which should do the best per core, is doing consistently badly, at about 200 points per day instead of around 1500 as it should. So I will just go with the Ryzen.
11) Message boards : Rosetta@home Science : Alphafold visits IPD (Message 91058)
Posted 24 Aug 2019 by Jim1348
Post:
Remarkable stuff. I wonder how that will affect our work? Presumably they will want to use something like AlphaFold for some types of predictions, and move off to other work here?
12) Message boards : Number crunching : Unrealistic expectations (Message 91050)
Posted 20 Aug 2019 by Jim1348
Post:
cc_config options require a client restart as it is only read at client startup. app_config can be read.

Thanks. I usually just reboot anyway.
13) Message boards : Number crunching : Difference between applications (Message 91046)
Posted 18 Aug 2019 by Jim1348
Post:
It is mainly a small difference in how they implement the science. But the "mini" doesn't mean anything. They run the same on our machines.
14) Message boards : Number crunching : Unrealistic expectations (Message 91044)
Posted 18 Aug 2019 by Jim1348
Post:
The initial BOINC estimates are way off, about 4 hours that I recall (I run the 24-hour work units). So until they correct themselves, I always get too many when first attaching.

To speed up the correction, you need a cc_config.xml file placed in the BOINC data folder, with this entry:
<cc_config>
  <options>	
  	<rec_half_life_days>1.000000</rec_half_life_days>
  </options>
</cc_config>

You then need to activate it by reading it in, or else restarting BOINC.
It should correct the time estimates in a couple of days.
15) Message boards : Number crunching : Rosetta 4.0+ (Message 91035)
Posted 16 Aug 2019 by Jim1348
Post:
Still nothing here. My machines are crunching, exporting, end of story. Windows 8.1 x64. Something strange is happening.

I wasn't seeing them here either. Then I got one (Ubuntu 18.04).
https://boinc.bakerlab.org/rosetta/result.php?resultid=1087794059

So they happen, just not very frequently. It only took 10 seconds, so I don't think the scientists will be spending much time trying to fix it.
16) Message boards : Number crunching : Good bye Rosetta !!! (Message 91013)
Posted 9 Aug 2019 by Jim1348
Post:
So, Roseta is not the most important ! .

There are other important ones, and I do them too. But your examples are so weak that the comparison shows the importance of Rosetta.
I just added an i7-9700. Have fun elsewhere.
17) Message boards : Number crunching : Good bye Rosetta !!! (Message 91002)
Posted 7 Aug 2019 by Jim1348
Post:
The only recent (16july19 by admin) is about "audacious projects". Nice, but is not a part of Boinc.

It is fabulous news, and great for Rosetta.
I think it is entirely possible that Rosetta is now the most important BOINC project there is.


As you may have heard, the Institute for Protein Design was recently selected as part of The Audacious Project. This large-scale philanthropic collaboration, which is the successor to the TED Prize, surfaces and funds projects with the potential to change the world.

As a result, we are expanding our Seattle-based team of scientists and engineers who will work together to advance Rosetta, our software for protein design and structure prediction. The funding will also allow us to invest in the equipment, supplies and lab space needed to design and test millions of synthetic proteins.

http://boinc.bakerlab.org/rosetta/forum_thread.php?id=13199&postid=90915#90915
18) Message boards : Number crunching : Rosetta 4.0+ (Message 90991)
Posted 6 Aug 2019 by Jim1348
Post:
From their point of view, it may not be that serious. They are still sending out work and accepting results, so it is only the people who look at "validation pending" who see the problem.
It may just be that they are upgrading their server software and don't see any reason to notify us. That is not ideal customer relations, but not a crisis.

EDIT: we are back in business, and my credits are up.
19) Message boards : Number crunching : Rosetta 4.0+ (Message 90989)
Posted 6 Aug 2019 by Jim1348
Post:
Mod.Sense is a great, but he is not an admin or a developer of the project.

Undoubtedly so. But if rjs5 can't get their attention, we have no hope.
20) Message boards : Number crunching : Rosetta 4.0+ (Message 90986)
Posted 6 Aug 2019 by Jim1348
Post:
In this thread we don't see admins since....i don't remember. I don't know even if they are reading this thread.

They hang out over here:
http://boinc.bakerlab.org/rosetta/forum_thread.php?id=1000


Next 20



©2019 University of Washington
http://www.bakerlab.org