World Community Grid

Last edited:
I was wondering when a Zika project would start; it's about time! Seeing photos of the babies afflicted with Zika absolutely breaks my heart. :( Quick glance at a couple of my boxes show some Zika WUs in the queue, so I'm ready to go! :D
 
Looks like OpenZika changed their badge. I like that they changed it as it was too similar to DDDT and DDDT2 badges
 
Also, in case anyone still needs to obtain a certain level in goals but haven't been paying attention.
~ days left
MCM - 71 days
UGM - 44 days
OET - 6 months
 
Discussion moved for 4P Intel system performance information. I've been mulling over the idea of getting an Intel 4P rig, and was wondering what thoughts you had on the performance of the following motherboard with either of the CPUs:

Intel S4600LTM S4600LH Server Board 2U Rack, Socket R, DDR3 New Board Only

4650L ES:

Intel Xeon E5-4650L ES 2.6 GHz QC6P 8 CORE

4650 ES:

Intel Xeon E5-4650 ES version QBED 2.7GHz 8 CORE Processor CPU LIKE SR0QR

System would likely be running naked as the board looks proprietary. 4 matched CPUs, a bit of RAM, and a good Corsair PSU. On-board video is present, as is IPMI. Does anyone have experience with Intel boards like this, or this specific model? I'm concerned their duds and that's why they're on sale, but I can't find any real reviews or complaints. Looks like they were sold in a matching 2U server. Anybody folding or crunching on one of these? If so, would the 4650 be a good choice? The ES chips seem to be fairly cheap when found. Might be able to piece together a system for under $1,500.00.
 
As an eBay Associate, HardForum may earn from qualifying purchases.
I have Dell R820's with:
E5-4607 v1
E5-4640 v1
E5-4640 v2
E5-4650 v1
​
One R830 with E5-4640 v4

One R930 with E7-8867 v3 (this box is pissing me off though; refuses to stay at 100% load. Might be a BOINC thing, not sure)

Tell me what you're after, and I'll see if I can get you some data.
 
Gilthanis said:
And don't forget that WUProp WUProp@Home has a lot of helpful information as well. Report limit
Click to expand...

Couldn't find any 4P Intel stats: CPU performance


Comparable?:

CPU model Number of computers Avg. cores/computer GFLOPS/core GFLOPs/computer
Intel(R) Xeon(R) CPU E5-2650 v2 @ 2.60GHz [Family 6 Model 62 Stepping 4] 12 13.25 2.98 39.49

The stats seem to bounce around wildly, and I have a sneaking suspicion they average in idle time and unused cores instead of basing it on a system designed for 24/7 crunching.

Actually I think I may have found one, it's not listed as an E5:

Genuine Intel(R) CPU @ 2.70GHz [Family 6 Model 45 Stepping 5] 10 44.80 3.31 148.20

3.31 GFLOPS/core 148.20 GFLOPs/computer might not be that bad. If the QBED ES chips rate the same,

The stats on the 2P 2620 systems I'm running seem to be on that page:

Intel(R) Xeon(R) CPU E5-2620 0 @ 2.00GHz [Family 6 Model 45 Stepping 7] 28 7.07 2.53 17.90


So I'd be looking at an 8-fold increase in GFLOPs/computer? That seems unlikely.

If I base it on raw GHZ, a 2P 2620 without HT yields 2 CPU X 6 cores X 2.0 GHZ = 24 GHZ
4650 at 2.7 per core 4P system without HT yields 4 CPU X 8 cores X 2.7 GHZ = 86.4 GHZ

The yield should be on or around 3.6 times a 2P 2620 V1 rig. A hefty increase, but not an 8X increase to be sure, right?

Given that they're both 1st generation, came out around the same time, wouldn't the technology be similar enough that the process should be consistent, given a standard margin of error?

This makes me think that the numbers have too many factors involved such as the priority of the work units, the number of hours per day put forth, what else the servers may be running simultaneously, or some other factor I'm just not thinking of. I'm no architecture expert, but the numbers on that site don't seem to add up to me. Especially when you factor in this:

Intel(R) Xeon(R) CPU X5675 @ 3.07GHz [Family 6 Model 44 Stepping 2] 19 23.37 4.02 94.00

A 2P Xeon 6-core X series absolutely wiping the floor with a 2P 6-core 2620 rig?

I could buy an entire server with two of those chips for $400! What makes them so special? I think this is why I avoided BOINC for so long. Too many factors, and I hate math these days anyway.

Another head scratcher. These were right next to each other on the list:

Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz [Family 6 Model 63 Stepping 2] 43 6.88 1.32 9.10
Intel(R) Atom(TM) x5-Z8350 CPU @ 1.44GHz [Family 6 Model 76 Stepping 4] 11 4.00 1.29 5.18

Am I to believe that an E5-2699 has the same GFLOPS/core as an Atom? That boggles the mind.


fastgeek,

I'd be interested in the WCG performance of the 4650 V1 4P system. It's essentially what I would be building, albeit using ES chips and hopefully they'd be compatible with that Intel mobo in question. I have no idea how to validate that. I don't think Intel would help me identify which ES chips would work on what board, so it might be a risky purchase. But if the numbers are tasty enough I might bite.
 
GFLOPS/core.... is tricky. The BOINC client does not take into account things like AVX and Turbo modes properly. So, you can't make a judgement call based on that. Yes some projects have used scoring systems based on those and you will find people that have pointed out the flaws. Some projects fix the issue with a different scoring mechanism and others do not. WCG uses the latest scoring mechanism (I believe a hybrid of it anyways) and doesn't really rely on the benchmarks too much.
 
The other thing is that, as I've recently noticed, not all WUs are the same. Some OZ WUs are finishing in ~3 hours and others in ~5. (despite the fact that they both have the same estimated GF per the WU properties.)

MGMCCALLEY, most of the time you'll find that a good number of other folks have ran ES chips on just about every board under the sun. I agree that it's a 4P system is a bigger risk though. Am sure you've done this, but I'd search for something like - MB MFG MODEL E5-4650 QS ES and see what comes up; or maybe drop the processor model from the query and just see what might be out there for the MB with ES / QS chips. Anyway, if you have some specific things you'd like me to check, tell me what they are and I'll do my best to get that info. :)
 
fastgeek said:
The other thing is that, as I've recently noticed, not all WUs are the same. Some OZ WUs are finishing in ~3 hours and others in ~5. (despite the fact that they both have the same estimated GF per the WU properties.)

MGMCCALLEY, most of the time you'll find that a good number of other folks have ran ES chips on just about every board under the sun. I agree that it's a 4P system is a bigger risk though. Am sure you've done this, but I'd search for something like - MB MFG MODEL E5-4650 QS ES and see what comes up; or maybe drop the processor model from the query and just see what might be out there for the MB with ES / QS chips. Anyway, if you have some specific things you'd like me to check, tell me what they are and I'll do my best to get that info. :)
Click to expand...


I'm curious about your Dell R820 performance with the 4650 V1 chips. I'm sure it would be comparable to the equivalent ES versions. Can you give me any stats on it running WCG?
 
Also, when you do the stats, you may want to run just one sub project at a time as they all score a bit differently
 
Do Beta tests offer an advantage in credit? If so, how do we take advantage?
 
in credit...no. They are just the hardest badges to acquire and so are heavily hunted. You have to actually check the box in your preferences at WCG in order to participate in them. Basically, to get them is right place/right time. You can only get 1 per thread typically at most for each beta application in progress. There are a few techniques for getting them easier, but it usually either means manually hammering away at the server, lots of wasted resources being sneaky, or just plain putting extra strain on their servers for a colorful icon in a badge.
 
I just added the Beta to my default device profile. I assume this will update all of my clients?
 
It doesn't need to update client side. This is a project preference and thus the server dictates what it will send based on that.
 
So this section on WCG has no effect? There is a check mark to participate in the Beta Testing Tab on the WCG page after you log in. Here's the description:

Beta Testing
white.gif

title_underneath.jpg



Beta Testing is available for those people who would like to assist World Community Grid in testing work for projects before they are officially launched on World Community Grid. Some projects may need debugging and you might recognize the name of the project; others will be run anonymously. In either case, you will receive credit and points for Beta Test work just as you would for any other project. We hope this will increase the number of users we have testing projects, which in turn will allow World Community Grid to launch projects that are stable and thoroughly tested.

How It Works
You can opt in any of your existing BOINC profiles to receive Beta Test work. When Beta Test work is available, you will receive that work before you receive work for any other project (i.e., Beta Test work will take first priority). If there is no Beta Test work currently available, you will receive work from the other projects in which you have chosen to participate.

As a beta tester, we may occasionally contact you via e-mail to gather information needed to help correct errors and fix bugs. Your assistance is greatly appreciated.

To find out about current Beta Testing, please visit the Beta Testing forum.


BOINC Device Profiles

Device Profile Name Participate in Beta Testing

Default
 
That section does have an effect. However, you said update all of your clients. The clients just ask for work and the server doles it out based on site preferences rather than client preferences. So, basically there is no updating of the client needed. It will pull based on server settings which is pulled from the setting you pointed out.

Saying it updates your clients implies a change client side. That is something an account manager like BAM! might do. There are some settings at WCG that do change client settings. Those are basically similar to the same options built into your client already. However, this particular setting does not.
 
Good to know, thank you. I haven't read up on all the ins and outs of WCG like I did for F@H.
 
Total n00b here, folks - but your feedback would be appreciated!

Last month I joined the World Community Grid ... the first PC I "recruited" is an i7-based Asus laptop (i7-4700HQ, 4C/8T, 2.4 GHz to 3.4 GHz, with an nVidia GeForce GT-840M GPU, which I allow for CUDA computations, as work units for GPUs become available). I have joined this laptop to all projects (Cancer Markers, FightAIDS, Zika, Genome Mysteries & Outsmart Ebola). I'm also a member of the University of Maryland WCG team (UMD = my old alma mater)

I realize that joining WWC with a laptop ain't such a red hot idea (given the sketchy / persnickety cooling requirements), but I use the "Balanced" power profile in Windows 10, and I restrict the CPU utilization to 95% max and 92% min. This keeps my i7 CPU's max core speed at 2.2 GHz (with very rare excursions to 3.2 GHz), and in the BOINC Manager, I max the overall computation utilization to 80%. This laptop runs 24/7/365, and the temps stay pretty much locked in at around 53°C min, to a very rare max of 62°C. I use CPUID HwMonitor. I gave the cooling ducts and CPU fan on this laptop a thorough cleaning before I started, and I clean it out every month. Over the last 3 weeks, using this laptop I've already received a Silver medal for Cancer Markers research, and a Bronze medal for FightAIDS.

A couple of days ago, I added a very solid, 5 year old Toshiba laptop (core i5-2410M CPU, 2C/4T) to my WWC account, using the same CPU usage parameters as my i7 Asus, and I'm getting the same temperature readings. I also gave this one a thorough cleaning, and is cleaned monthly.

I have two questions:

1. Will I be shortening the lifespans of these laptops?

2. How "SMT / Hyper threading aware" is the WWC - BOINC client? Is the code highly optimized for intel HT?

Thanks in advance!

-Trev
 
Trev, WCG does not really have optimized apps, so you will not see much advantage to using Intel vs. AMD if that is what you are asking outside just power efficiency. So, AVX extensions really don't come into play at WCG. It will utilize the HT in your CPU's and yes that will give you more output/efficiency.

As far as shortening the lifespan. This is tricky to answer, but most laptops will see a shortened lifespan if used heavily long term as they really aren't designed for it. With that said, I have had laptops that just sat in a corner for years with little cleaning while others seemed to die with extreme care. The main thing I can say to watch out for is certainly the temps. If you use the GPU (which WCG does not use right now), then that will probably be the #1 thing to wear out the cooling fan or wear out on the motherboard. Since they typically share the heat sinc with the CPU, this can do some damage if not properly managed.

Something else to watch is if you pick up any CEP2 work units. Last I looked the sub project was down and didn't have work available. However, when it does, laptop hard drives may struggle if you set your preferences to get more than a few at a time. they are very heavy on IO's. Especially at startup. If you have SSD's, that will help a bunch.

When you say you are limiting BOINC Manager's CPU utilization... are you setting 80% under User at most XX% of CPU's or Use XX% of CPU time? I would suggest just setting the % of CPU's so that it will just limit your i7 to 6 threads rather than running 8 threads at 80%. This will have similar effects on the CPU temps but will complete work units quicker and risking them not completing on time a bit less. It will also reduce the possibility of applications getting angry if they are easily affected by starting/stopping.
 
Gilthanis said:
Trev, WCG does not really have optimized apps, so you will not see much advantage to using Intel vs. AMD if that is what you are asking outside just power efficiency. So, AVX extensions really don't come into play at WCG. It will utilize the HT in your CPU's and yes that will give you more output/efficiency.

As far as shortening the lifespan. This is tricky to answer, but most laptops will see a shortened lifespan if used heavily long term as they really aren't designed for it. With that said, I have had laptops that just sat in a corner for years with little cleaning while others seemed to die with extreme care. The main thing I can say to watch out for is certainly the temps. If you use the GPU (which WCG does not use right now), then that will probably be the #1 thing to wear out the cooling fan or wear out on the motherboard. Since they typically share the heat sinc with the CPU, this can do some damage if not properly managed.

Something else to watch is if you pick up any CEP2 work units. Last I looked the sub project was down and didn't have work available. However, when it does, laptop hard drives may struggle if you set your preferences to get more than a few at a time. they are very heavy on IO's. Especially at startup. If you have SSD's, that will help a bunch.

When you say you are limiting BOINC Manager's CPU utilization... are you setting 80% under User at most XX% of CPU's or Use XX% of CPU time? I would suggest just setting the % of CPU's so that it will just limit your i7 to 6 threads rather than running 8 threads at 80%. This will have similar effects on the CPU temps but will complete work units quicker and risking them not completing on time a bit less. It will also reduce the possibility of applications getting angry if they are easily affected by starting/stopping.
Click to expand...




Thanks for your prompt reply, Gilthanis -


Heeding your advice, I made sure that the GPU-compute options are unchecked (my Asus laptop's GeForce 840M GPU does indeed share the same heat sink pipe with the CPU).

Also, as an FYI, I am running Romexsoftware's "Primocache" on the Asus laptop (core i7, 12GB DDR3-1600 RAM), which provides additional block-level caching for the 1 TB hard drive. I set up a 3,072 MB (3GB) cache for the hard drive (with 1 hour deferred writes - file system corruption not a concern: laptop batteries + APC UPS, so file systems are double-protected against sudden power outage), so that should mitigate most of the I/O stress resulting from issues like CEP2. The Toshiba i5 laptop has a Samsung 840 Pro 256 GB SSD, so I/O stress of the nature you described should hopefully not come into play (as you pointed out).

I revisited all of the relevant settings/parameters in both the BOINC manager and Windows 10 Power Plan profiles that I set up, and that rare 63°C max temp spike is locked on rails, for both laptops.They spend most of their duty cycle time at 49° - 53°C anyway. intel's Tj(max) ratings for both CPUs is 100°C, so on further reflection I honestly don't think the CPUs or the system boards are being unduly stressed, even given that they're running 24/7/365. I'm just glad they can be put to this good use!

Thanks again, Gilthanis!

- Trev
 
Last edited:
Gilthanis said:
You can also use eFMer's tthrottle to throttle the laptops down if they hit a certain temp and it does have settings for GPU as well. BOINC useful tools and download locations
Click to expand...
nk


Yep, downloaded & installed eFmer throttle app., thanks, Gilthanis! So far I've got 6 real cores and 6 logical cores crunching on the WC Grid. I'm tempted to throw another laptop (the one I'm using here) onto the WC Grid. It's got an i7-6700HQ quadcore CPU, 4C/8T, 2.5 - 3.5 GHz. That would bring my total Grid "compute node" to 10 real cores and 10 logical cores. I've turned my house into a regular Biomedical Research Compute-Cluster center.
 
For those trying to obtain a specific badge or goal, UGM has ~32 days worth of work but that could change drastically as people realize they have little time to hit their goals and up their contributions.
 
So I have the 8350 and 2500K up, but the 4790 went to the son since his 2500k has been unstable.
Would it be worth replacing the 4790 machine with a 390x that is collecting dust?
What would be some good work for it if so?
Also have it running on the 5820k for fun. Still a bit warm in the bedroom.
Thanks!
 
Some of those Beta units run into the 18 hour limit. Researches use data from that time so its still useful. If anyone gets WUs there are upload issues over the weekend with sending one of the files to Harvard. I have one that won't complete the upload.
 
With POEM going down, the only bio project for AMD GPU's would be FAH. If you want other suggestions, then it depends on if you care about the science or just wanting something to push. Projects using double precision and thus AMD outperforms nVidia is typically Milkyway or PrimeGrid's GFN work. Otherwise, we haven't had a lot of push at Einstein for a while.
 
Gilthanis said:
With POEM going down, the only bio project for AMD GPU's would be FAH. If you want other suggestions, then it depends on if you care about the science or just wanting something to push. Projects using double precision and thus AMD outperforms nVidia is typically Milkyway or PrimeGrid's GFN work. Otherwise, we haven't had a lot of push at Einstein for a while.
Click to expand...
Thanks! I popped the 390x into a 2500k system. I will put that system on Milkyway for the gpu power and I will run Einstein on my main machine.
 
I'm down to the 4P for now, but in a couple of weeks I'll have two 2P 2670 rigs to add to the mix.
 
Just curious. Do we plan to participate in next month WCG team challenge issued by SG? Something about WCG 12th birthday. Not sure if there is any interest in this. Maybe by participating will keep our members motivated. Well, I'll leave this up to our team captains to decide. :D
 
We will be participating. However, if we plan on asking anyone to move CPU hardware away from any other agenda that we make sure it is for the right challenge. We had traditionally participated in the WCG x-mas challenge which is almost the entire month of December. I don't mind running both, but want to make sure that if people are only willing to shift for one, that we focus on the same one. I personally feel that the Birthday challenge has a much bigger turnout these days and is probably the harder challenge. That would be my focus if I had to choose.
 
Gilthanis, I'll support the team if we want to participate in this challenge. We have 50+ active WCG crunchers though our recent two big heavy hitters are taking a break. Does all member contributions automatically count if the team enters this challenge? This is all I can find. Maybe start a new thread to see if there is an interest for this or the x-mas challenge or both. Thanks.
 
You must log in or register to reply here.
Back
Top