Just wondering if it makes a difference whether one or 4 WU's are running simultaneously. Also wondering what people generally do with their resource share option.
Thanks!
Printable View
Just wondering if it makes a difference whether one or 4 WU's are running simultaneously. Also wondering what people generally do with their resource share option.
Thanks!
I run one wu per core. I have an AMDx6 that I run six wu's on and I run eight wu's on my Intel Sandybridge 2600k. With some Intel chips you can run 2 per core. As far as settings I crunch all the time and use my GPU with computer is in use. Unless I'm gaming. ;)
Hope that helps.
Ya what Mike said. I just wanted to ass that Intel CPUs with Hyper Threading a.k.a. HT (info Link) allows a CPU core to process two threads at once. In general terms for crunching this works well but overall productions depends on the projects app and how well it is coded for optimization. Sloppy code works well with HT. For apps that are highly optimized for efficiency, it is recommended that the user disable HT in system BIOS to prevent the system from bogging down.
Personally I always have HT enabled on my two i7 boxes. Simply because its just a pain in the butt to keep enabling/disabling HT except for specific events like PrimeGrid Challenges. Although for the current challenge I forgot.
One other thing to note. Most projects, if not all, have a preference for a particular CPU brand as well as Operating System. This is dependent on how the app was coded as well as what type of work the app does. Though I do not keep track of such preferences, some of the Team Gurus like trigggl, zombie67, JPM and others are quite knowledgeable on such matters.
Thanks for the info!
I have and i5 2500k so that would be why my machine is running four tasks, one for each core?
Another question I have, that i haven't been able to find an answer for so far is; why do some tasks run to 90ish % completion and then start another task? Is there a way to force the BOINC manager to do a task to completion?
That may have to do with your computing preferences. Go to Tools then at the bottom select computing preferences, at the first tab (Processor usage) go to the bottom where is says "switch between applications every ____ minutes" I believe 60 minutes is default. I usually leave it at 200 minutes which is more than enough time for most projects. Also remember to check the box a little higher up on that same tab, "Use GPU while computer is in use" then you'll see the points fly. :cool:
Sorry about that :-x
I see that no one has mentioned that if you want some extra points you should attach to Freehal, WUprop, and Surveill as these are non cpu intensive projects and only use very little of your computing power. With Freehal you will have to go to the website and check the boxes to get the non cpu intensive work units. All 3 play very nice with the other projects.
Oooops. Forgot again...... :PQuote:
Originally Posted by c303a
Note: Some people have commented that Survill@Home can be network intensive at times.
We'll forgive you.....this time.
Awesome guys thanks for the info.
I've upped the time it takes to switch between projects so now they should finish haha. I'm also now crunchin Freehal, WUprop and Surveill. I think those projects will also be ideal for my old duel core lap top i'm setting up as a cruncher.
Any other suggestions for programs to crunch on an older machine?
You have to decide if you are going to go for as many points as you can or are you going to crunch for the science of it and forget the points. Go to the tools, add a project in boinc manager and attach to a project. As you scroll through the projects you can get a thumb nail sketch of what the project is. The info is also available on our home page. Pick out the ones that interest you and we can let you know which are best for your computer. Also, what type of video card do you have. They can also be used to crunch certain projects.
Well, if it has enough storage, I'd recommend Test4Theory. But that requires you install VirtualBox on the system and will require at least 512 meg of storage plus 256 Meg of RAM for the VM. The advantage for older boxes is that Boinc doesn't know how much work it accomplished in the VM, so all boxes, fast or slow, P3 or i7, get the same credit based on the time spent crunching. (At least that's what I've heard. ;) )Quote:
Originally Posted by shralper
again
I'd like to crunch a bit of both some for pure science, and some in accordance with team goals for max points, the competitive aspect appeals to me. My main computer has a good spread of such projects and i'm utilizing my GPU. I was looking into whether or not there were some specific projects that do better on older machines, which Mumps answered.Quote:
Originally Posted by c303a
I wish my signiture was more reflective of the current projects I was in so you guys could help guide my choices but my current signature itsn't representative of all the projects I'm crunching and points shown in my BOINC manager. I'm guessing this is a result of some of my projects and stats not showing up yet in my BOINC BAM manager yet. I'm assuming I'm just being to impatient haha.
I think sometime that is a BOINC problem. I have see where BOINC is working on a WU and you d/l more new ones and BOINC will stop working on the current one and jump on one of the new ones. It seems that BOINC "thinks" it won't finish the new ones on time so it jumps to them.Quote:
Originally Posted by shralper
BOINC has always had this problem but it seem to be getting better at not doing it.
I tried to make a Test4Theory account but it says I need an invitation code?
nm, I found where to sign up for them
Since I'm trying to play with maximizing output for a bit, I have a few questions here.Quote:
Originally Posted by Fire$torm
Running the I5 sandy bridge, so I am trying to run PPS seive till prima or dirt come back as well WU and freehal. Surveil has never been up since i came back. So if I remeber correctly I actually only have two cores and then HT.
So a few questions:
1) Is the code on pps seive optimized to where I will see a noticeable difference by turning off ht.
2) I only get around 8k credits on a good day 1500 of which are freehall, so turning off HT would cut that in half (I would think) is it still worth it at this point?
3) And lastly if I am going for max credits even though freehall and wu are nci they have to use some cpu, would running pps seive and only pps seive be more effective?
Thanks in advance,
N1
I cannot answer your question directly as I never disable HT on my i7s. I do know that Team SG will do so during PG challenges. And IIRC they also suspend FreeHal & WUProp for challenges.Quote:
Originally Posted by N1gh7m4r3
For us the catch is that we are always working on FreeHal & WUProp as we are in a fight on both projects.
FreeHal ---> http://stats.free-dc.org/stats.php?page=teams&proj=hal
WUProp ---> http://stats.free-dc.org/stats.php?page=teams&proj=wup
People have done comparisons in the past, i.e. 6 months or more ago, and at that time it was deterimed that hyperthreading gave 2-5% more credit than non-hyperthreading. So, an i7 running 8 tasks was better than an i7 running 4. If the question is whether running 4/8 WUs verses one multi-processor WU (e.g. an i7 runs a single WU on all cores at one time), it depens upon how the project grants credit.
I created a multi-processor CPU app for Collatz and using the same credit calculations, the overhead of multi-processing means it grants LESS credit than running multiple WUs on multiple cores/threads. How much? 2-3%. Also, the multi-processor apps often don't play nice if also running GPU apps. Running an OpenMP version of Collatz along side the GPU apps resulted in a 5% loss on the CPU and a 10% loss on the GPU.
AQUA, which is no longer active, only had multi-processor apps and paid more credit for apps using more processors since running those tasks meant dedicating more resources to AQUA ... and they ran a long time as well so they felt they should reward their crunchers accordingly.
If running both CPU and GPU apps, it may even pay to change your BOINC preferences to use N-1 CPUs (e.g. 87.5% on a hyperthreaded i7) in order for the GPU to never have to wait for a CPU. On some projects, the GPU more than makes up for the lost CPU credit.
In other words, it all depends. Ask for a specific project and someone has probably tested it both ways and will be able to help. But, in general, one task per thread/core tends to pay more.
Thanks for the info guys
---
I am here: http://maps.google.com/maps?ll=40.570168,-85.659670
Sent from my iPhone 4s using Tapatalk