| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: ud vs boinc, pros and cons hmmm switch or not? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 16
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
hmm i am playing with various setups, i have 3 puters, just for fun i am trying out boinc on an xp machine, hmm i have not unistalled the ud but i think i might have to otherwise there is a downloaded w/u sitting there, i guess iwill have to let it time out, seems a bit of a pain that i cant tell wcg about this - one of the nice things about boinc i can abort a w/u if need be or request no additional work etc..
so i have ud interface on my other xp machine, i think i will leave it there for interest and comparison sake.. 3rd machine has linux but i am testing suse instead of ubuntu - i will either let unbuntu install time out or let it complete work and swap the hard disks back to the suse harddisk,, not sure wot is best i am thinking to abort the w/u unless of course there is a big chunk of crunching already done, idont feel like taking the lid of the linux box tongiht iam tired after a bust day/weeek at work.. would be interesting to see wot others think, there are other mentions about "the joys of boinc" saying it is a pain waiting forothers to complete w/u b4 getting points, i am not fussed about this, however it is nice to get some points .. i see pros and cons for each setup, i am thinking of recommended the ud one for newbies as probably the simlest implementation etc but those on linux pretty much have to run boinc, i have heard of others doing workarounds with "wine" to run ud but as boinc clinet is here that seems rather pointless now... , hmm rambling on .. bfn |
||
|
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges: 
|
Hi retep57
----------------------------------------points allocation aside (which is now surreal) BOINC is the better client for the number crunching that needs to get done. If you are a newbie to crunching and you have a Windows Box then UD is the ideal introduction, pretty colours, pretty graphics, cool screen saver and your points are allocated in a nice consistant manner. BOINC is more of an aquired taste slightly more complicated to install and attach to the WCG, but they are working on it and not so much "instant feedback" but... You can run it as a service officially meaning that you can run it while not logged on or indeed invisible in a corporate environment (not suggesting that you should load it onto a works machine without permission, just saying if you are a systems admin and allowed to do such things then the end user doesn't need to know it's running) It has built in the capability to download work units in advance (like UDMon except it's built in) this means that even on the rare occasions that the WCG service is off air you can still continue crunching as BOINC provides a reserve. This is configurable through the device settings on the WCG website. Because of this it is slightly more efficient as you don't have to wait for the Work Unit to upload before you can continue crunching - this way more crunching gets done. You can configure BOINC to run while on Laptop Batteries. This may not be ideal but if you are able (I now crunch on the train with my laptop) you now can. With UD the crunching stops as soon as you are not connected to the mains electricity. BOINC is multithreaded. If you have more than one CPU in your PC/Server (or one of those Multicore CPU's) BOINC can take advantage of that 2nd or 3rd 4t etc CPU whereas UD cannot. BOINC runs on Windows, Linux and now MAC's with PPC Chips whereas UD is only Windows BOINC will run the FAAH program if your PC has less than 128MB of RAM whereas UD will only run HPF with more than this amount of Memory All those whirling Graphics in UD use up CPU cycles which could be producing results - the plain GUI of BOINC uses less CPU time to display. I think you can manage multiple crunchers through the BOINC Manager whereas UD is a one PC operation. I'm sure there must be even more reasons to switch (apart from the alarming number of extra points you get running BOINC) Anyone else like to contribute? Dave, a BOINC Fan  p.s. Don't worry too much about the Work Units you have running on your machines They are handed out to at least 4 others so no Work Units will be "lost" as you make the change. Enjoy your crunching  |
||
|
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18667 Status: Offline Project Badges: 
|
Dave provides a very good run down. I would recommend UD for folks that have old low-capability machines that are Windows only or for folks that are unfamiliar with computers and prefer to "push a button and forget about it". If you have a pretty recent machine, you'll probably do better to run BOINC. If you have a high end or multi-core machine, you definitely want to run BOINC. If you want to have the most flexibility in managing how you crunch, go with BOINC. As Dave points out, BOINC will run on battery power - very handy for mobile folks. UD will download a WU, crunch on it, upload when done and download a new one and so on. That doesn't give you much flexibility without getting involved in using the "unsupported" add-ons like UDMON. With BOINC, if you're going on holiday for a week and want to keep crunching but don't want to leave your machine connected to the net, you can relatively easily get enough work to cover that time before you leave. If you have to run an app on your machine that's CPU intensive and perhaps sensitive to other things running, you can easily suspend your crunching with BOINC (yes it runs low priority like UD too). If you're really into DC, you can crunch for multiple projects, not just WCG, with BOINC and you can "allocate" how much of your machine's time each project gets. I only crunch WCG myself on a single device. I'll agree to disagree with Dave about the "fairness" of BOINC points. You really can't do an apples to apples comparasion of BOINC and UD as they are fundamentally different. I'm convinced that BOINC's multi-threading helps on even a single machine. It's a more efficient cruncher because of that. As for the way WCG converts BOINC credits to WCG points, I'll trust the WCG admins to have it as fair as it can be. It seems that UD limits the points a machine gets on more machines than it doesn't because of the limits in it's benchmarking. I feel that UD was underreflecting the real amount of work being done by most machines rather than BOINC overreflecting it. That's just my opinion though. What really matters is what works best for you and what you're most comfortable with. Bottom line is keeping on crunching.
---------------------------------------- |
||
|
|
David Autumns
Ace Cruncher UK Joined: Nov 16, 2004 Post Count: 11062 Status: Offline Project Badges:
|
Nice one keithhenry
---------------------------------------- Thanks |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
the more i explore boinc the more i prefer it over UD. i do agree that UD is good and esp good for newbies.. is interesting to compare how others go crunching, see how long it takes for the same w/u on competing( collaborating prob better word) computers...
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I'll agree to disagree with Dave about the "fairness" of BOINC points. You really can't do an apples to apples comparasion of BOINC and UD as they are fundamentally different. I'm convinced that BOINC's multi-threading helps on even a single machine. It's a more efficient cruncher because of that. BOINC doesn't crunch. Neither does UD. Both are just managers for the program that does the actual crunching. I presume the program that crunches for BOINC is the same program that crunches for UD. I agree that BOINC has some nice features that UD does not have but I have been running an experiment that I feel totally disproves your assertion that BOINC is a more efficient cruncher on single CPU machines. I haven't concluded the experiment. It has run for about 48 hours and I plan to let it run for another 48 hours. However, I am convinced that the results even at this early date show that UD is just as fast as BOINC. I also feel it is quite evident there is a huge difference in points awarded though for me that's not a big concern. For my experiment I have setup 2 identical computers, 1 with BOINC, the other with UD. Both computers are the same model from the same manufacturer and they benchmark almost exactly the same. Both have been crunching for about 24 hours now, neither has been doing anything else, only virus protection running in the background. Both have hyper-threading, CPU is AMD Athalon64, 2.4 GHz. The machine running BOINC has returned 8 WUs and the Average Run Time Per Result reported on the WCG site is 5:08. The machine running UD has returned 10 WUs for an Average Run Time Per Result of 5:13. You might say I have just shown that BOINC is faster than UD but.... 24 hrs ago the BOINC machine led the UD by 30 minutes on the Average Run Time Per Result score. It seems safe to say that over the long run BOINC's lead will diminish to a negligible figure. The next 48 hours will either lend credence to that extrapolation or not. As regards points.... the BOINC machine has Average Points Per Result of 479, the UD machine has 189. Those figures have changed a little in the past 24 hrs but not much. Now, to be completely fair, I must admit there is one difference worthy of mention between the 2 machines in the experiment. BOINC is running under 32 bit Windows XP whereas UD is running under 64 bit Windows XP. UD is thus running under WoW (Windows on Windows) which is Microsoft's virtualization of the 32 bit environment for 64 bit platforms. I am not certain but I feel that WoW can only slow applications down and I feel that UD is thus running at a disadvantage compared to BOINC. Any comments? -- |
||
|
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18667 Status: Offline Project Badges:
|
Well, I haven't read source code from either. Depending on the machine you pick, you can show that UD is better and you can show that BOINC is better. There's a lot more than just CPU speed involved. Cache availability and size, RAM, paging and other things that are also outside my expertise. I've chosen to believe that multi-threading gives BOINC an advantage simply looking at it conceptually. If you can keep crunching in the CPU at the same time IO is being done for example, you'll do more in the same amount of time. That won't always be true in every case. You put a big enough turbocharger on a small enough engine and I would expect you could easily be wasting your money. On the other hand, do that right and you can have a riding lawnmower doing 50 MPH. There is no way to do a true apples to apples comparasion of UD and BOINC. All we can do is look at it in general. Based on experiences of folks posted in these forums, the low end machines tend to do better with UD and the high-end machines tend to do better with BOINC. Even here though, UD and BOINC are different enough fundamentally that we can't even agree on how to define "better". The best bottom line I can offer is go with what makes you happy as you choose to define that.
---------------------------------------- |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I have to agree with your bottom line recommendation to go with what makes you happy. Happy is good!
BTW, I haven't read the source code either but I don't think that's necessary. I have to question your assertion that one can show BOINC is better depending on which it is run on. Note I said question and not disagree. I question because I have seen more than a few claims that BOINC is faster but I have seen no solid evidence that it really IS faster on any platform. Can anybody point to any reasonably well designed test runs that compare BOINC with UD? The conceptual (rational) argument you offer is fine and I cannot disagree with it or discredit it from a conceptual position. However, I believe the results from the experiment I have reported show that inspite of BOINC's multithreading, UD is just as fast. I can't tell you why that is so but the results say it is so on the 2 identical computers I have used in my experiment. Those computers have HT which should, according to popular belief, give BOINC an advantage. But it does not, at least not on the computers in my experiment. Are there any case studies done on other platforms that demonstrate BOINC is faster? I mean case studies that look at no less than 10 WUs. -- |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
i would be delighted to run a test if the server could send me a few identical w/u i have 2 identical computers and have boinc on win xp on both so could not do the apples and oranges test but the HT or no HT is an easy thing to prove or disprove given equal wu to compare, even a few "test units " would be worth it to prove it one way or the other, i guess there are variations on cache and memory etc etc.. but it doe my head in as i am totally convicned that there is a simple way to find out, i suspect that a benchfull of varioulsly speced but pared-identical computers running identical wu could prove it one way or other ( or potentialy answer/s could be a table of : this spec puter is better with one wu / that spec puter better to AVOID doing 2 at same time..
where are the experiments, if all esle fails i will just do a bunch of sequential comparision myself, as i have sed b4 my HT pent 4 3.0 ghz worth 1gb ram seem to go ok ,, 1st result was about 20% in favour of doing 2 jobs at same time but alas couldhave been diff workunits and totally due to chance cheers.. impression was one wu 6.5 hrs and other puter did 2 wu's together in about 11 hrs.. .. if i did this again and again would tend to semi randomise and thus be more relialbe but could easuily be by chance.. we need real answers, this is a simple question and is easily solvable given opportunity to compare equal wu's in the context of boinc - HT computers.. cheers |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
i would be delighted to run a test if the server could send me a few identical w/u Hi retep57, From what I have read there is no chance of getting 2 identical WUs from the server. Even if you could then whatever conclusions you might form after the WU is processed would only apply to that WU. That WU may not be typical of most WUs. In fact there may be no such thing as an average WU. It's better to test a bunch of WU on both computers. That is what I have done with my 2 identical computers in my experiment. The idea is that if you test a bunch of WU on both computers then differences between WUs tend to average out. Yes, it takes several days of crunching to perform such an experiment but if you are interested then may I suggest something like the following... 1. Create a new WCG user. The results of the experiment will be recorded in that new account instead of being recorded in your retep57 account where the results of the experiment would be averaged in with the results you have already accumulated and thus get "lost". 2. Setup 2 computers in that new account from step 1. The 2 computers will be your 2 identical computers. One computer will use BOINC, the other will use UD. After the experiment is done you will have a good idea of which is faster, BOINC or UD. An alternate approach you might be interested in is to setup both computers with BOINC but 1 crunches 2 WUs simultaneously while the other crunches only 1 WU at a time. 3. Setup each computer's profile to process only FAAH WUs. Yes, that means the conclusions from the experiment will only apply to FAAH WUs but you can always repeat the experiment at a later date for the other kinds of WUs. 4. Start both computers crunching. The longer you let them crunch the better. I have been running the above experiment for about 48 hours now and I intend to let it run for another 48 hours, 4 days total. By then both of my computers will have crunched about 20 WUs each. The more data you collect the more things tend to average out and resemble real life conditions. 5. After about 48 hours, login to your account, observe the following data for each computer: a) number of WUs reported b)Average Run Time Per Result c)Average Points Per Result 6. Write the above numbers down and share your observations with the rest of us because we're interested too. -- |
||
|
|
|