Message boards :
Number crunching :
What is the % of S@H top computers with Lunatics apps/.exe installed?
Message board moderation
Author | Message |
---|---|
Stubbles Send message Joined: 29 Nov 99 Posts: 358 Credit: 5,909,255 RAC: 0 |
Is there a way to find out how many of the top 10,000 computers in RAC have Lunatics apps/.exe installed? My guess a month ago was that is was probably fairly high (whatever that means) ...but now, I would guess it is much lower than my original gut feeling. I don't remember seeing any data in host.gz or user.gz that would allow someone to figure that out, and the only way I know to get that info is to look at an individual computer's: Application details for host webpage to see if there's an anonymous platform installed. |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Even seeing "Anonymous platform" in someone app detail or task list doesn't tell you if they are using the optimized lunatics apps. It only tell you that host is using an app_info.xml. Someone might want to use the stock apps but limit which apps run or pick specific combinations of apps. Rather than run everything the server sends them. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Stubbles Send message Joined: 29 Nov 99 Posts: 358 Credit: 5,909,255 RAC: 0 |
Even seeing "Anonymous platform" in someone app detail or task list doesn't tell you if they are using the optimized lunatics apps. It only tell you that host is using an app_info.xml.Thanks HAL for that info. In most cases, I'm guessing "Anonymous platform" will be from having installed Lunatics since customizing app_info.xml is much more complex than simply installing Lunatics. Am I wrong? Since no one else replied, I did some preliminary stats. I looked at some individual computer's Application details for host webpage from a few of the pages from top 10,000 computers by RAC to see if there's an anonymous platform entry. Here are my findings for 3 pages: hosts with "Anonymous platform" host rank 1-20 : 18 host rank 241-260: 9 host rank 741-760: 6 host rank 1241-1260: 5 host rank 2541-2560: 1 host rank 5000-5020: 2 I started with the 1st and last entry and then filed in the ones in between from the lowest (2541-2560) to the highest (241-260). I'm a bit shocked that so few have installed Lunatics!!! |
Al Send message Joined: 3 Apr 99 Posts: 1682 Credit: 477,343,364 RAC: 482 |
In a way I am, and then again, actually I'm not all that suprised. It's been mentioned around here now and again for many years, that we here are the 1% of the 1%'ers, those that come to the forum and browse around, ask questions, try to wring the most out of our systems, build dedicated crunchers, etc. The vast, vast majority are what is colloquially referred to as "Set and forgetters". I sometimes lose sight of that fact, I get into the 'SETI mode', and pretty much forget that all those 10,000's of ppl are folks who came across an article, or a link, or a show that mentioned about searching for ET. They come here, think it's neat, and it's easy to install, so they do it. Run the installer, pick our project, and away they go. They may or may not even remember it's running, that is until maybe something happens that effects the performance of their system. Then, if they realize that it is BOINC causing it, they either uninstall it, or if they (hopefully) care, dig a little deeper and come here looking for answers. But those are the exceptions, def not the rule. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Taking into account that all GPU apps are "Lunatics" one there is much less incentive to go to anonymous platform now than few years before. Only for better CPU apps actually. SETI apps news We're not gonna fight them. We're gonna transcend them. |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Even seeing "Anonymous platform" in someone app detail or task list doesn't tell you if they are using the optimized lunatics apps. It only tell you that host is using an app_info.xml.Thanks HAL for that info. Some things to keep in mind: -The "stock" GPU app is a lunatics app. So every GPU system out there is actually running a lunatics app. -All the app tuning options for GPUs are available without using an app_info.xml now. The addition of app_config.xml & the specific GPU config files let all users tune the apps as they need. -There are some users that run GPU only systems. So the optimized CPU apps are not needed on their systems & they are OK with not having the most up to date build of the GPU apps. When I had ~35 machines running. They were almost all CPU only hosts. Most never made it into the top 10,000 hosts list, but they all had the best optimized CPU application for their architecture. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Shaggie76 Send message Joined: 9 Oct 09 Posts: 282 Credit: 271,858,118 RAC: 196 |
By request I wrote a few scripts to sort this out: Scripts uploaded to GitHub if you want to check my work. As pointed out by HAL9000 earlier: this is isn't necessarily accurate. Raw numbers if you want to graph it some other way: Rank,Anonymous,Stock 10,9,1 100,66,34 1000,327,673 10000,1284,8716 |
Stubbles Send message Joined: 29 Nov 99 Posts: 358 Credit: 5,909,255 RAC: 0 |
Wow! Thanks Shaggie 10,9,1 100,66,34 1000,327,673 10000,1284,8716So: ~87% (7/8) of the "Top 10,000 hosts by RAC" do nothing to optimize!!! ...and also about two thirds (2/3) in the top 1000 PCs. {I'm in shock} No wonder it took me almost no effort (and only ~$900can) for me to rise to the 1% club so quickly (even with RAC being as slow as it is to rise). And I thought it was because I had bought the best gear for the money! ;-} Hmmm...so... considering that optimizers usually have more than 1 power rig, it could very well be that over 90% (maybe even 95%) of the "Top 10,000 participants" only run stock. Thanks again Shaggie for crunching those stats!!! |
Shaggie76 Send message Joined: 9 Oct 09 Posts: 282 Credit: 271,858,118 RAC: 196 |
Out of curiosity I've switched a few of my rigs to lunatics -- I'm curious to see the relative difference. From the sounds of things it would be mostly CPU throughput but I've noticed my GPU running a bit hotter so that's a good sign too. |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Wow! Thanks Shaggie At my peak I was about #30 by RAC with my ~35 hosts running. I think only 4 or 5 were in the top 10,000 list. With our 138,101 active users, boincstats isn't showing hosts right now, I think we have something like 300,000 active hosts. So to say 0.5% of active hosts have actively installed the lunatics apps doesn't seem to far fetched to me. Considering <1% of users ever visit the fora & would not otherwise have any knowledge the other apps exist. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Stubbles Send message Joined: 29 Nov 99 Posts: 358 Credit: 5,909,255 RAC: 0 |
With our 138,101 active users, boincstats isn't showing hosts right now, I think we have something like 300,000 active hosts. On a small tangent: I think it depends on what criteria for "active" is being used. BoincStats seems to use something like: user/host has a "Valided task" during the last 60 days. As for: http://boinc.netsoft-online.com, it seems to use the same definition except the criteria seems to be: ...during the last 30 days. since their count is much lower: Active User Count 108,657 Host Count 162,640 My criteria for active is: ... during the last week (or possibly at the most: 2 weeks) has MORE than 2 validated tasks. The reason why I have "more than 2" is because: some hosts haven't reported a completed task for 3 months and yet they might still be considered active because the wingman never submitted the task before the deadline so it was resent out again to another host. My guess is that is why there is a diff of about ~30k between the two sites. |
Stubbles Send message Joined: 29 Nov 99 Posts: 358 Credit: 5,909,255 RAC: 0 |
Out of curiosity I've switched a few of my rigs to lunatics -- I'm curious to see the relative difference. From the sounds of things it would be mostly CPU throughput but I've noticed my GPU running a bit hotter so that's a good sign too. If my memory serves me right, NV_SoG is only available via v0.45 beta3 (which for me has worked very well and I haven't seen a bad comment or bug report from anyone) For your GPUs, I don't know which ones are better at processing more than 1 tasks at a time on NV_SoG. On Cuda, there is a huge advantage to running 2-3+ in parallel. |
Shaggie76 Send message Joined: 9 Oct 09 Posts: 282 Credit: 271,858,118 RAC: 196 |
Stats for 8026559 Stock Core i7-5960X, 751.646218607873 cr/hr (47 WU checked) GeForce GTX 980 Ti, 680.596911595191 cr/hr (53) Lunatics_Win64_v0.44 Intel Core i7-5960X, 903.245045606621 cr/hr (152) GeForce GTX 980 Ti, 340.334738185789 cr/hr (45) CPU up, GPU down? I didn't customize anything so it's still just running 1 GPU task at once. OpenHWM says GPU core isn't saturated at all -- maybe It's entirely possible my script has a problem -- I was wondering about that actually -- if a task gets interrupted because I'm using my machine it might stall for 12hrs until I'm done for the day -- I'll check my data later to see if this occurs. |
W-K 666 Send message Joined: 18 May 99 Posts: 19057 Credit: 40,757,560 RAC: 67 |
You might need to do a survey of the 'angle range' (AR) of the tasks on the GPU runs. The VLAR tasks those where AR = 0.00nnnn can take much longer on the GPU. Something like 4 * mid range AR (0.42nnnn) but the CR are often less than 50% more. So if the run contains mostly VLAR tasks then the total CR can be much lower, for short runs. You will notice VLAR tasks are annotated, that is because at one time the selection process diverted them from the GPU. That could also be the reason why your totals 47 & 53 and 152 & 45 are not balanced. |
Shaggie76 Send message Joined: 9 Oct 09 Posts: 282 Credit: 271,858,118 RAC: 196 |
I don't have records of my GPU utilization on the stock build but I dimly recall it being 80-90% most of the time. However, I checked the 0.44 cuda5 build running on my GeForce 980ti and I was surprised to see it barely using a quarter of the GPU. I lit up 3 tasks and it wasn't quite saturated so I tried 4 tasks which got it above 90. It's a little disappointing because I'm sure the context switches will eat into the overall performance. I'll let it cook for a few days and re-run the stats. I checked the other rig that I updated to the same build; it's only got a GeForce 780 and for some reason it's keeping above 90% utilization. |
Shaggie76 Send message Joined: 9 Oct 09 Posts: 282 Credit: 271,858,118 RAC: 196 |
You might need to do a survey of the 'angle range' (AR) of the tasks on the GPU runs. The VLAR tasks those where AR = 0.00nnnn can take much longer on the GPU. Something like 4 * mid range AR (0.42nnnn) but the CR are often less than 50% more. Hmm, this is interesting. I'll see if I can maybe incorporate this information into my GPU performance survey. Thanks! |
Shaggie76 Send message Joined: 9 Oct 09 Posts: 282 Credit: 271,858,118 RAC: 196 |
Following up on the stats below; same host running lunatics with 4 GPU tasks: Host, Device, Credit/Hour, Work Units 8026559, Intel Core i7-5960X @ 3.00GHz, 990.894680277032, 131 8026559, NVIDIA GeForce GTX 980 Ti, 459.128720715548, 41 Still getting less GPU CR/Hr with Lunatics and CPU CR/Hr average is even a bit higher. Note that this is considerably lower than the average of stock apps. I skimmed the validated tasks the script averaged and by eye-ball didn't see any GPU VLAR tasks. It's strange -- it could just be the the luck of the draw for easier/harder tasks or it could be that the context-switching overhead of 4 tasks at once is detrimental. |
W-K 666 Send message Joined: 18 May 99 Posts: 19057 Credit: 40,757,560 RAC: 67 |
Something is wrong there. I've just looked at your host, application details and for SETI@home v8 (anonymous platform, NVIDIA GPU) Number of tasks completed 119 Max tasks per day 162 Number of tasks today 11 Consecutive valid tasks 129 Average processing rate 79.14 GFLOPS Average turnaround time 0.20 days While my host ID 8048155 with GTX 670 I get SETI@home v8 (anonymous platform, NVIDIA GPU) Number of tasks completed 212 Max tasks per day 269 Number of tasks today 45 Consecutive valid tasks 236 Average processing rate 171.68 GFLOPS Average turnaround time 0.95 days Which indicates my 670 is twice as powerful as your GTX 980ti. You've been ripped off somewhere, or you have finger trouble. Don't wash your hands so often When this host was ID 7006214 its Average processing rate 174.33 GFLOPS, not much different. |
Shaggie76 Send message Joined: 9 Oct 09 Posts: 282 Credit: 271,858,118 RAC: 196 |
As I mentioned below I was running 4 tasks on the GPU concurrently; how many were you running? |
W-K 666 Send message Joined: 18 May 99 Posts: 19057 Credit: 40,757,560 RAC: 67 |
As I mentioned below I was running 4 tasks on the GPU concurrently; how many were you running? I'm only running one, as this the my most used computer when at home and even then will pause GPU processing when doing some other things. I have tried it on 2 tasks at once but then it took several seconds before I could get back control which I didn't like. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.