Message boards :
Number crunching :
I think you made your point !!
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5
Author | Message |
---|---|
Send message Joined: 1 Oct 08 Posts: 106 Credit: 24,162,445 RAC: 0 |
That is just too fast! It shows the wasted computation time. With the Windows stock app it takes close to 6 hours to calculate a long WU on a 3GHz Core2. But it can be done in just 2 minutes! Most likely even faster, if someone would take the time and effort of rewriting the complete app. |
Send message Joined: 9 Aug 08 Posts: 10 Credit: 46,370 RAC: 0 |
It shows the wasted computation time. With the Windows stock app it takes close to 6 hours to calculate a long WU on a 3GHz Core2. But it can be done in just 2 minutes! Indeed. But in all likelihood, these are just college kids. I can think of several things that would take precedence over an undertaking like that (some of them even related to school!). I was pretty proficient in Basic back in the day, and I think C looks a lot like it. But to sit down and just look at code without knowing much about the flow or end product, it would take me forever to get anywhere! I have other priorities too - too many. I think if it were me, I would have just given them the improved code, but that's just me. I'm sure the developers are very appreciative of the hints you have already given them and they will implement them when they find them. |
Send message Joined: 21 Oct 08 Posts: 3 Credit: 7,081,130 RAC: 0 |
I´m shaking my head while reading zed´s comments. Do you cut down a redwood tree just to get one toothpick? You don´t get the point, right? It doesn´t matter if Milksops account is deleted or whatever happens to it. Its just a proof of concept, that even unexperienced "programmers" can speed up the computation by the factor of 50-100 within hours. The average user doesn´t know that his ressources are wasted in such a shameful manner by the project, ignoring optimizations that can be mathematically proven to deliver the exact same result as the original application. even if the credits/day was specifically adjusted to 10 for the account of milksop, it won´t matter: the project still would get 100x correct results resulting in a huge leap for the completion of the project. |
Send message Joined: 4 Oct 08 Posts: 1734 Credit: 64,228,409 RAC: 0 |
No reply to the Optimisation thread asking how the recoding is going, if at all? |
Send message Joined: 1 Oct 08 Posts: 106 Credit: 24,162,445 RAC: 0 |
No reply to the Optimisation thread asking how the recoding is going, if at all? Maybe some people will now start to understand that just waiting for an improvement isn't the right tactic here ;) |
Send message Joined: 21 Aug 08 Posts: 625 Credit: 558,425 RAC: 0 |
No reply to the Optimisation thread asking how the recoding is going, if at all? Alternatively, maybe people who aren't in college, or as you all across the pond say, aren't in "uni" right now might perhaps understand that the past week has been mid-term...and there could be other things taking precedence that has to take precedence due to the academic calendar. I could be wrong, obviously, but I'm willing to give the benefit of the doubt... |
Send message Joined: 1 Oct 08 Posts: 106 Credit: 24,162,445 RAC: 0 |
Alternatively, maybe people who aren't in college, or as you all across the pond say, aren't in "uni" right now might perhaps understand that the past week has been mid-term...and there could be other things taking precedence that has to take precedence due to the academic calendar. I could be wrong, obviously, but I'm willing to give the benefit of the doubt... Me too. Let's wait another week. Btw., I'm working at a university and Travis has already graduated ;) |
Send message Joined: 4 Oct 08 Posts: 1734 Credit: 64,228,409 RAC: 0 |
I graduated, reading Chemical Engineering, just a tiny few years ago. Pity it's now been 35 years since I stopped programming a AIM65 for light control of continuous chemical process plant. |
Send message Joined: 1 Sep 08 Posts: 520 Credit: 302,524,931 RAC: 0 |
One thing to consider, when some folks use their optimized code to generate massive credit counts on limited CPU capacity, it eventually gets noticed by other players in the BOINC world. Even with the current unoptimized code, MilkyWay is awarding a LOT more credit per CPU cycle than other projects, even those which have available optimized code (SETI for example). Folks like milksop and cruncher3r show up toward the top of the overall BOINC stats for their very high RAC numbers -- and since they are single project folks, they shine something of a light on MilkyWay. At some point, perhaps the optimized code for Milkyway will find its way into public use rather than for those who have done the code modications only. When it does, I suspect the admins here will be *compelled* to drop their credit schema so that credit per CPU cycle doesn't make MilkyWay a pariah project in the BOINC world. But that might be something of a collective citizenship argument -- one which might well have no credence in this particular population. |
Send message Joined: 10 Aug 08 Posts: 218 Credit: 41,846,854 RAC: 0 |
I suspect the admins here will be *compelled* to drop their credit schema so that credit per CPU cycle doesn't make MilkyWay a pariah project in the BOINC world. Well MilkyWay is already being discussed on a number of forums, and it isn't all nice and rosy. For as long as this has gone on does it really appear that anyone in an administrative position really gives a damn what anyone or the other projects say or think? And you're right about the overall stats watchers. Those that have been quietly crunching along for years and generally moving up the stats at a regular pace suddenly find that someone who has only been on one project for less than 1 month had generated more credits than what a majority of crunchers have in years. I wonder how long it will be before David Anderson and the credits cops show up and put this project in the toilet or make them straighten it out. Or who knows when the stats sites finally get enough of this and pull MilkyWay out of the stats so that everything goes back to normal. But then more than likely after this project is screwed into submission these folks would just leave here and go somewhere else and start this all over again with another project. There isn't and won't be any winners here for either the crunchers, or the project science. |
Send message Joined: 28 Aug 07 Posts: 146 Credit: 10,702,809 RAC: 2,669 |
I wonder how long it will be before David Anderson and the credits cops show up and put this project in the toilet or make them straighten it out. Well, that already happened some time ago, look here. But there was no serious answer from the project admins, only a user debate, so it ended up in nothing (like it mostly does). Or who knows when the stats sites finally get enough of this and pull MilkyWay out of the stats LOL, that won't happen. Even a misleaded project like Predictor wasn't excluded from the stats sites because of it's behaviour against his users... ;-) Btw: MilkyWay isn't the project with the highest credit rate per hour. It is way beaten by PS3Grid, where you can make at least over 150 Credits/hour even with the slowest NVidia card! But the users there are still very less. Consider when they get some thousand users in the future, it will blow up the stats sites and will beat SETI for sure some day. ;-) Member of BOINC@Heidelberg and ATA! My BOINCstats |
Send message Joined: 27 Nov 07 Posts: 4 Credit: 9,698,836 RAC: 0 |
MilkyWay isn't the project with the highest credit rate per hour. But only with cards that run Cuda 2.0 compiled applications. My Geforce 8800 GTS support only 1.0. So even recognized by BOINC as CUDA device, it still get "Full-atom molecular dynamics is not available for your type of computer". :( |
Send message Joined: 29 Aug 07 Posts: 486 Credit: 576,548,171 RAC: 1,260 |
MilkyWay isn't the project with the highest credit rate per hour. You need at least a 8800GT or 8800M to run the CUDA 2.0 WU's, the 8800GT will run the Wu's in about 16 Hr's & the 8800M will run them in about 65 Hr's ... :) |
Send message Joined: 29 Dec 07 Posts: 16 Credit: 158,120,935 RAC: 0 |
exactly, and since the WU's run so long, the credit is not all that great per day. PS3 running this stuff only avg. 3500 points per day, where-as here its easy to get 9k per day on a decent pc running IX. |
Send message Joined: 27 Aug 07 Posts: 915 Credit: 1,503,319 RAC: 0 |
|
Send message Joined: 28 Aug 07 Posts: 16 Credit: 70,797,368 RAC: 0 |
Yea, but the only difference between PS3GRID/GPUGRID and Milkyway is that PS3GRID is using FLOPS counting. So you really get the credits for what is computed and not a fictional number of credits... The CUDA code on GPUs is faster than a something computed on a CPU. That's the way things are... ;) |
Send message Joined: 28 Aug 07 Posts: 31 Credit: 86,152,236 RAC: 0 |
Well, I think I was waiting long enough for an improvement. Setting 30+ boxes elsewhere - where I known it's not a 99% waste of CPU cycles and electricity. BOINC Project specifications and hardware requirements |
Send message Joined: 2 Jan 08 Posts: 79 Credit: 365,471,675 RAC: 0 |
I am out too with 36 cores.:( |
©2024 Astroinformatics Group