| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: OpenPandemics - COVID-19 Project
Thread: Work unit availability |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 822
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Yesterday, MIP's share was 64 years.
|
||
|
|
adriverhoef
Master Cruncher The Netherlands Joined: Apr 3, 2009 Post Count: 2346 Status: Offline Project Badges: 
|
you may thus also want to rethink the whole badge system to reward the points which actually is a better metric for contributed effort. If you need 24 hours to complete a task, that's your contributed effort. And so it follows that time is your contributed effort. Points or credits are the reward you get for your contribution, they aren't a contributed effort, they are the result of your contribution. If you wouldn't get points and you wouldn't get badges for runtime, you wouldn't be awarded anything. Suppose you wouldn't get points, then you would still get badges based on runtime. Now, the situation is that you get points AND badges for your contribution. They are two different things. You can't express contributed effort in points. They are two different things. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Looking at the statistics, yesterday the total run time was 546 years of which 34 was MIP (down from over 60 years per day since the announcement). As this was only 10% of total crunching the slack should easily be able to be absorbed by other projects. However, the next project to finish might be a different matter without new work. There is still supposed to be a project coming along but how long that will take is anyone's guess. Mike The problem is, once MIP finishes, you only have MCM to absorb the extra crunching capacity. HST is VERY limited, they are trying to throttle back OPN, ARP is limited due to size of the data, that only leaves MCM. If SCC comes back, they should come back using the GPU version [Edit 1 times, last edit by Former Member at Jun 14, 2021 3:52:41 PM] |
||
|
|
hnapel
Advanced Cruncher Netherlands Joined: Nov 17, 2004 Post Count: 82 Status: Offline Project Badges: 
|
you may thus also want to rethink the whole badge system to reward the points which actually is a better metric for contributed effort. If you need 24 hours to complete a task, that's your contributed effort. And so it follows that time is your contributed effort. Points or credits are the reward you get for your contribution, they aren't a contributed effort, they are the result of your contribution. If you wouldn't get points and you wouldn't get badges for runtime, you wouldn't be awarded anything. Suppose you wouldn't get points, then you would still get badges based on runtime. Now, the situation is that you get points AND badges for your contribution. They are two different things. You can't express contributed effort in points. They are two different things. No, the effort is in the points because it requires you to have the hardware and power requirements for it, you can gain 'time' by running on low spec hardware, my measly raspberry pi contributes lots of time but few WU's and points compared to a high end system, if I wanted to game the time based reward system I could do that easily by spinning up a lot of VM's and oversubscribe the underlying hardware, if you continue to challenge my point I may actually do this. |
||
|
|
Falconet
Master Cruncher Portugal Joined: Mar 9, 2009 Post Count: 3315 Status: Offline Project Badges: 
|
SCC uses AutoDock Vina, not AutoDock 4.
----------------------------------------But I guess if they did use the GPU version of AutoDock for future targets, based on how long past work lasted using CPU's, I'm thinking we would have only a few days of work, assuming no restrictions on workunit release.  - AMD Ryzen 5 1600AF 6C/12T 3.2 GHz - 85W - AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W - AMD Ryzen 7 7730U 8C/16T 3.0 GHz |
||
|
|
pututu
Senior Cruncher United States Joined: Jan 3, 2016 Post Count: 243 Status: Offline Project Badges:
|
Off topic: for credits versus run time, my impression is that this caters for two different groups: those without powerful hardware can still get recognition via badges and those with newer powerful hardware get recognition with higher credit score but no badges.
----------------------------------------I've been running OPNG tasks and I gain more credits since Apr of this year than what I put in the previous five years prior. However I didn't get any update on new OPNG badge since Apr. Also with limited gpu task availability (please consider increasing this in the not-to-distant future) and limited gpu cards (they are super-expensive now), it will take a long long time to get 50 year badge from where I'm now.  [Edit 1 times, last edit by pututu at Jun 14, 2021 9:49:39 PM] |
||
|
|
adriverhoef
Master Cruncher The Netherlands Joined: Apr 3, 2009 Post Count: 2346 Status: Offline Project Badges:
|
you may thus also want to rethink the whole badge system to reward the points which actually is a better metric for contributed effort. If you need 24 hours to complete a task, that's your contributed effort. And so it follows that time is your contributed effort. Points or credits are the reward you get for your contribution, they aren't a contributed effort, they are the result of your contribution. If you wouldn't get points and you wouldn't get badges for runtime, you wouldn't be awarded anything. Suppose you wouldn't get points, then you would still get badges based on runtime. Now, the situation is that you get points AND badges for your contribution. They are two different things. You can't express contributed effort in points. They are two different things. No, the effort is in the points because it requires you to have the hardware and power requirements for it, you can gain 'time' by running on low spec hardware, my measly raspberry pi contributes lots of time but few WU's and points compared to a high end system, You can gain time by dedicating lots of threads to the tasks that are sent to your devices and returning them in time. You get points for your effort, that's your reward. Higher specs, more credits (points). You get badges for (run)time, that's another reward. More time, more badges. They are however not the same. The effort lies in the combination of soft- and hardware that you use, they deliver computer time. For dedicating and delivering computer time you are awarded badges, based on time. For delivering (valid) results, sent back to the server, you get points. They (computer time and results) are different. They're just not the same. if I wanted to game the time based reward system I could do that easily by spinning up a lot of VM's and oversubscribe the underlying hardware, if you continue to challenge my point I may actually do this. It sounds to me as if there is a lot of unfairness, injustice or harm done. So what would be the harm? [Edit 1 times, last edit by adriverhoef at Jun 15, 2021 12:46:52 AM] |
||
|
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges: 
|
Off topic: for credits versus run time, my impression is that this caters for two different groups: those without powerful hardware can still get recognition via badges and those with newer powerful hardware get recognition with higher credit score but no badges. I've been running OPNG tasks and I gain more credits since Apr of this year than what I put in the previous five years prior. However I didn't get any update on new OPNG badge since Apr. Also with limited gpu task availability (please consider increasing this in the not-to-distant future) and limited gpu cards (they are super-expensive now), it will take a long long time to get 50 year badge from where I'm now. If AMD GPUs were credited the full elapsed time for each GPU task like Nvidia gets because of their poor OpenCl support instead of just the CPU time AMD users get many people would get to their badge goals quicker. I can only guess how many thousands of hours of run time I've lost on the 2 GPU projects here because I run AMD and not Nvidia. I've contacted support about this issue and their reply was BOINC only sends CPU time to the server and not GPU time. I pointed out that BOINC also sends elapsed time to the server and asked why that can't be used to grant run time to AMD users. I received no reply to that question.
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.
----------------------------------------  [Edit 1 times, last edit by nanoprobe at Jun 15, 2021 12:51:23 AM] |
||
|
|
gordonbb
Cruncher Canada Joined: May 14, 2019 Post Count: 19 Status: Offline Project Badges: 
|
If AMD GPUs were credited the full elapsed time for each GPU task like Nvidia gets ... AFAIK using NVIDIA GPUs for OPNG does not get any additional run-time. I run 3 Nvidia cards 24x7 on OPNG and another 8 occasionally and I get the same amount of run-time as if I only ran OPN, i.e. for both AMD and Nvidia GPUs the only run-time awarded is the CPU portion.I'd prefer to run AMD GPUs but I primarily run COVID19 WUs on Folding@Home and RDNA2 and GCN both under perform there as the OpenCL implementation on GROMACS is much less efficient than the CUDA implementation. This appears to come down to NVidia being willing to work with the developers to optimize the code. Maybe one of these days AMD will likewise put more effort into working with developers but given how long it takes for them to fix issues in their Graphics drivers I'm not holding my breath.  AMD - 2600x, 2 x 2700, 2700x, 3900x, 3950x, 2 x 5900x, 5950x Intel - E3-1231v3, 9900K NVidia - GTX 1060 6GB, 1660ti, 1070ti; RTX 2060, 2060s, 2070a, 5 x 2070s |
||
|
|
pututu
Senior Cruncher United States Joined: Jan 3, 2016 Post Count: 243 Status: Offline Project Badges:
|
I have both nvidia and amd cards. With 81,600 tasks per day available to run on OPNG and assuming each task takes 10 mins on average, the total cpu time (elapsed time) is about 567 days or 1.55 years (less for amd cards by about 1/3, based on my estimate) assuming running one task per gpu (more if you run more than one task per gpu) but the total work done in terms of credit granted is probably half a billion WCG credits!
---------------------------------------- |
||
|
|
|