Countdown to HCC Completion Thread!

[0ne.shot]

Ok, another poster mentioned here, as of the second of April, there are an estimated 25 days left of the HCC project until it is completed!

Here is a post from knreed, a tech at WCG, who posted this on the HCC forum at WCG.

To be honest, the combined production of the GPU's surprised us as well. We had anticipated that with the launch of GPU we would double the daily output. instead it increased by a factor of 16-17 (before GPU we were completing ~60k/results per day with 1 image per job, we are now >500k/results per day with 2 images per job. This led to some great adventures in handling greater capacity (which we call a 'success' problem).

http://www.worldcommunitygrid.org/fo...ad?post=417375

It's amazing what a group of people can do out of their own goodwill. I'm not a big cruncher by far, but with my 7950, I've stacked up the points since the introduction of the HCC GPU WU introduction.

As days go by, many of us may not be aware of the impending completion resulting in the return to normal PPD production, so lets keep everyone posted with a daily countdown. Of course there will be resends which may lengthen the time a little. I will continue to post the number of days left here in the OP.

UPDATE:

The researchers have identified additional batches for this project which need to be run (and weren't included in the original 25 day estimate). I'm in the process of getting them loaded into BOINC and should be able to provide a more accurate estimated number of days left once this is complete. At minimum though, there should be more than five days left of work before we exhaust the supply of non-resend work units.

Seippel

UPDATE #2:

Last of the new work for Help Conquer Cancer Sent
We have just finished distributing the last of the new work for the Help Conquer Cancer project. It will take about 7-10 days for the in progress work for the project to complete. During this time there will periodically be additional results sent out to finish up those in-progress workunits.

THANK YOU so much for your contribution and efforts on this project!

http://www.worldcommunitygrid.org/fo...st?post=420071

Big thanks to Bill P for posting Knreed's post.

Estimated Time Remaining:
0 Days

[bearcatrp]

I just bumped my 7950 to crunch 21 wu's. lets hope WCG has another project coming on board to use GPU.

[Conquistador SW]

I just bumped my 7950 to crunch 21 wu's. lets hope WCG has another project coming on board to use GPU.

Yeah, it would be really nice to at least know if there is something GPU related in development or planned. Otherwise I will sell all my AMD cards after this and go back to Nvidia; much less power consumption at idle with multiple monitors and I have some problems while gaming I never had with Nvidia cards.

[OldChap]

One of the reasons that I have kept beta included in my app_config is in the hope that something would show up as being in the pipeline but as was explained over at WCG, For the work to prepare a gpu version to be worthwhile it must be a very big project. I am sure it will happen....just not soon.

What will be interesting is to see how the end of this GPU project will effect the number of participants on WCG.

[Mumak]

What will be interesting is to see how the end of this GPU project will effect the number of participants on WCG.

Yep, that will be interesting... I'll stay with WCG on all machines, but with a lower priority. nVidia ones will continue GPUGrid + E@H. AMD ones will switch probably to MW@H.

[Gandalf]

... What will be interesting is to see how the end of this GPU project will effect the number of participants on WCG.

While I will be keeping my WCG CPU tasks working, I have been invited to take my GPUs over to the xtreme GPUGRID Team and my GPUs won't be coming back to WCG. Fool me once, ...etc.

[Johnmark]

Sorry to say but if WCG doesn't replace HCC with HCC2 or another cancer project that runs on GPU's I'm done.
I'll then give the cancer society a nice monthly check that I was spending on crunching.

Again I'm sorry but if WCG and their counterparts can't be bothered to build applications that can do the most work for the least cost (GPUvsCPU) then it speaks volumes about their commitment to the distributed computing community.

[D_A]

Sorry to say but if WCG doesn't replace HCC with HCC2 or another cancer project that runs on GPU's I'm done.
I'll then give the cancer society a nice monthly check that I was spending on crunching.

Again I'm sorry but if WCG and their counterparts can't be bothered to build applications that can do the most work for the least cost (GPUvsCPU) then it speaks volumes about their commitment to the distributed computing community.

While your commitment to cancer research is commendable the last statement is plain ignorant. It has been made perfectly clear over and over again that not all science applications are portable to GPU processing. Some of the algorithms are even slower on GPUs than CPUs simply because of the narrow functional parameters and capability of the hardware, it has nothing to do with lacking commitment to the community or the research.

[Snow Crash]

WCG does not write the apps so please stop slamming them. https://secure.worldcommunitygrid.or...ad?post=417467

They did a remarkable job building the most stable BOINC environment in the WORLD and handle way more traffic that any one else, provide better tech info than most, contractually require projects they host to make all results publicly available so saying they could not be bothered just isn't right.

Sorry to say but if WCG doesn't replace HCC with HCC2 or another cancer project that runs on GPU's I'm done.
I'll then give the cancer society a nice monthly check that I was spending on crunching.

Again I'm sorry but if WCG and their counterparts can't be bothered to build applications that can do the most work for the least cost (GPUvsCPU) then it speaks volumes about their commitment to the distributed computing community.

[bearcatrp]

WCG could try nVidia like gpu grid to see if that would be more portable. Just guessing though. Would be interesting.

[Movieman]

Deep breath guys, remember: "Friends first"..
Hey, I just dropped $200+ on a new vid card just for this but I'll live with this as part of life.

[Gandalf]

Patience, Hell. I'm going to kill something.

[Movieman]

Patience, Hell. I'm going to kill something.

[Snow Crash]

I'm all for the friends first ... we sometime need to help each other with the truth :-)

WCG does not write the applications, they only host the projects.

If you want to see a change then maybe try contacting the people who do write the applications.
A nice place to start looking into the details is the research overview page.
http://www.worldcommunitygrid.org/re...AllProjects.do Here you can find what the individual projects are really about who's project it really is. Sometimes you have go to the actual project's website to get contact information, but it is there and readily available. Let us know how that goes.

[[XC] Oj101]

TBH I'm looking forward to getting back to GPUGrid, my CPUs will stay on WCG forever though

[bearcatrp]

It will be weird not having GPU crunching on WCG if another project doesn't present itself. Kind of seen the writing on the wall so ordered a gigabyte 660ti 3gb card for my new build to try GPU grid. 1st time owning nVidia. Just hope the 660ti works good.

[D_A]

On an actual CUDA enabled project they apparently kick ass. nVidia's weakness is OpenCL, unfortunately.

[Mumak]

The 660 Ti produces very well and is probably the best perf/price model. For me it does a long Nathan WU in ~5.5h (70,800 points) or short Noelia in ~2h (23,700 points).

Just hope the 660ti works good.

[bearcatrp]

It will be here Monday and going in my 3770k build. Have no clue if those points are good or bad. Have done Milky Way and prime grid in the past besides WCG and climate prediction. Will have to go read up GPU grids info. Have heard they are low on wu's lately. Thanks for the reply.

[Mumak]

Currently there seems to be enough supply of WUs, they had just some issues with long Noelia ones, which they pulled.
I think those scores are quite good, probably not as much as you get with MW, but in total a machine produces a few times higher score than on WCG. I haven't started MW yet, but plan to join after HCC is finished.

It will be here Monday and going in my 3770k build. Have no clue if those points are good or bad. Have done Milky Way and prime grid in the past besides WCG and climate prediction. Will have to go read up GPU grids info. Have heard they are low on wu's lately. Thanks for the reply.

[bearcatrp]

Does GPU grid do one wu at a time or can multiple wu's be crunched like HCC1? I know MW is one at a time but a good GPU would eat them fast.

[Gandalf]

I just started at Einstein@Home and it grabbed one computer thread and a 0.5C + 1 ATI GPU.
I'd like to modify that configuration myself, because that is a waste of 0.5C, since I am running hcc1 at 0.2C + 0.2ATI.
Hold the phones, I just finished the ATI task and as soon as I forced a Report (to get a first completed task reported), Einstein start a half a computer thread task in place of a replacement ATI task. I'm confused as to their thinking. It's not a one for one task startup. OK, I'm a wizard not a physicist. Sue me.

[Embalmed]

been noticing the disappearance of the 7970 models from microcenter and newegg, bummer, I wanted to pick two more of these up for gaming.

maybe it's the easynews people trying to make a comeback

[Mumak]

Yes, that's possible, but doesn't make much sense.
Running multiple HCC1 WUs improves the yield, because of the nature of the WUs - each of them has 4 portions of pure CPU work, so the GPU is not utilized. Running multiple concurrently makes sure the GPU is fully utilized.
GPUGrid tasks on the other hand utilize the GPU at a very high level all the time + some CPU time to feed the GPUs (the amount of CPU time there seems to depend on GPU family, Fermi cards seem to require more CPU time).

Does GPU grid do one wu at a time or can multiple wu's be crunched like HCC1? I know MW is one at a time but a good GPU would eat them fast.

[Mumak]

Why do you think that those BRP4 tasks don't really need 0.5 CPU ? Each application is very different and demands different amount of resources.
On GPUGrid for example I reserve a full CPU thread for a GPU task, though they ask for much less CPU (however in reality it seems that on Fermi cards, the CPU resources required to feed GPU are much higher than originally proclaimed).

I just started at Einstein@Home and it grabbed one computer thread and a 0.5C + 1 ATI GPU.
I'd like to modify that configuration myself, because that is a waste of 0.5C, since I am running hcc1 at 0.2C + 0.2ATI.
Hold the phones, I just finished the ATI task and as soon as I forced a Report (to get a first completed task reported), Einstein start a half a computer thread task in place of a replacement ATI task. I'm confused as to their thinking. It's not a one for one task startup. OK, I'm a wizard not a physicist. Sue me.