Einstein: Einstein@Home GPU Application for ATI/AMD Graphics Cards

[RSS]

After more than a year of work by Oliver Bock, Bernd Machenschalk, Heinz-Bernd Eggenstein and other developers, we are pleased to announce the release of the first Einstein@Home application for ATI/AMD Graphics Cards.This OpenCL application, which searches Arecibo data for new radio pulsars, is about a factor of ten faster than the same search running on a typical CPU. The application is currently available for Windows and Linux computers with Radeon HD 5000 or better graphics cards. We hope to have a version for Macintosh (Apple OS X 10.8, Mountain Lion) sometime this summer, but there are still some problems that need to be fixed or worked around.Volunteers who wish to run this application will need to install version 7.0.27 or later of the BOINC client. Please see this thread (http://einstein.phys.uwm.edu/forum_thread.php?id=9445) for more information, or if you want to ask questions.Many thanks to the AMD/ATI for their support in the OpenCL software development effort.Bruce AllenDirector, Einstein@Homehttp://einstein.phys.uwm.edu/img/opencl.png (http://www.khronos.org/opencl/)

More... (http://einstein.phys.uwm.edu/forum_thread.php?id=9446)

[zombie67]

Nice to see another option for ATI cards!

[Mike029]

Nice to see another option for ATI cards!


I hope the credits are competitive.

[zombie67]

No change in the credits. 500/task. This is not a project for max credit hunters.

[Duke of Buckingham]

Nice is a project for me.

[DrPop]

No change in the credits. 500/task. This is not a project for max credit hunters.

Which is a real pity. If they offered better credits I would crunch Einstein all the time. Love this project...they actually do something useful with the data! :o

[Mumps]

Which is a real pity. If they offered better credits I would crunch Einstein all the time. Love this project...they actually do something useful with the data! :o

Um, excuse me, a GPU app that's 10 times faster than the CPU version and offers the same number of credits per WU? At least it gets a cruncher to random MM's 10 times faster. If not as fast as a Donate or Collatz, still faster than a CPU project...

[Slicker]

Um, excuse me, a GPU app that's 10 times faster than the CPU version and offers the same number of credits per WU? At least it gets a cruncher to random MM's 10 times faster. If not as fast as a Donate or Collatz, still faster than a CPU project...

10x faster for the same credit is 10x more credit per day than if crunching via CPU.

MM chaser's should not read the rest of this message as it goes against their main philosophy. But, from a strictly TopGun point of view....

That's like saying a Money Market is a much better invenstment than a savings account because it pays 1.5% vs. 0.5%. When you consider that inflation is 2.5% or higher, you are actually losing money by putting your cash into any type of bank run investment! By comparison, if you invest in bonds and stocks and you will see 5-8% increase on average and earn money rather than lose it. I see Einstein CPU app as a savings account. Their GPU app (Money Market account) is better, but why keep losing credits when other projects like DiRT, PG, MW, Moo, Donate, Collatz, etc. (Stocks and Bonds) are known to pay so much more?

Or....

Patient: Doctor, it hurts when I do this.
Doctor: Then stop doing it!

[spingadus]

Just received credit for my HD 6970 on this project and it is definitely NOT worth running on an AMD card.

Run time was nearly 8 hours for only 500 credit.

That comes out to 63 credits/hr or 1519 credits/day.

My Nvidia GTX 590 does 2 tasks in about 40 minutes.


As mentioned, this isn't a project you run for top credit. Stick with Nvidia here if you must crunch it. Perhaps it will become more efficient in the future for AMD?

[Rattledagger]

Just received credit for my HD 6970 on this project and it is definitely NOT worth running on an AMD card.

Run time was nearly 8 hours for only 500 credit.

That comes out to 63 credits/hr or 1519 credits/day.

Hmm, my only attempt on a 5850 lead to failure after 0.94 hours and if not mis-remembers around 80% done, so close to 8 hours seems much too high.

But granted, computer crashing multiple times and taking the checkpoint-file with it in the process can also have screwed-up the cpu-time, so... :-??

[spingadus]

Hmm, my only attempt on a 5850 lead to failure after 0.94 hours and if not mis-remembers around 80% done, so close to 8 hours seems much too high.

But granted, computer crashing multiple times and taking the checkpoint-file with it in the process can also have screwed-up the cpu-time, so... :-??

I've run 3 tasks and all took about 28,000 seconds. It's possible that if I freed up a thread or 2 it may speed up the task. I'll probably not test it any time soon though. Perhaps a third opinion?

[kmanley57]

I just looked at one of my Wu's that had been run by a NVidia and a ATI card. The Cuda ran it in 2,326 seconds and the ATI in 41,311 seconds, with the ATI card being the faster of the two in floating-point calculation speed! So the 2096 GFlop ATI to the 1405 GFlop CUDA looked bad! Both machines ran with a 4 core cpu, and near the same clock speed. :(

[DrPop]

This does not mean the CUDA card you have is faster / better than the ATI card. What it means is in this case, the code you are running on that project does not use the ATI card to its full potential. It is either poorly coded on their end, or using only a percentage of what the ATI card can do. Hope that makes sense? Some projects are really lousy on one brand and much faster on the other. For example, if you were on Collatz, you would see just the opposite happen. ;)

Sent from my MB860 using Tapatalk 2

[Rattledagger]

I just looked at one of my Wu's that had been run by a NVidia and a ATI card. The Cuda ran it in 2,326 seconds and the ATI in 41,311 seconds, with the ATI card being the faster of the two in floating-point calculation speed! So the 2096 GFlop ATI to the 1405 GFlop CUDA looked bad! Both machines ran with a 4 core cpu, and near the same clock speed. :(
Well, once-upon-a-time running SETI with a full set of cpu-tasks on my Ati-5850 wasn't a problem, but after a OS re-install the GPU-utilization often dropped to zero, and since didn't manage to hit the "correct" combination of drivers the only option to speed-up SETI again was to sacrifice a cpu-core for this.

So, wouldn't be surprised if Einstein@home also wants a cpu-core to keep the GPU busy...

[kmanley57]

The difference I would expect is a coding issue! It wants less than many other projects for core utilization, but I will try to half the cores used and see if that helps any.

[Rattledagger]

After swapping-around things that seems to have solved the crashes and finally having time to install a different OS is up and running on the Ati-5850 again. For the moment under win8-preview, so many things probably won't work, but atleast v7.0.28 BOINC finally installed without any problems.

Configured Einstein@home to run 2 tasks at once on the GPU, by setting "GPU utilization factor of BRP apps" to 0.5 in the Einstein-preferences. On client this equals 0.5 GPU, 0.5 CPU per task.

Started-up GPU-Z, verified OpenCL was installed, and started downloading Einstein-GPU-work. Then the 1st. task started it according to GPU-Z used roughly zero of the GPU, and both "Elapsed" and "Remaining" was counting-up.

Then the 2nd. task started on the other hand, GPU-usage jumped to roughly 75%, and one of the cpu-only tasks was paused. While "Elapsed" for both GPU-tasks counted up, "Remaining" now started counting-down, but granted not as fast as estimated.

Both finished after roughly 2h10m, meaning on average 65 minutes per task. Looking on actual cpu-time, this was roughy 30 minutes/task, so by running 2 at once this means 1h cpu-time for 2h10m run-time or Einstein@home uses 45% of a cpu-core.

Now no idea if so short run-time was just a fluke and neither if my GPU just computed garbage and will fail validation, but atleast these results indicates Ati is much faster than either spingadus or kmanley57 results indicated, as long as allows Ati the usage of a cpu-core.

[kmanley57]

Well I reserved a core for the ATI card and it did the WU in ~14000 seconds, but my NVidia still does a WU in ~7000 for the same amount of credits(500). So I see a CPU taking ~22000(~250 credits), the ATI ~14000(500 credits), and the NVidia about ~7000(500 credits). But with faster cards I am sure others will get better returns on time used. :D

[Al]

Configured Einstein@home to run 2 tasks at once on the GPU, by setting "GPU utilization factor of BRP apps" to 0.5 in the Einstein-preferences. On client this equals 0.5 GPU, 0.5 CPU per task.
Is an app_info file needed in addition to the preference change? My NVIDIA card shows .2 CPU and .5 Gpu, but only one task runs at 77% Gpu utilization.

I suck at app_info files so I'm hoping it isn't necessary.


Sent from my ADR6300

[zombie67]

No app_info.xml needed. Just change the setting:

1 = 1 task per GPU
.5 = 2 tasks per GPU
.33 = 3 tasks per GPU, etc.

FYI, the upper limit is the memory of the GPU. Per the admin, you can run 2-3 tasks per 1gb of VRAM.

[Al]

My bad. Had to wait until it completed one before it would do 2 at a time. Thanks

[denim]

Can I ask if anyone has seen a thermal difference between this and PG gpu work? I love the return on the pps gpu work, but my lappy will reach thermal even when the lappy cooler is on full tilt and both are a foot off the desk.

[Al]

Can I ask if anyone has seen a thermal difference between this and PG gpu work? I love the return on the pps gpu work, but my lappy will reach thermal even when the lappy cooler is on full tilt and both are a foot off the desk.

Einstein runs in the mid 40c for me, much cooler than PG. Then again, it doesn't utilize the gpu as well as PG. I have to run 3 wus at once to get gpu usage up to 85%.

Sent from my DROID RAZR using Tapatalk 2

[Slicker]

Can I ask if anyone has seen a thermal difference between this and PG gpu work? I love the return on the pps gpu work, but my lappy will reach thermal even when the lappy cooler is on full tilt and both are a foot off the desk.

Poem may be another alternative. With their newest version, after 10 minutes of running, the GPU was still at 0% utilization and running very cool. Then again, I'm not sure the WU would ever finish in my lifetime. However, prior to this version, the previous version still ran very cool because it was doing half the work on the CPU. From the looks of it, they now do ALL the work on the CPU making the GPU run very cool.;)

[rgathright]

Another one for you...

This laptop has AMD Radeon HD 6320 graphics but still gives this error:
http://einstein.phys.uwm.edu//host_sched_logs/5459/5459533

Here is the computer configuration:
http://einstein.phys.uwm.edu/show_host_detail.php?hostid=5459533

[Al]

I'm no expert but it looks like you dont have enough gpu memory.

Sent from my DROID RAZR using Tapatalk 2

[zombie67]

Remember, einstein apps take lots of VRAM. "per GB of VRAM you should be able to execute at least 2, possibly 3 instances." That works out to less than 500mb per task for sure. *Maybe* less than 333mb per task. So if you have 256mb, you are out. 512mb, and you can run one per.

You have 384mb, so that makes it borderline.

[Duke of Buckingham]

:D Finnally I could download some GPU tasks. The use of the GPU is quite funny and goes like a stairway. 10%, 25%, 70% and 100%. Is always jumping returns to zero. Well is this normal for Einstein or is my computer GPU having problems to do Einsten. :confused:

[zombie67]

That means your GPU task is CPU-starved. Change Tools -> computing preferences -> use at most -> 99%.

[Duke of Buckingham]

:D Nice what a litlle number can do. :-B Thanks Z. :-bd

Geez I don't know how to take this attached image. Sorry

[Duke of Buckingham]

I was looking for one app to Einstein and I could find one for cuda but I can not use it with the HD5770. Someone knows about one to use with ATI that could work fine? :p

[coronicus]

I was looking for one app to Einstein and I could find one for cuda but I can not use it with the HD5770. Someone knows about one to use with ATI that could work fine? :p

As soon i figure out how to tame my 5770 ill let ya know...

[Duke of Buckingham]

Thanks coronicus. The use of about 70% of the GPU is good but I think could be better. The system seems to float a lot also. The use of CPU changes in every second. I think the use of one core for two tasks could improve a lot the use of the GPU. Also the CPU can (with 2 cores) seems unable to run 2 CPU tasks and 1 GPU task. So to lose one core better run 2 GPU tasks at once. Only my thoughts about that, sorry if I said something very wrong.

The Einstein for GPU seems to error more than usual with all kinds of mistakes.Some tasks had been ran for 6 users and they keep on putting them to run again and the credit, well 500 points is not much (when they give them) for a GPU crunching, if we look at the electric bill. I am there because the team is in the Euro Crunch 2012. Of all the 6 tasks I made for GPU only one count as credit for the team.

I think this is really bad for a project to go anywhere, sometimes I think if the credits are paid with the scientists money. 500 points for more than 2 hours of GPU crunching is very few, even for me. I made some SETI beta and is almost the same thing, I hope they don't make a fashion out of this.

But many thanks coronius, you are always trying to help and that is very good. ^:)^

Very sad Duke with Einstein credits. :mad:

[EmSti]

To get Einstein GPU apps to work well, I had to do the following

1) Removed overclock setting on the GPU. I read multiple comments of people having issues. This removed nearly all of my invalid problems.
2) Decreased the % of processors used in the preferences. This frees up the CPU to feed the GPU and improves overall points. For my 8 core I am using 66%, which only allows 5 of the CPU threads to be used by Boinc.
3) To get more thruput I changed the "GPU utilization factor of BRP apps" to 0.5 in the Einstein account preferences. This will allow 2 tasks to run per GPU, but will also take more CPU (GPU credit is better). This setup needs 2 more CPU threads on my machine (0.5 CPU per task times 4 tasks). So that leaves me with 3 CPU threads to run CPU tasks. The GPU gets used more and overall is faster for completeing 2 tasks. (Note: only do this if your GPU has enough memory).
4) Unchecked the box for "Run CPU versions of applications for which GPU versions are available" - they take too long on the CPU.

On a 2 core system I would suspend CPU tasks and only run Eistein GPU tasks.

Also for my ATI cards, I had to upgrade the Boinc Client to 7.0.28. Looking at others results, cuda cards must be able to use 7.0.25

[Duke of Buckingham]

Thanks Emsti. The only issue there is I am overclocking the GPU. It worth while a lot. I will stop the overclocking the GPU meanwhile.
The use of GPU for 2 tasks needs an app once Einstein defines the WUs as as using one GPU for task. As my GPU is a simple one I would run only one GPU task and no CPU tasks. I would lose one CPU task and that is my main credit at Einstein for the moment.

Duke

[Al]

Good synopsis EmSti. I was going to write something about this and you saved me the time.

F$ - can that post be promoted to an article for future generations of Einstein crunchers?

[coronicus]

Thanks for the tip emsti that did help get my 6950 running full on but the 5770 eludes me as it is for duke... for some reason it refuses to run two task on the gpu even though it has 1g vram. but im still trying just a tad busy if anyone else has solved this puzzle please let me and duke know would love to see this thing running at 90+%.. so far i tried installing latest amd driver as well as the latest beta boinc but that did not solve it tried a few cc_config stuff but that did help any yet. so ill keep trying and let you know if i figure something out but so far i have had no luck..

oh the new driver from amd did solve the reboot issue it was having fyi.. updated 12.3 to 12.4

[Duke of Buckingham]

Try to take out one point for the GPU on the app. This system is very tied and as no space for defenitions. So I think. Taking 1% of the CPU was quite a lesson as Z told me. I think they didn't gave space to any mistakes.

Duke

[EmSti]

Thanks for the tip emsti that did help get my 6950 running full on but the 5770 eludes me as it is for duke... for some reason it refuses to run two task on the gpu even though it has 1g vram. but im still trying just a tad busy if anyone else has solved this puzzle please let me and duke know would love to see this thing running at 90+%.. so far i tried installing latest amd driver as well as the latest beta boinc but that did not solve it tried a few cc_config stuff but that did help any yet. so ill keep trying and let you know if i figure something out but so far i have had no luck..

oh the new driver from amd did solve the reboot issue it was having fyi.. updated 12.3 to 12.4

Another bizzare thing I ran into is trying to get the wus to follow the preference setups. Switching between work and home locations (one with .5 and the other with .33 for the gpu utilization) doesn't always take effect right away. I have forced update, aborted tasks, restarted clients, etc and had no luck. The log would say the location changed but didn't have an immediate effect. But after a few cycles of wus finishing, the change would suddenly kick in. A few times it was immediate. I even tried forcing it in the client state xml and restarting; it was ignored. Currently hand managing hydra6 as I try to get it back to 2 per gpu.

[coronicus]

Another bizzare thing I ran into is trying to get the wus to follow the preference setups. Switching between work and home locations (one with .5 and the other with .33 for the gpu utilization) doesn't always take effect right away. I have forced update, aborted tasks, restarted clients, etc and had no luck. The log would say the location changed but didn't have an immediate effect. But after a few cycles of wus finishing, the change would suddenly kick in. A few times it was immediate. I even tried forcing it in the client state xml and restarting; it was ignored. Currently hand managing hydra6 as I try to get it back to 2 per gpu.

well well that did the trick cancled the one it was working on and the que and now the 5770 is running 92% with two task.. thank you so much..

[Fire$torm]

Good synopsis EmSti. I was going to write something about this and you saved me the time.

F$ - can that post be promoted to an article for future generations of Einstein crunchers?

---> http://www.setiusa.us/showthread.php?3706-Einstein-GPU-app-How-To&p=42714#post42714

[Al]

Another bizzare thing I ran into is trying to get the wus to follow the preference setups. Switching between work and home locations (one with .5 and the other with .33 for the gpu utilization) doesn't always take effect right away. I have forced update, aborted tasks, restarted clients, etc and had no luck. The log would say the location changed but didn't have an immediate effect. But after a few cycles of wus finishing, the change would suddenly kick in. A few times it was immediate. I even tried forcing it in the client state xml and restarting; it was ignored. Currently hand managing hydra6 as I try to get it back to 2 per gpu.

Zombie clued me into this. If the tasks are downloaded when the preference setting at .33, they're tagged as such and your gpu will try to run 3 at once. If you want to change to 2 tasks you have to change the preference setting to .5, abort what you have downloaded at .33 and download more....which then get tagged at .5

@F$ - thanks for making that post a Sticky.

[Rattledagger]

To get Einstein GPU apps to work well, I had to do the following
Good writeup, but in my opinion it's missing one important thing, the minimum system-requirements then it comes to Ati-cards, meaning 5xxx or later GPU, Catalyst 12.x or later driver and 500 MB or more GPU-memory.

[EmSti]

Good writeup, but in my opinion it's missing one important thing, the minimum system-requirements then it comes to Ati-cards, meaning 5xxx or later GPU, Catalyst 12.x or later driver and 500 MB or more GPU-memory.

Good point, forgot that I read that.

[Fire$torm]

Good writeup, but in my opinion it's missing one important thing, the minimum system-requirements then it comes to Ati-cards, meaning 5xxx or later GPU, Catalyst 12.x or later driver and 500 MB or more GPU-memory.

Fixed.

[rgathright]

Zombie clued me into this. ...

As soon as I read this... my South Park mind went into overdrive. :p

http://static.fjcdn.com/pictures/CLUES_616ba2_61504.jpg

[EmSti]

As soon as I read this... my South Park mind went into overdrive. :p

http://static.fjcdn.com/pictures/CLUES_616ba2_61504.jpg

I am going to follow my clue over there

[rgathright]

I activated dual processing for both of the GTX 550 Ti graphics cards in my ASUS motherboard today. <:-P

Thanks team for sharing your experiences, it really helps many other people when you take the time to leave a note! :D

953953

[denim]

Finally have a card that can crunch Einstein GPU WU. :)