Einstein@Home has passed 800 Teraflops!
Submitted on 27 Nov 2012 10:23:15 UTC
Many thanks to the DC Russian Union for their ""Pulsar Rampage" computing challenge which helped to push Einstein@Home past the 800 Teraflop computing-power level for the first time in the history of the project!
Bruce Allen
Director, Einstein@Home
Language
Copyright © 2024 Einstein@Home. All rights reserved.
Comments
Einstein@Home has passed 800 Teraflops!
)
I would suggest thanking SETI@home as well. They are having major database problems and have restricted users to 100 WU per GPU. There are also bandwidth problems making connection to the SETI servers difficult. As a result, many SETI GPU crunchers including myself have moved (temporarily?) to Einstein to keep their crunchers busy.
My Team!
)
My Team!
How is the computing
)
How is the computing power/floating point speed computed? Does every work unit have a fixed and known number of floating point operations? Doesn't this depend on the system that is used? And why are there different numbers on different BOINC web pages?
Is that a lot?
)
Is that a lot?
With POEM ceasing WU
)
With POEM ceasing WU production, expect it to get even busier !
dunx
RE: I would suggest
)
This is a good point, but the fact that SETI@Home is having database and bandwidth problems isn't something I can feel happy about. Over the years the people behind SETI@Home have given us a lot of technical and intellectual support, and I am sorry that they are having these issues.
In any case, I am very grateful to the SETI@Home community for supporting Einstein@Home, and I hope that many of you continue to provide that support even after SETI@Home resolves its technical problems!
Director, Einstein@Home
RE: Is that a lot? Yes!
)
Yes! To put this in perspective, it was only four years ago (June 2008) that a computer broke the 1 Petaflop (1000 Teraflop) barrier. Einstein@Home is now very close to that computing speed. What's more, Einstein@Home is delivering this performance on a real-world problem, not on an artificial "Linpack" test code.
Director, Einstein@Home
RE: RE: I would suggest
)
I fully agree and my post was not meant to be taken too literally, just to point out that a lot of seti crunchers use Einstein as a backup project and that means that over the last 2-3 weeks they have been significantly increasing the amount of Einstein they crunch. I hope that seti resolves at least some of its problems soon so that I can take my main cruncher back to seti which is where I intended it to be. In the meantime it is hard at work crunching Einstein on CPU and GPU for you.
Hello, Bruce. You could, to
)
Hello, Bruce.
You could, to add to your 800 petaflops, have 75% of my true quadcore i7 when I'm not using it, and 25% when I am--if only someone will reply to my post on the Problems page.
Pieter Watson
Hello again. Never mind my
)
Hello again.
Never mind my previous post: Jord has replied to me. I assumed that replies would appear when I viewed my messages, but evidently they don't.
Sorry for hassling you.
Pieter
With Seti@home off-line until
)
With Seti@home off-line until next tuesday, we can make it to the 1Petaflop mark!
RE: In any case, I am very
)
Hi Bruce and thanks for all who remembers us, we are very happy to be here and give a hand to the E@H cause. Of course we are expecting the SETI problems will fix ASAP but until then you could count with all of our "firepower". And even some more from few new mates that are comming soon.
I just curious, and probabily that is a well know fact in the E@H comunitiy and i know could find the answer in some place, just don´t find where, there is any place that i could easely find whow many FLOPS (TERA i don´t expect but maybe in the range of GIGA) our team is doing for the E@H cause?
And congrats for the 800Mark, but that was part of the past, the last number show by the server status page is: 868.1 TERAFLOPS (maybe a new record?)
I expect our team will break the 2MM daily credit mark today (credit x FLOPS i have no ideia if have any conversion rate), hope that will add some more FLOPS to that mark. And that number could be better, only I have Pending credit: 10,176.39 in allmost 1900 WU, and i imagine the others are in the same range.
Congratulations on breaking
)
Congratulations on breaking 800 Teraflops!
And the kitties
)
And the kitties helped!!!
Meows.
RE: And the kitties
)
With some more help from the kitties, E@H reaches 872.9 TFLOPS (and rissing) showed now in the servers status screen!
I realy start to think E@H likes our hungry hosts...
I've a question, what would
)
I've a question, what would it cost roughly to operate an 800 TF supercomputer?
RE: I've a question, what
)
Megabucks per year in power costs ( including cooling ) for a start.
[ We're at 892.5 as I speak .... :-) ]
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
Although significant, I don't
)
Although significant, I don't imagine power is the 'biggie' in the mix. You need infrastructure to house it, people to manage/operate it, maintenance contracts to keep it all going, a hardware/software budget to keep it up-to-date and consultants fees to make sure you are able to convince those who supply all the funds that the money is being spent efficiently :-). I'm sure there are other 'things in the mix' as well.
This is what makes the concept of BOINC so brilliant. It turns out there are a whole host of people out there willing to work all sorts of hours, just to provide ALL the above for ... a bunch of 'cobblestones' :-).
Now who woulda thunk that!! :-).
Cheers,
Gary.
Hope we could help in the
)
Hope we could help in the final push tomorrow and break the 900 TeraFlops mark before SETI returns on next tuesday. We still trying to do our best for the E@H cause, today the production of our "litte" team reaches the 2,852,288 daily production mark and 2 more friends from our SETI main team join us now at the end of the night, so i hope we could reach the 3MM mark tomorrow.
Of course there is a list of
)
Of course there is a list of members here since the beginning and will continue to be here 24/7 as long as this project exists regardless what other projects are doing.
RE: Of course there is a
)
I am sure that is quite true.
But some of us are also quite happy to have chosen Einstein as our backup when our preferred project is down.
I suppose it really is a
)
I suppose it really is a "fair" deal. We give them widgets, they give us cobblestones/kibbles/popcorn..
Oh who am I kidding, but as long as it is useful why not? Hope to see you to 900 tomorrow.
Please don´t take my words
)
Please don´t take my words wrong, i realy like the time we pass here, and realy hope we gave E@H a little help, going to sleep now, hope to hear the good news of the new record when i wake-up, we are just 3.8 TF to break the 900TF mark!
Well one teraflop is as good
)
Well one teraflop is as good as any other! :-)
898.1 as I write .....
Cheers, Mike.
[ edit ] 898.7
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
According to the info here
)
According to the info here (date 24.02.2012), germans fastest supercomputer makes 831 teraflops with initial cost of 22,5 million euros and running costs of 2 million euros annual.
Don't know how excact the flops calculation based on RAC is, but in the end, a cool performance for some "Home-Computer" :-).
(at the moment 899.0 TeraFlops on status page)
Hi! Just for the records:
)
Hi!
Just for the records: since then, the supercomputer described in the article above has lost its throne as Germany's Top Supercomputer to this impressive installation:
http://en.wikipedia.org/wiki/SuperMUC
Cheers
HB
899.5 Cheers, Mike.
)
899.5
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
RE: Hi! Just for the
)
Out of curiosity, how does MPI-AEI's own Atlas compare on that scale?
We just reached it!
)
We just reached it! =D
Floating point speed
(from recent average credit of all users) 900.3 TFLOPS
Congrats friends from E@H the
)
Congrats friends from E@H the 900.8 TFLOPS mark was breaked!
I´m happy we are here with you all to celebrate!
Just too early here to open the celebration beer! But later that is guaranted!
Somebody needs to change the thread title...!!!
RE: Out of curiosity, how
)
Just go to the http://www.top500.org/ and find it in the Top 500 lists. They show the Flops, Power Hungry, and some more interesting features.
BTW a 900 TERAFLOPS is almost equivalent to the Rmax (TFlop/s) of the #26 world biggest supercomputer by that list and uses about 2800.00 kW to do it´s job.
Hallo! Amazing!!! Today at
)
Hallo!
Amazing!!!
Today at about 11:35(UTC) (interpeolated) we reached 900 TFLOPS and it´s still increasing at a rate of impressive 23,7 +/- 0,6 TFLOPS/day. If this holds, we will have 1 PFLOPS to the end of this week. Hopefully !!!!!!!
Kind regards and happy crunching
Martin
On that basis, we should be
)
On that basis, we should be past Russia's fastest TOP500 entry shortly (901.9 TFLOPs).
Yes, but you can´t realy
)
Yes, but you can´t realy compare two systems just by the TERAFLOPS, but that´s gives us "simple mortals" a good way to do that.
BTW i ask before but no one answer, is there any place i could see how many MEGA or GIGA Flops my hosts or team are realy doing? Something like a credit x flops conversion table? Thanks in advance for any help.
RE: BTW i ask before but no
)
The BOINC credit is the Cobblestone, whereby a 1,000 MIPS/1 GFlOp machine doing 24 hours of computation will earn 200 credits. This works out to 432 billion ops. = 1 BOINC credit. Computing is so cheap these days, no?
So you can get your GFlOps by dividing your RAC by 200... You're doing just over a teraflop.... nice. :^)
Happy to have brought my piddly fifth of a teraflop to this project while SAH is down. Seems to be the go-to secondary for the SAH crunchers (so similar... even uses Arecibo data.) We'll break a petaflop yet.
Thanks... I'll have a look at that. Thought it was universal >.
RE: BTW i ask before but no
)
E@H seem to use RAC / 100000:
http://einsteinathome.org/node/196514
RE: So you can get your
)
Thanks... so i´m happy i almost made my first TERAFLOP for the E@H cause!.
Still thinking somebody needs to change the thread title... 800 is past!
Congrats for 903.4 TFLOPS
)
Congrats for 903.4 TFLOPS !!!
RE: RE: So you can get
)
I'm quite excited about this -- I am thinking about adding another news items for > 900 Tflops. I hope we can pass one Petaflop (1000 Teraflops) before the end of the year!
Director, Einstein@Home
RE: RE: Out of curiosity,
)
Well, I think I was asking because even by the time of the Open Day and guided tours in July last year, Atlas had already outgrown its 2008 build specification of 30 TeraFlops. They were adding Tesla nodes to supplement the original CPU-only design around the time we were there, and were planning to double the cluster's size by taking over the empty basement room next door.
So my curiosity was more to do with the status of those expansion plans.
RE: I'm quite excited about
)
Don't forget that tomorrow (Dec. 4th ETA) SETI@Home will be reopening and a lot of the big crunchers will be returning to there.
Sorry to be a downer. :^)
RE: RE: I'm quite excited
)
But i´m sure our "pendings" (i already have more than 10k pendings in more than 2K WU > 1MM in pendigs credits) will continue to add some more TERAFLOPS to the E@H cause for at least few days... and with the 100WU limit, our battleships will sure continue to crunch a lot of more E@H WU... helping the E@H cause!
Maybe a new thread for the 900 Tflops is a better ideia.
RE: But i´m sure our
)
I have enough work from here for about a week (due to the 100 work share you indicated... heh) and I'll let it all complete.
Ideally, SETI@Home and Einstein@Home would be semi-unified as they are both using Arecibo data. It would be nice if every work unit on both projects was checked for everything (ie both narrowband ETI signals, pulsars and whatever else.) It's all good discovery!
RE: RE: I'm quite excited
)
Well, that is an estimate. And when things do come up, I would not be surprised at all to see some difficulties arise.
The servers and the bandwidth are going to be very congested with all the empty hosts clamoring for work.
All my rigs are on auto...and I suspect they shall still be doing some EAH for a week or two if they can't get enough Seti to keep them occupied.
RE: All my rigs are on
)
Now that I finally know about the zero workshare for secondary projects -- thanks Juan... only took me ten years to find that out. ;^) -- I'll be doing the same and leaving both projects active so I don't have to keep manually switching back and forth, oh, and running out of work!
Yeah that first couple of days back will be rough so it helps having a week's work here to keep busy with.
RE: RE: RE: Out of
)
I'd forgotten about that! Nearby the cage with the tape robot inside. Yeah, what's the go with that?
Can anyone remember what E@H's TFlops was back then? I think we've at least tripled since.
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
RE: I'd forgotten about
)
The [url=http://wayback.archive.org/web/20110623041318*/http://einstein.phys.uwm.edu/server_status.html]Wayback machine[/url] says:
[pre]15-Aug-2010: 273.4 TFLOPS
27-Apr-2011: 355.0 TFLOPS
25-May-2011: 351.8 TFLOPS
23-Jun-2011: 399.1 TFLOPS[/pre]
Edit - and to answer the other part of the question, Atlas was rated at #58 in the top 500 in June 2008, when it had 5,420 CPU cores. The Linpack benchmark hasn't been updated since, and Atlas last appeared at #415 in the November 2010 list.
So a tad over doubled ( since
)
So a tad over doubled ( since mid 2011 ) then, thanks Richard. :-)
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
I personally believe an
)
I personally believe an official news post would be very appropriate to mark your 900T threshold.
Floating point speed 915.9
)
Floating point speed 915.9 TFLOPS and still rissing!