I just queued up a bunch (24 to be exact). Hopefully that'll get this machine through the outage.
Well, 24 tasks could be a whole heap or not much at all depending on how fast/no. of cores your machine has :-). You need to talk in "days of cache" that you have stored away. Before my little power outage, about a third of my machines had 7 or more days of work stored up. About another third had about 3-5 days and the rest were due to get their fill this morning.
Very true. They take on average 8 hours a piece, two at a time. So that's about 4 days worth. I've also reattached to PrimeGrid and lowered my cache back down to .01/.01. That way if I do run out of E@H, I'll pick up PG.
But hey, in the grand scheme of thing, it's important to have fun... which is why I'm building the client and (maybe) the manager.
I didn't recognise the full "prior art" term because I was too focussed on just the word "prior" as in which event is "prior" to which :-).
I like the idea behind the various GNU public licenses - basically 'I am the author, I'd like to give this away to The World, but not so as to allow someone to pick it up and profit from it and/or strangle it' .....
Quote:
like some dominant large energy industry buying up potentially emergent rival technologies for the sake of throttling them at birth .... to take a purely random example :-)
I'll see if I score an S5R2. Ah, I see my best two machines just had a rash of deletion requests from server side and then more than a few downloads ( in the 1176.xx range ).
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
S5R3 search progress
Total needed Already done Work still remaining
7,369,434 units 7,369,434 units 0 units
100 % 100.000 % 0.000 %
311.9 days 311.9 days 0.0 days (estimated)
Work and Results
WORKUNITS Approximate #
in database 533,824
with canonical result 270,634
no canonical result 263,190
RESULTS Approximate #
in database 1,129,909
unsent 115,354
in progress 243,335
deleted 556,005
valid 542,039
valid last week 448,858
invalid 33
Oldest Unsent Result 0 d 6 h 50 m
Only incoming WUs from clients and then it is done?
S5R3 search progress
Total needed Already done Work still remaining
7,369,434 units 7,369,434 units 0 units
100 % 100.000 % 0.000 %
311.9 days 311.9 days 0.0 days (estimated)
Work and Results
WORKUNITS Approximate #
in database 533,824
with canonical result 270,634
no canonical result 263,190
RESULTS Approximate #
in database 1,129,909
unsent 115,354
in progress 243,335
deleted 556,005
valid 542,039
valid last week 448,858
invalid 33
Oldest Unsent Result 0 d 6 h 50 m
Only incoming WUs from clients and then it is done?
The unsent are the left WU and than it's done!!!!!111
The unsent are the left WU and than it's done!!!!!111
Yes, there is no more new work to generate for S5R3 and the 115,354 unsent represents the remaining new work still to be distributed.
Now, some 16 hours later, that figure has dropped to around 66,000 so in less than 24 hours, the primary work will all be "out in the wild". There will then be a significant flow of resend work that will become available minute by minute, principally as a result of work that was issued some 18 days ago that will be expiring. That work will probably be at a fairly constant rate for the subsequent 18 days and may even reach a peak in the latter stages if a significant number of people have stocked up over the last couple of days with more than they can actually handle. There may also be a component of "abandoned caches" as curious people decide to try out the new app. After all the grass is always greener ... :-).
Over the coming weekend, with the server shut down for the upgrade, probably quite a bit of resend work will accumulate so that there should be a bit of a burst of that work as soon as the new system is back on line. There will also be quite a scramble for hosts to return work completed over the weekend and for many of those same hosts to get the whole new kit for S5R4. Each time a host contacts the server, a decision on what work to send will need to be made by the server. Monday (or whenever startup time actually is) might be a good time to avoid trying to contact the new system :-). It should certainly get a baptism of fire :-).
So fellow Einstein crunchers,
does it mean that those of us with the Beta and Power applications, will have to delete the App_info.xml file for us to then get work with the new S5R4 application (or what ever it will be called)?
I take it that the Power Apps are only for S5R3 and can't just be merged with the new work.
The S5R4 then is a new application all together and therefore has not yet been optimised like the Power Apps ?? Is this correct?
Thanks and have a good day.
Time for me to look for a back up project for a while, those last few Lattice work units might fill the bill as they take about 130 hours or more, but the credit is not the best.
Might try QMC or Rosetta as Cosmology is giving Zero credit on many days worth of work units, wasting my time there.
Yeah might go Rosetta and try and get over 1,000,000 credits.
So fellow Einstein crunchers,
does it mean that those of us with the Beta and Power applications, will have to delete the App_info.xml file for us to then get work with the new S5R4 application (or what ever it will be called)?
I take it that the Power Apps are only for S5R3 and can't just be merged with the new work.
You hit the nail, that's correct! Removing the app_info.xml (after running out of work) is one way. Or if you feel like investing some time, you can edit the app_info.xml to support both new and old runs, follow the links in this message.
Quote:
The S5R4 then is a new application all together and therefore has not yet been optimised like the Power Apps ?? Is this correct?
Quite to the contrary. The new Linux/Windows/OSX apps for the S5R4 run are almost identical to the fastest power app versions of the S5R3 run. Modifications include a different output format for results and enhancements for compatibility with the fresh input files for S5R4.
Having said that, the apps are one thing, the workunit design and credit awarding system are another.
Bernd has announced that the workunits will be designed differently for S5R4, also to minimize the runtime-variations we see in S5R3. I guess the goal of the design is to have an overall runtime per workunit roughly identical to S5R3 workunits, but we will see.
In the past it has been usual to align the credits in a way so that credit output is comparable for identical PCs on the major projects, and then leave the credits untouched during the run even if the apps are optimized further. So do not expect that credits/h will remain the same for S5R4 compared to S5R3. Bernd will probably make an announcement on credits etc in the next days.
So fellow Einstein crunchers,
does it mean that those of us with the Beta and Power applications, will have to delete the App_info.xml file for us to then get work with the new S5R4 application (or what ever it will be called)?
I take it that the Power Apps are only for S5R3 and can't just be merged with the new work.
You hit the nail, that's correct! Removing the app_info.xml (after running out of work) is one way. Or if you feel like investing some time, you can edit the app_info.xml to support both new and old runs, follow the links in this message.
Quote:
The S5R4 then is a new application all together and therefore has not yet been optimised like the Power Apps ?? Is this correct?
Quite to the contrary. The new Linux/Windows/OSX apps for the S5R4 run are almost identical to the fastest power app versions of the S5R3 run. Modifications include a different output format for results and enhancements for compatibility with the fresh input files for S5R4.
Having said that, the apps are one thing, the workunit design and credit awarding system are another.
Bernd has announced that the workunits will be designed differently for S5R4, also to minimize the runtime-variations we see in S5R3. I guess the goal of the design is to have an overall runtime per workunit roughly identical to S5R3 workunits, but we will see.
In the past it has been usual to align the credits in a way so that credit output is comparable for identical PCs on the major projects, and then leave the credits untouched during the run even if the apps are optimized further. So do not expect that credits/h will remain the same for S5R4 compared to S5R3. Bernd will probably make an announcement on credits etc in the next days.
CU
Bikeman
Thanks Bikeman,
Ok as no more work will come for a while I will stop my computers asking for work and complete what I have then set to no new work, for a few days.
That's disappointing about the credit thing. I thought they were about right.
If they are adjusted then I hope they aren't hit too much, it mainly depended on how many of the shorter work units one got as to how much a person received in credits per hour.
The credits have been good though, which is why I went to the power applications, they were heaps better than the standard application which was on the low side.
Don't get me wrong, I didn't mean to say credits will definitely be lower, but that you should not expect them to remain the same, just because the apps are more or less the same. We will see next week. My personal view is that with the power/beta apps (and new S5R4 stock apps) E@H is now a bit over the average credits rate.
RE: RE: I just queued up
)
Very true. They take on average 8 hours a piece, two at a time. So that's about 4 days worth. I've also reattached to PrimeGrid and lowered my cache back down to .01/.01. That way if I do run out of E@H, I'll pick up PG.
But hey, in the grand scheme of thing, it's important to have fun... which is why I'm building the client and (maybe) the manager.
Kathryn :o)
Einstein@Home Moderator
RE: I didn't recognise the
)
I like the idea behind the various GNU public licenses - basically 'I am the author, I'd like to give this away to The World, but not so as to allow someone to pick it up and profit from it and/or strangle it' .....
I'll see if I score an S5R2. Ah, I see my best two machines just had a rash of deletion requests from server side and then more than a few downloads ( in the 1176.xx range ).
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: ... it's important to
)
Sure is ... :-).
Wot!! ... delving into makefiles, etc? Boy, that's impressive!!
I'd better dust off that platinum plated geek badge that I've hidden away somewhere because Jord was trying to purloin it when I wasn't looking :-).
Cheers,
Gary.
No new WUs i
)
No new WUs i suppose?
Only incoming WUs from clients and then it is done?
RE: No new WUs i
)
The unsent are the left WU and than it's done!!!!!111
RE: RE: unsent 115,354
)
Yes, there is no more new work to generate for S5R3 and the 115,354 unsent represents the remaining new work still to be distributed.
Now, some 16 hours later, that figure has dropped to around 66,000 so in less than 24 hours, the primary work will all be "out in the wild". There will then be a significant flow of resend work that will become available minute by minute, principally as a result of work that was issued some 18 days ago that will be expiring. That work will probably be at a fairly constant rate for the subsequent 18 days and may even reach a peak in the latter stages if a significant number of people have stocked up over the last couple of days with more than they can actually handle. There may also be a component of "abandoned caches" as curious people decide to try out the new app. After all the grass is always greener ... :-).
Over the coming weekend, with the server shut down for the upgrade, probably quite a bit of resend work will accumulate so that there should be a bit of a burst of that work as soon as the new system is back on line. There will also be quite a scramble for hosts to return work completed over the weekend and for many of those same hosts to get the whole new kit for S5R4. Each time a host contacts the server, a decision on what work to send will need to be made by the server. Monday (or whenever startup time actually is) might be a good time to avoid trying to contact the new system :-). It should certainly get a baptism of fire :-).
Cheers,
Gary.
So fellow Einstein crunchers,
)
So fellow Einstein crunchers,
does it mean that those of us with the Beta and Power applications, will have to delete the App_info.xml file for us to then get work with the new S5R4 application (or what ever it will be called)?
I take it that the Power Apps are only for S5R3 and can't just be merged with the new work.
The S5R4 then is a new application all together and therefore has not yet been optimised like the Power Apps ?? Is this correct?
Thanks and have a good day.
Time for me to look for a back up project for a while, those last few Lattice work units might fill the bill as they take about 130 hours or more, but the credit is not the best.
Might try QMC or Rosetta as Cosmology is giving Zero credit on many days worth of work units, wasting my time there.
Yeah might go Rosetta and try and get over 1,000,000 credits.
RE: So fellow Einstein
)
You hit the nail, that's correct! Removing the app_info.xml (after running out of work) is one way. Or if you feel like investing some time, you can edit the app_info.xml to support both new and old runs, follow the links in this message.
Quite to the contrary. The new Linux/Windows/OSX apps for the S5R4 run are almost identical to the fastest power app versions of the S5R3 run. Modifications include a different output format for results and enhancements for compatibility with the fresh input files for S5R4.
Having said that, the apps are one thing, the workunit design and credit awarding system are another.
Bernd has announced that the workunits will be designed differently for S5R4, also to minimize the runtime-variations we see in S5R3. I guess the goal of the design is to have an overall runtime per workunit roughly identical to S5R3 workunits, but we will see.
In the past it has been usual to align the credits in a way so that credit output is comparable for identical PCs on the major projects, and then leave the credits untouched during the run even if the apps are optimized further. So do not expect that credits/h will remain the same for S5R4 compared to S5R3. Bernd will probably make an announcement on credits etc in the next days.
CU
Bikeman
RE: RE: So fellow
)
Thanks Bikeman,
Ok as no more work will come for a while I will stop my computers asking for work and complete what I have then set to no new work, for a few days.
That's disappointing about the credit thing. I thought they were about right.
If they are adjusted then I hope they aren't hit too much, it mainly depended on how many of the shorter work units one got as to how much a person received in credits per hour.
The credits have been good though, which is why I went to the power applications, they were heaps better than the standard application which was on the low side.
Hi! Don't get me wrong, I
)
Hi!
Don't get me wrong, I didn't mean to say credits will definitely be lower, but that you should not expect them to remain the same, just because the apps are more or less the same. We will see next week. My personal view is that with the power/beta apps (and new S5R4 stock apps) E@H is now a bit over the average credits rate.
CU
H-B