Here's an excerpt from one of my very early posts here:
Quote:
Oh ..... and my clinic ( of which I only have a quarter share ) has two offices ( separate towns ) and there is a total of 12 workstations plus server and firewall beasties. Now there's some serious horsepower under that hood, huh? And it doesn't get used at night. The BOINC usage preferences are clearly the tool for that. So methinks after sorting out the squiggles at home I might put an item on agenda for the post Christmas practice management meeting! Get 'em to sign off when hungover ehh?
..... a similiar level of flippancy, not really electricity related, but permission was obtained for use of something I definitely had access to, but only fractionally owned.
It's a significant issue, quite worthy of mention, the price of energetic electrons. Bruce has quoted the saving to the project here:
Quote:
The numbers are amusing. We currently have around 70000 computers working on Einstein@Home. Typically a modern computer will use another (say) 40 Watts of power if it is working hard, versus idle. If it's turned on and working instead of turned off, this would be more like 150 Watts. So I estimate that Einstein@Home is using perhaps 4 MW of power on the host computers. Now one MW-hour of energy costs around $100 USD. So this is around $400 USD/hour or $9600 USD/day just in electricity costs.
That was 10 Dec 2005, now if you look here 246,697 computers ( approx host computers with credit ) I think (?) is the current comparative figure. The math is even more favourable to the project today, for which I can certainly say there is gratitude from E@H! There is a cost deflected from the project's budget for sure, but it ought be emphasised that actual cost paid by users will vary widely in different locations as per the local market. ( I recall Eastern Europe being quoted as particularly steep for instance ).
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 would like to congratulate both Gary Roberts and Mike Hewson for the excellent moderation they are performing on this board.
Thank you for putting things in the proper perspective, and for effectively bringing stray discussions back on topic before they degenerate further.
In fairness to JRenkar, though, he did bring to my attention that I was using more electicity than I had specifically asked permission for. When I cajoled and pulled rank for my installations, I had focused on the idle cycles and science generated to back my position, and inadvertently forgot to raise the issue of cost.
The decent thing to do, therefore, is to inform superiors and colleagues of the additional 40-odd Watts per hour that those machines need to crunch the work the various BOINC projects send their way.
Quote:
...there is gratitude from E@H!
Yes, the moderators (Gary and Mike), project administrator/developer/scientist (Bruce), and the generally decent posters around here certainly make the E@H experience both informative and inspiring!
It's a significant issue, quite worthy of mention, the price of energetic electrons. Bruce has quoted the saving to the project here:
Quote:
The numbers are amusing. We currently have around 70000 computers working on Einstein@Home. Typically a modern computer will use another (say) 40 Watts of power if it is working hard, versus idle. If it's turned on and working instead of turned off, this would be more like 150 Watts. So I estimate that Einstein@Home is using perhaps 4 MW of power on the host computers. Now one MW-hour of energy costs around $100 USD. So this is around $400 USD/hour or $9600 USD/day just in electricity costs.
You are lucky one.
My electricity bill says per 1KWh:
0.1558 Euro
electricity tax 0.0205 Euro
then VAT 16%
thats 0.2045 Euro per KWh or 204.5 Euro per MWh. The exchange rate from Euro to US-$ is today 1.2666, for one MWh I have to pay 259 US$.
It's a significant issue, quite worthy of mention, the price of energetic electrons. Bruce has quoted the saving to the project here:
Quote:
The numbers are amusing. We currently have around 70000 computers working on Einstein@Home. Typically a modern computer will use another (say) 40 Watts of power if it is working hard, versus idle. If it's turned on and working instead of turned off, this would be more like 150 Watts. So I estimate that Einstein@Home is using perhaps 4 MW of power on the host computers. Now one MW-hour of energy costs around $100 USD. So this is around $400 USD/hour or $9600 USD/day just in electricity costs.
You are lucky one.
My electricity bill says per 1KWh:
0.1558 Euro
electricity tax 0.0205 Euro
then VAT 16%
thats 0.2045 Euro per KWh or 204.5 Euro per MWh. The exchange rate from Euro to US-$ is today 1.2666, for one MWh I have to pay 259 US$.
Bruce was pretty close on the cost estimate for the USA.
After checking my last power bill, my charge per kWh is about $0.0848, or $84.80 per MWh. I guess that's around 0.067 Euro on average per kWh. We don't have VAT, but state sales tax in NC is 3%.
By my estimates (13.661% completed after 1:17:43) it will take about 9.5 hours for one of these work units (as opposed to approx. 38-40 minutes for S4 work units with AKOSF S41.07) Ouch, that's 14-15 times as long.
The question is, will the credit get 14-15 times bigger as well ?
If not then it's the same effect as the Enhanced had on the SETI project and
I'm quiting this until until shorter WUs are available.
For now I'll suspend the project and keep on eye on the new credits to compare.
By my estimates (13.661% completed after 1:17:43) it will take about 9.5 hours for one of these work units (as opposed to approx. 38-40 minutes for S4 work units with AKOSF S41.07) Ouch, that's 14-15 times as long.
The question is, will the credit get 14-15 times bigger as well ?
If not then it's the same effect as the Enhanced had on the SETI project and
I'm quiting this until until shorter WUs are available.
For now I'll suspend the project and keep on eye on the new credits to compare.
You will most likely never be able to get the credit rate you got when using Akos' optimized client, regardless of how short or long the work-units are. When you were using the optimized app, you were processing 3-4 times faster than the average user, and therefore getting 3-4 times as much credit. Now that (most) of the optimizations are incorporated into the base client, you can no longer get a credit rate 3-4 times average, and the average credit rate cannot rise, as that would make Einstein's credit system skewed compared to all other BOINC projects. Thus, you credit rate will drop to normal levels. This, of course, was inevitable, as the optimizations were destined to become part of the base application. While S5 made it convenient to make these changes, neither the work unit length, nor the transition to S5 have any true causal relationship to the perceived credit drop.
Further optimisation of the app is possible. Credit is now determined server side based on the properties of the WU. If Akos or others do produce a more optimised app, using it will give a credit advantage, even without a calibrating client. Because some optimisations have already been included in the official app, the advantage will never be what it was for S4.
Here's an excerpt from one of
)
Here's an excerpt from one of my very early posts here:
..... a similiar level of flippancy, not really electricity related, but permission was obtained for use of something I definitely had access to, but only fractionally owned.
It's a significant issue, quite worthy of mention, the price of energetic electrons. Bruce has quoted the saving to the project here:
That was 10 Dec 2005, now if you look here 246,697 computers ( approx host computers with credit ) I think (?) is the current comparative figure. The math is even more favourable to the project today, for which I can certainly say there is gratitude from E@H! There is a cost deflected from the project's budget for sure, but it ought be emphasised that actual cost paid by users will vary widely in different locations as per the local market. ( I recall Eastern Europe being quoted as particularly steep for instance ).
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 would like to congratulate
)
I would like to congratulate both Gary Roberts and Mike Hewson for the excellent moderation they are performing on this board.
Thank you for putting things in the proper perspective, and for effectively bringing stray discussions back on topic before they degenerate further.
In fairness to JRenkar, though, he did bring to my attention that I was using more electicity than I had specifically asked permission for. When I cajoled and pulled rank for my installations, I had focused on the idle cycles and science generated to back my position, and inadvertently forgot to raise the issue of cost.
The decent thing to do, therefore, is to inform superiors and colleagues of the additional 40-odd Watts per hour that those machines need to crunch the work the various BOINC projects send their way.
Yes, the moderators (Gary and Mike), project administrator/developer/scientist (Bruce), and the generally decent posters around here certainly make the E@H experience both informative and inspiring!
...back to lurking...
RE: It's a significant
)
You are lucky one.
My electricity bill says per 1KWh:
0.1558 Euro
electricity tax 0.0205 Euro
then VAT 16%
thats 0.2045 Euro per KWh or 204.5 Euro per MWh. The exchange rate from Euro to US-$ is today 1.2666, for one MWh I have to pay 259 US$.
faeshn
RE: RE: It's a
)
You can buy your electricity in france,
1kwh = 0,1057 € during the day, and 0,0644 € for 8 night hours
VAT 5,5%
http://particuliers.edf.fr/article494.html
Bruce was pretty close on the
)
Bruce was pretty close on the cost estimate for the USA.
After checking my last power bill, my charge per kWh is about $0.0848, or $84.80 per MWh. I guess that's around 0.067 Euro on average per kWh. We don't have VAT, but state sales tax in NC is 3%.
RE: By my estimates
)
The question is, will the credit get 14-15 times bigger as well ?
If not then it's the same effect as the Enhanced had on the SETI project and
I'm quiting this until until shorter WUs are available.
For now I'll suspend the project and keep on eye on the new credits to compare.
RE: The question is, will
)
no
bye!
Hi, RE: The question is,
)
Hi,
you are comparing apples and bananas.
I've just finished a long workunit on an AMD Athlon XP 2000+:
cpu time: 46,875.00 secs
claimed credits: 177.36 (up to now no granted credits, but most likely it will be the same)
credits per hour: 13.62
The same machine gives me an average of 7.18 granted credits per hour for standard S4 albert 4.37 results.
As you can see I get more credits per hour than before if I'm comparing the standard applications.
Regards,
Carsten
RE: RE: By my estimates
)
You will most likely never be able to get the credit rate you got when using Akos' optimized client, regardless of how short or long the work-units are. When you were using the optimized app, you were processing 3-4 times faster than the average user, and therefore getting 3-4 times as much credit. Now that (most) of the optimizations are incorporated into the base client, you can no longer get a credit rate 3-4 times average, and the average credit rate cannot rise, as that would make Einstein's credit system skewed compared to all other BOINC projects. Thus, you credit rate will drop to normal levels. This, of course, was inevitable, as the optimizations were destined to become part of the base application. While S5 made it convenient to make these changes, neither the work unit length, nor the transition to S5 have any true causal relationship to the perceived credit drop.
Further optimisation of the
)
Further optimisation of the app is possible. Credit is now determined server side based on the properties of the WU. If Akos or others do produce a more optimised app, using it will give a credit advantage, even without a calibrating client. Because some optimisations have already been included in the official app, the advantage will never be what it was for S4.
Dead men don't get the baby washed. HTH