The workunits are longer than previously most definitely, but I don't think it's all that bad..............the old S4 units on my PIII 1.0 took around 8 or 9 hours with the standard app......akos dropped that to a bit under 2 hours....
with S5 we are crunching more intensive data......the project peeps did that to utilize our increased power with the included Akos help....so we're looking at essentially a higher resolution......this is good!!!1 And, the longer the work units take, the more points will be awarded, to attempt to keep the average credit/crunching hour equal......can't remember the value.....10 to 30 I think.
So, we're not loosing anything as far as credit......just not turning over as many units, which has nothing to do with credit based ranking anyway.
And,......the completion times for me are not that different than other projects I participate in..........i.e. QMC on my PC is 60% done at 23 hours!!!!! Einstein is 67% done at 14 hours......so it's going faster than my average QMC, WCG usually seems to take me about 12 or more hours.........LHC takes upwards of 8 hours......and CPDN take months!!!!!!
as long as we get enough credit for the time taken, everything is ok in my book.
And I like thinking we're digging down deeper looking for those hidden waves.....
I do miss watching the target zoom across the screen, but that's just gimmicky emotion.......I can reason through that and get over it.
Pepperami has a valid complaint in why didn't they just adjust the times up to where the standard S4s were......with the slightly optimized app for S5 this would still enable a "resolution" increase........dunno why they didn't, may be time segments wouldn't work out that way.........
hypothetical.......say S4 used 30 minute segments..........with more speed they wanted to increase the segment to dig deeper, but could only move up to 45 minutes..assume 15 minute time chunk increases are all is available.......so in being forced to do this increase......the completion times jumped up ALOT!
numbers above are just made up.
I'm going to try this out........
I'm going to add up enough S4 units by adding the completion times until I get up to about 20 hours.......then I'll tally the granted credit........I'm then going to compare this with the granted credit from this Long S5 I've got and see what I get......
I'm hoping for equal, or more with S5.........grantedcredit/reported crunched time comparison.
I have noticed no change in my RAC. In another project when things were changed, you noticed a large change in RAC, quickly. Since this one has been running for several days, and now most of my machines are on S5, I am still getting the same RAC.
I believe that they have the numbers running well. No one else is complaining of loss of RAC, that I have seen, either.
No matter what our problems are with S5, it is nothing compared to the chaos that Akos optimizations have wrought on SZTAKI Desktop Grid. For instance, my SZTAKI WU times went from ~60 min to ~1 minute after Akos optimization! That kind of improvement seems to have brought the SZTAKI WU generation to a complete halt today. I am not sure the SZTAKI management regards Akos as a hero, like we do.
I'm going to add up enough S4 units by adding the completion times until I get up to about 20 hours.......then I'll tally the granted credit........I'm then going to compare this with the granted credit from this Long S5 I've got and see what I get......
I'm hoping for equal, or more with S5.........grantedcredit/reported crunched time comparison.
You'll need to do the comparison with the standard "stock" apps for S4 & S5. Putting an Akos app in the mix will throw the curve off quite a bit because the S5 is only as you said "slightly optimized". Even a C41.03 is a big boost to "stock" S4.
I think a lot of people are misrepresenting and/ or misunderstanding the comparisons with the time issues based on the figures posted by Einstein's project developers. Their comparisons are between the S4 & S5 "official" applications.
this is true nevermorestr!.....umm the best I can do is use the times reported with the akos apps using the trux calibrating client......I still don't have an S5 completed to compare.......but my Akos/trux S4 I got 17.81 credits per hour of crunch time....
how does this compare with others who have finished an S5?
and yes, some people seem to be comparing crunch times to the akos apps instead of the standard S4.....
S5's credit comparison - I was just considering crunch time in the previous post, since credit is credit is credit, regardless.
This is a slight different set of work units, since I didn't include any "pending":
Average crunch time for S4's (the last 20 validated) was 1:08:06 (akosf)
Average credit for above was 35.71 claimed, 31.21 granted.
Average credit per hour was 31.46 claimed, 27.49 granted.
Average crunch time for S5's (the first 5 validated) was 15:01:25 (standard)
Average credit for above was 154.48 claimed/granted.
Average credit per hour was 10.28 claimed/granted.
Hope it helps anyone studying such things - if it matters, these are from P4 (Northwoods) with CPU speeds from 2.4 to 3.4 GHz, with HT enabled on all but the 2.4 (not HT capable).
And yes, I realize I'm comparing apples and oranges somewhat in regards to processing time since the S4 numbers are using the optimized application - but the "granted" credit comparision should be reasonable unless almost all of the quorums were being made up of at least 2 optimized applications (I honestly didn't drill into them far enough to check whether they were or were not).
S5's credit comparison - I was just considering crunch time in the previous post, since credit is credit is credit, regardless.
crunch time yes.......I guess that is the title of the thread.......and I perhaps wrongly brought in some time/credit comparisons.
I was hoping some of the pain of slower wu turnaround time could be alleviated with the realization of doing more advanced work, and there being a minimal impact on overall credit......
am wondering if in the long run if longer crunch times might elicit more credit overall due to less time uploading/downloading a bunch of workunits and switching between them every hour......
still off topic a bit though......
long times hurt emotionally, but I think like you all have said, it's a subjective and relative thing......hopefully we can get used to it....didn't bother me when I was running F@H alot.....they also had long long crunch times, and granted set values for credit based on the specific unit.
Okay - first few S5's
)
Okay - first few S5's completed:
Average crunch time for S4's (the last 42 I finished) was 1:14:06 (akosf)
Average crunch time for S5's (the first 7 I finished) was 16:01:04 (standard)
Ouch! Purely subjective, I know - and I understand the rationale, but it still "feels" like turning a P4/3.4 into a P/233.
.
The workunits are longer than
)
The workunits are longer than previously most definitely, but I don't think it's all that bad..............the old S4 units on my PIII 1.0 took around 8 or 9 hours with the standard app......akos dropped that to a bit under 2 hours....
with S5 we are crunching more intensive data......the project peeps did that to utilize our increased power with the included Akos help....so we're looking at essentially a higher resolution......this is good!!!1 And, the longer the work units take, the more points will be awarded, to attempt to keep the average credit/crunching hour equal......can't remember the value.....10 to 30 I think.
So, we're not loosing anything as far as credit......just not turning over as many units, which has nothing to do with credit based ranking anyway.
And,......the completion times for me are not that different than other projects I participate in..........i.e. QMC on my PC is 60% done at 23 hours!!!!! Einstein is 67% done at 14 hours......so it's going faster than my average QMC, WCG usually seems to take me about 12 or more hours.........LHC takes upwards of 8 hours......and CPDN take months!!!!!!
as long as we get enough credit for the time taken, everything is ok in my book.
And I like thinking we're digging down deeper looking for those hidden waves.....
I do miss watching the target zoom across the screen, but that's just gimmicky emotion.......I can reason through that and get over it.
Matt3223: I agree with most
)
Matt3223:
I agree with most of what you say. But I just find it odd that the relative performance should change so dramaticly.
Pepperami has a valid
)
Pepperami has a valid complaint in why didn't they just adjust the times up to where the standard S4s were......with the slightly optimized app for S5 this would still enable a "resolution" increase........dunno why they didn't, may be time segments wouldn't work out that way.........
hypothetical.......say S4 used 30 minute segments..........with more speed they wanted to increase the segment to dig deeper, but could only move up to 45 minutes..assume 15 minute time chunk increases are all is available.......so in being forced to do this increase......the completion times jumped up ALOT!
numbers above are just made up.
I'm going to try this out........
I'm going to add up enough S4 units by adding the completion times until I get up to about 20 hours.......then I'll tally the granted credit........I'm then going to compare this with the granted credit from this Long S5 I've got and see what I get......
I'm hoping for equal, or more with S5.........grantedcredit/reported crunched time comparison.
I have noticed no change in
)
I have noticed no change in my RAC. In another project when things were changed, you noticed a large change in RAC, quickly. Since this one has been running for several days, and now most of my machines are on S5, I am still getting the same RAC.
I believe that they have the numbers running well. No one else is complaining of loss of RAC, that I have seen, either.
No matter what our problems
)
No matter what our problems are with S5, it is nothing compared to the chaos that Akos optimizations have wrought on SZTAKI Desktop Grid. For instance, my SZTAKI WU times went from ~60 min to ~1 minute after Akos optimization! That kind of improvement seems to have brought the SZTAKI WU generation to a complete halt today. I am not sure the SZTAKI management regards Akos as a hero, like we do.
RE: I'm going to try this
)
You'll need to do the comparison with the standard "stock" apps for S4 & S5. Putting an Akos app in the mix will throw the curve off quite a bit because the S5 is only as you said "slightly optimized". Even a C41.03 is a big boost to "stock" S4.
I think a lot of people are misrepresenting and/ or misunderstanding the comparisons with the time issues based on the figures posted by Einstein's project developers. Their comparisons are between the S4 & S5 "official" applications.
this is true
)
this is true nevermorestr!.....umm the best I can do is use the times reported with the akos apps using the trux calibrating client......I still don't have an S5 completed to compare.......but my Akos/trux S4 I got 17.81 credits per hour of crunch time....
how does this compare with others who have finished an S5?
and yes, some people seem to be comparing crunch times to the akos apps instead of the standard S4.....
S5's credit comparison - I
)
S5's credit comparison - I was just considering crunch time in the previous post, since credit is credit is credit, regardless.
This is a slight different set of work units, since I didn't include any "pending":
Average crunch time for S4's (the last 20 validated) was 1:08:06 (akosf)
Average credit for above was 35.71 claimed, 31.21 granted.
Average credit per hour was 31.46 claimed, 27.49 granted.
Average crunch time for S5's (the first 5 validated) was 15:01:25 (standard)
Average credit for above was 154.48 claimed/granted.
Average credit per hour was 10.28 claimed/granted.
Hope it helps anyone studying such things - if it matters, these are from P4 (Northwoods) with CPU speeds from 2.4 to 3.4 GHz, with HT enabled on all but the 2.4 (not HT capable).
And yes, I realize I'm comparing apples and oranges somewhat in regards to processing time since the S4 numbers are using the optimized application - but the "granted" credit comparision should be reasonable unless almost all of the quorums were being made up of at least 2 optimized applications (I honestly didn't drill into them far enough to check whether they were or were not).
.
RE: S5's credit comparison
)
crunch time yes.......I guess that is the title of the thread.......and I perhaps wrongly brought in some time/credit comparisons.
I was hoping some of the pain of slower wu turnaround time could be alleviated with the realization of doing more advanced work, and there being a minimal impact on overall credit......
am wondering if in the long run if longer crunch times might elicit more credit overall due to less time uploading/downloading a bunch of workunits and switching between them every hour......
still off topic a bit though......
long times hurt emotionally, but I think like you all have said, it's a subjective and relative thing......hopefully we can get used to it....didn't bother me when I was running F@H alot.....they also had long long crunch times, and granted set values for credit based on the specific unit.