yeah here's a decent suggestion. something i'd like to see happen. make BOINC store all the data in use in the RAM until shutdown.

i installed a RamDISK on my PC. it runs a underclocked i7 2600 at 2800Mhz. i underclock it because it doesn't heat up as much and still works fine, and i have better performance per watt. much less of a draw and just about the same results - mediocre

i also run a R9 270x underclocked for similar reasons. it come in at 1100 base clock and the memory is at 1425, but i lower the clock to 600 and leave the memory at 1425. it performs at just about the same amount, much less power draw as well tho.

anywho! after installing a 4GB ram drive and putting the BOINC data file in the ram drive, the cpu time got cut to about 2/3's or even 1/3 the time. all ops are pretty much just floating around the ram and the cpu cache so it helped quite a bit. worth mentioning it! crystaldiskmark came in at about 3000+ read/write. went down to about 1600 after a while tho.. probably cause it's a bit hotter and in use.

this project is neat but suffers from severe inefficiency. the fact that i pay my own power bill doesn't solve the issue of productivity per watt and tweaks like this could help a lot.

also, when shopping for a CPU - here is a good formula to keep in mind. it's rather simple.

Watts x (base clock x cores) ÷ cinebench score = (lower is better) a comparative score that could give you a rough idea of the efficiency per watt.. it's a rough estimate really and if you don't know the wattage, TDP could work but its not very precise tho easier to find.

right now one of the best i could find was the i5 8400t. with that formula, the generational differences in intel processors aren't that much different. some score worse.

anywho, RamDisk! try it out.

anywho! after installing a 4GB ram drive and putting the BOINC data file in the ram drive, the cpu time got cut to about 2/3's or even 1/3 the time. all ops are pretty much just floating around the ram and the cpu cache so it helped quite a bit. worth mentioning it! crystaldiskmark came in at about 3000+ read/write. went down to about 1600 after a while tho.. probably cause it's a bit hotter and in use.

Exactly which data file? And what did you have to change where to point to the new location of the "data file".

Thank you.

Tom

---

A proud member of the OFA (Old Farts Association).

First, I can understand your desire to cut down on your energy consumption by down-clocking (which is the opposite of what many are trying to do...)
But have you considered that your run-times are now more than double those one would expect for your processor, thus you are negating the energy saving. One of the most effective energy saving techniques without impacting too much on the performance is to turn off all the automatic "turbo modes" that many modern processors have.
Second SETI does not make many large disk accesses during the execution of a task - by default it only writes a few bytes once a minute per task running, and this can be safely reduced to once every five or even ten minutes for most tasks. Over the years many people have tried various RAM disk utilities and SSDs, only to find their use doesn't actually affect the processing rate due to the lack of disk use by SETI. I do know that some other projects make continuous disk use and so the SSD/RAM disks can have a positive affect on overall performance.
Another thing to consider is that there is a very large range in execution times due to the nature of the data we are processing which makes it very difficult to see the impact of a change over a relatively short period of time - you may initially hit a pile of slow tasks, then do a change and find you are getting a pile of fast tasks and think you've found the "golden egg", when in reality you've done nothing really to the average execution time. To get a realistic measure of how long "average" tasks are taking you need to run for a couple of weeks, noting the time for every task, then do a change and log for another couple of weeks.
And then of course we have the initial "time guess" that for the first day is very inaccurate - it can be hundreds of percent too long, and only settles down after 11 "standard" tasks have been validated, without error, just now these tasks are rather rare, so it can take some days to achieve this. However actual run-times are logged accurately on the "valid tasks" for your computer.
Finally I notice that you "abandoned" a couple of hundred tasks in the wee small hours of the day (GMT) - this may be a result of the RAM disk doing a re-sync - please keep an eye on this.

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

well when i installed BOINC it asked for two directories, one was the program files and the other the data. tho ramdisk is giving me problems, so i have to completely turn off BOINC, manually back up the files and completely uninstall and delete ramdisk or else i keep getting a blue screen on start up.. sometimes i just get a blue screen when i try to activate the ramdisk... but when it works, it's stable. then i just copy all the data files back onto it and restart boinc.

my average credit score went up a bit.. donno if it's related, i'm not sure how they are given out.

oh and Bob, yeah it's showing improvements. but yeah might be a bit to soon. and i have turned off many features on the CPU. for me it's not so much raw performance, it's performance per watt. and also a lower TDP. so underclocking tends to give decent results but while using less power. tho the results aren't as good, it's just a more bang for my buck sort of situation. anyways, thats just me. i only stated it because my PC would run slower then a comparable one with an i7 2600 & R9 270x

As SETI only accesses the hard disk for a total of a few seconds (less than ten) during a two or three hour run your are very unlikely to see any improvement as a result of using RAM disk - which, as you rightly say, you have to re-synch every time you start and stop the computer (and no doubt a few times in between) you probably actually loose more than you gain.
I run all my computers at "stock" clock speed, but with all the various "turbo" modes turned off, mainly to extend the life of systems. As I said a the outset, I can understand why you are down clocking, have you actually measured the power drawn from the wall - it would be interesting to see how much power saving you have managed as I think it could be a bit more than you expect - power draw vs clock speed for electronics is not a linear relationship, but a sort of inverse square above a certain clock rate.

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

yeah it's probably not a night and day difference, but the heat levels are a good indicator. lower voltage drop. and yea their might not be to much saving on a ramdrive compared to a SSD.. but maybe it's cause i was using an older HDD. it does seem to be cutting a bit of time off. tho i might have been getting smaller tasks.

i wouldn't mind setting up a little rig with a few low end GPU's like 1050's or even gt 1030's. it would probably get a decent amount of productivity for a small amount of power draw. plus maybe a i5 8400t. don't really have the money for it tho. but maybe someday. gtx750's are pretty cheap.

The change from HDD to SSD did nothing noticeable (less than 1 second in 2 hours) in terms of SETI processing time, but does make a big difference in the boot time. That is because SETI makes so few disk accesses and they are so small.

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

ok then. seemed to work. thx for the info.