Thoughts On Credits

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6591
Credit: 322197174
RAC: 262398

RE: How much do sequence

Message 83699 in response to message 83697

Quote:

How much do sequence numbers matter? .....

Absolute minimum runtime is 46 seconds,
absolute maximum is at 71 seconds,

But this is only an upper limit for runtime variation because in a WU, there will be several hundred skypoints and they cannot all be at the absolute minimum or absolute maximum.

I'd guesstimate that the real-life observed runtime variation in the frequency range near 140 Hz for complete workunits should be more like

(maximum runtime) ~ 1.4 x (minimum runtime)

Under the assumption that all WUs contain the same number of skypoints (as they did approximately in S5R4, no idea whether this still holds in S5R4 tho).

Don't we love your Tomcat! That min/max ratio is about what is was during S3 for many machines ( varying with integer/float architectures, etc ) across alot of the parameter space. You'd expect a different calibration of the runtime formulae, but I don't see a radical departure there.

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

Brian Silvers
Brian Silvers
Joined: 26 Aug 05
Posts: 772
Credit: 282700
RAC: 0

RE: Also, Brian I hope your

Message 83700 in response to message 83698

Quote:

Also, Brian I hope your dad's surgery went smoothly and successfully and that he has a speedy recovery

He's "ok". He's awake, but I don't know if I'd call him "alert". He can't eat anything, even jello, without it coming back up... He may or may not be coming home tomorrow. Depends on how much he improves overnight.

th3
th3
Joined: 24 Aug 06
Posts: 208
Credit: 2208434
RAC: 0

Dont have many comparable

Message 83701 in response to message 83700

Dont have many comparable results for Win SSE vs Linux SSE2 but here goes anyway:

WinXP guest OS on VMWare WS,
42315360 37,689.01 sec - 214.93 cr
42315382 36,221.45 sec - 214.93 cr (not validated yet)

The host OS, Debian Etch 2.6.18-6-bigmem,
42290744 33,105.41 sec - 192.84 cr

The VM was using single CPU, therefor Boinc on the host OS were set to use only 1 CPU also to make it more compareable (Intel E2140 has a very small shared L2 cache, performance impact is considerable when using both cores, and using both cores in a VM slows down the host OS for some applications)

Seconds pr credit:
Win XP SSE - 171.34 (avg)
Linux SSE2 - 171.67

To many unknown factors but i want to say that the Win SSE app is very fast on this CPU. My thought is that the SSE2 code is so sparse that it doesnt make much difference, very much like the S5R3 GNU/Linux app 4.49 SSE vs SSE2. Could be very different on other CPUs, and i have only a total of 3 comparable results from very different frequency range and sequence numbers, the validity of my little test here is indeed questionable =)

Brian Silvers
Brian Silvers
Joined: 26 Aug 05
Posts: 772
Credit: 282700
RAC: 0

RE: To many unknown

Message 83702 in response to message 83701

Quote:

To many unknown factors but i want to say that the Win SSE app is very fast on this CPU. My thought is that the SSE2 code is so sparse that it doesnt make much difference, very much like the S5R3 GNU/Linux app 4.49 SSE vs SSE2. Could be very different on other CPUs, and i have only a total of 3 comparable results from very different frequency range and sequence numbers, the validity of my little test here is indeed questionable =)

One such factor could be the CPU. The AMD penalty could have crept back in. The reference unit, run on an AMD, is still showing upwards of 20% difference between Windows and Linux.

Would be informative if someone ran the reference unit on Intel and see if the results are comparable to running it on AMD...

th3
th3
Joined: 24 Aug 06
Posts: 208
Credit: 2208434
RAC: 0

Looks like the Linux result

Message 83703 in response to message 83702

Looks like the Linux result in my previous post would have gotten a higher credit if it had been sent/generated a day or 2 later, making my testing even more invalid.

The next WU in that sequence gave 221.23 credits instead of 192.84, which would have clocked the GNU/Linux SSE2 app in at just below 150 sec/credit. Thats more than 14% advantage over the Windows SSE app.

More testing is needed when we have a clearer picture of the frequency and sequence number impact on runtime, but so far it looks like the Windows app is really considerably slower. On the bright side its slower for Intels also, so the AMD penalty theory seems less likelly now.

Quote:
Would be informative if someone ran the reference unit on Intel and see if the results are comparable to running it on AMD...

I can try out the reference WU, but need to know how to run it in Windows?

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6591
Credit: 322197174
RAC: 262398

RE: Some participants have

Message 83704 in response to message 83665

Quote:
Some participants have from time to time expressed quite an enthusiasm for some species of 'certificate of acknowledgment'. This is not as silly as it might first appear to some. It hasn't triggered much actual activity, though.


My apologies, I've just noticed such certificates as already existing - see one's account page near the bottom of the 'Computing and Credit' area on the left. How long have they been available? This is precisely what I had in mind. So well done whom-so-ever implemented it! Anyone own up to that? :-)

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

Alinator
Alinator
Joined: 8 May 05
Posts: 927
Credit: 9352143
RAC: 0

RE: RE: Some participants

Message 83705 in response to message 83704

Quote:
Quote:
Some participants have from time to time expressed quite an enthusiasm for some species of 'certificate of acknowledgment'. This is not as silly as it might first appear to some. It hasn't triggered much actual activity, though.

My apologies, I've just noticed such certificates as already existing - see one's account page near the bottom of the 'Computing and Credit' area on the left. How long have they been available? This is precisely what I had in mind. So well done whom-so-ever implemented it! Anyone own up to that? :-)

Cheers, Mike.

Hmmm...

They've had these available at SAH for a long time. I joined BOINC in early 2005 and you could generate certs for your SAH account back then.

However, it looks like that it wasn't made generally available until recently. The link doesn't show on projects which haven't upgraded far enough to have the Community area on the account summary page for example.

I don't recall seeing it on my SAH account page until this last big round of backend upgrades. OTOH, I can't say I was looking for it either, so it might have shown up somewhat earlier. ;-)

Alinator

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6591
Credit: 322197174
RAC: 262398

Re-aligning the thread

Re-aligning the thread vortices and posting pinions, we arrive here continuing on the polite ( congratulations all! ) discussion of general credit issues/whinges/praises/observations .... :-)

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

Alinator
Alinator
Joined: 8 May 05
Posts: 927
Credit: 9352143
RAC: 0

RE: Re-aligning the thread

Message 83707 in response to message 83706

Quote:

Re-aligning the thread vortices and posting pinions, we arrive here continuing on the polite ( congratulations all! ) discussion of general credit issues/whinges/praises/observations .... :-)

Cheers, Mike.

LOL...

Acknowledged...

I am Bender; Please insert girder! :-)

John Clark
John Clark
Joined: 4 May 07
Posts: 1087
Credit: 3143193
RAC: 0

Coming in a bit late on the

Coming in a bit late on the conversation, and not expressing a strong view. But, cross-project parity is a bluff.

I was recently crunching for SETI Main, using the optimised clients for my rigs. I am now crunching for Malaria with those machines, and the rest of my window box crunch here.

Making use of a fairly stable RAC as a, sort of cross-project parity, using the same hardware (but optimised in SETI and stock client in Malaria). My SETI crunching is almost 3 x more than my Malaria one.

Just shows cross-project equality of output is really kite flying.

When I reach my target total in Malaria I will shift them to crunch here.

Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.