Denis@home team challenge: 31st Oct - 7th Nov
- Thread starter hanluc
- Start date
I've tucked in behind Jon in 2nd place on the Top Hosts board 
Quite happy with that.
I'm trying to sweet talk the "wife" but she not having any of it with regards to adding my 5930k
"If its on its on" was the reply....Haha
Great to see more cores being added by others, thanks for joining in folks. I'm trying to recruit my bro to add some spare cpu time
Quite happy with that.I'm trying to sweet talk the "wife" but she not having any of it with regards to adding my 5930k
"If its on its on" was the reply....HahaGreat to see more cores being added by others, thanks for joining in folks. I'm trying to recruit my bro to add some spare cpu time
Last edited:
Associate
Other news, we need a sailor with his whistle as unlikely as it sounds a captain is coming aboard
I've only just seen this, superb
Ok think I've been a bit silly........
I installed Boinc client and added the Denis project, I was asked for existing or new account. So I did a new one thinking it was for Denis but now I think it might have meant Boinc. So is there anyway to merge my new account with my old one I created when I started SETI?
I installed Boinc client and added the Denis project, I was asked for existing or new account. So I did a new one thinking it was for Denis but now I think it might have meant Boinc. So is there anyway to merge my new account with my old one I created when I started SETI?
Ok think I've been a bit silly........
I installed Boinc client and added the Denis project, I was asked for existing or new account. So I did a new one thinking it was for Denis but now I think it might have meant Boinc. So is there anyway to merge my new account with my old one I created when I started SETI?
When it asked you for new account was this after you clicked add project? If it was after that's for the project login, if you were new to DENIS no need to worry. As long as its Team 10 you joined ofc !
I honestly can't remember but I thought it was after I added the project. However all my other efforts ( SETI, MilkyWay and Einstein ) appear as a combined score and I can't see how to add my new Denis effort to that.
Can I change the Cross Project ID associated with my Denis account somehow to tie in with the others?
Can I change the Cross Project ID associated with my Denis account somehow to tie in with the others?
Associate
Optimised apps are (hopefully) running on both machines now, also discovered HT was turned off on my 2xL5639 for some reason so turned that back on. They're both finishing off some yoyo WUs as well so not quite at full crunching power, but once they're done they'll be 24/7 to the end of the challenge 
Associate
Oooh, a challenge! I will dust off Boinc and get to work!
Associate
Optimised apps are (hopefully) running on both machines now, also discovered HT was turned off on my 2xL5639 for some reason so turned that back on. They're both finishing off some yoyo WUs as well so not quite at full crunching power, but once they're done they'll be 24/7 to the end of the challenge
8,612 sec per wu seems far too long for the Genuine Intel(R) CPU @ 2.00GHz [Family 6 Model 63 Stepping 1](28 processors)
the xeon not much better at 7000 secs , no idea whats wrong tho lol
Associate
Wish I could catch some of you guys up. I'm still near the bottom of the team rankings 
Associate
PDW takes number 1 spot on the team, pcs "hidden" prob hacked into the bbc weather forecast machine! lol
Hot and spicy, correct
Optimised apps are (hopefully) running on both machines now, also discovered HT was turned off on my 2xL5639 for some reason so turned that back on. They're both finishing off some yoyo WUs as well so not quite at full crunching power, but once they're done they'll be 24/7 to the end of the challenge
WOW your windows machine is flying!
I thought your 2 ghz chip turbo's up 2.3ghz when all cores are running, is that right? my haswell xeon turbo's up to 2.5ghz, I have it now at 2.576ghz with bclk slightly overclocked, but your runtimes are over a minute quicker than mine
What board are you running it on??8,612 sec per wu seems far too long for the Genuine Intel(R) CPU @ 2.00GHz [Family 6 Model 63 Stepping 1](28 processors)
the xeon not much better at 7000 secs , no idea whats wrong tho lol
He initally ran the stock app.
Associate
8,612 sec per wu seems far too long for the Genuine Intel(R) CPU @ 2.00GHz [Family 6 Model 63 Stepping 1](28 processors)
the xeon not much better at 7000 secs , no idea whats wrong tho lol
I think they're the tasks with the stock app, with the optimized app they seem to be around 1/10 of those times
WOW your windows machine is flying!
I thought your 2 ghz chip turbo's up 2.3ghz when all cores are running, is that right? my haswell xeon turbo's up to 2.5ghz, I have it now at 2.576ghz with bclk slightly overclocked, but your runtimes are over a minute quicker than mine
He initally ran the stock app.
2.3Ghz max turbo, yes. It's on a Gigabyte X99-UD4, no overclocks or other trickery, so not sure why it's so much faster than yours?
I dunno either, I'm on win 7 but I wouldn't have thought it was that, I need to have a tinker.
Associate
I was going to say, that's the only possibly major difference I can see, but I wouldn't have thought it'd be that either. One odd thing is my machine is reporting "Measured floating point speed" and "Measured integer speed" both as 1000 million ops/sec which seems unlikely to be true because it's too exact, so I'm not sure if that's some sort of error, possibly a clue of some sort
Only other thing I can think of is that possibly with all your cores and threads loaded, from what I've read elsewhere when using optimized AVX2 code it basically causes contention - because HT is basically only using "spare" resources for the extra logical cores, when using very optimized code there's less advantage with HT as there's not as much "slack" - if that makes any sense? So I'm thinking because not all of my cores/threads are running Denis yet, maybe it's running faster, and perhaps once I've finished these yoyo units and am only running Denis the runtimes might go up a little. Of course as long as the runtime doesn't double then HT still provides an advantage. I do only give BOINC 22 cores out of 28 as it keeps the machine responsive for other tasks without me having to pause it, too. Would be great if BOINC had an option to only use x cores when the PC is in use - I know there's an option to use x % CPU when the PC is in use but it's not quite the same.
I was going to say, that's the only possibly major difference I can see, but I wouldn't have thought it'd be that either. One odd thing is my machine is reporting "Measured floating point speed" and "Measured integer speed" both as 1000 million ops/sec which seems unlikely to be true because it's too exact, so I'm not sure if that's some sort of error, possibly a clue of some sort
Only other thing I can think of is that possibly with all your cores and threads loaded, from what I've read elsewhere when using optimized AVX2 code it basically causes contention - because HT is basically only using "spare" resources for the extra logical cores, when using very optimized code there's less advantage with HT as there's not as much "slack" - if that makes any sense? So I'm thinking because not all of my cores/threads are running Denis yet, maybe it's running faster, and perhaps once I've finished these yoyo units and am only running Denis the runtimes might go up a little. Of course as long as the runtime doesn't double then HT still provides an advantage. I do only give BOINC 22 cores out of 28 as it keeps the machine responsive for other tasks without me having to pause it, too. Would be great if BOINC had an option to only use x cores when the PC is in use - I know there's an option to use x % CPU when the PC is in use but it's not quite the same.
I'm still learning this machine, sometimes, for whatever reason, it gets stuck on the x12 (idle) multiplier, if I restart boinc it goes to x25 it stays there.
But I think you've hit the nail on head with you only using 22 cores out of 28 threads. I know that my (and yours) single thread turbo is more than all threads turbo. I have just now paused all my work units that hadn't started and let the others run out, as the work units decreased, they completed faster and faster.
That's something I noticed when I had to shut down my Xeon rig to swap over the SSD. When running 24 units all cores were at 100% load, but with 2 units processing at least 6 of the cores were under load (between 10 and 45%) indicating Windows processes data over multiple threads. It makes sense if you have more free resources (up to a point) then the system can distribute the processing over the most suitable cores resulting in decreased run times, albeit with a lower total output.