1)
Message boards :
Number crunching :
Race to the New Moon Challenge
(Message 76670)
Posted 3409 days ago by  ThrasherX-17
Ken_g6, but that happens only with LLRs. Other projects run well.
Q6600@3,08GHz
25.05.2014 10:58:42 | | Benchmark results:
25.05.2014 10:58:42 | | Number of CPUs: 4
25.05.2014 10:58:42 | | 2893 floating point MIPS (Whetstone) per CPU
25.05.2014 10:58:42 | | 9011 integer MIPS (Dhrystone) per CPU
|
2)
Message boards :
Number crunching :
Race to the New Moon Challenge
(Message 76667)
Posted 3410 days ago by ThrasherX-17
Look how single core mobile Celeron @1,7GHz easily beats overclocked quad-core Q6600 @3,1GHz in a single task performance.
One more interesting thing.
x6 1055T @3,2 GHz 4Gb (2x2Gb) DDR3 1600MHz -1WU/66 hours
x6 1055T @3,6 GHz 4 Gb (1x4Gb) DDR3 1700MHz - 83% of 1 WU/95 hours
x6 1045T @3,1 GHz 2 Gb (1x2Gb) DDR3 1550MHz - 51%/79 hours
That's insane difference between 1 and 3. More than 2 times.
|
3)
Message boards :
Number crunching :
Race to the New Moon Challenge
(Message 76587)
Posted 3412 days ago by ThrasherX-17
Wingless Wonder, I have 14 PCs with Win7 SP1 x64 and one with Win 8 x64.
Michael, I will test few hardware combinations right after the chelly.
If it's not about the size of the memory how can dual channel make 20-30% faster speeds? Almost all of my memory are DDR3 1600 and only one kit 1866.
It's empty words from me, I know that you need facts. I will write my results when my AMD x6 will complete the first lap.
Also I have a question. How does the integrated boinc manager's performance test make influence on PG computing? Do I need to check it every time when I change my RAM sticks or does PG refresh this info with every start of Boinc manager?
I'm making a small test, I reinstalled my ram sticks (2x2Gb) in non dual channels.
|
4)
Message boards :
Number crunching :
Race to the New Moon Challenge
(Message 76574)
Posted 3412 days ago by ThrasherX-17
Can anyone explain me how does the crunching speed depends on the quantity of memory (size, frequency, channels and latency).
X6 1055T @ 3,5GHz + 4Gb RAM completed 49% of tasks
X6 1055T @ 3,9GHz + 4Gb RAM completed 41% of tasks
X6 1045T @ 3,3GHz + 2Gb RAM completed 22% of tasks
(all of them are boinc dedicated).
Why??
I have an i7 2600k CPU. 8Gb(2x4) of RAM versus 4 Gb of the same RAM almost doubles crunching speed. Is it because of dual channel or is it because of double memory size?
How much RAM per core do I need to get maximum performance?
I think admins should post this information in the starting topic.
Need some good advice for better PG output.
|
5)
Message boards :
Number crunching :
The Year of The Snake Challenge
(Message 61297)
Posted 3913 days ago by ThrasherX-17
Michael, I've never criticized any recommendations from PG moderators since my very first challenge in 2010. And I used to respect all the positive and negative experience in PG crunching.
It's good to see how much you care but sometimes recommendations are very radical (in my opinion of course).
I have 4 types of NV cards running the GFN challenge (GTX470, GTX460, GTX560Ti, GTX570).
All of their cores are overclocked +120-200MHz.
My PSUs are not great. CoolerMaster 850W/FSP Epsilon 800W/Antec 550W.
Also I have 1 problematic GTX470 that sometimes gives errors in almost all BOINC GPU applications (PG, DiRT, Einstein etc) no matter with or without overclocking.
The other GPUs are feeling fine with GFN and PPS Sieve WUs.
I can't say the same thing about CPU overclocking.
Some CPUs are giving errors but I checked them and figured out that those CPUs are giving errors even at default frequencies.
Anyway I will be more confident when all the WUs will be validated. Time will tell.
|
6)
Message boards :
Number crunching :
The Year of The Snake Challenge
(Message 61288)
Posted 3913 days ago by ThrasherX-17
If you were a regular reader of these forums you would have your answer to your 2 questions.
There are pages of work unit from hundreds of users with failed work units.
The posted warning is a just one. I'm sure Mike will chime in with more detail.
Last year we've run the GFN challenge. I remember those very hot summer days. I kept all of my GPUs at the highly OC level. The GPU temps were insane. I could finish the challenge with almost all WUs validated. I was surprised and thought that all warnings were because of the GFN newness.
A lot of time has passed but we all still see the same warnings.
I wouldn't start this dialogues but some of my team members have read the official thread and started to downclock their GPUs.
Also I'm very sad about the fact that this discussion is too late and I feel myself guilty because I haven't seen that warnings before the start of the challenge. Probably it's better to stay with the tested settings during the event.
Happy challenge everyone!
|
7)
Message boards :
Number crunching :
The Year of The Snake Challenge
(Message 61281)
Posted 3914 days ago by ThrasherX-17
If you have never run GeneferCUDA before, please read these tips. GeneferCUDA stresses a GPU more than any other program I've seen and is incredibly intolerant of overclocking.
In my opinion, GFN is a regular crunching stress for GPUs. Almost every BOINC GPU app that can load GPUs at 99% gives almost the same GPU stress.
GFN on GTX470 - 76'C
DiRT on the same GTX470 - 79-80'C
GFN on GTX560Ti - 63'C
DiRT on the same GTX560Ti - 66'C
Room temperature is the same in all cases.
What is a source of information about problems with overclocked GPUs @GFN app?
Is it a single story or numerous reports?
I totally disagree with the statement like that.
|
8)
Message boards :
Number crunching :
Winter Solstice Challenge
(Message 60686)
Posted 3929 days ago by ThrasherX-17
Oh yeah! My pulse is still running out of time. What a great challenge! Maybe the greatest one in my short boinc life.
The biggest thanks to all the Keepers of the Fire Alive!!! And of course I want to mention and give my big thanks to our friends from other teams!
The whole year we've tried hard and did our best. We could make it to TOP-10 in the 2012 series of PG challenges. 9th place is a fantastic result for such a small team of fiery crunchers. And of course the 6th place@Winter Solstice with almost 31M/3 days. In 2011 it was 10M.
Thanks to all the Primegrid administration! In 2012 the challenges became more stable and efficient. Great job!
Thanks to all the rivals for giving us some crazy hours of competition.
Thanks to Ukraine team for the last challenge. It was a great battle.
P.S. Keep on crunching! Happy holidays!!! :-)
|
9)
Message boards :
Number crunching :
Winter Solstice Challenge
(Message 60524)
Posted 3933 days ago by ThrasherX-17
It's a very special event for me because 2 years ago Winter Solstice chelly was my and team's first experience in challenging. I was 405th and the team was 135th. A lot of things have changed since then but Primegrid is still one of our favorite projects.
Keep The Fire Alive is 2 years now! It's cool to celebrate this date in action!
Everybody have a nice crunching and let's make some noise!
|
10)
Message boards :
Number crunching :
321 Blast off Challenge
(Message 59125)
Posted 3975 days ago by ThrasherX-17
I have a box with AMD X6 1090T and more than half of workunits are getting "Completed, validation inconclusive".
Some say that this workunits will not get final scores. But I don't see any computation errors while crunching these tasks.
Here's a proof:
http://www.primegrid.com/results.php?hostid=275983&offset=0&show_names=0&state=2&appid=7
And this PC also is getting validation inconclusive.
http://www.primegrid.com/results.php?hostid=299995&offset=0&show_names=0&state=2&appid=
So what to do?
|
