Great Turnout @NFS@ :)

[Logic Labs]

I just wanted to say the last 2 days everyone participating has done a really great job.Im really impressed by dirk and terry there numbers have jumped very quickly.I think we have a very respectable position in the challenge.its too bad we haven't been doing this well the whole time it would be nice to take out the alliance in the second heat. !!! Id like to see what you guys are Running btw!!!!...... ive been out of crunching since GPU was in beta test testing.I am literally crunching every computer i can get my hands on anything dual core or above 1.5ghz and am still struggling to get good numbers on this project ... in fact i was with you guys for the very end of number fields im running twice as many CPU now then i was then and still not getting close to the credits....please correct me if im wrong but ive noticed for example on This project My Dual Socket Xeon Blade 3.2 is only getting about 500 credits average for all four threads but my C-60APU is getting about 700 and its only 1.33 .......now the blade is old but new enough to have pci express it suppose to be pretty bad ass i mean it was a $3000 blade 5 years ago. so i can only assume that some projects have additional instruction sets the c-60 has but the xeons don't ... I have no formal education on the matter so please do correct or enlighten me if this not the case.....however if this is the case can any of you vets show me IF a place exists that is where i can find such information so that i can give the appropriate hardware the appropriate task that its suited for......thanks,Gratz,and woot for us

[Dirk Broer]

If only NFS runs on the Dual Xeon then it ought to blast the C-60 out of the water
But when it doesn't, it is either a setting of the BOINC Manager that prevents the Xeon from getting the best performance, or the SSSE3 instructions that the C-60 supports and the Xeon 5060 doesn't
(that is the model I assumed you have)
Other questions could be:
Are the BOINC Manager settings of the two machines identical? Are both set to running all the time and taking 97% percent CPU when idle?
Have both machines the same amount of memory per thread? The C-60 is much newer and can have e.g 4GB for its two threads, while the Xeons, who run four threads each, are probably sticked on a mobo
with just 2GB in total, so only 250 Mb per thread. In that case you will see the WUs in the BOINC Manager with 'waiting for memory'.

[Logic Labs]

ah ha dirk you do raise a good point i have a lot of work to do today so i will have to look into it feather later on today but i will say this the blade is on 24 7 its only job to crunch and the c-60 is also a media center its not even on as much as the blade... that being said the c60 has 1 gig per core of ram to work with and the xeon has only 512 per core if you look at it that way . both sockets on the xeon have independent ram banks but they are nearly empty because its ecc and i don't think i have anymore. the amount of ram really did no cross my mind i was under the impression that wouldn't really use that much. i have not seen a waiting for memory notice besides one time and only for a second. also i guess i do need to recheck the settings and see if its aloud to use all the memory. as always thanks for your time dirk ill get back to you as soon as i have a more informed perspective .. one more thing i guess what i was asking about the instruction set is where can i find what instruction sets the particular project is using. or are they all bound to the same set because they are running threw the manager

[Dirk Broer]

The instruction sets used are pretty much up to the individual projects. The late Mersenne Project made almost full use of the available set per CPU, some other projects will see SSE as the holy grail and optimize no further.
I use 2 GB per running WU as a rule of thumb when building a new system, and some projects even need more (e.g. the notorious BURP project). Vaughan builds systems with 32 GB, as his mobos can take no more.
My latest FM1 mobos will gladly take 64 GB, but I still have to see a 16GB DDR3 RAM stick that i can afford.

BTW: You can use the tool CPU-Z to see both the actual type of CPU as its capabilities

CPU-Z printscreens via CPU-World, forums.overclockersclub.com and geekstogo

[Logic Labs]

Thanks Dirk. I do believe you are correct about the ram per core ratio. i logged into a few bots and found not only was the opteron 1.66 getting a better average but every bot with a higher ram to core ratio was. I have negotiated an additional four gigs of ecc slimline for the server. Wile its down and apart i do believe i will be replacing the cooling system its freekin LOUD even behind two doors and since i don't need the case to be closed since its not in a rack I think ill put two am2 fans on the Heatsinks. so that leaves just one last thought. i have 7 on there now. Since im rebuilding it anyway i don't mind to do a re install the os drive is a 10k SCSI 75gig i want to turn this thing over to crunch what os i have xp64 corp but ill try ubuntu if you think that it runs more efficiently.

[Dirk Broer]

In a thread about benchmarks we found out that the same CPUs under Linux have a higher Whetstone and Drystone MIPS ratings in the BOINC Manager than those CPUs under Windows.
This should theoretically result in running more efficiently, BOINC-wise.
If your system is a bit older and you do not have an abundant supply of RAM, you can also try Xubuntu, a more lightweight variant, or Lubuntu, an even lighter one.