the draw rate for a cc match btwn u and the cluster should be very high, at least 85% i'd imagine (probably closer to 90%+). if so, it means even in the extreme case where one side goes undefeated, his elo delta will be held to under +50elo.
i did not say that.at longer time controls i have better chance of a win.
Still trying things with Pauls comp, I had to increase the vcore a little more and although Paul feels happy enough with one of the cores hitting 70C when the machine is fully loaded I do not think I want to push it any further on air cooling, don`t get me wrong I feel there is still more under the hood but in a case 1 CPU will always run hotter especially with air cooling.
I may yet back it off to 4.2Ghz depending on how more tests go and after Paul tests the machine at my place on sunday.
I think I will save some profiles in the BIOS just in case :D
The pic below is only with 8 cores being benched as this is the max the old fritzmark can handle
Rybka 4 KNS from the start position below
anyone know the speed of lukas' cluster?
I was egged on to break a certain KNS with Rybka which I did below although the KNS was jumping around a bit, I had to increase the speed a little. I must admit Pauls comp was not stable at this point on all benchmarks and I had a few BSOD`s so the speed has been dropped back to 4.3Ghz ... To get it stable and go above 4.4Ghz you are going to need 1.45 Volts through the CPU which on air is prob a bit iffy especially as the machine will be in a room which already has a skulltrail and this new comp heats my kitchen quite well already lol :D ... On Water it is easy on my 6 core I am 4.5Ghz and now I have fitted the watercooling setup I am a full 15C lower temps wise fully loaded
On Pauls comp with windows 7 for some reason others may explain when I enable Large Pages the KNS actually drops
a request: to determine r4's scaling efficiency, can u disable all but 1 of the cores, and then tell us the kN/s for the single-core (at that clock speed of 4.37ghz)?
> to determine r4's scaling efficiency
Well, first Bigben should tell us if the million nps was obtained at each try (multi-core numbers are known to be non reproducible )
am i mistaken?
> but the kN/s remains fairly constant
Well, try it (if you have a Dual, a Quad): 5%, 10% variations are not unusual...
>Well, first Bigben should tell us if the million nps was obtained at each try (multi-core numbers are known to be non reproducible )
It is reproducible but you are correct in that at times the KNS jumps wildly about with Rybka 4 ..... At times when I have increased speed including memory timeings and increased the scores on all other benchmarks along comes Rybka 4 at the start position and the KNS drops .... So the figures should only be used as a guide but fast is fast in the 12 cores case.
Other scores posted is seeing Zappa Mexixo 2 running at 10500000 KNS
If I get some spare time I should show the CPU scores in the Everest Benchmark where of course the 2x 5680 trounce my 980x .... I only have today (saturday) as Paul is picking the comp up tomorrow but I have the gym today and some of my friends are fighting tonight at http://www.circustavernnightclub.com/events/sporting-events/517/iba-boxing-september-4th-2010- .... I do not have much testing time left
so how close was i?
I posted a reply with quite a few benchmarks including what JP wanted with Arena2+cyclone extreme benchmark but I had included a bench of one of the clones for comparison between 2 machines and the whole post was deleted by the moderator rather than taking out a couple of the benchmark pictures (I simply forgot I was on the main forum)
It is annoying as I put a lot of work and time into the post before I left for the boxing venue ... If you did not get to see the post before it was moderated\deleted my apologies
Paul has just left with his computer so my work is done, it is a shame the post was totally deleted.
The machine is running at 23x187 = 4.3Ghz .... I had been testing 22x196 but it seems around 190 is the max for the motherboard and although it ran and completed some benchmarks it was not totally stable.
For anyone wanting to try the same kind of setup with watercooling or phase etc my only hint of caution is that you cannot exceed the x25 multiplier (unless you use Turbo mode and that brings other probs which may be corrected with a new BIOS hmmnn or maybe something alters if you choose the extreme cooling enabled option?) and if the max the motherboard can run at is 190 that equals 4.75Ghz, you are going to need the board to get to 200 to make 5Ghz ..... Wish now I at least enabled the extreme cooling option to see what happened but time was of the essence
I read also that Paul has taken his new toy to home..so maybe later when Paul has time he can retry the test..and if he want he can
post the results on Xtremesystems!
Thanks for your time..and see you later!
Did you get the SR-2 motherboard in the UK or from abroad?.
At the moment there does not seem to be a local supplier here in the Netherlands :-(
I assume this set up is better then I7980x?
have also posted on the edge thread
>I assume this set up is better then I7980x
It certainly is better but overclocking is better with a single chip as the motherboards are in many ways more advanced especially with memory support.
I think and I have not got there yet but a 980X at 5Ghz will push Paul fairly close but a lot depends on scaleing
Since I have never used any of AMD Phenom processors I really cant comment but there must be members of the forum who have some knowledge
have a great holiday.
my Skully gave 35.03
That must be built into the chessbase deep rybka 4 software ... I was running the one from http://www.jens.tauchclub-krems.at/diverses/Schach/fritz9_benchmarks.html .... Which is as I thought an out of date one.
Thanks for that Paul, I think it puts you top of that benchmark list in the link above lol
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill