I've been running MilkyWay@home for a little while now, and I've been noticing that the multi-threaded tasks tend to run incredibly fast - as in TeraFLOPs/Second, as listed in the "properties" window after the task has been completed. Most seem to hover around 1 TeraFLOPs/Sec, but this one recently hit 11 TeraFLOPs/Sec. I run a MacBook Pro with an Intel i7 2.7Ghz CPU that has 2 physical cores and 2 more "virtual" cores (total of 4 recognized cores). In Geekbench, the maximum FLOPS-style measurement for my machine was 15.4GigaFLOPS, and that was (I think) for the Dot-Product test. I very seriously doubt that my machine is actually capable of 11 TeraFLOPS, so is there a known issue in eiher BOINC or the multi-threaded tasks that would produce such a result? I searched through the archives on this board, and I couldn't find any reports of similar occurrences, though I could have missed something.

I am running BOINC 6.12.35.

Here are BOINC's benchmark results:

Number of CPUs: 4
3497 floating point MIPS (Whetstone) per CPU
6055 integer MIPS (Dhrystone) per CPU

Totals are:
13,988 floating point MIPS Whetstone
24,220 MIPS integer MIPS Dhrystone


Here's a screenshot of the "Properties" readout for the 11TeraFLOPS example:
http://imgur.com/aD72C

You're right: although you've got fine CPU, 11 TFlops can not be extracted from this hardware even under optimal circumstances. My best bet would be that the FLOPs estimation of the WU is wrong.

MrS

---

Scanning for our furry friends since Jan 2002