Ignore the Gflops.
Your test is also custom so I cant do a comparison.
Not sure if serious. Come on stulid have you not read any of my posts on this topic?
The GFLOPS is a measure of the intensity of the stress test, which increases with the problem size (memory usage). One must use enough memory to achieve sufficiently high GFLOPS. My rule of thumb is 80% of peak.
In this chaps case, he wants to be seeing at least 0.8*32*4.4 = 113 GFLOPS.
The number you're getting is around half that, which means one of two things. Either it's not using AVX, however this is unlikely as you're using win7 and IBT v2.52. Therefore, this discrepancy is because you only used "standard" level memory of 1 GB.
The standard profile is way out of date these days with 8 GB systems. Essentially by only giving it 1 GB memory you're choking it and it's not stressing fully.
I suggest running 5 or 10 loops at 4 GB memory usage. You'll probably see higher temps, as it's FAR more stressful.
There are loads of claims of IBT not showing instability, but really the fault is with the user, and thus with the programmer for not releasing documentation/an smarter interface.
(using 9000MB)
