err, Athlon 64 3500+ chewing through prime95 probably faster than most CPUs out there

[mikeymikec]

I'm a bit confused by a machine I've come cross - Athlon 64 3500+ (AM2) processor, I wanted to stress-test it, and prime95 (latest version, as well as an old app I tried called 'orthos' which is pretty much the same thing) 's CPU testing (small FFTs) went through about 100 tests in a minute. When I stopped the test, it said zero errors or warnings.

I've installed an old nvidia demo on it to give the CPU a good kicking, and the demo is working correctly as well as the CPU hitting 100%.

Has anyone seen anything like this before?

 

[BTRY B 529th FA BN]

depends on the size problems

 

[jvroig]

100 tests a minute even for small fft is 100x faster than it should be.

Does blend and large fft behave the same way?

 

[mikeymikec]

I didn't try that out when I saw the computer, it might be a while until I see it again.

It seemed to be acting normally otherwise, it's just that it had a board swap (for a different board, newer nvidia chipset) recently, so I ran it through its paces (memtest86+, prime95, finished up with the 'clear sailing' nvidia test).

 

[Plimogz]

What you're describing does ring a bell -- unfortunately I can't exactly recall the exact CPU or stress test, so this *might* only be a figment of my imagination, but...

As best as I can remember I have seen this and the clear indication that nothing was actually getting done was that the chip wasn't even heating up as the results just flew by. That and the fact that I had faster machines around which provided a reality check. I expect that I just wrote it off as a bug and changed applications.

Or it may all just be one of those sleep-deprived, caffeine-induced, 04:15 AM OC session linked memory malfunctions that I sometimes suffer from.