I use a number of single thread legacy code number crunching programs and am trying to confirm whether the C2D processors w/4Mb L2 cache or x2/FX processors w/1Mb L2 cache have any significant advantage over other A64 x2s. So far, it looks like a push for the C2Ds (up to 6300), but I've not been able to test on a C2D with the 4Mb L2 cache (6600, 6700, 6800). Haven't found an FX or x2 5200+ to test either. Results so far at toward the end of viewtopic.php?t=35599&highlight=.
The programs are a lot of floating point math on arrays held in memory, and the AMDs have historically been much better (relative to CPU clock speed) than the Intels. I think the on-chip memory controller of the AMDs still gives them an advantage over Intel, but haven't been able to test if the larger cache size on the higher end chips might change things.
If there's anyone out there with one of these machines who is willing to take about 15 minutes to run a couple of benchmarks I've set up (batch files with timers that run in Command Prompt windows), I'd greatly appreciate it. Please let me know (here or PM). I can email a .zip file. The program resides in the folder where installed, doesn't do anything to the registry, etc.
TIA to any volunteers. I'll update any further findings here and on the other thread.
Bench help? FX6x, x2 5200+, or C2D 6600 or higher?
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
-
- Posts: 88
- Joined: Mon Jun 20, 2005 1:11 pm
- Location: California
-
- Posts: 88
- Joined: Mon Jun 20, 2005 1:11 pm
- Location: California
Not directly related to the thread subject, but may be of interest:
I downloaded the trial version of the Intel Compiler, and compiled x87 single threaded code, optimized for C2D.
Then I compared with the VC6 compilation: the benefits were 0 to 5%, so I'll have to go full SSE2 and dual-threaded to see improvements.
I downloaded the trial version of the Intel Compiler, and compiled x87 single threaded code, optimized for C2D.
Then I compared with the VC6 compilation: the benefits were 0 to 5%, so I'll have to go full SSE2 and dual-threaded to see improvements.
-
- Posts: 88
- Joined: Mon Jun 20, 2005 1:11 pm
- Location: California
The answer is in. Thanks to pyogenes for running a number of tests on an E6600 at both stock, over- and underclocked speeds, as well as varying FSB and memory speeds. See the link above for details.
Short version: Run times are exactly inversely related to CPU frequency, regardless of cache, FSB or memory speed, and the AMDs are about 22% faster in througput per CPU cycle than the C2Ds. This makes an OC'd E6600 or non-OC'd x2 5200+ about the best options on price/performance. Still looking for a 5200+ and/or FX60 or FX62 volunteer to see whether the L2 cache size helps there.
Again, thanks to pyogenes!
Short version: Run times are exactly inversely related to CPU frequency, regardless of cache, FSB or memory speed, and the AMDs are about 22% faster in througput per CPU cycle than the C2Ds. This makes an OC'd E6600 or non-OC'd x2 5200+ about the best options on price/performance. Still looking for a 5200+ and/or FX60 or FX62 volunteer to see whether the L2 cache size helps there.
Again, thanks to pyogenes!