Question regarding performance tests.
Just a quick question - I've always used passmark cpu benchmarks to compare cpu's for rendering etc.
We have a dual-cpu 2.33ghz xeon 5345 system, the benchmark listings have recently changed but to clarify:
Passmark (single CPU): 2628
(below the 5355 (2.66ghz) as would be expected. (2950)
E5345 x 2 (current rig) 7004
Passmark (dual CPU)
(now above dual E5355, (at 6933) which seems a bit strange, (i've used both in dual operation and the 5355 is definitely quicker)
So it goes from 2628 to 7004, more than double its original score, is there a different rating system in place for dual setups?
The reason I'm curious is because I've just bought an i7 rig, according to the single cpu charts it should be around the same power as my current 2x5345 system, but based on the dual cpu charts it looks more like being significantly less powerful.
I would have expected a dual CPU system to add less than 100% extra processing power, or am i completely misunderstanding things here?
(I'm guessing that I am but I would welcome any thoughts!)
There are lots of factors that can increase or decrease the CPU score. Usually in the charts these differences average out however in this case due to the small number of samples the results had been thrown off a bit.
The CPU charts are made up of results from both version 6 and 7 of PT. Although we tried to keep the CPU mark roughly equivalent between the versions PT6 tends to score a bit higher on more modern CPUs. What had happened in this case was the dual cpu systems were mostly represented by version 6 results whereas the single CPU systems had more version 7 results.
We've removed some of the higher version 6 results from our database to make the scores more representative (dual CPU is 5700 instead of 7000 now). As PT7 has only been out a few months oddities like this will appear from time to time. Going forward however version 7 results are going to far outweigh the version 6 results making this no longer a problem.
Aha I see - thanks for explaining that.
I'm guessing benchmark scores from previous/older CPU's are based on older versions of the software - are those scores indicative of how those processors would perform on v7 of the software?
Or in other words, does the passmark score, in general equate to a linear and continuous approximation of CPU power from the oldest to newest CPU's?
The CPU benchmark tests in V6 and V7 are not the same. But we scaled the results to be the same on typical average CPU. But different CPUs show different amounts of variation between V6 and V7.
The results for the older CPUs contain relatively more V6 results, than the results for the newer CPUs, which are made up of almost entirely V7 results.
Tags for this Thread