Tweet
Our IT department has given us two identical servers, every component is reported as identical. However the memory tests scores in Perfroamnce Test are consistently around 10% lower on one machine. Our IT department says that Performance Test is not suitable for Enterprise servers but cannot provide any alternatives. Is there a reason why Performance Test would not be suitable for a large enterprise server?
-
-
It should work fine.
Maybe your hardware isn't really identical. What are the individual test results from the memory test. Are all other results identical (CPU & disk)? Is the RAM installed in the same slots on both machines? Are both BIOS versions the same? Do they have the same cooling (one isn't venting hot air into the 2nd one?). Do they have exactly the same software installed on both machines?
You might also be seeing a NUMA effect. We especially coded in features to cover cases like this. In particular we added Processor group support and NUMA support for servers that have multiple CPUs and multiple memory busses. You can find this in the Advanced memory test (under the Advanced menu).
NUMA = Non-Uniform Memory Access
There is also some additional discussion of NUMA support here,
https://forums.passmark.com/performa...d-threadripper
Comment