Sandbo

The database test ... How much did Anandtech receive for it?

Sandbo

引用:

原帖由 qcmadness 於 2017-7-13 01:27 發表

I don't think so.
The database / transaction server is still in Xeon territories due to latency issues.

If you read the top comments, you will see (as in the article they also mentioned),
they chose "A small database that can be mostly cached in the L3-cache", which is largely not true for realistic scenarios.

Sandbo

引用:

原帖由 qcmadness 於 2017-7-13 01:33 發表

But even the database is stored in memory, the latency issues still remain.
You could think of some bias but the latency difference is still too large to ignore.

Not sure, but if as comments mentioned (I don't know much about database) a real-life model requires data stored mostly in RAM,
EPYC has a much better bandwidth comparing to Intel's.

引用:

So the dual socket AMD system should theoretically get 307 GB per second (2.4 GT/s* 8 bytes per channel x 8 channels x 2 sockets). The Intel system has access to 256 GB per second (2.66 GT/s* 8 bytes per channel x 6 channels x 2 sockets).

So it is interesting (and probably more useful) to test a large database.