Memory Performance Tests

Okay and then why was the 2990WX so disappointing in a lot of these tests? Well as I alluded to earlier it is 100% down to retentiveness bandwidth, much more than and so than core to core latency or memory latency.

Here nosotros can see the sustained memory bandwidth for each processor. You'll observe that the 2990WX is a trivial downward on the 2950X and that's due to the added latency the dies without memory controllers incur. It'southward a seven% drop in bandwidth merely that lonely doesn't explicate the performance problems we've seen.

Simply to confirm those results I did also test with AIDA64 and here is the memory copy performance, over again the 2990WX was down 7% on the but that doesn't explain the miserable performance in the encoding, compression and encryption benchmarks.

For that nosotros need to await at memory bandwidth per core, not for the entire processor but rather the individual cores. Arranging these results by a unmarried thread or core we come across that with just i cadre active the Ryzen CPUs enjoy tremendous bandwidth.

At present please note the performance of each core inside the CPU is measured individually and issue you see here is the boilerplate bandwidth across all the private cores. And so the 2700X and 2950X, both 2nd gen parts deliver the same 29 GB/s. Then the 1st gen Ryzen parts deliver between 24 - 25 GB/due south and then we have the 2990WX at 20 GB/s. This is why we saw a slight drib in total memory bandwidth in the previous test, the margin is amplified here showing the 2990WX to be virtually 30% slower equally were non limited by the DDR4 memory in this case.

The reason the single cadre bandwidth is downwardly is due to the fact that 16 of the 32 cores aren't connected direct to the retention and therefore suffer increased latency.

Finally we see that almost all the Skylake-Ten parts are limited to just 14 GB/s, though this is less of an issue as xiv GB/southward per core is substantially overkill, and here is why.

If we rearrange this graph by the 'all-threads active' result, the arrangement changes quite a flake. Now for these results all CPU cores are actively accessing organisation memory and nosotros're showing the average throughput of an individual core. Essentially with the CPU running at full steam in a memory intensive workload, this is the typical amount of bandwidth each core has at its disposal.

This here is the trouble. The 2950X enjoys a bandwidth of 4.4 GB/due south per core when maxed out and this is why the fourteen GB/s we saw with but a single core active on the Intel CPU'southward isn't an issue, since the maximum sustained bandwidth of the Skylake-X CPUs is effectually 64 GB/s, with 5 cores active in an extremely retentiveness intensive workload you lot're going to apply up all that bandwidth and once yous start adding more cores yous start to see a drop in efficiency equally they aren't feed information fast enough.

Naturally the more cores y'all have the worst of yous're going to be in this exam without increasing the overall memory bandwidth. With octa-channel memory the 2990WX would indeed be able to match the four.4 GB/southward per core of the 2950X. But with simply quad-aqueduct memory that effigy is halved, well a picayune over halved due to the increased latency so it's a bit of a double whammy. In the end merely shy of 2 GB/south of bandwidth per core just isn't plenty and we see the problem this causes when running retentiveness sensitive applications like VeraCrypt for example. Okay so before we move on to overclocking, power consumption and a few other tests, let's rapidly go over gaming performance.