Memory subsystem performance
AIDA64’s basic tests of memory bandwidth and latency don’t mean much on their own, but they can give us some insight into why some of our later results come out the way they do.
These graphs all tell an interesting story, but that last one is the most curious. The revised “Zen+” cores in the second-generation Ryzen parts boasted significant improvements in memory latency over the original Ryzen CPUs—as you can see here—so the major regression in this metric is a bit disappointing. It was likely unavoidable in light of the chiplet-based construction of these CPUs, though. We’ll be keeping a close eye on our more real-world benchmark results to see whether Zen 2’s big L3 cache can mask the effects of its poor memory latency.
Synthetic math tests
Much like the memory tests above, AIDA64’s math micro-benchmarks can tell us a lot about the processors’ potential despite being entirely synthetic. As before, don’t focus overmuch on these tests in a vacuum; they’re mostly for reference when analyzing other benchmarks. We’re using the latest AIDA64 version 6.00.5100, so these tests use all available SIMD instructions (including AVX-512 on our Core i9-7900X).
The next three charts measure more specific capabilities. The CPU Queen test is a chess simulation that heavily stresses the CPU’s branch prediction capabilities; it uses MMX and SSSE3, and favors CPUs with lower misprediction penalties. Queen is similar to a typical single-core gaming workload spread over many cores simultaneously. Meanwhile, the CPU SHA3 test measures each CPU’s cryptographic capabilities.
Since the CPU Hash benchmark has been upgraded to use the SHA-3 algorithm, the Ryzen CPUs can’t flex their SHA acceleration extensions. Now let’s take a look at some web performance benchmarks.