| Multi-threaded Performance |
|---|
In the single threaded results it was necessary to artificially reduce performance in some tests (such as the browser tests) in order to reflect the capabilities of a single core. Here, we compare like-for-like core and thread counts.
We will see the full power of the Ryzen 8-core in the SMT/CCX page. For now, I only compare four cores - without SMT.
| 3Ghz - Relative to Excavator | |||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
With Ryzen showing a 56.72% benefit over Excavator in single threaded performance you could be forgiven for not expecting the story to change with multi-threaded workloads... but then you'd be neglecting the fact that Ryzen doesn't share its precious resources the same way Excavator did. This stinginess on the behalf of Ryzen has led to a massive 75.66% performance boost compared to Excavator! This partly demonstrates just how detrimental the dual-core module design of Excavator really was. This is also the first hint we get to Ryzen's cache performance benefit - as Excavator lacks an L3 and has poor cache design and policies. I had to remove the hashing and AES benchmarks - Ryzen contains special hardware just for those uses... and this chart then extends into the 400% territory. Including them results in an 85% uplift in performance over Excavator. |
 | ||||||||||||||||||||||||||||||||||||||||
| 3Ghz - Relative to Sandy Bridge | |||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
Finally, we can see Ryzen showing some weak spots. While still much better than prior AMD efforts, it falls short of Sandy Bridge in numerous benchmarks. However, it no longer falls short in a few benhcmarks when using lower latency memory. Overall, however, Ryzen delivers 10.24% better performance than Sandy Bridge. Using CL14 timings Ryzen beats Sandy Bridge by a healthy 15.54% in these same tests... and it's important to remember that the Sandy Bridge system is using DDR3-2133 CL9 memory! If I were to include hashing and AES tests this would move the results to a whopping 30.55% improvement over Sandy Bridge. |
 | ||||||||||||||||||||||||||||||||||||||||
Continue...