When AMD’s brand-new core architecture launched in the form of the Ryzen 7 processors a couple of weeks ago, they were very well received. In the majority of benchmarks and productivity tests by reviewers the Ryzen 7 octa-cores easily outperformed similarly priced Intel processors and matched or even beat parts costing twice as much. While this marked the return of AMD to high-performance desktop computing there was a concern: a few scenarios where Ryzen 7 didn’t quite perform as many had expected. One such example of this is in gaming.
Ryzen 7 issues, or lack thereof, get an update
While Ryzen 7 offered incredible leaps over previous FX parts in gaming, many reviewers published some outlying scenarios where they couldn’t match even lower-end Intel parts. Typically these were when reviewers tested the CPU at lower resolutions to attempt to push the bottleneck to the CPU. There were even some scenarios where the Ryzen processors performed better with their SMT (two threads per core) disabled.
Many theories began to arise, from cache latency to communication between Core Complexes. One such theory alleged that the Windows 10 thread-scheduler was to blame for the reduced performance. The basis of this theory was due to the apparent reduced bandwidth / increased latency between CCX units, the thread scheduler should try to group lightly threaded workloads to a single CCX.
In a recent blog post on their website, AMD has responded to several points of interest about the Ryzen architecture’s performance. According to AMD, the Windows 10 thread scheduler is working correctly.
In addition to the thread-scheduling query, AMD also pointed out that the 1800X and 1700X processors have a temperature offset of 20 degrees Celsius which is apparently designed to allow for a consistent fan policy. This offset is apparently not applied to the 1700, which explains why people were getting higher temperatures on the two higher-end models.