I suspect the gamers in this thread are using Windows, so the rest of my post might not be meaningful.
Over at STH ("Serve The Home"), where they look at server hardware and not just for home use, I read some interesting benchmarks and reports from their testing. Here's the highlights:
- SMT on Linux needs a kernel patch. AMD handles SMT slightly different than current Linux kernel code expects. AMD has released that patch to the public and Linus has "pulled" it in his tree.
- Running the latest public versions of Ubuntu LTS code still isn't enough. It needs to be patched to improve/correct Linux kernel handling of Ryzen SMT.
- In STH benchmarks, single-threaded apps are performing worse compared to Intel parts on standard STH benchmarks, but Ryzen multi-threaded performance is similar to dual CPU setups using Intel E5-2620/2630 chips.
- The "sweet spot" of the Ryzen 7 series family, according to STH, is the 1700 not the higher priced 1700X. I did not see mention of the Ryzen 1800 part in their testing.
Over at Phoronix there is another batch of Linux-oriented testing being done. One of the forum threads from a user that appears to be associated with AMD in some role indicates that some motherboards for Ryzen processors might not be capable of running DDR4 memory faster than 2133 specs. One specific post suggests that some boards may even operate DDR4-3000 memory down at DDR4-2133 speeds. The same poster went on to post a graphic showing how memory speeds can affect benchmarks.
So a lot fo the current commentary on Ryzen performance on Windows might evolve around the following:
- motherboards for Ryzen processors not being able to run some DDR4 memory modules at their rated speeds;
- Ryzen SMT handling requires an update to Windows 10 so that OS handles Ryzen SMT behavior properly;
- games have to be "tuned" for Ryzen CPUs;
- the need for better coding practices when writing multi-threaded apps for Ryzen processors.
Oh it must be nice to live life on the bleeding edge of technology....
