AMD, despite negative comments to the contrary, have released a technological masterpiece that mostly exceeds expectations when it comes to performance and it’s design even though it competes well, though not quite up to par with Intel’s counterpart in all scenarios. There are some issues, some initial issues that are rather alarming. At first.
Ryzen will only get better with age
And there’s a very good reason for that. The micro-architecture is new, after all. And with something so new, there seems to be a few teething issues with the platform. Mostly there are issues pertaining to performance at 1080P resolutions, in some single-threaded applications and in other fringe situations. The Ryzen 7 1800X, and its ilk, just don’t seem to be as good as it should. Something seems fishy and, as a result, it would seem that the consensus is divided in regards to AMD’s new Ryzen CPUs. Consumer confidence is also teetering on the edge, despite a swift response from AMD on the matter (spoiler: optimization).
So what’s actually going on? Well, it’s somewhat convoluted but it’s fairly simple to explain. In general, a lack of optimization on the part of developers combined with unfinished BIOS’ from MSI and ASUS and poor scheduling from the operating system itself all contribute to a level of performance in certain situations that just isn’t right.
Though my GIGABYTE AX370 Gaming 5 had a more complete BIOS that had no effect on performance (though we’ve only tested at 4K for now, which it’s more then on par with Intel at), a new MSI BIOS that was just released has supposedly increased performance by a large margin for those reviewers that were given MSI motherboards. This revelation highlights how important that aspect of hardware actually is. Supposedly these board makers only had a few weeks to prepare their BIOS files.
What about the other issue, why would scheduling really matter so much? Well, there’s an underlying architectural decision that may be the reason behind a lot of the performance disparity as well. The different CCX’s, or core complexes, can communicate with each other and share their data, though AMD has told me that they have to share this data and keep coherency via the system RAM instead of through a faster cache system or through the Infinity Fabric itself. That alone adds on quite a bit of latency and reduces communication bandwidth by a significant amount. That coherency and a lack of intrinsic development can cause performance to wane further than expected.
So, optimizations are actually needed so that applications send the appropriate data concurrently to the correct CCX for the best results. So, again, given time and work from the developers (with AMD helping quite a bit in this regard) we’ll see improvements in games and all applications.
So, sure, there are issues at lower resolutions with some games, but for the most part it’s still an efficient and very well engineered CPU. That said, we’ll see improvements in performance and efficiency as time goes on, an attribute seemingly inherent in some of AMD’s GPUs. AMD is also quite aware of that particular communication limitation between CCX modules and will likely have something new designed for Zen 2.0. Perhaps a full-blooded fabric that allows bi-directional communication at higher speeds than available now.
Part 2 of our analysis of Ryzen is coming soon, though it may be delayed a week or so while we wait for developers to play a bit of catch up. That way we can see just how far it’s come in performance compared to when it was released.