AM5 X3D CPU Insights Updated (7800X3D vs. 9800X3D vs. 9950X3D). Januar…
 | Hey, kids. I'm back with an update to that thing I did before. I got my hands on the 9950X3D, so I decided to see what a 16-core CPU can really do. If you just want the tl;dr, here it is: the 9950X3D is indeed the best X3D for Cities:Skylines II if you don't care about money. As before, I monitored average sim speeds in my ugly but CPU-efficient city at various populations. Everything was tested with no mods, Developer Mode enabled and the sim speed set at 4X (the default max speed in the UI, or "three ticks"). This build runs a Gigabyte 3080 Ti and 64 GB of DDR5 @ 6000 MHz (30-36-36-76); all OS and game files are on NVME storage. I did not re-run the 7800X3D due to logistical constraints, but I did sample the 9800X3D again to get a new baseline because my last test was conducted before the January patch, which included some sim optimizations. So, did doubling our core count double the sim speed? Oh, boy, not even close. The 9950X3D keeps the sim speed floored all the way to 600k, but between there and 800k, it begins to drop off, but it still maintains a commendable 40% lead over the 9800X3D. By the time we're at a million, the 9950X3D's lead shrinks to a margin of about 10%. Meanwhile, several of the 9950X3D's 32 threads were less than fully utilized. What's most interesting to me here is how much of a performance improvement we got in the Q1 patch this year. If you compare the 9800X3D results from my first run (purple) and today's testing (blue), you see a dramatic bump, especially at 600,000. Interestingly, whatever issue is causing my performance to dip @ 800,000 but rebound at 1 million on the 7800X3D/9800X3D does not result in the same dip with the 9950X3D. Somehow, my city of 800,000 is trying to do more at once than my city of 1 million. I'm betting it's related to poorly-managed service districts in the 800k version. Thoughts: Time for some unsolicited commentary. If you're planning to upgrade your hardware with C:S in mind (either the first or second game, really), you need to be mindful in setting your expectations. Even if you're upgrading from low-end hardware, consider what doubling, tripling, or even quadrupling your sim performance really means. If you're currently running a city of half a million that is creeping along at 0.25X, even a hypothetical CPU that quadruples your sim performance is only going to give you enough headroom for 1X, which is the equivalent to default game speed on an empty map. In other words, don't drop $700 on a 9950X3D and then yell at me because your city's still slow. Note to the Intel crowd: I've seen the same benchmarks you have, and I'm just as skeptical of them as I was of the claims that the 9950X3D would come anywhere close to doubling the 9800X3D's sim performance. If somebody out there has a spare (and healthy) 14900K lying around that they'd like to see tested side-by-side on a clean install of Windows, let me know; we can probably work something out. I now have a spare 7800X3D on my hands... Usual disclaimer: Like many benchmarks designed to highlight CPU performance, this test is unbound by the game's usual constraints. This city is ugly and uses virtually zero transit systems. It has almost no traffic. What you're seeing here is the difference in performance headroom offered by these CPUs and is not representative of what you should expect from upgrading. [link] [comments] |