games experience changes due to PKTI on version 10
games experience changes due to PKTI on version 10
Video games appear to be influenced by the meltdown workaround according to my tests. I conducted some basic amateur tests and hope someone can perform more detailed professional evaluations. My main concerns were: 1) Phoronix ran on Linux, while Windows uses a different architecture—possibly due to the modern OS book by Andrew Tannenbaum, which mentions that the Win32 API contains many graphics-related system calls. This suggests Windows might generate more syscalls, possibly causing more frequent invalidation of MMU caches in games. I’m not an expert on the Win32 API, but this seems like a reasonable assumption. 2) Phoronix only tested single-player locally; online multiplayer wasn’t included, which is unusual since most benchmarks use server-side load. This makes it hard to generalize. Many are warning that IO-heavy applications will struggle, and online multiplayer relies heavily on network I/O. If the workaround triggers MMU cache invalidation, it could significantly affect performance. It seems like a flawed approach to evaluate games in realistic conditions.
In my experiments, I used an i5-3570k with a GTX 1060 (3GB, 1280x960). Before the patch, my average FPS was 259 on the benchmark map. After applying kb4056892 and rebooting, my online multiplayer FPS dropped to about 234. In deathmatch mode, FPS fell by roughly 30% on a 64-tick server (128-tick servers could involve more network traffic). I suspect these drops might be inconsistent—sharp spikes followed by recovery—which would mean the actual lag is worse than the average suggests.
Here’s my setup:
- Processor: i5-3570k
- GPU: GTX 1060
- Resolution: 1280x960 (4:3)
- RAM: 3GB
- OS: Windows 10
- Benchmark: CSGO map with offline and multiplayer modes
I noticed a noticeable dip in performance after the patch, especially during intense gameplay. The results indicate that what appears to be a stable average may mask significant fluctuations. If these patterns hold, online multiplayer games could face more serious challenges than currently understood.
Let me check the section you posted to see if it matches correctly.
In the context of Win32, 3D applications rely on Direct3D (or OpenGL/Vulkan), not Win32 APIs which were mainly for Windows interface elements. For more accurate data collection, consider using a real-time FPS counter like FRAPS. Run several tests to average out fluctuations. Focus on keeping only the patch status consistent—anything else could skew results. Keep in mind the advice from the article: https://arstechnica.com/gadgets/2018/01/...us-patches
It was what I anticipated—it primarily impacts multiplayer titles. The latest update also changes writing speeds and reading performance on SSDs.