To say that Cyberpunk 2077 has gotten off to a troubled start is putting it mildly. Bugs and glitches abound, performance on previous-gen consoles is quite terrible, and even on PC, you'll want a fairly potent rig. We've looked at GPU performance, CPU scaling, and even RAM requirements. Meanwhile, the v1.05 patch arrived for PC on Dec. 20 and could potentially invalidate previous benchmarks, so we wanted to do a quick sanity check. There's no sense in putting in dozens of hours of testing if things haven't really changed, right?
Based on our limited testing, and depending on your hardware and settings, the performance differences between 1.04 and 1.05 range from negligible to pretty decent.
If you're running with an AMD CPU and haven't already used the SMT hack to enable the extra threads, v1.05 fixes that for you. How much that matters varies based on your CPU. 4-core/8-thread and 6-core/12-thread Zen 2 and Zen 3 CPUs appear to benefit the most, and 8-core/16-thread chips can also see a moderate boost. However, the 12-core/24-thread and 16-core/32-thread CPUs appear to have enough cores to go around, so SMT doesn't really benefit the game (and may actually drop performance slightly).
What about GPUs? We originally tested the RTX 3080 and RX 6800 XT with version 1.03. On the Nvidia GPU, 1.05 performance is largely unchanged—basically within the margin of error for live frametime captures using OCAT. We did see improvements in 99.9th percentile frametimes, which should mean slightly fewer stutters/hitching, but there's far more variability in 99.9th percentiles (i.e., a few 'bad' frames can quickly skew the results).
On the AMD side, our 1.03 results are higher than 1.05, which in turn is slightly to moderately higher than 1.04. However, I think I may have inadvertently had AMD's Rage Mode overclocking enabled on the 6800 XT during the first round of 1.03 testing, which could account for the differences we're now seeing. That or the stars were in alignment on those first tests, but that's no longer the case. (In other words, I've tried to recreate the original test results on 1.03 and cannot, despite cleaning out all GPU drivers and reinstalling. To be fair, it was a crazy night of benchmarking Cyberpunk 2077 at launch.)
Otherwise, looking at current 1.04 and 1.05 results, 1.05 does appear to improve AMD's performance at 1440p and 4K. 4K performance is 6 percent higher at medium settings and 9 percent higher at ultra settings, while 1440p medium performance is only 1 percent faster, but 1440p ultra is 5 percent faster. There's still no support for ray tracing on AMD's GPUs, but given the number of other bugs and glitches, we imagine that's a lower priority than getting the game to work properly.
What about lower-tier GPUs? Some of those may see similar improvements, but we're going to hold off on retesting any other cards. Besides, it's not like most people would want to run with a buggier 1.03 or 1.04 build. Then again, there's some entertainment in the various things that can go wrong in Night City—though we'd just google those. It can almost be more fun watching the best glitches than playing through the actual story.
