Testing CPU scaling in Resident Evil Requiem — and why we weren’t able to finish the job

The Resident Evil franchise has a storied history with CPU performance and DRM. Just this month, a few weeks before the release of the latest entry, Resident Evil Requiem, Capcom removed the controversial Denuvo Anti-Tamper DRM and replaced it with a new DRM in 2023’s Resident Evil 4 Remake, resulting in up to a 40% drop in overall performance primarily focused on the CPU. In 2021, with the last mainline entry Resident Evil Village, Denuvo could cause massive stuttering — an issue only uncovered due to the fact that pirated versions of the game ran better than legitimate copies. It was eventually removed.

Given Capcom’s recent history, I wanted to look into the CPU performance in Resident Evil Requiem, but I wasn’t able to complete my testing. You can probably guess why; Denuvo Anti-Tamper. Although I had planned to gather data for around 35 CPUs, I had to cut testing short and settle for 13. There are still some interesting patterns in that data, which we’ll go over here. But it’s worth discussing Denuvo Anti-Tamper in more detail, the limitations it has, and how it can stonewall your gameplay if you’re not careful.

Overall, Resident Evil Requiem performs well. I’ve played through about eight hours of the game outside of benchmarking, and I haven’t encountered the notorious stuttering from Village. Performance is where I expect it to be given that the game is built on the widely-used RE Engine. The game even includes path tracing, which looks incredible, though I avoided it for CPU testing.

Specifically looking at games from which RE Engine gets its name, it has proven itself time again to be a highly scalable engine, and even flashing back seven years to the Resident Evil 2 Remake, you can see how well Capcom’s tech holds up. Thankfully, Requiem follows in the footsteps of previous Resident Evil releases, unlike other titles we’ve seen on RE Engine, including Monster Hunter Wilds and Dragon’s Dogma, that have put unnecessary strain on the CPU.

Before getting into the benchmarks, however, we have to address the elephant in the room Denuvo. Although it presents a unique problem specifically for this type of testing, it can also pose a hurdle in normal usage.

Denuvo Anti-Tamper, and why we had to cut Resident Evil 9 testing short

Denuvo makes a few products, but the one that’s relevant for Resident Evil Requiem is Anti-Tamper. It’s a third-party DRM that’s embedded within the executable, launching at runtime to authenticate your copy of the game with one of Denuvo’s servers before generating a token that grants you access. Critically, it isn’t kernel-level anti-cheat along the lines of Riot Vanguard. It doesn’t provide low-level access to the hardware, which is a common misconception when “Denuvo” is used to talk about many different forms of copyright protection.

Like many games with Anti-Tamper, Requiem allows you to authenticate on five PCs within a day. Denuvo itself says its five PCs within a 24-hour period, though my testing suggested that there’s some universal reset time in which all of those devices are reset. In the case of CPU testing, Denuvo flagged a CPU swap as a different device. Every five chips I tested, I would essentially be locked out of the game for 24 hours, severely limiting my ability to get through dozens of CPUs and hit a reasonable publish time.

As your system changes, a previous Denuvo Anti-Tamper token will become invalid, forcing the game to authenticate once again to generate a new token. The issue is that it’s not clear what constitutes a system change. Anti-Tamper is opaque by design, and that design means it’s difficult to pin down what exactly forces a new token to be generated. Some suggest something as small as a Windows or GPU driver update will trigger a new authentication, while others say using different Proton versions on Linux will trigger new tokens. Clearly hardware changes force a new token, as well. But we don’t know for sure and likely never will.

There’s also a performance question with Denuvo. We’ve seen examples in the past, most notably with Resident Evil Village, of the performance impact of Anti-Tamper when combined with other DRM. But this changes from game-to-game and the specific implementation of copy protection. Without a pirated, cracked copy of the game, you can’t drill down on the exact performance overhead (if any) that Anti-Tamper imposes.

Anti-Tamper isn’t always permanent. Its stated goal is to “secure the launch window,” delaying a cracked copy of the game from making the rounds while publishers are rolling out initial copies. We’ve seen many times before, including in multiple Resident Evil titles, where Denuvo Anti-Tamper is removed a year or two after release. It’s possible we’ll see the same rollout with Requiem.

CPU performance in Resident Evil Requiem

With Denuvo out of the way, let’s get into the interesting bits of this testing: the benchmarks. Requiem has both first- and third-person camera perspectives, but I didn’t notice a significant performance gap between them. Still, I tested with the first-person camera as the majority of the opening hours of the game are in the first-person perspective. I chose a scene early on, shortly after you’re let loose from scripted sequences and handed the first major escape room-style puzzle of the game.

I used the CPU test bench I use for reviews for these test passes, which includes some specific settings and hardware choices I’ll detail below. For now, the important bit is that I used an RTX 5090 FE to isolate CPU performance as much as possible and prevent any GPU bottlenecks. I also ran multiple passes (three to five) of the same sequence with each chip to make sure the results were solid. The results below aren’t an average of those; I instead chose the median result.

As expected, AMD’s X3D chips top the charts, but there’s a little more to the story with Requiem. You can see the Ryzen 7 7800X3D, Ryzen 7 9800X3D, and Ryzen 5 7600X3D all posted nearly identical results, both across average frame rates and 1% lows. Requiem favors extremely fast data access and punishes small differences in latency. There’s great evidence of that with the Ryzen 9 9950X. Both the Ryzen 5 9600X and Ryzen 7 9700X posted clearly higher performance, exposing the inter-CCD latency of AMD’s 16-core flagship.

You can see similar results on the Intel side of things. With Raptor Lake and Raptor Lake Refresh, you see about a 10% jump in performance when moving from the 14-core Core i5s to the 16- or 20-core Core i7s. The Core i7s both have larger cache pools, notably a larger shared L3 cache. That seems like the linchpin for CPU performance in this game. The Core i7-14700K and Core i7-13700K posted almost identical performance, despite the Raptor Lake Refresh chip boasting a higher core count and higher boost clocks.

Driving this point home are the Alder Lake chips: the Core i5-12600K and Core i7-12700K. A mistake on my part — or a “happy accident” if you’re of the persuasion of Bob Ross — exposed a lot about how Requiem responds to memory speed and cache sizes. With the Core i7-12700K, I tested with stock 4,800 MT/s, while all other Intel chips were run with memory at 7,200 MT/s. And, as a result of that, the Core i5-12600K actually beat the Core i7-12700K, despite the latter chip having access to a larger L3 cache, more cores, and higher boost clocks.

Cache is king in Requiem, which shows up everywhere from multiple generations of Intel chips to the top three X3D chips, where we’d expect to see scaling, but otherwise see a flat performance wall.

Image 1 of 4

Looking at other aspects of performance, we can see that higher clocks don’t translate into better performance. Higher power consumption as a result of those clocks, and boosted core counts, doesn’t account for much, either. Finally, looking at efficiency, you can see the X3D chips absolutely run away with performance, with the Ryzen 5 7600X3D standing out as the true star with chart-topping results.

Although I would’ve liked to test more chips and drill down on memory speeds, in particular, that wasn’t possible due to Denuvo in the short window I had to test before launch. Unfortunately, there’s no way to tell if you need a new token with system configuration changes without running into the five-device limit and locking yourself out of the game for a day.

The Test Bench

If you’ve read any of our CPU reviews, the test bench should be familiar here. We used the same bench with nearly identical OS images across the Intel and AMD platforms, short of the drivers specific to those platforms. We also tested with the RTX 5090 FE running at stock settings.

In addition to the hardware, there are a handful of software tweaks we make. Virtualization-Based Security is disabled, Resizeable BAR is turned on, and any automatic overclocking features like AMD’s PBO are explicitly disabled. We test warrantied, out of the box performance. In the case of AMD’s X3D chips, that also means leaving the X3D gaming boost settings on newer motherboards turned off.

You may be able to get slightly higher performance with these features, but as they aren’t warrantied, we leave them off to level the playing field. On the Intel side of things, we also stuck with the default power settings, not the Performance or Extreme power profiles that have caused issues for Raptor Lake chips, in particular.