Gears of War E-Day is the rare Unreal Engine 5 game whose frame rate scales with the CPU

In this game, the graphics preset decides where the bottleneck sits. Between the Ryzen 7 9800X3D and the Ryzen 5 9600X, Tom's Hardware measured a 6.8% gap at the Ludicrous preset and nearly 29% at Ultra, and the gap grew again at High ^[11]. One step down from the top preset widened the spread about 4.3 times ^[1]. "The dynamic is binary," the publication wrote ^[15].

It ranked the chips at High because High is the visual floor at which the game becomes CPU-bound ^[12]. Medium adds very little performance and slightly exaggerates the differences between chips. Tom's Hardware called High a compromise that shows near-maximum scaling without being an unreasonable test case ^[12].

So the 42% gap comes from one specific workload: an RTX 5090 at 1080p, the High preset, and a single in-game scene ^[2][10]. To reproduce it on another machine, the processor has to stay the slower component at whatever settings the player picks. A weaker graphics card or the Ludicrous preset moves the result toward the 6.8% end ^[11].

The method behind the number is careful. All 25 Intel and AMD chips ran on the publication's CPU-review systems ^[3]. The frozen OS images were identical, and the only change was the GPU driver, because E-Day will not launch without the Game Ready driver ^[9]. The one variable that moved was one the game insisted on. The testers also checked the built-in benchmark against gameplay before trusting it. The benchmark ran slightly slower than two in-game scenes, likely because its sweeping camera draws more than a player sees, so testing used scene 2 ^[10]. The CPU load held through the opening hours regardless of scene complexity or interactions, so one scene is a fair sample ^[7].

Core count is the other variable Tom's Hardware flags. On average it expects less than a 5% gain from a Ryzen 7 7700X over a Ryzen 5 7600X. For E-Day, it calls the core-count scaling the interesting result ^[14].

Does this extend to Unreal Engine 5 as a whole? The same test bench says no. Even the UE5 outliers the publication names, Marvel Rivals and The Blood of Dawnwalker, scale less than Starfield or Baldur's Gate 3 ^[5]. Tom's Hardware does not say what in E-Day's code produces the load. I think the defensible hardware rule is per title. On this evidence, budget CPU for E-Day at 1080p at High or Ultra, and treat UE5 in general as GPU-led until other titles show the same spread on the same bench ^[4].