The PS4 and Xbox One are commonly referred to as eight-core systems. While this technically describes the number of CPU cores in each console, it’s not an accurate way of measuring how many of those CPU cores are actually available. Both Sony and Microsoft reserve a certain number of cores for the operating system to use while games are running — on the Xbox One, MS reserves a single CPU core, while Sony historically reserved two cores on the PlayStation 4. That’s now changing — a changelog from November 17 patch notes refers to a new ability to address a previously locked seventh core on the PS4.
That’s interesting news from a technical perspective, since it means developers will have access to somewhat more horsepower than they did previously — but are we likely to see dramatic improvements in frame rate or gameplay? Despite what some have predicted (and with all respect to my Geek.com colleague Matthew Humphries), I’d be surprised if we did.
CPU performance in the Xbox One and PS4
As Geek notes, Microsoft has long shipped the Xbox One with an accessible seventh core and a higher overall clock speed (1.73GHz for the Xbox One vs. 1.6GHz for the PS4). Since both consoles use the same CPU, the combination of a higher CPU clock (8% faster) and an extra CPU core (1.16x more cores) should give the Xbox One a clear advantage in gaming scenarios if the CPU is actually the bottleneck.
Instead, we’ve seen the opposite. In launch after launch, the Xbox One has been confirmed to run at lower resolutions than the PS4. Sometimes that means a static 900p vs. 1080p, while some games use dynamic resolution and scale between various settings to maintain higher frame rates. Even in these cases, the Xbox One is invariably the platform dropping to a lower overall resolution.
Adding more CPU cores to a system only improves performance if the system is able to take advantage of them — and that becomes trickier the more cores you have. There’s a general law of computing, known as Amdahl’s law, that describes how much theoretical performance can be gained by adding additional CPU cores.
Amdahl’s law states that the serial portion of a workload (that part of the task that must be executed on a single CPU) ultimately controls how useful parallelization can be in any given scenario. If 95% of a workload can be parallelized, then moving from 2 cores to 8 cores will improve performance by 3x (from 2x to 6x). If just 50% of the workload can be parallelized, quadrupling the number of CPU cores only has a modest impact on performance.
Modern game rendering engines are designed to keep multiple jobs in flight, as discussed in this presentation on Killzone: Shadow Fall. This slide illustrates the difficulty of keeping the CPU properly balanced: Hurling more threads at a task isn’t helpful in and of itself; workloads and memory allocations must all be properly balanced. This presentation is old enough that I wouldn’t use it to draw hard conclusions about Sony’s thread scheduling or execution behavior, since those areas of the chip have likely evolved over time — but it’s enough information to illustrate that just hurling more cores at a console doesn’t guarantee better performance.
We’re not saying that no one will take advantage of the new CPU resources, but it’s unlikely to make a major difference in the performance delta between the two consoles. All available evidence suggests that it’s the Xbox One’s limited memory bandwidth and smaller GPU that hamper its performance compared to the PS4. It’s important to remember that Microsoft’s original vision for the Xbox One put a great deal of resources into Kinect 2, assuming that gamers would value the add-on enough to charge a $100 premium for including it. When that bet failed to deliver, MS was left with a console that had received a much smaller share of development dollars than its rival.
The base stats on the PS4 and Xbox One make it unlikely that either console will fundamentally one-up its rival this late in the game, but with rumors of new consolesappearing as early as 2018, we might see the two companies jockeying for different positioning in the not-too-distant future.