For a long time, well before the release of the new consoles there was a great deal of debate about just how much more powerful the PlayStation 4 was, when compared to the Xbox One. People like to compare specs; and vocalised confirmation bias, much like penis measuring is a reality of the internet. It’s starting, however, to really look like the the Xbox One has a significant performance deficit. Here’s why, and why it’s only going to get worse.
As you’ve not doubt heard, it appears that Tomb Raider: the Definitive Edition runs at a resolution of 1080p averaging at 60 frames per second on the PlayStation 4, but averages closer to 30 frames per second on the Xbox One. This is still unsubstantiated, but it’s increasingly likely. And it just comes down to raw power. While both the Xbox One and the PlayStation 4 use very similar, very familiar architecture, their implementations and specifications differ.
Let’s look at the Xbox One. It uses an 8 Core AMD custom CPU, running at a clock speed of 1.75 GHz. The PlayStation 4’s CPU is essentially the same thing; both based on AMD’s Jaguar APU. Though Sony’s been mum, most reckon it runs at a base clock speed of 1.6ghz, and increases as needed, though nobody knows what speed it caps out at. When it comes to pure CPU cycles, the differences are negligible, with the Xbox One perhaps having a very slight advantage. When it comes to the GPU, however, the differences are much starker.
The folks at Chipworks have taken high-res images of the chips, and counted everything, and they say that the specs of each of the consoles’ GPU differ quite significantly.
1.18 Teraflop GPU with 12 active Compute Units (2 disabled) – (Currently 1.18, maximum of 1.31 TFlop)
48 Texture mapping units
16 Render output units
2 Asynchronous Compute Engines with 16 queues
1.84 Teraflop GPU with 18 active Compute Units (2 disabled) // 56% higher than Xbox One
1152 Shaders // 50% more
72 Texture mapping units // 50% more
32 Render output units // 100% more
8 Asynchronous Compute Engines with 64 queues // 400% more
In pure, raw numbers, the PS4 has more than 50% of the GPU muscle than the Xbox One. That doesn’t translate in to 50% better performance in the real world, it does mean there are significant advantages. While it’s worth noting that both of these are complete SoC’s – so they’d work much more efficiently than their desktop counterparts. with the console’s closed systems, also making them much more efficient, they’d perform much better than you’d expect from PC GPU’s with similar specs. If, however, you were to compare those with commercially available desktop GPU’s from AMD, the closest would be AMD’s 7790 for Microsoft’s Xbox One GPU, though it has disabled Compute Units, fewer TMU’s and a lower clock, bringing the performance down from from 1.79 TFLOPS to 1.18. In truth, it sits somewhere closer in performance to the 7770.
The PS4’s GPU is also seemingly based on a mid-range card; the Radeon 7870, though it too has disabled compute units, fewer texture units and a lower clock speed, giving a rating somewhere between the 7870 and the 7850. A big thing differentiating it from the desktop card though is the 8 ACEs, which is architecture it has more in common with AMD’s flagship R9 290X. It’s a bit of a hybrid, and it’s something I really believe gives the PS4 a graphical edge. ACE’s involve a ton of technical mumbo-jumbo, but they help with graphical multitasking by allowing tasks to be executed out of order.
This is very nearly meaningless, but here’s a 3DMark score comparison between the 7770 and the 7850 – bearing in mind that the Xbox One’s GPU is technically better than the 770, and the PS4’s is better than the 7850’s..but by around the same margin
There’s also the issue of memory. As you know the PS4 uses more expensive GDDR5 memory, which gives it a significant memory bandwidth lead over Microsoft’s system. The Xbox One instead uses slower DDR3, coupled with 32Mb of incredibly fast ESRAM on board to make up the deficit, and act as a framebufffer, which shuttles the generated imagery from the card to the display. It’s fast, but I think MS underestimated how much they’d need, or overestimated its speed gains. The PlayStation 4 uses GDDR5 8GB, running at 1375MHZ ( 4 banks, 5500MHZ effective) on a 256 bit memory bus, giving you a total memory bandwidth of 176GB/s. The Xbox One memory system uses DDR3-1066 (2 banks, 2133MHZ effective) also on a 256 bit memory bus, providing 68GB/s. ESRAM has a theoretical maximum speed of 218GB, but that’s only in one direction, effectively halving the effective speed of the ESRAM. Coupled with the size of it, it ends up being quite a bottleneck, requiring clever programming tricks to make up for it.
There’s really no way of getting around it; there’s no magical secret sauce lurking in the Xbox One, the cloud isn’t going to make the Xbox One faster. The PS4 is the more powerful system, and that power is going to be more evident as this generation goes on. Yes, developers will learn new tricks, and as they do they’ll be able to do more with the Xbox One, and it will get better. There’s no point in denying that the Xbox One is a weaker console anymore, and unless there are dodgy dealings, I expect third party multiplatform games to continue to run and look worse on the Xbox One.
Of course, it’s all about games, and the question of which system has the better ones at the moment is still up in the air – but if you’re a console gamer and things like resolution and frame-rate are important to you, there’s a clear choice.
Last Updated: January 24, 2014