Teraflops: These are all everyone wants to talk about when the upcoming Xbox Series X or PlayStation 5 consoles are mentioned. That’s because Microsoft and Sony brag about big performance increases thanks, in part, to an increase in teraflops.
Console versus PC
The Xbox Series X GPU is based on AMD’s RDNA 2 architecture and will be capable of supporting 12 teraflops. Meanwhile, Sony’s PlayStation 5 (also based on AMD’s RDNA 2 architecture) will have a GPU with 10.28 teraflops.
That’s a lot of flops going on, and it’s comparable, or better, to what high-end PC graphics cards offer right now.
As of April 2020, the Radeon RX 5700XT (around $ 400 at the time of writing) is one of the best AMD cards, with a 9.75 teraflop GPU. The NVIDIA GeForce RTX 2080 Ti ($ 1,300 to $ 1,500 as of this writing), meanwhile, is capable of 13.4 teraflops. There’s also NVIDIA’s RTX Titan, with a whopping 16.31 teraflops. But, at well over $ 2,000 as of this writing, this is out of reach for most players.
We do not anticipate ourselves, however. Let’s talk about what teraflops are and why this specification is important for graphics.
What are the FLOPS?
FLOPS stands for floating point operations per second. Floating point arithmetic is the common way to process numbers in game development. Without getting too lost in the weeds, floating point operations make it easier for computers to work more efficiently with a wider range of numbers.
The most common way to express flops is in the single precision floating point format, FP32. This means that the computer uses 32 bits to store data in that format. There is also a half-precision format that occupies 16 bits (FP16) instead of 32. The most common way to express teraflops for GPUs right now is single precision. However, AMD has used FP16 in its Vega GPUs and RDNA 2 allows FP16.
In the real world, floating point is much easier for game creators with 3D graphics to use. If games were based on fixed point operations, like the original PlayStation, there would be a lot of problems. In-game visuals would have poor appearance and behavior, and the code would generally be less efficient.
So, cheers on floating point operations!
FLOPS Inflation led to TFLOPS
Games have to process a lot of data, which is why flops are such an important benchmark. The more flops a GPU can do, the faster the data can be processed and the more computing power available to run games.
The original Sega Dreamcast (1999) had 1.4 Gigaflops, meaning it could process up to 1.4 billion floating point operations per second. A few years later, the original Xbox (2002) was rocking 20 gigaflops (20 billion flops). The PlayStation 3 (2006) had nearly twelve times that, at 230.4 gigaflops.
Each console has improved dramatically over its predecessor, largely thanks to the graphics processing power. The flop specification is a quick way to get an idea of how much power there is under the graphics processor hood of a particular console or GPU.
The 12 teraflops of computing power of the upcoming Xbox Series X means it can perform up to 12 trillion floating point operations per second. The PlayStation 5, meanwhile, peaks at 10.28 trillion flops.
If we relied solely on flops as a measure, we would conclude that Xbox Series X will be better than PlayStation 5, which would be a mistake.
How important are TFLOPS?
Flop count is important across console generations, but not so much when the gap is smaller.
Comparing teraflop counts for modern graphics cards, such as the AMD Radeon 5700 XT and GeForce RTX 2080 Ti, can also be misleading. The new consoles will use AMD’s new RDNA 2 architecture. The new architecture usually means better performance than previous cards, even with similar hardware specifications.
Like anything else in computing, though, it’s all a matter of implementation. CPU specs, RAM, and even software make a difference. When you put it all together, the consensus is that new consoles should outperform most PC gaming platforms currently out there.
Xbox Series X and PlayStation 5 will have eight-core and sixteen-thread processors. This reaches incredible levels of PC gaming and it’s been a long time to get to the set-top boxes. Both consoles also plan to use NVMe SSDs, which means faster load times for games and better overall responsiveness.
The new console GPUs will also have an impressive number of high-clocked processing units: 52 at 1,825 GHz for Xbox and 36 CU at 2.23 GHz for PlayStation. For comparison, the Radeon 5700 XT has 40 CUs at 1.6 GHz.
Of course, AMD’s RDNA 2 won’t live exclusively inside the new consoles. Once it hits PC graphics cards (along with NVIDIA’s expected Ampere architecture), any advantage consoles have over PCs will disappear.
TFLOPS aren’t the only thing that matters
There is no doubt that the new consoles will be powerful beasts. Microsoft and Sony say their consoles will hit 60 frames per second at 4K resolution in AAA titles (typically the most graphics-intensive games).
Microsoft is also looking to hit 120 frames per second at 4K for esports games, which are usually less demanding in terms of graphics. However, higher refresh rates mean a smoother picture and an easier time to understand what’s happening on the playing field. Given the chaos that ensues in esports, smoother graphics are a big plus.
In addition to improved performance at higher resolutions, the new consoles will also support ray tracing. We first saw this new technology in NVIDIA graphics cards. Ray tracing increases the lighting effects within a game, often with noticeable improvements. It also offers a more dynamic and realistic game environment where shadows and reflections are more realistic. The computing power (teraflops) available within upcoming GPUs will also help these new features.
Teraflops aren’t the only spec you should pay attention to. However, it will give you a general idea of how a console’s graphics power compares to other hardware, past and present.