How to Measure Teraflops: Understanding the Metrics of Computational Performance

Teraflops have become a key metric for measuring the computational performance of modern gaming consoles. But what exactly are teraflops and how are they calculated? In this article, we will break down the process of calculating teraflops and what they actually mean for system performance.

To put it simply, a teraflop is a metric used to measure the amount of computational power a device can produce. This metric is particularly important for gaming consoles, where the graphical capabilities of the hardware are a crucial component of the gaming experience.

The process of calculating teraflops is relatively straightforward when it comes to AMD graphics hardware, which is found in both the PlayStation 4 and Xbox One consoles. The calculation involves multiplying the number of individual shader cores in the GPU by their clock speed, and then multiplying that result by 2, since the GPUs can carry out two instructions per clock simultaneously. Finally, the result is divided by 1 million to get the final teraflop figure.

For example, the PlayStation 4 has 1152 shaders running at 800 megahertz, which results in a total output of 1.84 teraflops. The Xbox One, on the other hand, has 768 shaders running at 853 megahertz, resulting in an output of 1.31 teraflops.

The leak spec for the PlayStation 4K, also known as the Neo, showed a significant boost in GPU power, with 2304 shaders running at 911 megahertz, resulting in an output of 4.2 teraflops. This represents a 2.3x boost over the original PlayStation 4 GPU and a 3.2x increase over the Xbox One GPU.

Microsoft’s Project Scorpio, the successor to the Xbox One, is rumored to have a six teraflop output. While the exact makeup of Scorpio’s GPU is still unknown, it is likely to be based on AMD’s next-gen processor, Vega. One possible scenario involves using 3584 shaders at 840 megahertz to achieve the six teraflop output, or alternatively, using 3840 shaders at 800 megahertz.

While teraflops can be an important metric for measuring GPU performance, they are not the only factor to consider. Other aspects such as drivers, memory bandwidth, and CPU power can also affect system performance. Teraflops are most useful for comparing two cards from the same GPU family, and even then they are not the only factor to consider.

The increase in GPU power between the PlayStation 4 and Xbox One did not necessarily translate to a 40% increase in performance, as seen in benchmark tests of games like Hitman. Other factors such as memory bandwidth and CPU power can also limit gaming performance.

In conclusion, teraflops are an important metric for measuring the computational performance of gaming consoles, but they are not the only factor to consider. The comparison between different families or manufacturers does not hold much value, and other factors such as memory bandwidth and CPU power can also affect system performance.