Can a Switch based on Maxwell with just 1 Tflop of compute performance compare to the Xbox one? (Graphical and mathematical analysis)

I have seen and argued many times that Switch's rumored 1 Tflops of performance could be on par with the Xbox One 1.31 Tflops due to Nvidia's edge over AMD in the real performance of their GPUS. Nonetheless I wanted to make sure it wasn't just bias so I made an analysis to try and illustrate that.



  1. AMD and Nvidia GPU performance was gathered in Wikipedia, links are as follow. Geforce 900 Series (Maxwell). Geforce 1000 Series (Pascal). Radeon HD 7000 series. Radeon RX 400 Series. Radeon RX 300 series.

  2. Real performance was obtained from UserBenchmark "Effective 3d Speed". This measure its gathered from user benchmarks using their software.

  3. Consoles compute performance (In Tflops of FP32).

PS4 Xbox Switch (Rumor) Tegra Parker Tegra X1
1.84 1.31 1 0.75 0.5



  1. I Put the numbers of each series together and created this graph. Which seems to indicate that on average Nvidia cards perform better than their AMD counterparts.

  2. Then I created a graph that only contained the two relevant series (Maxwell from the recent rumor and HD7000 from the architecture PS4 and Xbone are based).

  3. I then drew the average line of each technology curve. Thus getting this as result.

  4. Next I zoomed in on 0~2 Tflops and drew the positions in which PS4, Xbox one, Switch (Rumor), Tegra Parker and Tegra X1 should appear. Thus getting this as result.

  5. However I noticed a poorly drew average line would result in too many errors, so I decided to do a mathematical analysis using a method to extrapolate an average curve based on a series of points I have used before on experimental physics found here on page 13.

  6. Then I estimated the performance each console based on the average lines formula calculated and their compute performance. The results are as follow (Based on UseBenchmark's Effective 3d Speed).

PS4 Xbox One Switch (Rumor) Parker X1
36.4 29.7 30.5 25.7 20.9
  1. Lastly I created a graph to illustrate better the levels of performance. Said graph can be found here.


TLDR: Yes, here is the graph with the estimated performance for each console based on the performance of their desktop architectures.


Please note:

  1. This estimation considers that the performance of the consoles stays on par with the performance of their GPU lines.

  2. This estimation considers that the average line method is a good estimation to presume the performance of a GPU as the graph indicates.

  3. This estimation was made with the only intent of demonstrating that at least in PCs Nvidia has and advantage over AMD, and if the trend remains on Switch the rumured 1 Tflops would be equivalent to Xbox Ones 1.31 Tflops of computing performance. Thus this is in no way a real indication of what performance the Switch will actually have.

  4. The data I used to compile all graphs can be found here (It is a bit disorganized).

Discussion started at here by americosg

Share this post