Analyzing GPU Tiers: GPU Performance Hierarchy


Since it’s a new year, we’d like to introduce something new from Back2Gaming: some of you may probably know or heard about the GPU Performance Hierarchy chart made by Tom’s Hardware. According to Tom’s Hardware, it helps in judging if a graphics card is of reasonable value. They grouped the GPU’s based on performance and the topmost row contains the fastest GPU’s and performance decreases as you go down the chart.

As of August 26, 2016

We appreciate Tom’s Hardware’s effort in helping the gaming community because we know how hard and how time-consuming to benchmark PC hardware is. However, we have observed something with the chart: though the chart shows how a user’s existing GPU fares against a prospective GPU upgrade, it doesn’t give any meaningful information as to how a prospective GPU upgrade will perform in the current games. For example, you currently have a GeForce GTX 660 graphics card and you are looking for an upgrade. The Radeon RX 480 is way above the GeForce GTX 660 based on that chart but you don’t have an idea how it performs in the current games running at a specific resolution.

That being said, we came up with a solution and grouped GPU’s according to their average frame rate at a specific resolution and AT A SPECIFIC APPLICATION. A bit of background, games put varying loads on GPUs and systems in general hence the System Requirements we see with many other software. These System Requirements whether it be minimum or recommended, are the reference specifications that the developers intend their software to run on. In our case, games, which are oftentimes the most demanding applications that will ever put on our PCs which true enough, are the primary driving force for GPU sales and development for the consumer front. The cold hard fact is that not all games put the same load on the system.

Play resource demanding video games virtually, with a lag free CPU performance along with new GPU acceleration on the cloud by using a cloud based virtual machine. For other innovative cloud solutions visit http://www.apps4rent.com.

Let’s break it down further: developers like Blizzard want their games playable on as many systems as possible which is why their games are so light yet Overwatch has become such a hit that many gamers are now using it as their primary factor in deciding for a GPU purchase. While it is true that Overwatch does have support for advanced graphics detail and other ways to improve in-game visuals, Overwatch on Epic will never put the same load on a system than say Deus Ex: Mankind Divided or even an older title like Crysis 3 on their High settings.

This method of GPU tiering may be helpful in comparing theoretical performance relative to market segment but doesn’t really say much about which is the best for which applications. Again we go back to the topic of people actually deciding GPU purchases based on games they play not on how they want their games to look. And this says a lot what gamers really want: eSports players will prefer higher FPS/low frametimes than decent FPS on max settings on say QHD or 4K.

The performance levels were divided into seven: 40-45 FPS, 46-50 FPS, 51-60 FPS, 61-75 FPS, 76-90 FPS, 91-120 FPS, and 120+ FPS. We have set the lowest performance level to 40-45 FPS so there is enough allowance for possible frame rate drops during game play scenarios where there are lots of effects like smoke, fire, flying debris, and etc. If the average frame rate is already at 30 FPS, any reduction in frame rate is likely to have a noticeable negative effect on gaming experience.

We are still in the process in completing our own data but we will show you now a sample output of our proposed format for GPU performance hierarchy. The following data came from Guru3D‘s review of MSI GeForce GTX 1070 Quick Silver 8G OC.

You might have noticed that a GPU’s performance level changes depending on the game. In Grand Theft Auto V, both RX 480 8GB and GTX 1060 6GB are in the performance level of 1080p @ 76-90 FPS while in Deus Ex: Mankind Divided, the RX 480 8GB is in the performance level of 1080p @ 61-75 FPS and the GTX 1060 6GB is in the performance level of 1080p @ 51-60 FPS. That is another flaw of the chart by Tom’s Hardware which we have addressed with our proposed format.

In addition, we don’t agree that GPU’s should only be benchmarked using maximum settings. Yes, it’s good for showing the performance difference among GPU’s in worst case scenarios but it does not show the true value of a GPU nor PC gaming itself. For example, benchmarking the GeForce GTX 1050 Ti using maximum settings on the latest games will make it seem like a bad product. It’s a GPU meant for gamers who don’t want to spend a lot for playing games at 1920 x 1080 resolution. Its a fact that tweaking the individual image quality settings can and will noticeably improve gaming experience in the majority of games.

Once we complete our own performance data, our GPU performance hierarchy chart will show performance when using maximum image quality settings and performance when using tweaked image quality settings.

Our proposed format is not perfect but we believe it gives more insight in making an informed decision and gives helpful information to consumers and gamers. Factors like power consumption, heat, memory size, etc. all are common questions asked by users so we’ll grow from here. Furthermore, our format for the GPU performance hierarchy will be useful for those looking to buy a gaming monitor. We would continuously find ways to improve our presentation of data so feel free to leave your questions and suggestions at the comments section below.

More Stories
How To Know If An Item is Fake or Counterfeit (Lazada & Shoppee)