ReLive (AMD) vs ShadowPlay (Nvidia)

As of launching this video, AMD have just released their Radeon Settings Crimson ReLive driver, an update to the Crimson driver launched last year. There are going to be loads of other, much more detailed reviews on the features such as Radeon Chill that drops the core speed dynamically to save power, but this video is specifically on the ReLive recording feature.

It allows you to record gameplay easily, add a webcam, custom png overlay, change the output bitrate, livestream directly from the driver, add your system info, and a nice enough toolbar as you can see here with a relatively  limited amount of settings available on it.

To test the performance difference, I’m using Unigine Heaven and the built in benchmarks in GTA V and Dirt rally, as I wanted a repeatable result that when I hit benchmark I’d get the exact same number every time so as to rule out my lack of skill when playing games and show the actual performance differences.

To test the AMD driver, I’m using an RX 480 on a Windows 10 PC. I repeated the results multiple times to make sure my analysis is correct, so the footage you’ll see of the benchmark results is representative of the results I got for both cards. Since I’m not directly comparing GPUs here, only the performance difference between recording and not, I’m using a GTX 1060 for the Nvidia side.

As you can see – my initial benchmark came out at 83.9 – basically 84 FPS. The little recording timer is visible in both the recordings and on screen, although I’m not sure if you can turn that off or not. This is a clip of the recording while Unigine was benchmarking – I love how the footage came out and the ability to alter the bitrate is brilliant too. The second benchmark shows a roughly 2FPS drop from recording during the benchmark – something I’d say is perfectly acceptable and perhaps even could be labelled pretty insignificant.

As expected, GTA V shows a similar result. The footage of this too looks really nice, this is set to 30mbps recording, while Twitch has a limit of only 3.5mpbs so your recordings will look much better than the stream to twitch.

The results we are looking for here are the bottom right number as that is what I’d consider the most realistic run the benchmark does, and as you can see there is around a 3FPS difference. Again, that’s pretty small, especially when you are talking about 110FPS average.

Of all the footage, Dirt Rally is my favourite here. I love how it looks with the trees and the gravel rushing past, and I think I’ll be using some of this footage to update my normal benchmark footage! In terms of benchamarks, as you can see here there is about a 4FPS difference – through my multiple runs I saw this be as low as 2FPS but once again it’s still pretty small.

I just want to take a second to be annoyed here – WHY DO YOU HAVE TO SIGN IN TO USE THIS STUFF? Sure, you could have it as an option if you do want to save your preferences or connect to a shield, but otherwise there is no reason why you have to sign in. Gah. Anyway, since ShadowPlay has been around for longer, the UI for me is a bit better. There are more options in the toolbar too which I like, and the footage is pretty much the same quality as you can see here.

As far as I’m aware, you can’t change the bitrate of this footage meaning you would be stuck with fairly large file sizes, but I could be wrong there.  Performance wise, it’s basically like it isn’t recording. Of the multiple tests I did, they all came back basically the same or within margin of error, which is really impressive. GTA V is the same, the footage here looks awesome, although I’d need to do a blind A B test to really see much of a difference. Looking at the bottom result you see 131 when recording versus 132 when not, but if you look at some of the ones above, you might say the game runs faster when it’s recording!

Once again dirt rally is the same story too, beautiful footage – this time without a timer at the bottom – and next to no performance hit at all – at least from my testing anyway. This could potentially be limited by what CPU you have too, but I’m using a socket 1151 i7 so I’m all good.