I decided that I wanted to go down the rabbit hole and explore something that I’ve never really done before because I have been using NVIDIA GPUs for so many years. You can read articles, blog posts and watch countless YouTube videos about AMDs Crossfire technology but none of it means anything unless you can see it with your own eyes.
The only problem was. I didn’t have 2 identical AMD GPUs and I didn’t know whether or not you could still use Crossfire. Lets find out what happened. Spoiler Alert. It works but you’re going to want watch to see how well it works. Also before you leave stupid comments, watch the whole video first.
Lets talk about how were testing this. Were using a similar test system to the one we used when we tested the ASRock Phantom Gaming RX580. The motherboard is the ASRock X470 Fatality Gaming K4 with a Ryzen 5 2600 and 16GB of Team Group TForce memory at 3000MHz.
I threw in the ASRock Phantom Gaming RX580 8G OC that we tested a few weeks ago you can check that video out right now in the top right hand corner along with a Gigabyte Aorus RX580 4GB as well. To be honest. I didn’t know what to expect and I thought that it probably wasn’t going to work but to my surprise, it worked without me having to even enable anything. I’m guessing most people will say in the comments that I should have known it was going to work, but seriously I never even thought about trying it.
We tested the Crossfire setup with a 2 titles that I know should scale well just to see how big a performance difference can be had. In every single test the GPUs were running at its factory settings. Just keep in mind that both of these cards are factory overclocked as well.
Lets start off with Rise Of The Tomb Raider Benchmark. We did single card testing in our ASRock Phantom Gaming RX580 video so we will use those result so we don’t have to retest. First set of tests with everything set to High in DirectX 11.
With a single card at 1080p we saw an average of 75FPS and a maximum framerate of 116 frames per second.
In crossfire at 1080p we saw an average frame rate of 85FPS and a maximum framerate of 217 frame per second.
Already were seeing a trend and were only 1 test in. Because were using multiple GPUs we also added some 4K tests in as well
In 4K we saw an average framerate of 48fps with a maximum framerate of 141 frames per second.
From tests with some other titles in DirectX12 with cards with above 4GB I’ve found that DX12 performance is usually better than its DX11 counterpart, but again its dependent on the game. So we tested Rise Of The Tomb Raider in DX12 as well just to make sure.
With a single card with DX12 in 1080p we saw an average framerate of 75fps with a maximum frame rate 117 frames per second. That’s pretty close the the DX11 result
With a single card with DX12 in 4K we saw an average framerate of 26fps with a maximum framerate 40 frames per second. Here’s what I wasn’t expecting
With Crossfire enabled with DX12 in 1080p we saw an average framerate of 114fps with a maximum framerate 225 frames per second.
With Crossfire enabled with DX12 in 4k we saw an average framerate of 59fps with a maximum framerate 94 frames per second.
That’s a pretty solid increase. That’s almost double the framerate. I think its because crossfire generally scales better in DX12. I would like to visit this again in another video though.
Okay lets get into the quake champions test. You knew it was coming. Come on its me. Everything needs to be tested with Quake. The reason isnt just because I love Quake Champions, its that iD markets Quake Champions as being optimised for AMD GPUs, where we know that in truth it isn’t. We did a video about that and its in the top right hand corner right now if you want to check it out.
Going off the results in our ASRock Phantom Gaming RX580 Benchmark video we know how it should perform with a single card. Lets go over those results really quickly to refresh our memory.
With a single card In 1080p high we saw an average framerate of 114fps with a maximum framerate of 134fps.
With a single card In 4k high with 90 percent texture scaling we saw an average framerate of 62fps with a maximum framerate of 73fps.
We upped the anti and did all of the rest of the tests in ultra not high because, well why not.
With Crossfire enabled In 1080p ultra we saw an average framerate of 116fps with a maximum framerate of 119fps.
With Crossfire enabled In 4K ultra we saw an average framerate of 56fps with a maximum framerate of 61 frames per second.
Given that its not a direct comparison between single card performance you can get an idea of how well Crossfire scales. We upped the quality and texture scaling at it still technically performs better than the lower setting.
Overall Crossfire performed better than I expected it to. In fact it worked without me having to switch AFR modes like with SLI, although you can do that if you want, I didn’t need to. We got the best results in all of the testing from using the default settings. So yeah just keep that in mind.
Now should you buy 2 RX580s and put them in crossfire instead of buying 1 really powerful GPU. Probably not, but if you have 1 RX580 and wanted to take the power up another level without forking out for a whole other GPU its an option. But probably not an option id personally choose, but man is cool.
I’d love to explore this a bit further. If you have some titles or benchmarks you think we should use that scale well with Crossfire, let me know in the comments. I’m keen to play around with this a bit more. This was one of those, I’m bored on a Saturday afternoon type of videos.