I had the chance to catch up with Terry Makedon, AMD’s Sr. Director for Software Strategy, and Pete Vagiakis, Radeon Link Lead, on the sidelines of the Radeon Software Briefing in Bangkok. We talked about Radeon Software, Radeon Link, and multi-GPU, among other areas. Here’s a transcript of the interview.
HWT: Having looked at the presentation today and the addition of integer scaling, we’re getting the feeling that AMD and Nvidia are playing catchup to each other, with regards to feature parity. Nvidia introduced integer scaling a while ago, and before that, AMD introduced Radeon Image Sharpening, which was followed by Nvidia revamping its Freestyle Sharpen. So are these feature additions based primarily on user feedback or are they because of the need to meet parity? How is that decision made?
Terry: I mean, it depends on the features. In particular, the integer scaling one, if you open up our user feedback, you can see that it is our number one most user-requested feature. And so, sometimes people ask me when will we get this feature, when will we get that feature. Of course, due to the nature of what we do, we can’t really say what’s coming in the future. It’s not something that we typically do. But I do have a tip: People can usually see our voting page. People can see what end users are voting for. It’s pretty much guaranteed that what people are voting for 8 times out of 10 will be coming out that particular year. For that one, in particular, it was a feature that I missed out on because we didn’t understand the need for it. But our users spoke loud and clear so we delivered it to them.
HWT: Pete, I think we talked about this before, but basically you’d mentioned that streaming is the future of Radeon Link. And right now, I think the way that it works is a lot more like Steam In-home streaming, on the same wifi network. But what’s your vision here long-term? Is this something we can expect to work remotely like Stadia or Geforce Now?
Pete: So first of all, there are two things. There’s the type of streaming that we do, which is based on your Radeon PC. And of course, as you said, there’s the type of streaming that Stadia does, that Geforce Now does, that it’s somewhere on some sort of server and you just connect to that. So for now, our focus is your Radeon PC. So you have your games, as I said. Streaming is becoming more and more important. You see many players trying to get a piece of the pie.
And of course, it’s still at its infancy. You see that there are issues here and there. You read about them I’m sure. But from what I know, this think is actively being worked on and you will see the technology get better as time progresses, so our focus here is basically the user. So, we are software for Radeon, right. So we allow Radeon users to be able to game whether that’s you Tablet, your TV, whether you’re home or now on the go. We allow you to do that. The users speak loudly. And the way I see this going is that if there is a need in the future for something server-based, we will take a look if our users ask for it. But for now, our focus is the Radeon PC, your Radeon PC.
HWT: The image and video upscaling solution that was shown today, that appeared to be leveraging machine learning. There’s a lot that the competition says can only be done using dedicated hardware. But for today’s use cases, would you really say that that’s really the case? How are you approaching it?
Pete: As we said, we’re using machine learning and DirectML. So basically what we’re leveraging here is Microsoft’s libraries and our hardware, essentially. So our hardware makes sure that these functions are accelerated so you can have a very good result whether that’s denoising or upscaling. So that’s the focus, we’re using our hardware that you know has a very good compute performance from the actual gaming to do all these things.
HWT: Where are we going in terms of multi-GPU? Where are we going in terms of driver support? Both AMD and Nvidia have kind of quietly dropped SLI and Crossfire. But Nvidia recently added support for checkerboard multi-GPU rendering at the driver level. And the whole industry is moving towards having MCM modules. So are there any plans to rework crossfire in a similar way?
Terry: Crossfire, I’m surprised you said it went away quietly. I didn’t actually consider that viewpoint. We were very very clear that Crossfire went away and what crossfire means is driver-level control of rendering. And the reason it went away was that the introduction of DirectX 12 took away the control of the rendering from the driver and into the game itself. So, you know, crossfire going away meaning that our driver no longer needs to do rendering alternate frame type of rendering between two GPUs does not mean that multi-GPU support for gaming has gone away. It depends on the game developer if they want to introduce it. And some games support it and some games don’t.
Conclusion: A more user-oriented AMD
From the looks of things, AMD’s putting a much greater emphasis on user interests and issues going forward, listening to user feedback and adding features that are voted for. Adrenalin 2020 introduced a ton of useful features, and we look forward to seeing what new goodies Radeon users will get next.