A resurgent AMD sees huge potential in VR: An interview with David Bennett

Exciting times for AMD

As the year rounds off to a close, we're on the cusp of a series of major hardware launches that should have enthusiasts on the edge of their seats. Other than the impending announcement of AMD's Vega GPUs (which we hope will give NVIDIA's Pascal cards a run for their money), we're also close to the roll out of AMD's Ryzen processors and Intel's Kaby Lake desktop processors. 

But the hardware is only interesting because of the new experiences and capabilities it unlocks, and few things are more enthralling than a virtual reality experience done right. 

We speak to David Bennett, Corporate VP, Worldwide Commercial/ Asia-Pacific Japan Mega Region for AMD, to find out his thoughts on the burgeoning virtual reality space and how AMD is poised to drive developments in this area.

How do you think VR is shaping graphics development?

If you look at the advancements in GPUs and CPUs, you’ll notice that CPUs have for the most part improved only in modest, incremental steps, going up by between seven and 12 per cent from generation to generation. On the other hand, GPUs have managed far more massive improvements between generations. These performance gains don’t come from just die shrinks on the silicon and owe a lot to actual architectural improvements.

Until recently, these advances have been driven primarily by gaming, but with the advent of VR, things have suddenly become a lot more demanding. VR requires 90fps, minimal latencies, and very high resolutions. And in order for the hardware and technology to more closely approximate reality, we’ll require at least a 4K or 8K resolution per eye, which we’re still a long way off from achieving.

Ultimately, I think VR is going to drive GPU development for at least the next eight to 10 years, and performance increases are going to be centered around making VR experiences more realistic.

Related read: AMD's Radeon VR-ready initiative.

What are your thoughts on how mixed reality stacks up against VR?

They are very different. At AMD, a lot of the technology and effort is being put into VR now. Microsoft is obviously going in a different direction with augmented reality, which I think is actually an easier entry point than VR for the general consumer. It is less invasive, and we might see the pick-up rate for AR be a lot quicker.

But what VR has going for it is price. The price of a headset is coming down, and if you go to China or Taiwan, there are literally over 80 headsets being made. What’s more, the PlayStation VR just redefined a new price point for VR, and AMD also recently entered a partnership with Oculus to lower the entry-level price point for a VR-ready PC to just US$499. In comparison, we haven’t heard too much yet about an affordable price for AR.

In your opinion, what’s up next for VR?

What we want to do is enable VR for everyone. But now that we’ve got the base price of the PC down to US$499, I think the ball’s in the HMD manufacturer’s court to see what they can do. Obviously, the next step is to get the HMD price down. Then there’s the issue of providing a seamless experience at 90fps with no tearing or stuttering, because once you start to experience nausea or feel sick, VR goes from being a phenomenal experience to a really poor one. Ideally, we’ll be able to keep things affordable without sacrificing the quality of the experience.

AMD is also poised to take advantage of a lot of the things that are driving VR. If you look at things like asynchronous compute, asynchronous spacewarp, and DirectX 12, the fact is that a lot of these lend themselves very well to AMD.

Furthermore, Sulon has managed to put an AMD FX-8800P APU with integrated Radeon graphics in the Sulon Q, a tether-free VR headset that doesn’t require you to be hooked up to a PC. Sulon has got a prototype that’s working right now, and there are also all these VR backpacks from MSI and Alienware that really suggest that untethering the HMD from the PC is going to be the next big step in terms of VR experience and adoption, especially in the commercial verticals as opposed to gaming.

Speaking of VR applications other than gaming, can you talk more about what these might be?

Gaming is going to continue to drive VR, but if you look at the applications for VR, we’re seeing a lot of interest from a whole bunch of commercial verticals. I think the obvious ones are telepresence applications for things like meetings, trainings, and simulations. But we’re seeing a lot of work being done for the medical field as well, including everything from remote surgery to treatments for patients with dementia or schizophrenia.

Telepresence VR Demo: 2016-02-17 from Human-Engine on Vimeo.

Other major applications include construction and 3D modeling. Before VR, you had to rely on your imagination to visualize what a bunch of 2D models look like in 3D, but once you put on the VR goggles, you’re immediately able to see it. We think that’s going to open up the industry to a lot of people who don’t have the ability to do that kind of spatial translation in their heads.

In fact, the revenue forecast for non-gaming applications is actually starting to be bigger than it is for gaming applications, although gaming is always going to be the lead driver.

What is AMD’s biggest competitive advantage?

If you look at the x86 CPU space, there’s only two companies that can design really high-performance CPUs, and that’s AMD and Intel. In the GPU space, the two major players are AMD and NVIDIA. So our biggest competitive advantage is the fact that we’re the only company that can design both high-performing CPUs and GPUs and put them together.

Furthermore, AMD has shown a great willingness to form partnerships with other technology companies. For example, we’re in Sony and Microsoft consoles, and are very flexible as to the kinds of solutions we create. We want to work with customers to create the precise solutions they want, and our main goal is not just to create the most powerful or most expensive product. Instead, we want to enable the best possible experience for users at a particular price point.

How has AMD benefited from being in both the CPU and GPU business?

In terms of our architectural improvements, we’ve stayed at the same semiconductor node for quite a few generations now. In contrast, Intel has been introducing new nodes, going from 28nm to 14nm and soon to 10nm. They’re driving a lot of their performance increases from shrinking the process node. On our end, we’ve increased performance and decreased power over two generations on the same 28nm node. It is virtually unheard of to achieve these levels of improvements, and we did this by applying lessons learned from our GPU business to our CPU development.

It seems like there has been a lot more excitement in the GPU rather than the CPU space lately. Why is this so?

When Apple released the Mac Pro, it had two CPUs and one GPU, and that was the case for quite a number of years until it suddenly reversed course and went with two GPUs and one CPU. Many people asked why, but the fact is that the increases we were seeing on the GPU side of things were much greater than what we saw with CPUs. CPUs produce incremental upgrades, and there is only so much improvement you can squeeze out while sticking to the same power envelope and clock speeds.

On the other hand, we’ve been able to get a lot more from GPUs in the last couple of generations. That’s why AMD has bet heavily on GPUs, and it is no accident that graphics comprise a bigger portion of our APUs now. GPUs used to be capable of only accelerating a couple of applications, but now it can accelerate Windows 7, 8, and 10 and even the entire Office suits and Netflix. Today, faster and higher performing GPUs will almost surely net you a better overall experience.

Play

That said, Zen is our first new CPU architecture designed in almost a decade. This is the single biggest increase in performance that AMD has had from generation to generation. This kind of performance increase is unprecedented in the industry, and we’re very excited about the Zen CPU core. It brings AMD back into the high performance segment, and it brings competition back into the market. That’s good for the industry, but it’s even better for the end user.

What do you think people want from their hardware now?

If you look at the trends, everyone wants higher performance chips, lower power envelopes, and longer battery life, especially for the notebook side of things. We expect this will continue, and I think AMD is poised to capitalize on this with our APUs, which are essentially combinations of CPUs and GPUs.

The important question is just how to strike the right balance between how big the CPU portion is relative to the GPU portion of the chip. You want to have a powerful CPU, but because a lot of users are also running applications accelerated by GPUs – for instance playing games, watching videos, or using other graphically intensive applications – it becomes key to include a powerful GPU as well. AMD does exactly this, and we’ve actually dedicated a greater amount of our silicon to the GPU because we believe that is going to be a crucial part of the user experience going forward.

Why are AMD APUs not being more widely adopted? How are you working to encourage adoption?

AMD is a technology first company. We spend a lot of our money on technology and not other bells and whistles. We are also a company of firsts. For instance, we were the first to combine CPUs and GPUs on the same chip, and also the first to debut a graphics card with High Bandwidth (HBM) memory. However, the industry hasn’t followed our lead as quickly as we’d like. As a result, a lot of benchmarks that people were using didn’t really give credit to the raw performance that was available to the APU by way of the CPU and GPU balance.

Moving forward, we expect things to improve with new benchmarks. PCMark 10 is releasing soon, and there are a growing number of applications that are better able to take advantage of the GPU part of the APU. It took a while for the software and the ecosystem to match what we’ve done on the hardware side of things, but we’re getting there. What’s more, with Zen coming out, we’ll once again be competitive in the high-end space and will no longer create system bottlenecks.

As things stand, there are already plenty of spaces where people interact with AMD hardware but aren’t aware of it. We definitely need to do a better job of telling that story. AMD hardware is incorporated in Times Square signage and Boeing jets, and I think they are only going to increase in popularity.

Finally, we need to work harder to ensure that AMD chips end up in platforms that people want to buy. This means getting our hardware into high-end systems where they have a chance to shine, which equates to systems with dual-channel memory, FHD displays, and SSDs.

Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.