Roy Taylor on APUs, gaming, and the end of CUDA
AMD’s Roy Taylor (@amd_roy) talks APUs, gaming and writes the eulogy for CUDA and PhysX.
The PC market is shrinking, and Roy Taylor knows that.
But AMD’s VP of channel sales doesn’t let it get to him. The chief B2B salesman of the underdog in the CPU and GPU wars has quite the analogy to explain why he thinks AMD will do just fine in a market that just keeps shrinking.
“If the water becomes shallow, and you’re a whale, you’re gonna get grounded and you’ll die,” he said. “If the market goes down by 50 percent, it’s awfully dangerous if you’re a whale.”
VR-Zone recently sat down with Taylor when he was in Taipei to discuss talk about the state of the APU, and AMD’s push into the gaming market.
VRZ: How are APUs doing on the market?
Any conversation we have about an APU needs to begin with an agreed upon definition. An APU, to us, is any processor which includes a serial processor — a CPU and GPU in one package. By that definition, not only are Trinity and Richland APUs but so are Ivy Bridge, Sandy Bridge and Haswell. Those comparisons I think are starting to be made. But I don’t know why more aren’t making them.
How we’re doing? Well it’s growing really, really well. We’re really pleased with the growth, but of course we want more.
But how do you know? What I mean by that is, how many people who buy APUs use an APU and how many have added a [discrete] graphics card? Because a graphics card effectively makes it a CPU because you’re disabling the GPU on the APU. On those grounds, I think we are making progress in establishing the strength of the GPU.
A recent review on AnandTech went into a depth of detail between Haswell, GT2 and Richland. I think it’s kind of been a landmark in getting people to start having a narrative around how powerful GPUs are. That’s been really helpful.
Interestingly, the CPU team at AMD when I joined did not know the 3DMark score of the GPU portion of our APUs. Now everyone knows an A10 performs at just under 1100 on 3DMark Fire Strike. I’m very comfortable now that every one of our sales guys knows that. It’s important, because if you look at a GPU stack for our business, 1100 3DMark’s is more powerful than Nvidia’s Geforce GTX 620 or 630 (entry-level discrete) and it’s also more powerful than some of our lower Radeons. If you want to establish a value point for your APU, that’s an important reference.
We’re very pleased with the progress we’ve made, but we want more.
VRZ: Do you see the APU eroding the lower-middle end discrete graphics card market?
I think we’re already seeing that’s the case. For enthusiast gamers, graphics cards will never go away. Unlike our competitor, who’s obsessed with launching consoles in the mobile market, we still love PC gamers and we’re absolutely committed to them. That’s never going to go away. Nobody should have any doubt that we’re committed to GPUs.
Now for the mainstream market, those who can’t afford to have separate discrete graphics cards, yes.
VRZ: How are you ensuring that AMD doesn’t compete against itself?
It doesn’t make any sense to add a graphics card that performs less well than the integrated GPU you have. Unless, of course, you want to have them running together because [AMD] has the ability to run them together in Crossfire.
If you like at Sandy Bridge, 17 percent of the die was GPU. When they got to Ivy Bridge, it went up to 27. Haswell’s now, I think, 31 to 32. Clearly there’s a trend.
We think the reason they’re doing that is because of GPGPU. It’s not because of games. I think they see that HSA is an absolutely unstoppable force. I just don’t know why they don’t call [Haswell] an APU… it seems just like pride. If you remember [ATI] tried to join the coin term VPU… ‘No, no, no, it’s a VPU not a GPU,’ they would say. GPU just became widely adopted they just quietly adopted it, and I believe Intel will do the same. Look [Intel] it’s an APU, why are you protesting?
I do think that imitation is the sincerest form of flattery.
VRZ: Moving to CUDA, does this model have a future?
I think CUDA is doomed. Our industry doesn’t like proprietary standards. PhysX is an utter failure because it’s proprietary. Nobody wants it. You don’t want it, I don’t want it, gamers don’t want it. Analysts don’t want it. In the early days of our industry, you could get away with it and it worked. We’ve all had enough of it. They’re unhealthy.
Nvidia should be congratulated for its invention. As a trend, GPGPU is absolutely fantastic and fabulous. But that was then, this is now. Now, collectively our industry doesn’t want a proprietary standard. That’s why people are migrating to OpenCL.
VRZ: What’s the future of big cores? Does the rumor that the end of big cores is near hold any weight?
I’m going to frustrate everyone by pleading the tech industry equivalent of the fifth, meaning that we can’t comment on unannounced products. What I can, however, say is this: I’m a gamer and in many ways the [FX-9590] is my baby. I wanted to do that and I pushed very hard internally for that. It’s a great success. I did that because I love what we’re doing there.
On the APU side, I’m pushing to do something sexy and exciting there too. Watch that space. Personally, I’m a huge fan of doing sexy and exciting stuff.
VRZ: Moving on to gaming, what’s next for the Never Settle bundle?
It’s coming in August, and it will be called Never Settle Forever. I can’t yet tell you what the games are.
VRZ: What are some upcoming titles that are going to be exciting from a technical perspective?
Well the obvious one is Battlefield 4. We’re doing some stunning things with that.
VRZ: If there’s one “stunning” thing that’s lacking in gaming, what would it be?
I want to see more done on physics. When Nvidia bought PhysX, it was pretty exciting and there were lots of things that when I was at Nvidia we wanted to do but they never came to fruition.
VRZ: Like what?
PhysX breaks down into four or five types: there’s rigid bodies, particles, fluids are some of them. Some of the things we wanted to do then couldn’t be done. There were just too many compromises. There was a lot of complexity in switching with the processor on physics or rendering.
We’ve moved on a lot since then and we can do a lot of things that we couldn’t do previously. We still don’t have truly indestructible everything. For example, in World of Tanks there are things your tank can destroy then suddenly there aren’t some things. It’s illogical. I think there’s still a ton of work we can do around physics.
I also think there’s a lot of work that can be done around artificial intelligence as well. I think single player games are really challenged by the “uncanny valley” of artificial intelligence.
I would also like to see us break out of the sandbox. More diversity in multiplayer.