By Dayna McCallum
AMD has announced new additions to its high-end desktop processor family. Built for demanding desktop and content creation workloads, the 24-core AMD Ryzen Threadripper 3960X and the 32-core AMD Ryzen Threadripper 3970X processors will be available worldwide November 25.
AMD states that the powerful new processors provide up to 90 percent more performance and up to 2.5 times more available storage bandwidth than competitive offerings, per testing and specifications by AMD performance labs. The 3rd Gen AMD Ryzen Threadripper lineup features two new processors built on 7nm “Zen 2” core architecture, claiming up to 88 PCIe 4.0 lanes and 144MB cache with 66 percent better power efficiency.
Prior to the official product launch, AMD made the 3rd Gen Threadrippers available to LA’s Blur Studio for work on the recent Terminator: Dark Fate and continued a collaboration with the film’s director — and Blur Studio founder — Tim Miller.
Before the movie’s release, AMD hosted a private Q&A with Miller, moderated by AMD’s James Knight. Please note that we’ve edited the lively conversation for space and taken a liberty with some of Miller’s more “colorful” language. (Also watch this space to see if a wager is won that will result in Miller sporting a new AMD tattoo.) Here is the Knight/Miller conversation…
So when we dropped off the 3rd Gen Threadripper to you guys, how did your IT guys react?
Like little children left in a candy shop with no adult supervision. The nice thing about our atmosphere here at Blur is we have an open layout. So when (bleep) like these new AMD processors drops in, you know it runs through the studio like wildfire, and I sit out there like everybody else does. You hear the guys talking about it, you hear people giggling and laughing hysterically at times on the second floor where all the compositors are. That’s where these machines really kick ass — busting through these comps that would have had to go to the farm, but they can now do it on a desktop.
As an artist, the speed is crucial. You know, if you have a machine that takes 15 minutes to render, you want to stop and do something else while you wait for a render. It breaks your whole chain of thought. You get out of that fugue state that you produce the best art in. It breaks the chain between art and your brain. But if you have a machine that does it in 30 seconds, that’s not going to stop it.
But really, more speed means more iterations. It means you deal with heavier scenes, which means you can throw more detail at your models and your scenes. I don’t think we do the work faster, necessarily, but the work is much higher quality. And much more detailed. It’s like you create this vacuum, and then everybody rushes into it and you have this silly idea that it is really going to increase productivity, but what it really increases most is quality.
When your VFX supervisor showed you the difference between the way it was done with your existing ecosystem and then with the third-gen Threadripper, what were you thinking about?
There was the immediate thing — when we heard from the producers about the deadline, shots that weren’t going to get done for the trailer, suddenly were, which was great. More importantly, you heard from the artists. What you started to see was that it allows for all different ways of working, instead of just the elaborate pipeline that we’ve built up — to work on your local box and then submit it to the farm and wait for that render to hit the queue of farm machines that can handle it, then send that render back to you.
It has a rhythm that is at times tiresome for the artists, and I know that because I hear it all the time. Now I say, “How’s that comp coming and when are we going to get it, tick tock?” And they say, “Well, it’s rendering in the background right now, as I’m watching them work on another comp or another piece of that comp.” That’s pretty amazing. And they’re doing it all locally, which saves so much time and frustration compared to sending it down the pipeline and then waiting for it to come back up.
I know you guys are here to talk about technology, but the difference for the artists is the instead of working here until 1:00am, they’re going home to put their children to bed. That’s really what this means at the end of the day. Technology is so wonderful when it enables that, not just the creativity of what we do, but the humanity… allowing artists to feel like they’re really on the cutting edge, but also have a life of some sort outside.
As you noted, certain shots and sequences wouldn’t have made it in time for the trailer. How important was it for you to get that Terminator splitting in the trailer?
Marketing was pretty adamant that that shot had to be in there. There’s always this push and pull between marketing and VFX as you get closer. They want certain shots for the trailer, but they’re almost always those shots that are the hardest to do because they have the most spectacle in them. And that’s one of the shots. The sequence was one of the last to come together because we changed the plan quite a bit, and I kept changing shots on Dan (Akers, VFX supervisor). But you tell marketing people that they can’t have something, and they don’t really give a (bleep) about you and your schedule or the path of that artist and shot. (Laughing)
Anyway, we said no. They begged, they pleaded, and we said, “We’ll try.” Dan stepped up and said, “Yeah, I think I can make it.” And we just made it, but that sounds like we were in danger because we couldn’t get it done fast enough. All of this was happening in like a two-day window. If you didn’t notice (in the trailer), that’s a Rev 7. Gabriel Luna is a Rev 9, which is the next gen. But the Rev 7s that you see in his future flashback are just pure killers. They’re still the same technology, which is looking like metal on the outside and a carbon endoskeleton that splits. So you have to run the simulation where the skeleton separates through the liquid that hangs off of an inch string; it’s a really hard simulation to do. That’s why we thought maybe it wasn’t going to get done, but running the simulation on the AMD boxes was lightning fast.