AMD Interview

During the Beijing launch event, we were lucky enough to spend some time chatting with Travis Kirsch, who is the of Director Client Product Management on the Ryzen series at AMD.

PCPP: How are the CCX modules arranged in the Ryzen 5 chips?

Travis: It’s a symmetric design. We have the same number of cores per CCX module in all Ryzen variants. With the six-core R5 1600X and R5 1600, we simply have two CCX modules, and disable a core on each of these.

What about the quad-core R5 1500X and R5 1400? Are you disabling an entire CCX unit?

We’re using two CCX units, but disabling two cores per unit. That maximises the L3 cache.

So on that note, what are the cache amounts for the Ryzen 5 units?

For the six-core, it has 512Kb L2, so you get 3MB L2, then you get 16MB of L3. The highest four-core also has 16MB of L3 cache, while the R5 1500 has 8MB.

In that case, I’m guessing the transistors are identical, it’s just that they’re not all active?

That’s correct.

Given the lower thermal output of the Ryzen 5 chips as there are less cores active, are you seeing better overclocking results than the Ryzen 7 series, which tends to max out at around 4.1GHz?

Yeah, we’re seeing a little bit better. It depends on the cooling solution obviously; we’re not getting significantly higher, more like in the order of a hundred or two hundred MHz.

So which Intel models are you anticipating will go head to head against the new Ryzen 5 chips?

The R5 1600X would be matched against the 7600K, and then the R5 1600 would go up against the i7-7600. The R5 1500X will go up against the 7500, and then the R5 1400 would go up against the 7400.

During the XFR demo in Cinebech, you showed the 200MHz boost but this only works during the single threaded test. Can you explain why that is? Is it possible to hit that 200MHz boost with all four cores active?

No, we can do up to two cores with a 200MHz boost using XFR. There are a couple of things we’ve done with that part because of the market we’re going into. The first one is we put a 95W cooler with the 65W part, it’s a super quiet cooler. We’ve spent 2.5 years developing this so we can give more value to the customer.

On that note, will the new Ryzen 5 all be coming with coolers?

All but the R5 1600X. This doesn’t come with a cooler, but the rest all have a version of the Wraith cooler. The 1600 comes with a Wraith Spire, the 1500X comes with a Wraith Spire and the 1400 comes with a Wraith Stealth, which is a 65W cooler.

When we first saw the Ryzen 7 released, you guys showed a shitload of benchmarks comparing it to Intel’s comparable chip. However, with the Ryzen 5 launch, we didn’t see a single comparative Intel benchmark, which seemed a little odd. Why was this?

It’s illegal in China. It’s illegal to show comparative benchmarks in China. We’re not trying to hide anything, just given the venue we couldn’t do it. We’d love to show you benchmarks.

So how do you think the R5 1600X compares against Intel’s 7600K?

What we see with our competitors is that when they drop down from eight threads to four threads, they lose a lot of performance. If you look at that 1600X part, a lot of the specs are the same as the 1800X minus the two cores. So we’re still delivering 12 threads at very high frequencies, so that part is going to run very similar to an 1800X as far as gameplay goes, but the competition loses a lot when going from eight threads to four. There’s a lot of concern about 1080p gaming coming out of the community and we learned a lot about that. There’s a lot of things we can do with game developers to improve this experience. But the gap that you see between the 1800X and those eight core competition components, really goes away when you go to a four-core part. I guess what I’m getting at is that the 1600X will be much more competitive in those scenarios to a 7600K.

The extra cores combined with the amazing price makes the Ryzen 7 series a really compelling product versus Intel’s competition. What is it about the Ryzen 5 that you think will make them compelling versus Intel’s quad-core chips?

There’s two things. One is we’ll have double the thread count, which is always good. We like to look at those four-core, eight thread parts as having basically the specs of an i7 but we’re giving it at a price point under US$200. So one is the main value proposition. The single thread performance gap between the i5 is going to be much much closer, secondly the entire stack is multiplier unlocked. If you want to overclock an i5, you have to buy the K version and a Z-series motherboard. With us, you can do it with our lowest end Ryzen 5 1400 and a B350 motherboard. We’re trying to stretch that enthusiast mentality out to more of the market, because we want desktops to be fun.

You hinted briefly today at Zen 2 and Zen 3, with one slide that mentioned them. Is there any additional info you can provide on these – will the improvements be focused on frequency increases? Or architectural changes?

I can’t talk to the future roadmap, sorry.

Anything else you’d like to add?

I’m excited. I think it’s really important for to understand what we’re trying to do. We feel like the CPU ecosystem has been beat down for a really long time, and boring as hell, nothing to write about. This entire ecosystem needs competition, and needs to make the market fun again. That’s our goal. Our team has been working on this stuff for four or five years, and that’s our goal, is to make it fun, and to give people to have a livelihood in the ecosystem.

When you’re talking to your partners, such as motherboard manufacturers, can you sense a palpable change in the air and excitement?

I can just say that product pitches are a lot easier now (laughs). But you can tell that we’re not just helping ourselves, we’re helping them too. There’s a mutual feeling of confidence that is growing.