AMD unveils Ryzen brand for Zen processors, base clocks, and additional performance details

By Joel Hruska

• By Joel Hruska on December 13, 2016 at 4:05 pm

Last week in Sonoma, AMD unveiled new details on its upcoming CPU architecture, codenamed Zen. For the past 16 years, AMD has relied on a combination of Greek (Athlon, Phenom) and Latin-derived (Duron, Opteron). More recently, it’s used the old “FX” moniker from its high-end Athlon 64 consumer hardware to market Bulldozer and Piledriver-derived CPUs. Going forward, Zen will be marketed under the “Ryzen” brand name (pronounced RYE-zen).

AMD didn’t just unveil the branding — it gave some additional details on the CPU’s minimum clock frequency, total L2/L3 cache, and new capabilities as well.

When AMD tested Zen against the Core i7-6900K back in August, it locked both chips to 3GHz. This raised concerns that AMD might run at a significantly lower clock than Intel does, further harming the company’s ability to compete effectively. We now know that’s not going to happen — AMD isn’t revealing exact SKUs yet, or even its absolute minimum base clock, but the company has confirmed that it will offer an eight-core chip with at least a 3.4GHz base clock. The 20MB of cache doesn’t specify between L2 and L3, but that’s easy to break down — we’ve already been told that each CPU complex (CCX) consists of four CPU cores backed by 8MB of L3 cache. Two CPU complexes = eight cores and 16MB of shared L3, with 512KB of L2 for each CPU core. Each L3 cache appears to be unique to its core complex; it’s not clear what the penalty hit is for retrieving data stored in a different CCX’s L3.

There’s a new platform update coming as well, with support for PCIe 3.0 and USB 3.1 Gen 2. AMD has been contracting with Asmedia for its southbridge designs and we expect that will continue. AMD continues to insist that Zen represents a 40% improvement in IPC over Excavator, but it still hasn’t shared details on how it arrives at that calculation, or whether that includes the performance boost from its new symmetric multithreading technology SMT.

The following slideshow steps through some of the new features AMD announced in Sonoma and their impact on performance and power consumption. AMD is collectively referring to these technologies as SenseMI, where the MI refers to “Machine Intelligence.” It’s a clever bit of branding meant to capitalize on the recent popularity of AI and deep learning research. All slides can be clicked to open and enlarge them in a new window.

AMD is still being coy about launch dates, but Lisa Su has publicly stated that we’ll see Zen in Q1 2017. I expect the company to hold that date, but if AMD was launching Ryzen at CES it probably would’ve already said so. Since there’s always a few weeks of post-CES decompression, this puts a Zen launch somewhere between late January and late March. The 3.4GHz core clock for an eight-core chip is also good to see — it compares well against Intel’s overall product lineup, and it gives AMD’s own eight-cores an important boost against previous Piledriver chips. That might not seem important, given how poorly regarded Piledriver is, but it’s a chip with a base clock of 4.7GHz at the top end of AMD’s product stack. That’s 38% higher than Ryzen’s now-specified 3.4GHz minimum base clock, and 56% higher than the 3GHz clock AMD was showing in August. We always expected Ryzen would beat Piledriver, but it was never clear by how much once clockspeed differences were accounted for. A 3.4GHz base clock should give AMD more breathing room.

The following slideshow is from one of our previous deep-dives on Zen, but will give you more details on the architecture if you’re playing catch-up. Our overall opinion on the chip’s likely competitive positioning and our thoughts on positioning vs. Intel are available here.

It’s no exaggeration to say the entire future of AMD hangs on this launch. AMD’s primary businesses have effectively collapsed, and the company’s CPU and graphics revenue is 27% of what it was in Q3 2011. Zen represents an opportunity to reinvigorate the companies server and consumer fortunes. Historically, these markets were far more important to AMD’s bottom line than GPU sales, even though we’re also hoping that Vega has what it takes to offer Nvidia some competition in the mainstream GPU and even HPC markets. Enthusiasts have been waiting a long time for an AMD CPU that could compete with Intel’s Core i3 / i5 / i7 families — hopefully in a few months more they’ll have one.