As a jaded AMD fanboy, I’m not only getting on the hype train for their Ryzen CPUs, I’m the conductor taking tickets from everyone else. I will shovel hype coal into the hype train, and send it hurtling down the hype tracks with all available hype speed. For once in the last ten years or so, it looks like AMD is going to be able to toe-to-toe with the best chips Intel has to offer, and might even come out on top.
Is the Ryzen architecture a breakthrough for AMD? Early signs certainly point that way. But some context is important.
First, their previous Bulldozer microarchitecture was basically a dog from the time it was released in 2011. AMD did what it’s done since Intel released the first Core 2 Duo chips: competed on price, superior integrated graphics, and make their meager margins on the bottom end of the market. This is their third architecture since the successful K8 to try and top Intel.
Secondly, Intel has basically been standing still on overall per-clock performance for a few years. This isn’t to diminish what Intel has done to optimize for better power consumption, die shrinks, and allow for higher overall clocks. There’s a reason they’ve been the undisputed performance leader for a decade. But AMD has had a few years to hit the same target.
The last time AMD had a performance edge, Intel used marketing and some legally “aggressive” rebate programs to hold AMD from gaining too much market share. We’ll see how AMD can do this time around. Given that the market for high-end desktop performance is shrinking, we’ll see if they try to make a data center play. The real success of this architecture will be if AMD can see some adoption in the laptop market, where it has never had a compelling product.
During AMD’s “New Horizons” live stream, which served as a coming out party for the Ryzen CPU, the company showcased a brand new performance demo. It pitched a stock Intel Core i7 6900K processor—an eight-core Broadwell-E chip that retails for just shy of £1000/$1100—against an eight-core Ryzen CPU. Both chips were tasked with transcoding a short video clip into the “Apple TV 3” preset in Handbrake, a notoriously CPU-heavy workload that scales well across several cores. The result, according to AMD’s demo, was a completion time of 54 seconds for Ryzen and 59 seconds for Broadwell-E; Ryzen was about 10 percent faster.
Rich has been a tech enthusiast since he first used the speech simulator on a Magnavox Odyssey². Current areas of interest include ZFS, the false hopes of memristors, and the oral history of Transmeta.
We need more competition in the CPU space, I think. This is great news. With the PowerPC/nvlink stuff coming out too, I hope we see a new wave of innovation.
We need more competition in the CPU space, I think. This is great news. With the PowerPC/nvlink stuff coming out too, I hope we see a new wave of innovation.