Ryzen coming in 2017

AMD reveals the brand for Zen, a clock speed of 3.4 GHz and some new technical features.

As much as we might want it to be, today is not the day that AMD launches its new Zen processors to the world. We’ve been teased with it for years now, with trickles of information at event after event…but we are going to have to wait a little bit longer with one more tease at least. Today’s AMD is announcing the official branding of the consumer processors based on Zen, previously code named Summit Ridge, along with a clock speed data point and a preview of five technology that will help it be competitive with the Intel Core lineup.

The future consumer desktop processor from AMD will now officially be known as Ryzen. That’s pronounced “RISE-IN” not “RIS-IN”, just so we are all on the same page. CEO Lisa Su was on stage during the reveal at a media event last week and claimed that while media, fans and AMD fell in love with the Zen name, it needed a differentiation from the architecture itself. The name is solid – not earth shattering though I foresee a long life of mispronunciation ahead of it.

Now that we have the official branding behind us, let’s get to the rest of the disclosed information we can reveal today.

We already knew that Summit Ridge would ship with an 8 core, 16 thread version (with lower core counts at lower prices very likely) but now we know a frequency and a cache size. AMD tells us that there will be a processor (the flagship) that will have a base clock of 3.4 GHz with boost clocks above that. How much above that is still a mystery – AMD is likely still tweaking its implementation of boost to get as much performance as possible for launch. This should help put those clock speed rumors to rest for now.

The 20MB of cache matches the Core i7-6900K, though obviously with some dramatic architecture differences between Broadwell and Zen, the effect and utilization of that cache will be interesting measure next year.

We already knew that Ryzen will be utilizing the AM4 platform, but it’s nice to see it reiterated a modern feature set and expandability. DDR4 memory, PCI Express Gen3, native USB 3.1 and NVMe support – there are all necessary building blocks for a modern consumer and enthusiast PC. We still should see how many of these ports the chipset offers and how aggressive motherboard companies like ASUS, MSI and Gigabyte are in their designs. I am hoping there are as many options as would see for an X99/Z170 platform, including budget boards in the $100 space as well as “anything and everything” options for those types of buyers that want to adopt AMD’s new CPU.

What about this SenseMI Technology? This is the brand that AMD has assigned to a collection of five different features that improve the performance and efficiency of the Zen architecture as it exists in the Ryzen consumer product. I’ll be upfront – I don’t know if these items needed an encompassing brand umbrella though the features themselves are worth noting.

First up is Pure Power, an integration of a temperature and voltage monitoring system that helps optimize the clock speeds and voltages the CPU will run at in real world conditions. By keeping track of die temperatures and voltages running through it, Zen will utilize adaptive control schemes to lower power usages while maintaining performance and clock speeds at expected levels. There is a lot of jargon and terms on this slide above including “infinity control fabric” and “distributed embedded sensors” but this is technology we have seen before in Intel’s processor systems, both major vendors GPUs and in mobile devices. It helps keep performance levels at a static state but improves efficiency to keep power and thermals at lower levels.

Precision Boost is the other side of the same coin, using the same sensors and the same logic but for a different purpose. This is AMD’s answer to Intel Turbo Boost technology, dynamically adjusting the clock speed of the processor cores to improve performance for as long as the die can remain within the preset thermal constraints AMD (or a user) puts on it. We don’t have much detail on this feature yet but I would assume it would behave very similarly to Intel’s version – running at higher clocks during lower threaded workloads but lower clocks at highly threaded workloads. One interesting improvement with AMD Precision Boost is that it can operate in 25 MHz increments. Intel Turbo Boost only moves in steps of 100 MHz. While this will add more complexity to the system, in theory it should allow AMD to stretch that last little bit higher to get additional perf.

Extended Frequency Range has the potential to be very interesting. If you have better cooling, your processor will be able to run at higher clock speeds. AMD says this will scale with air cooling, water cooling and even liquid nitrogen! NVIDIA and Intel have both claimed that this capability exists for their products (NVIDIA GPU Boost), but in practice, we usually find that process technology creates a different limit than thermals for processor performance. For example, on NVIDIA GPUs, artificial voltage limits keep the GPU clock down well before thermals when running extreme cooling solutions.

The amount that XFR impacts consumers will be measured once we have samples in hand. If the difference between a basic air cooler and water cooling is ~100 MHz of usable, repeatable clock speed, then the value will be…debatable. If we can get an extra 300-400 MHz by just utilizing high end enthusiast coolers, that’s a great feature to add.

Now things getting a little more interesting. What AMD is calling Neural Net Prediction, and what they dub as a “artificial network”, offers improved prediction instructions and paths through the processor pipeline. If that sounds familiar to you, and it should, you’ll recognize this as a branch predictor.

The goals put forward on this slide exactly describe the goals of prediction units on a processor, and have for as long as I have been reading architecture documents. “Building a model of decisions driven by software code execution” is an interesting way of describe using the history of previous code results to attempt to predict the next code result early. A prefetch can grab the next instruction ahead of time and, if it predicted correctly, is ready to execute immediately without time decoding.

The description of this as a “neural net” is just one of semantics. A quick look over at Wikipedia shows a variation of the predictor called a Neural Branch Predictor. The differentiator between this and a standard unit is that a neural unit can exploit long histories of previous data with only a linear growth in required resources. That comes at the cost of added latency though. When I have more time with AMD’s engineers we can get a better explanation of what is going on, but for now, just understand that Zen has an improved branch and path prediction unit to improve performance and power efficiency.

Smart Prefetch has a similar setup. A prefetch unit attempts to guess which data the application will need from cache before the request is made, improving performance by reducing wait times for memory access. The same argument can be made as to if this constitutes a neural net, or smart network or whatever – the important aspect is that AMD claims the improvements in this area of the architecture result in performance gains for IPC, helping to add up to that impressive 40% improvement over the last generation.

More Performance Previews

Back in August, AMD held a quick press conference in San Francisco to show off the first official performance metric for Zen. It compared an 8-core Zen processor against an 8-core Broadwell-E, both running at 3.0 GHz. The Zen CPU finished a run through Blender first, indicating that the IPC of Zen was higher or on-par with that of Broadwell-E.

For today’s Zen information blast AMD had another pair of data points that worth discussion. First, AMD pitted a Ryzen processor running at a fixed clock speed of 3.4 GHz, with no boost features enabled yet, against a stock Core i7-6900K, with boost enabled. For your reference, a 6900K is an 8C/16T part that runs at a base clock of 3.2 GHz and can clock as high as 3.7 GHz with Turbo Boost. When operating at full thread capacity, the 6900K is in the 3.4 GHz to 3.5 GHz range. The benchmark was Handbrake, a very common piece of software that most review sites use for testing and that many consumers use for media transcoding.

While not the most riveting video coverage, this will show the transcode running on both systems, fully utilizing 16 threads. The AMD processor finished a few seconds ahead of the Intel part, again indicating that it is leading or competitive against Intel’s current enthusiast class processors.

A power demonstration was run, though this time with Blender, showing total system power draw during a render.

While we don’t know the rest of the components on the system, and that this is total system, not CPU only power draw, so we must be careful about making definitive conclusions. The AMD Zen PC was running at ~188 watts, the Intel Core i7-6900K machine at ~191 watts. Even if you give the platforms +/- 15 watts for variances in motherboards and the like, that means AMD is within range of Broadwell-E.

More to come

Now we know the brand, Ryzen. Now we know the clock speed, a minimum of 3.4 GHz on a flagship part. We know something new about the architectural improvements for power, speed and efficiency. We do not know the release date (1H 2017) or pricing. And we don’t know the full story in terms of performance and where Ryzen will compete against an Intel lineup that will be shifting in the next 2-6 months substantially.

I’m more excited for AMD new processor than I was in August when they first show a performance demo that was very quickly followed by doubt and scrutiny on the workload. One more benchmark does not a CPU story make, but it’s clear that AMD remains committed to its stance on processor IPC and competitiveness in the market. Those of us that have been waiting for true competition to Intel’s Core i5/i7 processors from AMD, without the overarching caveats, will not have to hold out much longer. If you were ever doubting that 2017 was going to be a pivotal year for computing and PC gaming, this is yet another slap in the head with a large trout.