AMD just gave a few more details, juicy details at that, about their upcoming Zen processor and it’s AM4 platform. The new brand, Ryzen, will be the name that the whole package goes by when it releases. Release is still somewhat confirmed for Q1 2017, meaning sometime after or around CES 2017. The entire platform is being marketed, and rightfully so, at the many content creators, streamers and gamers who need the power not just for gaming, but for rendering and producing content.
The New Horizon looks good from AMD’s point of view
Once released we’ll see a whole slew of different SKUs covering many different price ranges. AMD is exited about the prospects of their flagship, the 8-core 16-thread monster. This high-end SKU is rumored to cost much less than Intel’s top offering while performing very similar. It’s being positioned as a much better value with everything that Broadwell-E has plus even more future-proofing. AMD is saying that it has the “ultimate upgradability”. Though we weren’t aware of the specifics, the demonstration today used Handbrake to transcode a video. The multi-threaded test was done with an 8-core Zen against an Intel i7-6900K, the comparable SKU. Not only did Zen finish the transcoding task faster, it also did so using less power. The last demo we saw had the 6900K going against Zen in a Blender rendering test, of which they were very close. We were promised near Skylake IPC, and it seems that AMD is meeting those targets. Those these are just two tests. Though two that generally have favored Intel in the past.
In terms of the new AM4 platform, we’ve learned a great deal about what it’ll offer. Of course the chipset will support USB 3.1 Generation 2, NVMe SSD’s, SATA Express and likely Thunderbolt 3 natively. And it’ll have a wealth of PCIe lanes with which to use for various GPUs and other peripherals. Server Zen will have 64 PCIe lanes per CPU, so we’ll likely see a bit less then that, however. That said, AMD is indeed looking at the new horizon to ensure that AM4 is a stable upgrade path for the foreseeable future.
One of the better points that AMD was quick to point out today was their power saving features. Zen is going to be the most efficient CPU they’ve ever created. And they’re using Intel’s Skylake as a benchmark. SenseMI is the new technology from AMD that has 5-stages that allows the processor to be as efficient as possible in every situation. Zen will be able to scale in realtime with this 5-stage network. SenseMI will work in concert with embedded sensors that measure current, voltage, temperature and speed in order to make changes in real-time. They use what’s known as Infinity Fabric in order to control everything. There’ll be an integrated Infinity System Management Unit that’s controlling everything behind the scenes. There are five stages that control
The first such stage with SenseMI is known as Pure Power, which does precisely as the name implies. It allows the processor to run with the most efficiency possible. There will be optimizations that allow for the lowest power usage possible for a given task. This controls the tangible and physical aspects of the processor
Precision Boost is the next stage within SenseMI. This portion controls the clock speed itself. It can tune using 25MHz increments and can do so on the fly and without having to finish work in the queue. The core frequency adjusts to whatever requirements there are. Need more? It can reasonably adjust the speed within a given power envelope. Why 25MHz and not the usual 100MHz increments that current CPUs have? Good question. But it could imply that the base frequency is now 25MHz on Zen.
The third stage of SenseMI is called Extended Frequency Range. This will allow those with better than usual cooling to increase their clock speed well above the usual precision boost limits. Zen can expand with small increments to whatever limit you happen to have. It sort of takes away those limits and gives enthusiasts the chance to do amazing things. Or blow things up, too. It almost sounds as if there’s some sort of automatic overclocking utility within the CPU itself. Overclocking with no effort required. Which isn’t necessarily bad, but does take the fun out of it.
The next stage of SenseMI is a revolutionary new technology. They’ve added a DNN, of sorts, into the CPU that can do better, more efficient predictions than a normal pre-fetch algorithm. Their Neural Network Prediction and Smart Prefetch are a model that should be able to speed up all manner of operations simply by learning what you execute the most. The general feeling on the desktop when launching or conducting business will feel as if it’s much more swift. Theoretically this neural network will attempt to accelerate any future instruction fetching to improve your experience. Prefetch, too, is much improved. You’ll see a massive speed increase at it anticipates more intelligently and makes better use of the L3 victim cache.
Zen is getting more and more exciting, especially since their tests do, despite being very minutely controlled and first-party only, confirm that it can perform some tasks as well as Broadwell-E but with less power consumption. They’re accomplishing nearly all the objectives they’ve discussed and it seems that they’re still on target for their Q1 launch. For consumer Zen, they’re saying this is a gaming CPU, one made for the gamer and creator out there. There are nearly 1.7 million streamers on Twitch alone, and they’re marking this as a very suitable CPU for that crowd. And for anyone really. Zen is looking good and we can’t wait to actually get to take it through a thorough spin that explores the depths of it’s performance. The new horizon is definitely looking quite exciting.