I bet you weren’t expecting this article to start with that word, but there’s a good reason - here's a fun fact for you. AMD’s Ryzen is actually named after NASA’s little New Horizons spacecraft, which zipped by Pluto not long ago, giving us those magnificent pictures of the planet that turned into a dwarf after Horizon had launched. AMD wanted to originally use the architectural codename “Zen,” but legalities intervened preventing this, so - inspired by Horizons - AMD tweaked the name a bit to keep the zen, leading to “Horizen”, which gave way to “Rizen” - which the company liked - but to avoid confusion with the word “risen,” it added the Y, giving us the Ryzen name we know today. Neat eh?
Alright, onto the juicy stuff.
AMD has been handing out NDA-ed slides to the press, giving some more detailed information about the Ryzen CPUs and what makes them tick. Today that NDA lifts so we’d like to go over some of the details included within these slides - and some of what’s not.
First things first, a recap of the upcoming Ryzen CPU models:
- Ryzen 7 1800X – 8C/16T, 16MB L3, 95W TDP, 3.6GHz base - 4.0GHz turbo, US$499
- Ryzen 7 1700X - 8C/16T, 16MB L3, 95W TDP, 3.4GHz base - 3.8GHz turbo, US$389
- Ryzen 7 1700 - 8C/16T, 16MB L3, 65W TDP, 3.0GHz base - 3.7GHz turbo, US$319
- Ryzen 5 1600X - 6C/12T, 16MB L3, 95W TDP, 3.3GHz base - 3.7GHz turbo, US$259
- Ryzen 5 1500 - 6C/12T, 16MB L3, 65W TDP, 3.2GHz base - 3.4GHz turbo, US$229
- Ryzen 5 1400X - 4C/8T, 8MB L3, 65W TDP, 3.5GHz base - 3.9GHz turbo, US$199
- Ryzen 5 1300 - 4C/8T, 8MB L3, 65W TDP, 3.3GHz base - 3.6GHz turbo, US$175
- Ryzen 3 1200X - 4C/4T, 8MB L3, 65W TDP, 3.4GHz base - 3.8GHz turbo, US$149
- Ryzen 3 1100 - 4C/4T, 8MB L3, 65W TDP, 3.2GHz base - 3.5GHz turbo, US$129
Now of those, only the top three are launching immediately. The 1800X, 1700X and 1700 can all be pre-ordered right now for $699, $569 and $469 respectively in Australia. That’s still a lot of money for a CPU but when you consider that the $700 1800X is going toe-to-toe with Intel’s i7 6900K, which currently sells for over double the price at $1500, you'll start to appreciate the magnitude of the new price/performance curve AMD is creating here.
Even the 1700 (non-X) model, at $469, is $15 less than Intel’s latest i7 7700K, a CPU with half the core and thread count! Save some money and double your core count? Mind, blown. We’ve waited a long time for this and AMD is so far delivering in spades, exposing Intel’s price gouging for what it is, and showing why many enthusiasts and industry professionals wanted AMD to do well with Ryzen. Lack of competition leads to monopoly behaviours, like charging whatever you like.
Over $2300 for the ten core CPU? Ha! No thanks!
From very early on AMD was up front in stating it wanted the Zen architecture to obtain a 40% uplift in instructions per clock (IPC) performance over its latest Bulldozer iteration “Excavator”. Since launching, AMD announced it actually exceeded this target and achieved a greater than 52% increase; which when you think about Intel’s typical 5-10% increase each generation, AMD has pulled off something special. Mind you it has taken them far too long to get here, but that’s another story.
Even with this IPC boost it only just pulls AMD up to par with Intel’s latest, highlighting just how far behind the performance curve AMD had become. AMD is claiming IPC in Ryzen, in single threaded tasks, is just 2.5% below Broadwell and 6.5% below Kaby Lake. However, looking at AMD’s own benchmarks, and those that have leaked around the internet, once multithreading comes into play AMD easily makes up and even exceeds this slight IPC gap. This is something to keep an eye on and learn more about down the track, as AMD appears to have implemented Simultaneous Multithreading (SMT, or Hyperthreading in Intel parlance) better.
IPC is one thing; there’s no point doing huge amounts of instructions each clock cycle if you’re clock speed is no faster than your Zimmer-frame-using-granny at walking pace. And that is where Intel still has the advantage. Where Ryzen still lives in the 3GHz range, Intel has products that have base clock speeds starting in the 4GHz range, before you even talk about turbo speeds. Against Ryzen’s main competitor, Broadwell-E, that’s not a problem, but facing off against the likes of the 7700K, single- or lightly-threaded tasks will easily make use of those extra MHz.
Another exciting part about Ryzen is the power consumption and, by extension, boost clocks. AMD’s SenseMI technology employs the use of over 100 voltage, amperage and temperature sensors embedded directly into the CPU die to help AMD shut off various areas of the CPU as and when they’re needed. This tech is being dubbed Pure Power by AMD, and is part (but certainly not all) of the reason why AMD’s Ryzen 7 1700, an 8-core 16-thread part, has a TDP of just 65W compared to Intel's beastly i7 7700K 4-core 8-threaded CPU, which needs 91W!
Using this sensory platform AMD has also implemented its Precision Boost tech, AMD’s equivalent to Intel’s Turbo Boost. Offering clock speed increments as small as 25MHz at a thousand times a second, Ryzen can increase core clocks (so far) upwards of 700MHz, and of course all the way down to idle speeds.
Those things are all well and good but it basically only matches Intel’s current variable clock speed feature-set; however Ryzen has an extra ace up its sleeve. To understand it best, we oddly need to look to Nvidia as a reference.
XFR (eXtended Frequency Range) is a technology AMD is using to raise CPU clock speeds even higher than the boost clocks, if there is sufficient cooling headroom available. Sound familiar? It should, because Nvidia’s Boost 3.0 on the likes of the GTX 1080 already does a similar thing, wherein you have a base clock and a higher turbo clock speed, but depending on loads, temperatures and power usage, Nvidia’s Boost 3.0 can push core clocks even higher than what they’re rated at out of the factory.
XFR is essentially AMD’s version of this for CPUs. At this stage, Ryzen’s implementation of this doesn’t seem to offer as much of a boost as Nvidia’s Boost 3.0 does, but it’s still early days. Disappointingly we’ve only seen at best an extra 100MHz increase when XFR kicks in, but how far it’ll go in the real world, with some hefty cooling applied, will have to wait until reviewers get their hands on actual samples.
Interestingly, is seems the models with the X on the end of their names won’t be the only ones that have XFR support (as was previously thought). The above slide shows that models with the X simply have a doubling of how much more XFR will boost clock speeds, should conditions permit.
It seems Ryzen is AMD’s equivalent of Intel’s HEDT CPU lineup. Lots of cores but no integrated GPU. Enthusiasts won’t care about such things as they’ll be sticking in their own dedicated GPUs anyway, but for mainstream users this will be a sticking point.
However, it looks like the venerable Athlon brand name will live on in some form again when AMD bring its Zen-powered APUs online in the second half of this year. Speaking of release dates of other AMD products, Zen-powered mobile products will come to market in the second half of this year too. Sooner than that, in the second quarter of this year, we’ll get Vega (AMD’s GTX 1080 competitor) and Naples, AMD’s upcoming Ryzen-based Xeon competitor for servers (32 cores anyone?).
Motherboard chipsets have received a much-needed spruce up. AMD has segmented them much like, and against, Intel’s mainstream lineup with the X370, B350 and A320 facing off against Intel’s Z270, B250 and H210 boards respectively.
As an interesting side note, I wonder if Intel’s next-gen motherboard range will move into the 300 numbers, seeing as AMD has pulled a quick one and started its platform in that number range. B350 motherboards could be a thing from both vendors otherwise. Sly AMD, very sly.
Native USB 3.1 Gen2 is a big selling point for AMD’s motherboards. Even Intel’s latest Z270 requires an ASMedia USB hub chip to provide this capability.
Somewhere Ryzen apparently can’t compete directly against Intel is on the PCI-E lane-count front. Ryzen has 24 PCI-E 3.0 lanes that come directly from the CPU, which is indeed more than Intel’s mainstream i7 7700K for example, with 16 lanes, and far less than Intel’s HEDT line-up that sport some 40 lanes. The trick, however, is that Intel has its own proprietary chipset interconnect which uses DMI for communications, meaning the CPU’s PCI-E lanes don’t get touched.
On Ryzen however, those 24 lanes are used by the entire system. So, 16x lanes is reserved for use with graphics cards (yes that means only two 8x is available in dual-card setups), 4x is reserved for or given priority to NVMe applications (ie M.2 slots), and the remaining 4x lanes are used to connect the X370 chipset to the CPU, leaving a serious bandwidth gap compared to the Z270.
Having a quick look at two high-end Asus motherboards for the two chipset types, the Z270 effectively makes 22x PCI-E 3.0 lanes available to the user, whereas the X370 only gives 16x PCI-E 3.0 lanes, and around 7x PCI-E 2.0 lanes (as a reminder, 2.0 is half the speed of a 3.0 lane). Quite a difference.
On top of that the X370, when a 4x NVMe device is plugged in, only supports four SATA drives, and just two on the B and A series chipsets!
It seems AMD will severely need to increase PCI-E lane counts next year, or whenever the next Ryzen iteration comes, if it wants to remain competitive on the connectivity front. Z270 beats it handily and X99 is simply just laughing at this point.
Something to note, SLI and Crossfire will only be supported on the top tier X370 chipset.
One of the saving graces of AMD's new platform is that overclocking is available on the entire Ryzen CPU range, top to bottom. The only limiter is which motherboard chipset you use, with the lowest end A320 and A300 chipsets disallowing CPU overclocking. So, avoid the A-series motherboards and get yourself a B- or X-series chipset and you’ll be ready to go.
It seems AMD’s overclocking will be getting the Radeon Settings graphics driver style software treatment, too, with official Ryzen overclocking software to be available, meaning no reliance on sometimes sketchy motherboard manufacturer overclocking software.
Something to keep an eye on is memory support. While Ryzen supports up to 64GB in dual-channel mode, like Kaby Lake does, at least in this slide there could be a speed problem when you actually do fully populate the RAM slots, with four dual-rank DIMMs slowing down to just 1866MHz. Two single-rank DIMMs, however, exceed Intel’s latest supported RAM speeds at 2666MHz, compared to 2400MHz; however Intel motherboards support XMP profiles well into the 3GHz range. We’ll have to wait and see if AMD’s boards will be able to match that.
AMD’s revamped stock air cooler, dubbed the Wraith, has had a facelift, with a smaller version created for the less demanding processors in the line-up. The two higher end coolers have also had a lick of RGB included in them too.
Not all CPUs will come with a cooler though. While the 1800X and 1700X can come with the Wraith Max cooler (rated for 140W), there will also be versions of those CPUs sold without the cooler, which should be around $30 cheaper. The 1700 gets the Wraith Spire cooler (rated for 95W).
Aside from those three CPUs it’s unclear at this stage which models below those will get the Spire or Stealth coolers included. A safe bet would be that all quad core Ryzen CPUs (1400X and down) will have the Stealth cooler.
Ryzen is certainly shaping up to be the disruptive force we're all hop[ing it can be. All of AMD’s channel partners are gearing up to strongly support AM4 and Ryzen. Now, all we need to do is get our hands on some products and get testing! Something we aim to do ASAP.