AMD is at CES 2018 in Las Vegas, and the company took the opportunity to reveal their product road map for the next year. At CES 2017, AMD delivered one of the most prophetic presentations in recent memory, nailing almost every goal they set for themselves that year (except Vega, which is still a mess). 2018 is even busier for the company, and it is now under pressure to perform just like it did last year, delivering as many hits as possible to bring the attention back to them.

Desktop and Mobile


We start with the overall roadmap for Ryzen. 2018 is going to see some punches being thrown by AMD early on, starting tomorrow, 9 January, with the launch of Ryzen 3 processors with Vega Graphics for mobiles. That’s a big win for AMD in the cheaper segments of the notebook market, and those Vega APUs will give Intel and NVIDIA a reason to worry about their partnerships and mind share in these areas. For years this space has floundered, occasionally served by the odd NVIDIA chip here and there, or AMD’s older APUs. This is a much needed boost.

Straight afterwards is Ryzen desktop APUs on 12 February 2018. Up to now, AMD’s socket AM4 played host to Ryzen processors without any internal graphics, as well as the old A-series of APUs using Bristol Ridge-based processors and graphics from GCN 1.3. Those get chucked out the window now for Ryzen desktop APUs. These are on the same 14nm process as current products, so there’s no chance that these will overclock higher. These APUs match the capabilities of their mobile counterparts, and offer better performance with higher power draw. You can see the above sneak peak in select benchmarks by AMD, showing how much further these APUs can get you on integrated graphics than Intel’s chips can.

Q2 2018 is the interesting bit for most enthusiasts. AMD is targeting April 2018 for the launch of the refresh of their second-generation Ryzen processors, based on the revised 12nm LP FinFET process which nets them a 10% boost in efficiency. How AMD chooses to play this is anyone’s guess. They could take the 10% improvement in the form of an overclock, boosting every chip by 10% at the same power draw, or reduce heat levels and aim for better efficiency. For mobile form factors, I would have expected the latter, but for the desktop market I think AMD will pick up that performance boost. With Intel already suffering from bad press as a result of the Meltdown memory bug, this is an opportune time to have AMD shouting that their products are technically faster and not vulnerable to Meltdown. Midway through Q2 2018 we’ll get Ryzen Pro with Vega Graphics processors, which was expected.

The second half of the year isn’t as crowded. AMD will space out the launch for their second-generation Ryzen Pro processors as well as the Threadripper refresh, both on 12nm LP FinFET technology. With all these improvements, AMD is also adding in a new generation of motherboards based on socket AM4 using the X470 chipset. These boards are forwards and backwards compatible with AMD’s socket AM4 products, including Bristol Ridge APUs.

That’s not all, though. AMD also just announced price cuts across the board for all their existing Ryzen products, with some chips seeing as much as a 30% discount. Beating down their competition isn’t AMD’s style usually, but the price cuts, along with the Meltdown PR fiasco over at Intel, allowed this perfect storm to come into being. If you were on the fence about Intel’s latest platforms versus AMD’s offerings, the red team just got significantly cheaper and more attractive.

Graphics

On the graphics side, this year things are muted. There are no real improvements in the pipeline for the discrete GPU market that AMD is willing to share yet, save for two announcements. The first is the introduction of an optimised version of Vega for AMD’s Radeon Instinct product family. This will be fabbed on an IBM-licensed 7nm production process with FinFET technology, and its only job is to accelerate machine learning workloads. AMD has Vega-based Instinct GPUs already available, but this will be the first focused product for this application, like a NVIDIA Tesla GPU which is similarly purpose-built.

The telling part is the use of the word “sampling” in 2018. The 7nm process is nowhere near complete, and AMD is reserving it for Zen 2 and Navi, their next-generation products. The use of “sampling” here is to dampen any possible rumours that AMD will be launching a new family of graphics cards this year, especially any products on a new production process. The Vega launch in 2017 was a controlled failure from start to finish because of a number of issues AMD ran into, and it cost them a lot of brownie points with their customers when Vega wasn’t delivered on time with all its features enabled. This time, AMD is being as vague as possible.

As if things weren’t vague enough already, we also had an announcement from AMD about a mobile version of Vega packed with HBM2 memory, available in a separate form factor to be included in mobile designs of different shapes and sizes. We never got this with Fiji previously, even though it was completely possible and probably on the cards before something changed. But soon, perhaps quite soon, we’ll have gaming laptops with Vega graphics, completely separate from whatever the platform is and unrelated to the Intel and AMD partnership.

AMD also isn’t talking clock speeds, hardware counts, or even pricing. It won’t even say which partners are on board yet. Hmmm. It’s almost as if one party ended up buying out enough stock for their space grey fruity-themed machine for the next few months that it drained AMD’s ability to deliver the same solution to consumers.

Also, here’s a sneak peak at AMD’s redesign of their premium Wraith cooler. These will be shipping separately for $50 sometime later this year, and replace the existing Wraith coolers that also have RGB LEDs, but don’t have the translucent fan or the extra colour zones this one does. The Wraith is about as effective as a Cooler Master Hyper 212 Evo and costs more than the Evo, but it just looks too damn cool.

More stuff like this: