In the CPU space, AMD is firing on all cylinders currently, having reasserted itself into the performance sector with its mainstream Ryzen processors and high-end desktop (HEDT) Threadripper chips. Both are now in their second generation. So, what’s next? AMD is transitioning to a 7 nanometer manufacturing process, for both a new generation of CPUs and GPUs.
We have known this for a long time, of course, but what is interesting about this shift is that it is taking place while Intel is still trying to get its 10nm chips out the door, in volume. Cannon Lake has seen numerous delays, though according to Intel, everything is now on track. We will have to wait and see if that is actually the case.
Meanwhile, AMD is forging ahead and is banking on its 7nm chips paying big dividends by keeping the momentum going. AMD is not just hopeful of this, it needs 7nm to work out as anticipated.
“We knew 7nm would be a big challenge, so we made the bet, we shifted our resources onto the new node,” AMD CTO Mark Papermaster told CRN in an interview. “We didn’t just dip our toe in the water. We went all in.”
That “all-in” approach seems to be working, at least as it pertains to AMD’s roadmap. According to Papermaster, the Sunnyvale chip designer originally intended to release 7nm Vega GPUs in 2019. However, they are now scheduled to arrive sometime this year with the release of AMD’s upcoming Radeon Vega Instinct GPU. AMD’s “immense focus” on 7nm is what allowed the company to shorten the timeline, Papermaster said.
In addition, AMD is readying 7nm Epyc server chips for this year, followed by a 7nm Ryzen refresh sometime after. AMD sees a window of opportunity to roll out products on advanced manufacturing processes, while rival Intel continues to tweak its 10nm process node.
“I think we do have the opportunity to be positioned much stronger than we originally anticipated,” Papermaster added. “But I have to say our original plan was to be positioned very strongly, so any delay from our competitor could simply strengthen the value that AMD brings to the market.”
It takes a lot of planning and coordination to get where AMD is currently at. In going all-in with 7nm, the company had to make sure its design team, foundries, and fabrication partners were on the same schedule. Juggling the latter is one of the more difficult tasks with this sort of thing, because it means widespread changes to even the automation tools that they use. Unlike Intel, AMD is a fabless company.
“To really leverage what it takes to be successful in a semiconductor foundry industry, it needs 100 percent focus,” Papermaster said.
While it may sound like a gamble for AMD to put all of its eggs into a 7nm basket, it might be more out of necessity than choice. Intel contents that its 10nm process is more advanced than the competition, as it is packing more transistors and is doing a better job with scaling. In fact, Intel has even stated a desire to introduce a new metric to more accurately reflect where chip design technologies are really at.
“What is really needed is an absolute measure of transistors in a given area (per mm2). At the other extreme, simply taking the total transistor count of a chip and dividing by its area is not meaningful because of the large number of design decisions that can affect it—factors such as cache sizes and performance targets can cause great variations in this value. It’s time to resurrect a metric that was used in the past but fell out of favor several nodes ago. It is based on the transistor density of standard logic cells and includes weighting factors that account for typical designs,” Intel said last year.
No matter how you shake (or measure) it, though, 2019 is shaping up to be a very interesting year in CPU design and technology in general.