Over the past four months, Intel achieved a nearly 2x improvement in the rate at which 10nm products moved through their factories. Later during the call, Swan noted that the improvement in 10nm yields was better than expected, which prompted Intel to increase its outlook for 10nm volume for 2019.
Client systems with 10nm processors are expected to be on retail shelves for the holiday season. Swan did not comment on the recently leaked roadmap, which suggests a large part of Intel's product mix will remain on 14nm up until at least 2022. According to that leak, mobile will move to 10nm first, with desktop processors remaining largely on 14nm until 2022. Ice Lake-U is expected to be the first 10nm mass market product from Intel, while the desktop market will most likely get another "new and improved" 14nm respin of Skylake.
Swan revealed that 10nm server products will come to market a bit earlier than the historical 12 to 18 month gap between client and server. The 10nm Ice Lake-SP based Xeon processors are slated for 2020, with a release rather earlier than later in the year.
First, on 10-nanometer for server, what we've indicated is that we would have client systems on shelf for the holiday season, and our expectation is that server CPUs would be a fast follow. Historically, I think it's been more of a 12- to 18-month gap between client and then server. On 10-nanometer, that gap would be much shorter. And again, what we said is fast follow after client systems on shelf, so sometime in 2020, earlier versus later. In terms of just competitive positioning, with the -- we've been -- our Xeon Scalable product that we launched last year and then enhanced with our Cascade Lake launch just a few weeks ago I think, that has increased performance. It's got AI acceleration built into it. It's coupled with Optane memory. It's got a 56 core count. And the performance of that product, coupled with our knowledge of the environment in which -- our customers' environments, we think we really have demonstrated differentiated performance, a product leadership performance, even though it is still on 14-nanometer.
So we have a good product that really pulls together those 6 pillars that I talked about earlier. We launched it a few weeks ago. We said we expect it to ramp as fast as any launch that we've done in the past. And we think it positions us competitively in the second half of the year, despite increased competition prior to launching 10-nanometer in 2020.
So we feel good about 10-nanometer in general. We're going to be a fast follow with server in 2020. In the meantime, the Cascade Lake product has some real performance enhancements based on our deep domain knowledge of our customers. And it'll be an increased competitive environment, but we feel pretty good and we tried the best we could to capture that in our outlook for 2019.