Intel’s 10nm is broken and unfixable, two strong economic arguments show why. SemiAccurate has been saying this for a while and now it is time to lay out why 10nm is unlikely to be viable at Intel.
Authors Note: This is not a piece on Intel’s financial outlook, just a piece using financial arguments to analyze a technical point.
Cannons Misfire:
Lets once again start with a bit of history about the 10nm process. SemiAccurate first told you about the problem when the first 10nm CPU, Cannon Lake, came back in 2016. Yields were spectacularly awful, 8-10% according to our sources but that isn’t the 8-10% fully working that you normally would expect to see, it was for chips that were said to, “do something”. During the life of that part, which technically did ‘ship for revenue’ in 2017, there was, to the best of our knowledge, never a fully working Cannon Lake CPU publicly demonstrated.
Yes there were laptops that made it to the public like this one, but it was nothing more than a PR stunt that all used discrete AMD GPUs. Several OEMs told SemiAccurate the same story, they were given a limited amount of Cannon CPUs and ‘asked’ to make a single laptop out of them and, “make it look real“. Note to those not in the PC business, when making a product there is rarely any need to make it look like anything other than it actually is but we will leave it up to you to figure out why Intel asked this of the OEMs. Needless to say once our report came out, many of the models didn’t show up for some reason.
Cannon was over two years late and shipped in quantities that never came close to the intended six digits, this in spite of all the rosy talk about processes improving rapidly, problems being located and solved, and volume ramps proceeding according to plans. Go back and listen to the quarterly analyst calls from 2016 to 2018, they all say the same incredibly optimistic things. It is worth repeating that in the real world, Intel never showed or shipped a single working Cannon Lake CPU, emphasis ours. Compare and contrast that with the official messaging.
Iced Over:
Then we come to Intel’s Ice Lake, the new architecture that succeeded Cannon Lake. It ‘shipped’ in September 2019 after a last minute three month delay. The June ‘launch’ at Computex was supposed to be a hard one but that became a soft August re-launch without any products actually shipping. The messaging that came with it bordered on outright lying but using skewed benchmarks, hidden power consumption figures, and more ‘normal’ practices, they managed to show Ice beat the five-ish year old 14nm Skylake derivatives by a hair.
Once again in the real world those ancient 14nm parts clock to 4.9GHz, 5.1GHz since that release, and the newer 10nm Ice Lake CPUs climb to an amazing 3.9GHz. But they take ~40% more power to get there which is why Intel was so questionable in their press briefings. Would you want to show this off? Could you sell this product on store shelves? Does this explain why only some OEMs have a single Ice Lake laptop on sale, SemiAccurate can’t find any with a second SKU as of this writing. On the up side, the architecture of Ice Lake has a much higher IPC than the 14nm Skylake derivatives so it manages to claw back much of that deficit albeit at a high energy use cost.
Volumes are still non-existent however and many new fall/back to school/holiday notebooks have launched with Comet Lake, yet another 14nm Skylake derivative. It is notably faster and more efficient too, plus it is manufacturable, all things that OEMs like in a component. In short we are five months after the June launch at Computex, 10nm Ice Lake CPUs are ‘out’ in ‘volume’, and OEMs have essentially no SKUs on the shelves.
Talk It Up:
Intel is seeding a furious astroturfing whisper campaign to ‘counter’ this view, but again, the numbers don’t lie. 10nm Ice Lake CPUs are simply not available in any volume, and those that are run at painfully slow clocks, something SemiAccurate told you would happen in early 2017. Again compare and contrast that to Intel’s official messaging. In their Q1/2018 analyst call they directly implied that the problems were known and solutions were being implemented. (Note: SemiAccurate outright mocked Intel for these statements because they were directly contradictory to what the company was saying internally.)
According to the Motley Fool transcript of the call, CEO Brian Krzanich said, “We continue to make progress on our 10-nanometer process. We are shipping in low volume and yields are improving. Though the rate of improvement is slower than we anticipated. As a result, volume production is moving from the second half of 2018 into 2019. We understand the yield issues and have defined improvements for them. But they will take time to implement and qualify.“
That is pretty specific and it mirrors the tactics that Intel has been taking to cover up the monumental failure of it’s 10nm process. Then again the fact that it is 2H/2019 now and the chips are not in volume production is not an arguable fact. Intel knows the process will never work right, knows it will never yield acceptably, knows that it will never be financially viable, and simply won’t admit this. SemiAccurate is quite simply at a loss to explain why Intel is not acknowledging what everyone knows and moving on. Instead they keep digging.
Two Arguments:
That is where we once again come to the real world and the two financial arguments that we mentioned earlier. If you have any doubts about the technical problems Intel has and cannot solve that sunk the 10nm process, follow the links above for all the gory details. If you want the reasoning from a financial standpoint that the company both knows and can do nothing about the issues, read on. Intel’s actions are pretty telling when it comes down to the dollars.
Note: The following is analysis for professional level subscribers only.
Disclosures: Charlie Demerjian and Stone Arch Networking Services, Inc. have no consulting relationships, investment relationships, or hold any investment positions with any of the companies mentioned in this report.
Charlie Demerjian
Latest posts by Charlie Demerjian (see all)
- What silicon is in the new Sony Handheld? - Nov 26, 2024
- Surf Security Puts Deepfake Detection In A Browser - Nov 20, 2024
- Asus’s ROG9 phone has a lot of interesting engineering - Nov 12, 2024
- What platform is Intel’s Diamond Rapids on? - Nov 7, 2024
- What Is Diamond Rapids +1 Called? - Oct 31, 2024