Intel: 10nm Yield Is Ahead Of Expectation, 7nm Ponte Vecchio GPU On Track

Jan 24, 2020
Submit

Intel has had a fantastic quarter and the company is literally selling every chip it can make. Turns out there were a few additional nuggets of information buried in the slides as well. Intel has officially stated that their 10nm yields are ahead of expectations and that they are planning at least nine 10nm products for release in 2020. The 7nm lead product, Intel Xe Ponte Vecchio GPU remains on-track for Q4'21.

Intel planning nine 10nm products in 2020, yields ahead of expectations

Intel is incredibly skilled at improving architecture while staying on the same node, but considering 14nm has been around for a very long time, the shift to 10nm is eagerly anticipated and will likely be a turning point in the company's history. While there were some rumors earlier on (that Intel readily debunked) it is not clear that 10nm HVM is expected to make landfall in 2020. Intel has further revealed that not only are yields above expectation but they will be rolling out a total of nine 10nm products in 2020.

Here are the key takeaways:

  • 10nm yields are ahead of expectation.
  • Nine 10nm products are expected to release in 2020.
  • Intel will increase its wafer capacity by 25% in 2020.
  • 7nm Ponte Vecchio GPU on track for Q4 2021.

It would also appear that work on Xe-based Ponte Vecchio is going great - which is good news for gamers because this tech would eventually trickle down in the gaming space as well. Intel's Ponte Vecchio GPU for 7nm would be the first significant milestone in the company's ambition to become a GPU player and would constitute the first major diversification. Intel's DG1 GPU is likely one of the 10nm products that will be launched later in the year. As for the other eight products, these are likely going to consist of a mix of the CPU side as well as FPGA/AI based products.

Recap: Intel CEO reaffirms commitment to 7nm and explains what went wrong with 10nm

Bob Swan has previously answered a question about how Intel got into a position where it has lost a massive chunk of its CPU market share to AMD and is in a position where it is unable to meet demand (this is in contrast to its old philosophy of prioritizing err-ing on the side of caution and always having spare fab capacity):

How we got here is really kind of threefold, one we got a lot faster than we expected and the demand for CPUs and servers grew much faster than we expected in 2018. You’ll remember we came into 2018 projecting a 10% growth and we grew by 21% growth so the good news problem is that demand for our products in our transformation to a data-centric company was much higher than we expected.

Secondly, we took on a 100% market share for smartphone modem and we decided that we would build it in our fabs, so we took on even more demand.

And third, to exacerbate that, we slipped on bringing our 10nm to life and when that happens you build more and more performance into your last generation for us – 14nm – which means there is a higher core count and larger die size.

So those three – growing much faster than we thought, bringing modems inside and delaying 10nm resulted in a position where we didn’t have flexible capacity. - Intel CEO Bob Swan

While most of this is old news, this is the first time that Intel has given a solid reason for why it is not able to meet capacity - namely that it decided to produce smartphone modems in-house which in return meant that they were not able to focus on the CPU side of things. It is also a fairly plausible explanation for why Intel cannot even meet the demand for 14nm anymore and instead has to resort to extending 22nm products.

When asked specifically to explain what went wrong, Bob Swan candidly responded with an admission that Intel had gotten overconfident in its ability to beat the industry standard and suffered the consequence. Calling it "Scar tissue" here is the explanation Bob gave:

The scar tissue really started with Moore's Law. Two times scaling factor every two years and that's kind of the simple rule of thumb. That's worked for a very long time. And the transition from 22 to 40 nanometers and then 14 to 10 we decided that despite the fact the physics was getting more challenging we decided to set a higher bar for ourselves in terms of performance. So the 22 to 14 is not a 2 times density, it was 2.4 and it was bumpy along the way but it worked and that working gave us the confidence that for 14 to 10 why don't we take the scaling factor up to 2.7 when you do that the implications of trying to get more and more density and more and more performance [ you start to see the problem]

Secondly, we're not going to try to do 2.4 scaling or 2.7 scaling as we think about 7 nm, you know, we put 2.0 back in line with historical trends as we think about 5nm, which would be our competitors 3 nanometer...

In our first 7nm product in the fourth quarter of 2021... I would also argue that another positive of some of the challenges around 10 nm is that you kind of learned how to make 14 better along the way and you know, we have 14+,  14++ and and despite the fact that you've been on the same node for four years now performance of the chips that continue to improve as we go to 10. - Intel CEO Bob Swan

With Intel chasing a 2x scaling factor for 7nm and shifting to EUV as well, it seems like the company is all set to introduce its first 7nm products (equal to TSMC 5nm) in the fourth quarter of 2021. Bob also further stated that he expects to hit 5nm (equal to TSMC 3nm) by 2H 2024.

Submit