قالب وردپرس درنا توس
Home / Business / Intel's 10nm is broken, delayed until 2019

Intel's 10nm is broken, delayed until 2019



Intel today announced its financial results, and although it recorded another record quarter, the company revealed serious production issues with its 1

0nm process. As a result, Intel announced that it will deliver even more 14nm iterations this year. They will come as Whiskey Lake processors for the desktop and Cascade Lake Xeons for the data center.

Intel achieved a near-perfect quarter of record highs, with record sales of $ 16.8 billion in the first quarter, up 16 percent from a year earlier. The company raised its guidance for the full year to $ 67.5 billion, an increase of $ 2.5 billion from its previous forecast, but we will return to these figures.

The 10nm Problems

Overall, Intel had a star quarter, but it originally promised that it would deliver the 10nm process in 2015. After some delays, the company assured that it would launch 10nm processors in 2017. This was further refined in the second half of this year.

At today's profit distribution, Intel announced that it has postponed the high volume 10nm production for an indefinite period in 2019. Meanwhile, its competitors like TSMC are starting mass production of 7nm alternatives.

The current naming conventions for semiconductor nodes are not based on traditional measurements. They are therefore more of a marketing exercise than a science-based metric. This means that TSMC's 7nm is not quite on par with Intel's 10nm process. However, the continued process-node shrinkage at other Fabs indicates that other companies are successfully outmaneuvering the production challenges of smaller lithographs while Intel remains at 14nm.

Intel CEO Brian Krzanich has repeatedly urged that the company supply Cannon Lake on a small scale, but the company has not referred to specific customers or products. And we asked. As we mentioned earlier this year, the delay seems to be minor, but Intel has been selling processors based on the underlying Skylake microarchitecture since 2015 and has been stuck in the 14nm process since 2014. That is, Intel is the fourth (or fifth) iteration of the same process that has hindered its ability to bring new microarchitectures to market. This does not bode well for a company that regularly claims its process-node technology is three years ahead of its competitors.

Krzanich said the company had "bitten off" a bit too much on this thing by increasing the 10nm density to 2.7X across the 14nm node In comparison, Intel increased the density by only 2.4X as it moved to 14nm, and although the difference may be small, Krzanich pointed out that the industry average for density improvements is only 1.5-2X per node transition At 10nm, Intel has reduced its density target to 2.4X for the transition to the 7nm node, and Intel will also focus more on heterogeneous architectures with its EMIB technology (which we've covered here).

10nm is Intel's latest process, the based on traditional photolithography, and although Krzanich is not deeply immersed in details, he has listed the lithography technique as a significant contributor to the low 10nm yields will switch to EUV at 7 nm. Currently, Intel's multisampling process generates too many defects, thus reducing the yield to produce 10 nm at low cost. Krzanich says the company has recognized the problem and will correct it, but the fixes will take indefinitely to significantly impact returns. Intel was unwilling to commit to high production in the first half of 2019, so it is possible that 10nm will be postponed until the second half of the year.

We continue to make progress on our 10 nanometer process. We deliver on a small scale and yields are improving, but the rate of improvement is slower than we expected. As a result, series production will move from the second half of 2018 to 2019 . We understand the revenue issues and have defined improvements for them, but they will take time to implement and qualify them. We have leadership products on the roadmap that continue to use 14-nanometer, Whiskey Lake for customers, and Cascade Lake for the data center, which will come later this year.

Krzanich's statement says that although Intel has defined the fixes, they are not actually testing them in mass production. This means that the company could soon return to the drawing board if the fixes are not effective.

This is probably the reason for Intel's confirmation that it has reoriented its critical technology and manufacturing group that manufactures the company's silicon under new leadership. Intel also announced in good time today the announcement of profit sharing that the famous chip architect Jim Keller has led the silicon design initiatives. This was probably a step to reassure investors that Intel's 10nm production issues are paying close attention.

Unfortunately, process technologies require extended incubation periods, so it can take some time for leadership changes to have a significant impact on Intel's roadmap. Intel is obviously bringing the pieces together quickly, but its competitors, like AMD, are making good progress on their future architectures. AMD has already installed 7nm GPUs in its labs and projects and will be testing 7nm EPYC 2 processors this year.

Thoughts

Intel's late 10nm process has led to stagnation on the microarchitecture development front, and the problems are even greater than they appear on the surface. As the financial result clearly shows, the company diversifies among others successfully in AI, the data center, autonomous driving, 5G, FPGAs and IoT. It even added GPUs to the list.

Unfortunately, Intel's process technology touches every segment of this technology and the chips that drive it. The late arrival of 10nm could affect Intel's ability to compete in almost all segments, and all of this is due to Intel expanding new segments that already have dominant and entrenched deep pocket players. Krzanich has pointed out that Intel has improved the performance of 14nm since its debut in 2014 by 70%, but the company is sure to reach a declining point of return soon.

Intel's relatively flat R & D spending (+ 3%) is certainly not encouraging given the current climate. The company is likely to switch to 10nm + second generation due to its 10 nanometer yield problems, but did not confirm the change during the payout. Krzanich said the company will not jump to the 7nm node. Instead, it applies its learning from the 10nm node to the 7nm process.

Krzanich also acknowledged that the company's density is shrinking compared to competing fabs. Intel has long been the keeper of Moore's Law flame, and the company has continued to insist that the law will live long after other companies have acknowledged that it has expired. We'll have to see if Intel changes its messaging, but we're a long way from the tick-tock cadence. Considering that Intel has not delivered a smaller process in significant quantities since 2014, it is fair to say that Moore's original law is officially dead.

The Financials

Intel's Client Computing Group (CCG) posted strong financial results. Focusing on processors for laptops and desktops, the group has been targeted by AMD, which continues to grow dynamically. Intel's CCG group posted sales of $ 8.2 billion, up 3% year-over-year. However, this is mainly due to higher average selling prices and strong demand from enthusiasts.

The company saw a decrease in notebook processor sales of 8% and a 6% decrease in desktop CPU revenue year over year. Intel cites a shrinking addressable market as the main contributor, but a recent market report from IDC says that PC volumes were flat while competitor Bean Counter Gartner says PC shipments have dropped 1.4% compared to the first quarter of last year are. Intel has grown 4% over last year.

It is unclear how much of this decline is due to AMD's ongoing success, but AMD told us that it has gained as much as 50% of home-based CPU sales from retailers like Amazon and Newegg. AMD also reported a 95% increase in CPU and graphics business over the past quarter, but it's hard to say how much of that is due to processor revenue alone.

Intel's Data Center Group (DCG) increased sales by 24% on the sale of expensive Xeon Scalable models, but Intel reduced its server sales forecast in the second half of the year and speaks of "tougher competition in the second half of the year". This is likely in relation to the combination of AMD's EPYC processors and new Qualcomm ARM competitors.

Intel spent the last year transitioning to "data-centric" companies, which are mostly data center processors. This is because the company relies less on the PC segment for bread and butter. DCG contributed 46% to Intel revenue in the first quarter, so the plan is on schedule.

Intel's nonvolatile memory group produces NAND and 3D XPoint. NSG revenue also increased 20%, while the Programmable Solutions Group, which manufactures FPGAs, grew 17%.


Source link