| Manuel Calavera |
22.02.2018. 10:54 |
Evo malo info o nadolazećim nodeima
Citiraj:
Summary
-Intel is struggling to figure out 10nm while TSMC, Samsung and GlobalFoundries push on to 7nm.
-SRAM cell size points to 7nm superiority.
-AMD's fabless business model puts the company in prime position to take advantage of Intel's manufacturing woes.
Advanced Micro Devices (AMD), which has been fabless for almost a decade now, might soon be the beneficiary of Intel's (INTC) fabrication woes as the latter company struggles to figure out its 10nm process. With AMD's upcoming processors likely to be manufactured on the 7nm processes of Intel's competitors, the power dynamic in the CPU market looks poised to rebalance in AMD's favor.
Intel originally planned to have 10nm products enter production in 2015, but the release was then postponed to 2H 2017. Then when nothing arrived in 2H 2017, Intel said it would introduce 10nm in late 2017 and begin volume production in 2018. But, as of the writing of this article, we have still yet to see or hear any significant, official updates from the company in terms of products manufactured using this 10nm process.
In the meantime, Samsung (OTC:SSNLF), Taiwan Semiconductor (TSM), and GlobalFoundries have been pushing full steam ahead toward 7nm processes that boast higher SRAM cell density by significant margins. Here's an excerpt from an EE Times article discussing Intel's presentation at the 2018 International Solid-State Circuit Conference ("ISSCC") a few days ago:
The Intel design shows 0.62–0.58x scaling compared to its 14-nm SRAM, maintaining Moore’s Law and “within 15 percent of the smallest reported 7-nm cell,” said Intel’s Zheng Gui, pointing to smaller 7-nm SRAMs from Samsung this year and TSMC at ISSCC 2017.
Intel is essentially saying "hey, we're not THAT far behind" whereas, before the years-long delay in its 10nm process, the company had an unquestioned manufacturing and technological edge. And sure, if they had come out with this process node in 2015 Intel would be sitting pretty, but instead, we're two months into 2018 and we still haven't heard much regarding 10nm products. Meanwhile, TSMC expects to begin volume production on 7nm in 1H 2018, Samsung is on a similar schedule, and GloFo is set for trial production in mid 2018 and volume production soon thereafter. Here's a summary table for comparisons between the different nodes:
https://static.seekingalpha.com/uplo...922_origin.png
As we can see, 6T (six-transistor) SRAM cell size does not look good for Intel in the slightest. While the 7nm processes are all in the same ballpark in SRAM cell density, Intel is at 0.0312 (well at least they're within 15%). SRAM cell density is not the be-all, end-all for determining which processes will ultimately perform better, but Intel's multi-year delay for 10nm and the inferiority of SRAM cell density don't appear to bode well as the company matches up with competitors' processes.
What does this have to do with AMD? Unlike Intel, which manufactures its own chips, AMD decided to spin off its fab business (which is now GlobalFoundries) back in 2009. While Intel has to worry about each new node in addition to each new architecture, AMD can play the field and use whichever process fits best with the company's strategy. Therefore, AMD appears to have an opportunity that it hasn't had in quite some time: To compete on a level playing field. More specifically, without Intel's usual significant process advantages, the battle of microarchitecture vs. microarchitecture becomes much more interesting.
This isn't to say AMD will surpass Intel in every single processor performance metric - the process node is not the only determining factor of performance. However, Intel's current situation seems precarious and could snowball into a more significant disadvantage if the company doesn't get its act together soon. But even now, what we know about Intel's 10nm compared to other 7nm processes indicates that AMD's next generation of CPUs could very well be more competitive on a process basis, which would narrow the gap in performance between the two companies' chips (provided AMD doesn't drop the ball on the microarchitecture).
Investor Takeaway
Whereas AMD has had to fight against the tide of a process disadvantage for years, now it appears that the company will have access to an on-par or possibly better process than Intel. This is huge. This won't impact the current Zen processors (14nm node) or Zen+ processors (expected 12nm), but once Zen 2 hits the market, likely in 2019, the CPU space will become very interesting.
What's especially exciting for AMD investors is that regaining the technological edge in the fab business isn't easy once you've fallen behind. That Intel has now delayed its 10nm process for close to three years should be of substantial concern to INTC shareholders and should be a source of optimism for those holding a stake in AMD, considering the two companies might be on equal footing in the process game for the foreseeable future. And yet, in the face of losing the competitive advantage in the fab business, Intel still hasn't begun production of 10nm processors. I don't see any way an objective observer could see this development as anything else other than a boost for AMD and a blow for Intel.
The challenge Zen 2 and later generations of CPUs must overcome is one of mindshare: Intel still has the brand, the entrenchment, and the prestige. In my opinion, AMD can make a significant dent in Intel's mindshare (and market share) with solid execution on Zen 2. Currently, AMD is a competitor on price/performance alone, which is good for consumers but bad for gross margins. However, I think Intel's missteps on 10nm and other foundries' advancements on 7nm will allow AMD to more ably compete on a head-to-head basis, which will increase mindshare, market share, and revenue while also providing a boost to gross margins as well.
And yes, it's possible Intel will get its 10nm troubles figured out and have products manufactured on the process hitting shelves before the end of March. But realistically, the three-year delay has changed the balance of power in the industry: Intel is following, not leading. As a fabless company, this scenario benefits AMD. Further, I think this rebalancing of power gives AMD an opportunity with Zen 2 to truly challenge Intel's dominance, and if other foundries continue to outpace Intel, that opportunity will only expand.
|
|
