Home Hardware Opinion: Intel chip advancements show they're up for a competitive challenge

Opinion: Intel chip advancements show they’re up for a competitive challenge

Why it issues: If you comply with developments within the semiconductor trade, it will be straightforward to see how some folks might imagine Intel is in deep trouble. Several indicators haven’t been promising in any case. However, simply if you start to suppose that Intel’s slipping, the corporate has seemingly returned to its roots in semiconductor manufacturing to ship what seems to be very promising enhancements in primary transistor expertise that ought to translate in aggressive merchandise in a comparatively quick window.

First, the corporate introduced robust second quarter earnings, however its inventory tanked when it revealed yet one more downside and delay for its 7nm manufacturing course of. Second, associated to that, Intel not too long ago misplaced its crown as the most important and most respected semiconductor firm—a spot it held for many years—to Nvidia. In reality, as of the market’s closing on August 17, the hard-charging Nvidia surpassed a $300 billion market capitalization, versus a $208 billion capitalization for Intel.

Finally, from a CPU efficiency perspective, it’s been AMD (which crossed over the $100 billion market cap not too long ago and briefly) that has been handily beating Intel practically throughout the board on PC desktop, cell, and even server benchmarks. For an organization that has historically dominated in all these areas, this has been fairly a change.

However, simply if you start to suppose that Intel’s slipping, the corporate can shock you. And so it did, each at Architecture Day and at this week’s Hot Chips convention. The Architecture Day occasion held final week provided a complete have a look at a variety of improvements that Intel is engaged on throughout course of expertise, chip packaging, core chip architectures, high-speed chiplet-to-chiplet interconnect, safety and rather more.

Importantly, whereas a few of these developments fall into the extra theoretical realm, the overwhelming majority are going to carry sensible, real-world efficiency enhancements that ought to translate into choices that will likely be extremely aggressive with AMD’s newest.

Intel returned to its roots in semiconductor manufacturing to ship what seems to be very promising enhancements in primary transistor expertise.

Intel returned to its roots in semiconductor manufacturing to ship what seems to be very promising enhancements in primary transistor expertise. Until only a yr or two in the past, Intel was ready to make use of its lengthy historical past of experience in chip making to its clear benefit, bringing out common enhancements that different chip makers would usually take 18-24 months to catch as much as. As course of delays and different manufacturing challenges began to pile up, nevertheless, it began to appear like the corporate’s in-house manufacturing was changing into extra of an albatross round its neck.

With the debut of 10nm SuperFin expertise, the corporate is clearly reminding the remainder of the trade that it has a variety of completely different tips up its sleeve relating to core semiconductor capabilities. More importantly, these advances can nonetheless be used to make vital progress in areas that firms who don’t make their very own chips can’t get entry to till their manufacturing companions develop them as effectively.

By leveraging a brand new metallic capacitor expertise together with the corporate’s enhanced FinFET transistors, Intel claims that the 10nm SuperFin expertise allows it to drive increased currents, scale back resistance, and thereby considerably enhance efficiency over conventional transistor designs. In sensible phrases, the primary Intel half to make use of the brand new expertise (a CPU codenamed Tiger Lake set to be unveiled in early September) is anticipated to have a efficiency enhance that the corporate claims is the most important “intranode” (on the identical course of dimension) enchancment it has ever had.

In reality, Intel says it will be “comparable to a full-node transition”, which is a fairly large assertion—although it’s additionally meant to considerably counter continued delays all the way down to the 7nm node. Real-world efficiency stays to be seen, however preliminary commentary and evaluation from those that totally perceive the main points of chip architectures are promising.

At the identical time, the corporate additionally acknowledges that, as transistors get smaller and smaller and course of applied sciences get more durable and more durable, it is sensible to separate, or “disaggregate” chip design advances from manufacturing enhancements, in order that the 2 components can transfer ahead on separate timelines. Combining this with the truth that Intel not too long ago brazenly mentioned the concept of utilizing exterior chip manufacturing crops, or foundries, to construct a few of its components highlights a really sensible, arguably much more mature method to the scenario. It additionally displays a really completely different perspective for Intel—one that ought to assist the corporate progress ahead on superior chip designs, no matter the place (or by whom) they’re finally constructed.

Intel additionally supplied an excellent deal extra details about its Xe GPU structure, with a specific concentrate on how it is going to be capable of scale from higher built-in graphics—the primary iteration may even be a part of Tiger Lake—to a number of various kinds of discrete GPUs focused at completely different markets. Its first standalone GPU, codenamed DG1, is anticipated later this yr, as is a brand new server-focused GPU. Both of those chips, in addition to the built-in GPU, will likely be primarily based on what they’re calling the Xe-LP (low energy) structure. The firm additionally unveiled a brand new Xe-HP (excessive efficiency) GPU structure that’s destined for different discrete chips centered on extra superior datacenter functions, similar to AI acceleration, in addition to a gaming-optimized half that may help ray tracing.

Both at Hot Chips and at Architecture Day, the corporate additionally talked an excellent deal about its software program efforts, notably its OneAPI structure, which is designed to make the method of programming for any sort of non-CPU accelerator (from GPUs, to FPGAs to different AI-focused accelerators and past) a lot simpler to do. This is clearly a monumental aim and a difficult process. However, the corporate appears to be making stable progress and highlighted how software program enhancements alone can usually result in vital efficiency enhancements.

The backside line is that it’s clearly means too early to depend Intel out. If nothing else, the considerably extra aggressive surroundings through which it finds itself appears to have impressed it to be doing a few of its greatest work in years. Final particulars are nonetheless to come back, however the excellent news is that the corporate is clearly up for aggressive challenges, and no matter what distributors you select to help, we are going to all profit ultimately.

Bob O’Donnell is the founder and chief analyst of TECHnalysis Research, LLC a expertise consulting agency that gives strategic consulting and market analysis providers to the expertise trade {and professional} monetary neighborhood. You can comply with him on Twitter @bobodtech.

Most Popular

Recent Comments