Intel teases plan to speed up chip advancements

Illustration by Alex Castro / The Verge

Intel has struggled recently with getting its more advanced chip architectures out the door, but that might be changing soon. CEO Pat Gelsinger announced today on Intel’s Q2 earnings call that “7nm is progressing well,” and the company is gearing up to announce news on how its “accelerating its annual cadence of innovation with new advancements in semiconductor process and packaging,” at a webcast on Monday, July 26th.

The announcement comes alongside Intel’s Q2 earnings, where the semiconductor company managed to beat expectations and bring in $18.2 billion in revenue, up 2 percent year-over-year. Intel is feeling pretty optimistic about the future, too. It raised its 2021 outlook by an additional $1 billion to $73.5 billion, and now expects to achieve revenue growth year-over-year compared to 2020.

Intel’s earnings have been trending upwards for the past several quarters, despite its frequent delays and misses on production of both its 10nm and 7nm chips. That growth has been buoyed by spikes in demand for both personal computers and data-center chips in 2020 due to the COVID-19 pandemic, which saw millions buy new laptops and computers to aid in remote work and education and an increased reliance on cloud services.

And Intel seems to still be riding that wave, recording $10.1 billion in revenue for its client computing group (up 6 percent year over year, and a Q2 record for the company), despite the global semiconductor shortage. Gelsinger commented during the call that he “expect[s] the shortages to bottom out in the second half [of the year, but] it will take another one to two years before the industry is able to completely catch up with demand.”

The boost for the client computing group also helped Intel make up for slower revenue from its data center group, which was down 9 percent year-over-year with $6.5 billion in revenue. There’s signs that that the impressive growth for the PC segment is starting to slow, however: both IDC and Gartner noted that growth of PC sales has slowed down considerably in Q2, which could indicate that the boost in sales there will diminish in the coming months.

The past year has been a tumultuous one for Intel. Last summer, the company was forced to admit that it would be severely delaying its upcoming 7nm architecture for its next generation of processors. That news was followed by the departure of the company’s hardware chief, Dr. Venkata (Murthy) Renduchintala, and the replacement of CEO Bob Swan with Pat Gelsinger (a former chief technology office and hardware engineer at the company) in an attempt to right the ship.

Gelsinger has wasted no time in trying to establish changes to get Intel back on track, announcing a new “IDM 2.0” initiative in March that aims to see Intel return as an industry leader in chip manufacturing. As part of that plan, the company will be outsourcing production for “products at the core of Intel’s computing offerings” to competitors like TSMC and Samsung starting in 2023.

Intel is also working to launch a new branch of its business, Intel Foundry Services, which will see the company take on manufacturing projects for third-party companies, including a $20 billion investment in expanding its manufacturing facilities in Arizona.

Given the overwhelming demand for more chip making companies in light of the global semiconductor shortage, Intel Foundry Services could be a big win for the company down the line. Intel announced today that it has over 100 potential customers interested in the service, and promised more news on Monday about the new initiative, but it’ll still be some time before that business ramps up.

Comments

still no 5nm? Smartphones have moved to 5nm.
i thought they are going to announce a 5nm chip ready to be purchased in few months.
or is it hard to build a 5nm chip compared to smartphones?

5nm is just marketing term, although TSMC is indeed ahead of Intel.

different OEMs use different rulers

Intel is 2+ years away from TSMC 5nm’s peak transistor densities (171 million). Intel 7nm is fairly more dense than TSMC 5nm, but it’s only launching in 2023.

Which probably means late 2023, like all Intel nodes.

Maybe by 2024 we’ll have a more competitive market then. But, to be fair, this is just the node. Intel CPUs also will need be more efficient designs, which looks essentially impossible next to Qualcomm’s NUVIA or Apple’s cores, not to mention Arm’s X1 cores.

Even comparing node-for-node (as close as we can), Intel x86 designs just aren’t up to snuff next to AMD, Apple, nor Arm. Guzzling 2x the power for what? Today, Intel is only succeeding in volume—thus compatibility and availability—not on its technical merits.

This is incorrect. Intel’s peak density at 10nm is 100m transistors. TSMC’s at 10nm is 52m so just a little over half and at 7nm 92m (still less than Intel’s 10nm). The 5nm process (EUV) has a density of 170m which Intel’s 7nm will match.

What now? Read what I wrote.

Intel 10nm = 100 million
TSMC 5nm = 171 million
Intel 7nm = 200+ million

Intel 7nm launches in 2023 and is higher than TSMC 5nm. This, Intel will need 2+ years to beat TSMC 5nm, which is shipping in high-volume today.

Yes, but right NOW Intel is at 10nm and TSMC is at 5nm, with 70% higher density.
By late 2023 or more likely 2024, when Intel reaches 7nm, TSMC will be at 3nm, and again have much higher density.
In other words, whatever the metrics are, Intel is way, way behind, now or in 2023.

We are saying the same thing. I don’t understand your "Yes, but", though we can fill in the chart. For everyone else wondering, these are peak transistor densities, in millions of transistors per square millimeter, which are a rough guide for how small the nodes can be.

By high-volume release year:

  1. Intel 14nm = 38 million (late 2014)
  2. TSMC 16nm = 28 million (late 2014)
  3. TSMC 10nm = 53 million (late 2016)
  4. TSMC 7nm = 108 million (late 2018)
  5. Intel 10nm = 100 million (late 2019)
  6. TSMC 5nm = 171 million (late 2020)
  7. Samsung 5nm = 127 million (late 2020)
  8. TSMC 3nm = 292 million (late 2022)
  9. Intel 7nm = 200+ million (late 2023)

TSMC 7nm density is an average between N7FF, N7FF+, and N6.

//

OP asked, "still no 5nm? Smartphones have moved to 5nm." The reply is that Intel will only reach "smartphone 5nm / TSMC 5nm" in 2+ years; Intel is absolutely quite significantly behind.

Intel CPUs also will need be more efficient designs, which looks essentially impossible next to Qualcomm’s NUVIA or Apple’s cores, not to mention Arm’s X1 cores.

We’ll have to see how Alder Lake shapes up. The shift to big.LITTLE could give significant efficiency improvements to their chips. But this being Intel it’s even odds of just being yet another false dawn.

Intel doesn’t have very efficient cores right now, atom is just bad. Making big-little combination out of current parts wouldn’t really work. Alder does bump up both the atom and core parts and perhaps it improved both segments enough to make a good chip in the end. We shall see, hopefully this year already.

Well the scuttlebutt has the new Gracemont LITTLE cores performing like a Skylake big core (at least in terms of IPC) but at a fraction of the power consumption. Of course if they achieve that measure at all it will probably only be under some very specific workloads for very brief period but who knows, maybe that combined with shifting to 10nm, which will hopefully help make that power fraction small enough, will actually result in a worthwhile efficient core finally.

Wrong, atom has industry leading data transfer of efficiency, but intel did not bother to develop it further

Intel 7nm expected to be way ahead of TSMC’s 5nm or even 3nm. They are not comparable terms. However, currently TSMC 5nm is ahead of intel 10nm that’s for sure.

Why do Intel fans keep perpetuating that myth? It applied 10 years ago but the gap has shrink significantly. If it was an imaginary 5 versus 5 comparison or 7 versus 7,i would agree that Intel is better. But there is no way Intel 7 is way better than TSMC 5 or 3.

You’re right that claiming Intel’s 7nm will be better than TSMC’s 3 is ridiculous, but, based on history and Intel’s claims of doubling density from 10nm to 7, it seems absolutely possible they will beat their 5nm in terms of transistor density.

Now obviously transistor density isn’t everything but it’s a far better metric than just using nm which everyone measures differently.

more ticks less tocks

Intel is also working to launch a new branch of its business, Intel Foundry Services.

Correction: this is not a new branch. Intel’s CFO was bragging to Reuters in 2011 about their foundry services:

https://www.theregister.com/2011/05/27/stacy_smith_talks_intel_arm/

> Understandably, Smith would prefer to provide his customers with Intel-designed cores. "If Apple or Sony came to us and said ‘I want to do a product that involves your IA core and put some of my IP around it’, I wouldn’t blink. That would be fantastic business for us," he said.

> But he wasn’t as firm as Otellini when it came to non-IA cores. "Then you get into the middle ground of ‘I don’t want it to be an IA core, I want it to be my own custom-designed core,’ and then you are only getting the manufacturing margin, [and] that would be a much more in-depth discussion and analysis," he told journos after the UK investor event.

Intel has opened its foundry up for 10+ years. The kicker: everyone who tried it hated it because Intel always has treated it like a third-tier offering: poor support, overpriced, slow updates, and the risk of stolen designs. Intel could competes with almost anyone that would use their foundry.

https://semiengineering.com/intel-dethrone-foundry-giants/

Most prominently, Intel tried to be Apple’s foundry in the early 2010s, even hinting Apple could use their custom Arm cores. Steve Jobs refused.

https://www.cnet.com/news/steve-jobs-knocked-intels-chip-design-inflexibility/

>But Jobs implies in the biography that Intel wasn’t keeping up with the times. He explains why Apple didn’t select Intel chips for the iPhone.

>"There were two reasons we didn’t go with them. One was that they [the company] are just really slow. They’re like a steamship, not very flexible. We’re used to going pretty fast. Second is that we just didn’t want to teach them everything, which they could go and sell to our competitors," Jobs is quoted as saying.

>On one level that last statement is rather remarkable. Jobs, of course, was saying that Apple would have to teach the world’s premier chipmaker how to design better chips.

It’d be like Samsung sending its phones to Apple for manufacturing. It was stupid then and it’s stupid now.

Intel really doesn’t understand the consumer market. RISC-V sounds just like Arm and Itanium plays that Intel botched spectacularly in the 1990s and 2000s.

Looking at the M1 chip from Apple, it seems they know what they’re doing in that department.

Yep. Back in 2011, though, when CNET wrote that article, it was so rather laughable from all the public data.

Steve Jobs, however, had much more insight.

2011 iPhone: iPhone 4S
2011 Intel CPU: Sandy Bridge

True, but it could also refer to non-CPU parts. CPU is just one piece of the whole SoC and it might not even be the most important for phones.

That said, it is more meaningful if you consider who those competitors were. Intel had good relationship with Nokia – they were an important piece of Nokia’s fall from grace.

I could not care less about promises, promises until the product is literally stocked at a location that I can buy with US dollars.

Only about 10 years late…

Odd that Intel never talked up their plan to slow down chip advances ten years ago. That one was a huge success.

Who wants to buy consumer microprocessors that could be described as second or third best? Nobody. But, since a) the release of Apple’s M1 last year and b) AMD’s steady rollout of processors that offer embarrassingly superior throughput and energy efficiency to the parts offered by Intel second or third best is the best that computer manufacturers buying silicon from Intel can hope for at the moment.

Now, shifting away from Intel as a silicon supplier is no doubt quite tricky for obvious reasons – TSMC is already at the limit of its capacity turning out existing orders from AMD and Apple, so there is no prospect of satisfying current demand for x86 laptops, say, by moving all production to AMD silicon. TSMC simply couldn’t satisfy demand running at that level any time soon.

So, what happens in this situation? All of the players in the x86 market grit their teeth and put a brave face on things. In public the computer manufacturers adopt a nothing to see here business as usual posture. Behind the scenes they explore how quickly products can be moved to AMD inside. All the while computer manufacturers in this difficult situation will be hoping that consumers don’t notice what is happening and in particular don’t notice that they’re are being sold a second or third class bill of fare. Meanwhile Intel is forced to accelerate its own efforts in order to hold on to the consumer market – Alder Lake is the tangible evidence of that. And, the x86 computer manufacturers, too, hope that Alder Lake will save them from turbulent times ahead.

But will Apple or AMD give up so easily after having stung and bruised Intel so badly? Of course they won’t. It seems to me this is a situation in which the best technologies and products will win out and nothing I see from Intel indicates to me that they have the best of anything. I’m pretty sure that they will make a great chip foundry, though.

The end is nigh for intel and slightly slower for amd. The only thing keeping them in business is gaming and a non-ready port of windows to apple silicon. The day Microsoft gets greedy and ports it to apple silicon (remember Windows already has arm versions), it’s game over for intel. Amd will be around because of their pricing.

View All Comments
Back to top ↑