News Intel 3Q21 Results

[Hitman928]

Intel EPS beats by $0.60, misses on nnon-gaap revenue

Intel (INTC): Q3 Non-GAAP EPS of $1.71 beats by $0.60; GAAP EPS of $1.67 beats by $0.61.Non-GAAP Revenue of $18.1B (-1.1% Y/Y) misses by $170M.Shares -3.6%.Press Release Achieved...

seekingalpha.com

• EPS beat by $0.60
• Revenue very slight miss by $170M
• Guidance: Raising full-year 2021 EPS and gross margin guidance. Now expecting GAAP EPS of $4.50 and non-GAAP EPS of $5.28 from prior guidance of $4.80 vs. $4.11 consensus. and GAAP gross margin of 55% and non-GAAP gross margin of 57%.
• Q4 Guidance: Revenue of $18.3B vs. $18.26B consensus, EPS of $0.90 vs. $0.94 consensus.
• Q4 gross margin guidance weak at 51.4% compared to 56% Q3.

Edit:
Revenues

	3Q21​	Relative 2Q21​	Relative 3Q20
CCG​	$9.7B	96%	98%
DCG​	$6.5B	100%	110%
IOTG​	$1.0B	101.6%	154%
Mobileye​	$326M	99.7%	139%​

 

[DisEnchantment]

As an ISA, I don't see a whole lot of reason for people to specifically choose x86 unless they're trying to cater to an existing installed base. Maybe someone else here can give a technical reason as to why a small firm looking for custom/semi-custom would pick a clean sheet x86 design over ARM?

Why would a small firm pick custom/semi custom x86 instead of ARM. Why would they want to design anything at all besides some specific customizations and special IP blocks.
Just go to AMD and say I want this ISA (x86 or ARM, AMD will do both) and that IP and this amount of SRAM and that amount of Cache etc.
Intel is also offering the same here, they will give their customer the IP blocks and let them assemble the Lego and they fab it for them
No fuss, no patent/IP risks, and have your own custom chip.
Pretty much what Microsoft did when they "designed" their XBox SoC and when Google "designed" their Tensor SoC

Unless you are Amazon/Google/Microsoft, doing any custom high performance CPU is anyway going to be a very tough undertaking. ISA does not matter.
For Amazon/Google why would they buy a server CPU from some ARM CPU vendor when they can design their own using the same reference design from Arm, and in fact they can even do it better.
Unless you have deep pockets and can undertake long term yearly investment in microarchitecture, run of the mill ARM CPU offerings are in very tough spot.
They compete vs x86 vendors to a certain extent (because customers who need x86 will need x86, no other way) but in fact their main competition are their own potential customers who can also industrialize the chips from the same reference Arm designs with their own customizations.

The threat to Intel and AMD is not some startup ARM CPU vendors, they can out spend and out innovate them simply by being in the market for much longer.
The real threat are in house designs from Amazon, Microsoft etc.
Both Intel and AMD are so keen to do semi custom for such customers because of this reason

That is basically what Pat was saying and basically what Lisa was saying
Lisa went on to say they will do ARM too if the customer wants.
Latest ER Lisa went further


For your typical MCU and IoT applications, RISC-V is starting to gain share from ARM and the trend will accelerate.

 

[DrMrLordX]

That is basically what Pat was saying and basically what Lisa was saying

If that's the case, though . . . that is, if Intel is going to offer up IP blocks with a "lego" approach, then are they really in the foundry business at all?

 

[DisEnchantment]

If that's the case, though . . . that is, if Intel is going to offer up IP blocks with a "lego" approach, then are they really in the foundry business at all?

I think you have not gone through the interview
They are offering up x86 and even RISC-V IP to their potential customers as part of thier strategy

 

[moinmoin]

If that's the case, though . . . that is, if Intel is going to offer up IP blocks with a "lego" approach, then are they really in the foundry business at all?

I don't understand your question. All foundries offer up IP block with a "lego" approach for easy use in their foundries. For Intel to be able to include x86 specific IP blocks can be seen as a competitive advantage.

How come you think doing that is not part of the foundry business?

 

[Doug S]

x86 may not be a "big deal" for much longer. Just look at how well Apple can run x86 on ARM under emulation. When Qualcomm bring out their laptop SoCs using the custom CPUs they got from Nuvia, I think we're going to start seeing the same on Windows too. ARM chips that can run x86 binaries almost as fast as an Intel chip, or can outperform them in native code.

Not much chance of that, I'm afraid. The Nuvia cores may be as fast as Apple's, but that won't let them emulate Windows as fast as M1 emulates x86 Mac binaries. Apple accomplished this using static translation, which they were able to do because they control the hardware, the software, the API, even the compiler that essentially all Mac binaries are built with. They knew they were going to do this for YEARS and were able to drop older APIs and tweak the compiler as necessary to make the job easy. Well not easy, it was still extremely complex, but easy enough they made it work.

Qualcomm doesn't have any control over Windows, or its compiler or API, and hasn't had any time to prepare for this. The performance they'll get is whatever it is now, multiplied by whatever percentage faster the Nuvia core is. Qualcomm's x86 emulated performance will be no threat to Intel, even if Nuvia gives them cores that beat Intel's in native code.

 

[DrMrLordX]

All foundries offer up IP block with a "lego" approach for easy use in their foundries.

TSMC has fabbed chips for many companies that haven't taken IP blocks from TSMC; in fact, their largest-profile customers tend to either a). license their IP from ARM and/or b). develop their tech in-house.

Apple, Huawei, Qualcomm, AMD, Rockchip, nVidia, etc. have all used TSMC in the past and/or use TSMC today.

They did not license their designs from TSMC, at least in the best-known examples of ICs they have fabbed with TSMC. They might use some of TSMC's packaging tech, granted . . .

Regardless, that is a defining characteristic of a silicon foundry.

For Intel to be able to include x86 specific IP blocks can be seen as a competitive advantage.

Is that the only thing Intel will be doing through their foundry business?

 

[moinmoin]

TSMC has fabbed chips for many companies that haven't taken IP blocks from TSMC;

I don't think you are aware what an IP block consists of. Even something as basic as preconfigured SRAM can be an IP block. Licensing ARM cores doesn't preclude using IP block. Rather TSMC etc. offer standard ARM cores as ready made IP block, saving the customer the work to port and verify that core on that TSMC node again. Same happens with plenty other stuff customers may want to have in chips.

 

[DrMrLordX]

I don't think you are aware what an IP block consists of. Even something as basic as preconfigured SRAM can be an IP block. Licensing ARM cores doesn't preclude using IP block. Rather TSMC etc. offer standard ARM cores as ready made IP block, saving the customer the work to port and verify that core on that TSMC node again. Same happens with plenty other stuff customers may want to have in chips.

While that is true, my point still stands.

Is Intel going to fab clean sheet designs for anyone or not?

Rockchip has already played footsie with Intel in the past:

Newsroom Home

Intel Newsroom home page

newsroom.intel.com

If Rockchip wants to go to Intel and produce an ARM core that they licensed and developed in-house on 10ESF, can they do it? Would they even want to?

 

[BorisTheBlade82]

As per my understanding, yes, that is exactly what they could do. Intel will offer basically the same service as TSMC. And on top of that the ability to include Intel x86 cores if anyone wishes.

/edit: Otherwise how would it be possible to get Qualcomm as a prospective customer?

 

[DrMrLordX]

As per my understanding, yes, that is exactly what they could do. Intel will offer basically the same service as TSMC. And on top of that the ability to include Intel x86 cores if anyone wishes.

That is what I thought.

Yes, there may be some interest in Intel's unique IP blocks that they can offer to their customers; however, Intel really needs to attract more customers than just those if they want to compete. And they will very much be in competition with other foundries for those customers.

Otherwise how would it be possible to get Qualcomm as a prospective customer?

Intel gave up on their modem business some time ago, but there's nothing saying Qualcomm can't do it better on an Intel node than Intel could.

 

[BorisTheBlade82]

They are interested in the 20A process. They will hardly want to use that only for modems.
Rather they realized that it will get harder and harder to acquire leading edge capacity at TSMC.

Intel’s First High-Profile IFS Fab Customer: Qualcomm Jumps on Board For 20A Process

www.anandtech.com

 

[DrMrLordX]

They are interested in the 20A process. They will hardly want to use that only for modems.

Intel is going to need as much of that as they can get for their own products. Wonder how much extra capacity they'll have for Qualcomm?

 

[Doug S]

They are interested in the 20A process. They will hardly want to use that only for modems.
Rather they realized that it will get harder and harder to acquire leading edge capacity at TSMC.

Intel’s First High-Profile IFS Fab Customer: Qualcomm Jumps on Board For 20A Process

www.anandtech.com

There is nothing in the Intel/Qualcomm announcements that say which products will be built in Intel fabs or how many. Qualcomm will enter the ARM server market via its Nuvia acquisition, if for example they made those at Intel and still made Snapdragon at TSMC it would fulfill the terms of the announcement without being a whole lot of volume.

 

[DrMrLordX]

There is nothing in the Intel/Qualcomm announcements that say which products will be built in Intel fabs or how many. Qualcomm will enter the ARM server market via its Nuvia acquisition, if for example they made those at Intel and still made Snapdragon at TSMC it would fulfill the terms of the announcement without being a whole lot of volume.

It is doubtful that Qualcomm will use Nuvia to enter the ARM server market. Feel free to surprise me, Qualcomm, but I suspect that they'll use Nuvia's innovations to bolster their mobile offerings (particularly laptop SoCs). Furthermore, they'll probably tap Samsung nodes for those products well in advance of Intel making their 20A node available to foundry customers.

 

[DisEnchantment]

It is doubtful that Qualcomm will use Nuvia to enter the ARM server market. Feel free to surprise me, Qualcomm, but I suspect that they'll use Nuvia's innovations to bolster their mobile offerings (particularly laptop SoCs). Furthermore, they'll probably tap Samsung nodes for those products well in advance of Intel making their 20A node available to foundry customers.

I see the same thing as well, with everyone and their dog doing run of the mill Arm SoC designs, Samsung, Huawei, new entrants Oppo and Google etc., (not to mention Mediatek ) it is clear where the threat is. If not for their 5G IP the situation could have gotten worse
As someone who deployed Centriq before, doubtful we will do it again unless something big happens.

 

[Doug S]

It is doubtful that Qualcomm will use Nuvia to enter the ARM server market. Feel free to surprise me, Qualcomm, but I suspect that they'll use Nuvia's innovations to bolster their mobile offerings (particularly laptop SoCs). Furthermore, they'll probably tap Samsung nodes for those products well in advance of Intel making their 20A node available to foundry customers.

If they don't they won't keep the team past whatever (usually 18 to 36 month) lockup to stick around post acquisition.

They didn't leave Apple over wanting to pursue a server CPU, only to be acquired by a company that doesn't intend to pursue servers either. Sure, the millions in their pocket made them willing to sell out regardless of Qualcomm's plans, but they'll leave and try again the moment their lockup is over if all Qualcomm is doing is laptops and mobile SoCs. At least Apple had the Mac Pro for them to look forward to, which isn't exactly a server but is closer than what Qualcomm will be doing if they don't do servers.

 

[NTMBK]

If they don't they won't keep the team past whatever (usually 18 to 36 month) lockup to stick around post acquisition.

They didn't leave Apple over wanting to pursue a server CPU, only to be acquired by a company that doesn't intend to pursue servers either. Sure, the millions in their pocket made them willing to sell out regardless of Qualcomm's plans, but they'll leave and try again the moment their lockup is over if all Qualcomm is doing is laptops and mobile SoCs. At least Apple had the Mac Pro for them to look forward to, which isn't exactly a server but is closer than what Qualcomm will be doing if they don't do servers.

Independent ARM server business doesn't seem to be a viable pursuit. Only the hyperscalers want enough volume, and they have their own internal teams.

Laptop seems like a good target for Nuvia tech.

 

[jpiniero]

Independent ARM server business doesn't seem to be a viable pursuit. Only the hyperscalers want enough volume, and they have their own internal teams.

Yup, that's why nuvia sold.

 

[DrMrLordX]

If they don't they won't keep the team past whatever (usually 18 to 36 month) lockup to stick around post acquisition.

Why??? Do you imagine they all want jobs elsewhere now that they have a chance to work with Qualcomm on their consumer-level SoCs?

 

[moinmoin]

Intel is going to need as much of that as they can get for their own products. Wonder how much extra capacity they'll have for Qualcomm?

The whole reason for a semi-independent IFS is going for scale which helps recuperating all the necessary investment and R&D in fabs and new nodes. For this Intel wants and needs plenty customers for its foundries. If Intel proper again gets preferential treatment they can shutter IFS again right away.

 

[Doug S]

Why??? Do you imagine they all want jobs elsewhere now that they have a chance to work with Qualcomm on their consumer-level SoCs?

If they weren't happy working for Apple on their consumer level SoCs why would the "chance" to work for Qualcomm on consumer level SoCs be compelling to them?

They will either take their loot and retire or leave for somewhere that gives them the opportunity to build the server SoCs they wanted to. Whether that's another startup or they go to work for a hyperscaler like Amazon, they won't be working for Qualcomm a minute longer than they have to if Qualcomm has no server plans.

 

[NTMBK]

If they weren't happy working for Apple on their consumer level SoCs why would the "chance" to work for Qualcomm on consumer level SoCs be compelling to them?

They will either take their loot and retire or leave for somewhere that gives them the opportunity to build the server SoCs they wanted to. Whether that's another startup or they go to work for a hyperscaler like Amazon, they won't be working for Qualcomm a minute longer than they have to if Qualcomm has no server plans.

Maybe some of the senior leadership will depart, sure. But I'm sure enough of the team will find the idea of stealing the laptop market from Intel compelling, and if Qualcomm have any sense they will be spending the 18-36 months of lock in to transfer in their own people as successors.

 

[DrMrLordX]

The whole reason for a semi-independent IFS is going for scale which helps recuperating all the necessary investment and R&D in fabs and new nodes. For this Intel wants and needs plenty customers for its foundries. If Intel proper again gets preferential treatment they can shutter IFS again right away.

Doesn't inspire much confidence in future customers if Intel can or would shutter their foundry service at the drop of a hat. I do think Intel will find use for their 10ESF node going forward, though, so it won't hurt them to use Foundry 2.0 as an excuse to build out more capacity than the shaoreholders/board would authorize otherwise.

If they weren't happy working for Apple

How do you know they weren't happy working for Apple? Maybe they just wanted a bigger payout, which is what founding members of the company presumably just got.

 

[TheELF]

x86 may not be a "big deal" for much longer. Just look at how well Apple can run x86 on ARM under emulation. When Qualcomm bring out their laptop SoCs using the custom CPUs they got from Nuvia, I think we're going to start seeing the same on Windows too. ARM chips that can run x86 binaries almost as fast as an Intel chip, or can outperform them in native code.

Yeah I would love to look at that, where is the link? Last time I checked arm couldn't even run x64 at all and qualcom/snapdragon was pretty slow at emulating x86.

The times are-a-changin, though, and ARM is showing that it is even more of a Swiss army knife ISA than x86. ARM can be used in anything from cell phones to servers; in fact, the same core design (with tweaks made to branch prediction) can be used in a cell phone or a server! Look at all the places A76 cores find themselves thesedays.

That's not times are-a-changin, that's the basic idea behind creating risc, make a CPU that has "less" (only those you need) instructions and executes those in a single cycle, hence making it more efficient and faster but only in the instructions that it has natively.

The only reason for using ARM is because smartphone batteries aren't good enough yet to support real CPUs, and servers basically the same they can't use as much power as they want to.

 

[NTMBK]

Yeah I would love to look at that, where is the link? Last time I checked arm couldn't even run x64 at all and qualcom/snapdragon was pretty slow at emulating x86.

The 2020 Mac Mini Unleashed: Putting Apple Silicon M1 To The Test

www.anandtech.com

Apple Silicon M1 Emulating x86 is Still Faster Than Every Other Mac in Single Core Benchmark

The first native benchmarks of Apple's M1 chip appeared on the Geekbench site last week showing impressive native performance. Today, new...

www.macrumors.com

Introducing x64 emulation in preview for Windows 10 on ARM PCs to the Windows Insider Program

Updated 11/16/2021: x64 emulation for Windows is now generally available in Windows 11. For those interested in experiencing this, a PC running Windows 11 on Arm is required. Today, we're releasing the first preview of x64 emulation

blogs.windows.com

What Windows 11 means for Windows on Arm, and why it will bring more big name apps | TechRepublic

Making it easier for developers to port apps to Arm at their own speed without losing the convenience of emulation should make Windows on Arm more credible.

www.techrepublic.com

On Windows 11 you can now emulate x64. Snapdragon was pretty slow, and that is partly because the CPU cores just weren't up to it. Nuvia have brought in all new custom server-class CPU cores, which should make them much, much faster.