Let’s see what they deliver. If we look in Intel’a history, IDM 2.0 isn’t too new. Just “old strategies, but now with vigor!”
It doesn’t solve Intel’s management culture woes, its market arrogance (that will always bite any new initiative until it’s weak), and its utter lack of consumer understanding on mobile / thin-clients.
Foundry Services is just a rebrand of a decade-old failure. Intel has been looking for foundry customers (both with in-house and custom IP, including IA!) and they’re always claiming they’re doing great. Intel’s CFO was bragging to Reuters in 2011,
> 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.
Nothing new. Same-old strategy, nearly the same people, same corporate code problems.
Even RISC-V isn’t new; Intel has known for decades that its x86 designs have inherent baggage and will not work everywhere, especially in power constrained environments.
Intel repeatedly tried to pivot to Arm in the 1990s and they still hold an Arm v6 arch license. It’s literally the 1990s and early 2000s in Intel.
The RISC-V deal is again just…so what? Intel literally owned StrongARM, one of the highest performing Arm CPUs of the day, and flailed with XScale, selling it off and killing the division. StrongARM was developed by Dan Dobberpuhl, who promptly left once Intel took over.
Where did Dan go? Founded PA Semi and then went to Apple. 😂
Intel doesn’t care about anything but its own x86 cores. Trying to make IFS succeed while maintaining that “we’re a neutral foundry! We won’t steal your designs and you have the Intel Promise” sounds just not enticing.
Any IP firm that goes to IFS today is making an enormous risk and it is not only node timelines.
The latest node: Lakefield, Celeron, Pentium, Atom and Core-Y are also fabricated on the newest nodes. None are able to supplant the primary Core U / H / S / K / X CPUs. :-/
Intel foundries building non-x86 CPUs: It's been done before in the late 1990s / early 200s, for the exact same reasons, and it flopped so hard, Intel had to sell the unit off to Marvell.
— 1997: Intel "acquires" StrongARM, a RISC CPU IP firm, for $700m ($1.2b in 2021 dollars) — 1997 to 2006: Intel fabricates the newly-acquired Arm-based CPUs at Intel foundries, up to 45nm (PXA935) — 2006: Intel, having done incredibly poorly and floundering a once-cutting-edge Arm team, sells the XScale's PXA division to Marvell for $600m + liabilities
You can still read Intel's old press releases on their site, which is wild to me:
"The Intel® StrongARM® SA-1110 Microprocessor (SA-1110) is a highly integrated communications microcontroller that incorporates a 32-bit StrongARM® RISC processor core, system support logic, multiple communication channels, an LCD controller, a memory and PCMCIA controller, and general-purpose I/O ports."
These are genuinely ancient strategies. Buy StrongARM in 1997, buy RISC-V in 2021. Think they are "too weak" compared to your big cores, so you refuse to grow their market each year like everyone else (and how your own x86 CPU teams do). Figure out that it's simply losing money compared to all other competitors, so you cut your losses and proclaim, "IDM 3.0 is coming soon.... Intel acquires Qualcomm!" /s
LMAO, their Sapphire Rapids on 10nmSF which is going to take on Milan is not even out, which is where real money is at. And the Tiger Lake is another source of their money, but DC market is what kept Intel alive.
TSMC 3nm is too far away and we do not even know what processors they would be making, on top of that, Intel making a TSMC based silicon that is gen1 is going to be trash no question, a completely new set of manufacturing process for Intel since how many years ? It never happened.
That said, what are those CPUs ? Server processors which are Xeon based or the Laptop silicon ? Since Biglittle Intel, the Silicon will be made different unlike the usual XCC / HCC lineup that Intel has been doing since decades. Because there's no Server x86 processor which is going to be made like this, they won't be wasting the die space on Server sockets with that sort of joke of power saving BS. Esp when AMD Zen 4 is pegged at 80+ Physical x86 cores.
Intel is only going to play safe game, TSMC Tech for their CPUs and trying to beat AMD in contracts, until they can answer to Milan Intel is nothing but a corporate Goliath only fit for corruption.
Too far away ?? are you sure about this?? with Intel and Apple hardly pressing about ?? :) come on be realistic. After all, looking carefully TSMC 3nm is only less than 26% more dense than Intel 7nm (at the cost of some expensive double exposures with EUV), so Intel want it to have more volume, definitively 3nm is not for performance and average power reduction, they can have the same doing a 7nm+ or 7nm++ spending less money thanks to the single exposure.. Intel want to charge the huge amount of bad wafers on TSMC instead on itself. Unfortunately Amd is out of this big boys game and even on plain 7nm it can not deliver to customers. Funny enough in these days being an IDM is an advantage. You can deliver, your competitors NOPE.
Too much of talk and no actions to prove is what Intel nowadays, their talk is essentially BS with latest corporate policies infused in absolutely shameless Intel.
Rocket Lake was sandbagged by the slow cache ruining it's IPC gain and runs way hotter than 10900K, what a joke to make it worse the IMC has been castrated now, CML can run 4000MHz with ease on 10th gen but 11th gen cannot even run stable at 3733 awful trash. HEDT has been dead since ages.
As for me the only advantage Intel has is, Stable clock speed. No matter which CPU you pick up it will run at the Clock you set, no dynamic BS drama like AMD no core recycle BS no Windows Powerplan drama (I know this is due to scheduler for AMD CCD chiplets and UEFI BIOS CPPC etc). Second thing is most of the Intel HW can run Win7 without issues (yeah I want that OS, personal preference as it runs all stuff very reliably and looks much nicer than 10 and 11 and no telemetry BS, it has Updates, go to update catalog and check for yourself). Third is no USB Stability sporadic problems like AMD. Yes they have some HW BS like Foxville disaster needing 3 HW revisions to fix.
Keeping that aside, looking at the 10nm TGL is maybe fine, but its a BGA product, useless junk in 3-4 years and the BIOS on those machines is absolutely pathetic, Yep it can run games at high FPS no question but it's not a great product to point out as wow, wasted potential. Should have released an LGA1200 10C variant on Desktop.
Now Intel is going even more anemic garbage like BigLittle. Horrible shame. First copied AMD's chiplet with that Intel Xe, later Gear memory crap, now from ARM inferior small pathetic cores. While Apple is ramping up big cores in their own Silicon production. All because $$$$$ low key junk laptops sell a ton more so why not collude with M$ and push for junk irreparable trash with inferior performance. Muh Battery life. But that is now even in Desktops because unable to compete against AMD's SMT technology leadership.
Gracemont may be 1/3 the area of the Golden Cove cores. Still runs x86. Runs avx2 simd, like AMD cores. Have to wait and see on the performance ... looks like they will have SKUs with no Gracemonts, so no one is forcing you to buy the small cores.
12900K ES jst got a leak. It's scoring 11K marks on CBR20. That's stomping 5950X but probably it's at 5.3GHz, dunno the power draw or the max TDP and how much role does Gracemont cores play a role and how their clock speed varies.
AMD boldly put out V-Cache Zen3+ already at 15% gaming boost, with confidence. And Zen 4 AM5 will launch in 2022 Q3-4, by that time Intel Raptor Lake is also coming, just based on this rumor of ADL leak, Intel is somewhat back in the game..Still gotta wait and see.
Engineers in American and European companies can never spend the man hours Chinese and Taiwanese do. Don't think Intel can again lead in semiconductor manufacturing like it used to do.
Taiwan has the expertise. China has dumped ungodly amounts of money and gotten nowhere near samsung/tsmc/intel even after stealing IP and subsidizing every loss. fact that TSMC has swung out in front just tells you it's not as simple as throwing man hours at the problem.
What will happen at this stupid event is Intel will tell us what they plan to do for the next ten years (including tons of stuff they have barely working in the lab). And THEN, since these claims have been made public and could be considered legally actionable, they will be locked into that roadmap even if it makes no sense.
Just like happened with 10nm, and is now happening with 7nm. They will make promises about density, about GAA, about new materials, about buried power rails. And because those promises are in the open, they won't be able to walk them back when something goes wrong.
WTF do they do this? Who knows? But THAT is the essential difference here. TSMC STFU until they have something working. When things go wrong, they modify the plan, they don't continue digging a hole. And they can do this, because they haven't promised the world a ten year roadmap.
A SENSIBLE Intel would treat this like a TSMC event: - we're working on 7nm. We'll tell you more about it when it's ready. - meanwhile we've slightly tweaked 10nm to create our 9nm node, which we will use to test this ONE (and only one) modification to our existing technology stack
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
28 Comments
Back to Article
ikjadoon - Monday, July 12, 2021 - link
Do the asterisks mean to point to a footnote?Ian Cutress - Monday, July 12, 2021 - link
It was meant to be a line defining Dr. Kelleher's role, but I ended up integrating it into the paragraph anyhow.dullard - Monday, July 12, 2021 - link
Ian, you corrected 1 of the 3 missing footnotes.ikjadoon - Monday, July 12, 2021 - link
Let’s see what they deliver. If we look in Intel’a history, IDM 2.0 isn’t too new. Just “old strategies, but now with vigor!”It doesn’t solve Intel’s management culture woes, its market arrogance (that will always bite any new initiative until it’s weak), and its utter lack of consumer understanding on mobile / thin-clients.
Foundry Services is just a rebrand of a decade-old failure. Intel has been looking for foundry customers (both with in-house and custom IP, including IA!) and they’re always claiming they’re doing great. Intel’s CFO was bragging to Reuters in 2011,
> 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.
https://www.theregister.com/2011/05/27/stacy_smith...
Nothing new. Same-old strategy, nearly the same people, same corporate code problems.
Even RISC-V isn’t new; Intel has known for decades that its x86 designs have inherent baggage and will not work everywhere, especially in power constrained environments.
Intel repeatedly tried to pivot to Arm in the 1990s and they still hold an Arm v6 arch license. It’s literally the 1990s and early 2000s in Intel.
The RISC-V deal is again just…so what? Intel literally owned StrongARM, one of the highest performing Arm CPUs of the day, and flailed with XScale, selling it off and killing the division. StrongARM was developed by Dan Dobberpuhl, who promptly left once Intel took over.
Where did Dan go? Founded PA Semi and then went to Apple. 😂
Intel doesn’t care about anything but its own x86 cores. Trying to make IFS succeed while maintaining that “we’re a neutral foundry! We won’t steal your designs and you have the Intel Promise” sounds just not enticing.
Any IP firm that goes to IFS today is making an enormous risk and it is not only node timelines.
JayNor - Monday, July 12, 2021 - link
There was a story here on an Intel 7nm risc-v project for 2022. Doesn't fit the "doesn't care about anything but its own x86" theme.ikjadoon - Monday, July 12, 2021 - link
It's less exciting than it seems.The latest node: Lakefield, Celeron, Pentium, Atom and Core-Y are also fabricated on the newest nodes. None are able to supplant the primary Core U / H / S / K / X CPUs. :-/
Intel foundries building non-x86 CPUs: It's been done before in the late 1990s / early 200s, for the exact same reasons, and it flopped so hard, Intel had to sell the unit off to Marvell.
— 1997: Intel "acquires" StrongARM, a RISC CPU IP firm, for $700m ($1.2b in 2021 dollars)
— 1997 to 2006: Intel fabricates the newly-acquired Arm-based CPUs at Intel foundries, up to 45nm (PXA935)
— 2006: Intel, having done incredibly poorly and floundering a once-cutting-edge Arm team, sells the XScale's PXA division to Marvell for $600m + liabilities
You can still read Intel's old press releases on their site, which is wild to me:
https://www.intel.com/pressroom/archive/releases/1...
https://www.intel.com/pressroom/archive/releases/2...
"The Intel® StrongARM® SA-1110 Microprocessor (SA-1110) is a highly integrated communications microcontroller that incorporates a 32-bit StrongARM® RISC processor core, system support logic, multiple communication channels, an LCD controller, a memory and PCMCIA controller, and general-purpose I/O ports."
These are genuinely ancient strategies. Buy StrongARM in 1997, buy RISC-V in 2021. Think they are "too weak" compared to your big cores, so you refuse to grow their market each year like everyone else (and how your own x86 CPU teams do). Figure out that it's simply losing money compared to all other competitors, so you cut your losses and proclaim, "IDM 3.0 is coming soon.... Intel acquires Qualcomm!" /s
serendip - Tuesday, July 13, 2021 - link
Intel needs to do something with mixed cores like Lakefield or ARM big.LITTLE without the hype.lemurbutton - Monday, July 12, 2021 - link
Reports are that Intel and Apple will be TSMC's biggest 3nm customers. Sell AMD stocks if you have any.Silver5urfer - Monday, July 12, 2021 - link
LMAO, their Sapphire Rapids on 10nmSF which is going to take on Milan is not even out, which is where real money is at. And the Tiger Lake is another source of their money, but DC market is what kept Intel alive.TSMC 3nm is too far away and we do not even know what processors they would be making, on top of that, Intel making a TSMC based silicon that is gen1 is going to be trash no question, a completely new set of manufacturing process for Intel since how many years ? It never happened.
That said, what are those CPUs ? Server processors which are Xeon based or the Laptop silicon ? Since Biglittle Intel, the Silicon will be made different unlike the usual XCC / HCC lineup that Intel has been doing since decades. Because there's no Server x86 processor which is going to be made like this, they won't be wasting the die space on Server sockets with that sort of joke of power saving BS. Esp when AMD Zen 4 is pegged at 80+ Physical x86 cores.
Intel is only going to play safe game, TSMC Tech for their CPUs and trying to beat AMD in contracts, until they can answer to Milan Intel is nothing but a corporate Goliath only fit for corruption.
JayNor - Monday, July 12, 2021 - link
Sapphire Rapids is 10esf. broadly sampling since Dec 2020.johannesburgel - Monday, July 12, 2021 - link
Broadly sampling? There are still no Sapphire Rapids ES2 chips around.yeeeeman - Tuesday, July 13, 2021 - link
Lets not forget that this kind of thinking towards Intel has been predicated in 2015 towards AMD, that it couldn't make any comeback and boy they did!Gondalf - Wednesday, July 14, 2021 - link
Too far away ?? are you sure about this?? with Intel and Apple hardly pressing about ?? :) come on be realistic. After all, looking carefully TSMC 3nm is only less than 26% more dense than Intel 7nm (at the cost of some expensive double exposures with EUV), so Intel want it to have more volume, definitively 3nm is not for performance and average power reduction, they can have the same doing a 7nm+ or 7nm++ spending less money thanks to the single exposure..Intel want to charge the huge amount of bad wafers on TSMC instead on itself.
Unfortunately Amd is out of this big boys game and even on plain 7nm it can not deliver to customers. Funny enough in these days being an IDM is an advantage.
You can deliver, your competitors NOPE.
Spunjji - Tuesday, July 13, 2021 - link
It's a fun game triangulating troll accounts on different sights by taking note of which ones post this same un-sourced shite.Silver5urfer - Monday, July 12, 2021 - link
Too much of talk and no actions to prove is what Intel nowadays, their talk is essentially BS with latest corporate policies infused in absolutely shameless Intel.Rocket Lake was sandbagged by the slow cache ruining it's IPC gain and runs way hotter than 10900K, what a joke to make it worse the IMC has been castrated now, CML can run 4000MHz with ease on 10th gen but 11th gen cannot even run stable at 3733 awful trash. HEDT has been dead since ages.
As for me the only advantage Intel has is, Stable clock speed. No matter which CPU you pick up it will run at the Clock you set, no dynamic BS drama like AMD no core recycle BS no Windows Powerplan drama (I know this is due to scheduler for AMD CCD chiplets and UEFI BIOS CPPC etc). Second thing is most of the Intel HW can run Win7 without issues (yeah I want that OS, personal preference as it runs all stuff very reliably and looks much nicer than 10 and 11 and no telemetry BS, it has Updates, go to update catalog and check for yourself). Third is no USB Stability sporadic problems like AMD. Yes they have some HW BS like Foxville disaster needing 3 HW revisions to fix.
Keeping that aside, looking at the 10nm TGL is maybe fine, but its a BGA product, useless junk in 3-4 years and the BIOS on those machines is absolutely pathetic, Yep it can run games at high FPS no question but it's not a great product to point out as wow, wasted potential. Should have released an LGA1200 10C variant on Desktop.
Now Intel is going even more anemic garbage like BigLittle. Horrible shame. First copied AMD's chiplet with that Intel Xe, later Gear memory crap, now from ARM inferior small pathetic cores. While Apple is ramping up big cores in their own Silicon production. All because $$$$$ low key junk laptops sell a ton more so why not collude with M$ and push for junk irreparable trash with inferior performance. Muh Battery life. But that is now even in Desktops because unable to compete against AMD's SMT technology leadership.
JayNor - Monday, July 12, 2021 - link
Gracemont may be 1/3 the area of the Golden Cove cores. Still runs x86. Runs avx2 simd, like AMD cores. Have to wait and see on the performance ... looks like they will have SKUs with no Gracemonts, so no one is forcing you to buy the small cores.Silver5urfer - Tuesday, July 13, 2021 - link
12900K ES jst got a leak. It's scoring 11K marks on CBR20. That's stomping 5950X but probably it's at 5.3GHz, dunno the power draw or the max TDP and how much role does Gracemont cores play a role and how their clock speed varies.AMD boldly put out V-Cache Zen3+ already at 15% gaming boost, with confidence. And Zen 4 AM5 will launch in 2022 Q3-4, by that time Intel Raptor Lake is also coming, just based on this rumor of ADL leak, Intel is somewhat back in the game..Still gotta wait and see.
Dizoja86 - Monday, July 12, 2021 - link
Tell us how you really feel....sseemaku - Monday, July 12, 2021 - link
Engineers in American and European companies can never spend the man hours Chinese and Taiwanese do. Don't think Intel can again lead in semiconductor manufacturing like it used to do.whatthe123 - Monday, July 12, 2021 - link
Taiwan has the expertise. China has dumped ungodly amounts of money and gotten nowhere near samsung/tsmc/intel even after stealing IP and subsidizing every loss. fact that TSMC has swung out in front just tells you it's not as simple as throwing man hours at the problem.name99 - Tuesday, July 13, 2021 - link
It's not about man hours, it's about management!What will happen at this stupid event is Intel will tell us what they plan to do for the next ten years (including tons of stuff they have barely working in the lab).
And THEN, since these claims have been made public and could be considered legally actionable, they will be locked into that roadmap even if it makes no sense.
Just like happened with 10nm, and is now happening with 7nm. They will make promises about density, about GAA, about new materials, about buried power rails. And because those promises are in the open, they won't be able to walk them back when something goes wrong.
WTF do they do this? Who knows? But THAT is the essential difference here.
TSMC STFU until they have something working. When things go wrong, they modify the plan, they don't continue digging a hole. And they can do this, because they haven't promised the world a ten year roadmap.
A SENSIBLE Intel would treat this like a TSMC event:
- we're working on 7nm. We'll tell you more about it when it's ready.
- meanwhile we've slightly tweaked 10nm to create our 9nm node, which we will use to test this ONE (and only one) modification to our existing technology stack
But Intel is not run by sensible people.
Machinus - Monday, July 12, 2021 - link
Looking forward to Intel's 13nmgrant3 - Monday, July 12, 2021 - link
Intel is skipped from 14nm -> 10nm in their process naming. is "13nm" a typo or some kind of obscure joke?Machinus - Monday, July 12, 2021 - link
Intel's marketing for the last 6 years has been one big typoArcadeEngineer - Tuesday, July 13, 2021 - link
Looking forward to these people finally having to come up with another bloody jokeMachinus - Tuesday, July 13, 2021 - link
K. Intel 12nm coming up!prime2515103 - Tuesday, July 13, 2021 - link
LOLInchesmorethanm - Wednesday, July 14, 2021 - link
Looking forward to hear about Intel's 24" wafer process that will make all those nanometer wars meaningless.