The sustained performance is formidable this gen. despite what is still likely to be handicap of Samsung's 4nm process vs TSMC's 5nm.
Page 19's projections for Adreno GPUs performance in PC laptops is 15+ TFLOPs which greatly exceeds even the M1Max's throughput, so that number seems wildly ambitious for now, although I'm sure the M2Max will be in that neighborhood by 2023.
Odd to see no power nor thermal data on the chart. Looking at last year, Qualcomm’s Adreno 660 (SD888) had very high power draw, so is this next-gen “Adreno 670” actually more efficient or are they letting it run hotter for longer?
Cooling design is also key: identical test bench or two phones?
“The Snapdragon 888’s peak performance state is pretty absurd, as at its 840MHz GPU frequency I’ve measured average power of around 11W.”
11W pains me. I really hope Qualcomm can teach Apple, the king of GPU throttling, a lesson in sustained perf this year.
Agree w/ your points, it's just a tease. We'll see in about 2 weeks what they have in store, and it's been rumored that this "Adreno 730" is an entirely new architecture after the relatively long in the tooth but prolific 6xx line all the way from the Snapdragon 845.
You obviously don't know anything. Scaling up is hard. Interconnect and synchronisation is very hard, harder than the cores even. ARM even resorted to uncore boosting because they can't scale up their geometry and command pipeline.
There's nothing wildly ambitious about that since Andreno in SD 888+ reaches almost 2TFLOPS at under 10W peak power.. and Qualcomm already reached 2.6TFLOPS in an OLD 8cx designed for 15W TDP. As Apple's M1 demonstrated, arm chips can scale pretty much linearly.
GPUs (not ARM) can scale very nicely with clock-speed or die area due to the embarrassingly parallel nature of rendering. However, pushing clock speed doesn't linearly scare with power consumption, adding GPU blocks increases power, and die area for their likely design is shared with multiple other blocks and caches:
If you want to do 15 TFLOPS in a reasonable die area and target ~15-30W for a laptop, it would be very ambitious compared to current offerings. The 6XX Adreno is already very efficient relative to other designs iso-process, but scaling the 6XX to 15 TFLOPS would put them in the 80W range (on Samsung's current leading edge process.) They'd likely need the whole 8cx die for just the GPU as well.
Unfortunately this comment didn't age well. The new adreno GPU is indeed much much stronger than previous gen and fights shoulder to shoulder with the a15 GPU but power and thermals are so bad that Qualcomm is forced to switch back to tsmc (plus also yields are a big issue with Samsung 4lpe).
My big question is: December 2022 Qualcomm-Nuvia's chips will be better than Apple's… November 2020's M1… October 2021's M1 Pro/Max… 2022's M… (for MacPros)… or even 2022's M2 chips?
When they’re targeting a product release for 2023, why would they plan to target a product from 2021 let alone 2020? They need to be better at perf/watt against Apple, Intel and AMD. I’m sure those smart folks would know that and have a competitive silicon by end of next year. Ever though they come with great credentials, matching/beating Apple silicon for Nuvia with their v1 outing seems pretty monumental. Will be delighted to see more competition though.
Because the major microarchitecture for Phoenix was finished in 2020 likely—their performance simulations were beating A13 in perf/watt. Just like Zen1 couldn’t beat Skylake. It’s not like AMD tried to fail: they just needed another few generations. And then Intel came back.
I hope NUVIA does well, but let’s wait for benchmarks first. 🤝
High perf + perf/watt is the goal. Anyone can get higher perf/watt than AMD/Intel, while you can play with numbers to beat Apple’s perf/watt.
IMHO you are looking at the wrong company, their competitor is mainly Intel, not Apple. If they soundly beat Intel cpus in performance-per-watt and are comparable or better in terms of absolute performance, they have a winner on their hands.
Like essentially all other ARM chips, it will likely require custom images to boot. So, I would expect Linux support to take a while. Also, lets not forget Qualcomm's terrible support strategy. They will probably require companies to pay for driver support just like on Android. So, the Linux community will also have to reverse engineer all the drivers before it is usable.
Do we have any idea of what fab and node do they intend to use for the chip ?
Will it be Samsung 3GAE (which could nice if Samsung GAA transistors implementation does really happen to be more energy efficient than 3nm FinFet) or will it be TSMC 3nm ? Any of those 2 nodes should provide a better energy efficiency versus Intel 7 or TSMC 5nm (AMD)…
I hope it won’t be Samsung 5nm, neither TSMC 5nm because by end of 2022, Apple will very likely begin to transition to TSMC 3nm node and it is likely the Nucia designed chip would be at a disavantage versus Apple…
Looking at their timeframe, of only delivering limited samples to customers in Q3 2022 and not starting mass production for consumers until 2023, I would assume they are targeting 3-nm. If they want to have any hope of matching Apple, they know 3-nm will be needed because Apple has jumped to each new process node as fast as it can for a number of years now. Qualcomm uses both Samsung and TSMC, so it is anyone's guess as to which they might use.
More Hot Air BS claims by Qcomm, lets recall how the Centriq was heralded as next generation and marketed as well. And where is that ? Dead and they even axed the whole thing down along with Staff, chief people.
Nuvia is now the next miracle of the Computing era ? yeah sure Qcomm. Fool us once but fool us second time ? Apple M1 got destroyed by all Mobile laptops in the general workloads yeah it's fast on the Cinebench and when loading applications which are not made optimized for Apple silicon is not great. And gaming performance was slaughter fest, a 3060 destroyed it.
Qcomm cannot even beat Apple, all the CPUs are like a generation behind, yeah application performance is similar apart from SPEC leadership but still they are slow. Apple destroys them.
Alder Lake despite it's massive flaws like E cores being only relegated to SMT workloads and not gaming, and new DDR5 etc. The performance is great on P cores which is what really matters and they will slaughter Apple processors. Even at 57Billion transistors in SoC they are marginally at best, and this joke Qualcomm will beat all that ? And PC is Desktop LGA not some BGA crapware with limited power and limited everything.
Finally Software. Where is 8cX ? Dead, nowhere to be found and Win11 is now 64-bit x86 emulation, and with Intel Big little and Zen 4c it's not possible for this junk to exist., esp the whole nonsensical emulation part.
GPUs are one of the easier things to scale up and down. If Qualcomm had some miracle GPU tech one would think their phone SoCs would get it first. That's their bread and butter! But here I have this suspicion they're just spamming silicon on a low volume part. Easy to win the benchmarks, hard to commercialize.
Your reference to Centriq is specious. Centriq was not abandoned because it was technologically untenable. To the contrary. It was cut in what became a successful effort to save the company as an independent entity.
Apple had ordered the contract manufacturers to withhold all royalties from Qualcomm, at the same time that Apple was itself refusing to enter a PLA, and was instead instigating the FTC v Qualcomm case.
Broadcom tried to steal the company for $79/sh, bottom feeding the company that Apple was trying to destroy.
Centriq was reluctantly abandoned as part of the effort to withstand the existential threat, a major buyback of shares was implemented, CFIUS was lobbied, and the multi-frontal litigation onslaught was defended. Nuvia gets to fulfill their server ambitions, among other products.
Thanks for sharing that - it's a perspective I've not heard.
I thought the Centriq point was pretty far out there, and was just waiting for someone knowledgeable to step in and pour some cold water on it. I'm glad you did.
Centriq was in 2017 and Broadcomm and Apple take over was in 2018, I know Apple wanted to steal Qcomm and eat their LTE patents it was out at EETimes for that, Apple is a scum garbage. That's not new. They robbed Imagination Tech, they bankrupted GT Advanced tech as well and ruined Dialog Semi as well.
Centriq was heralded as next thing and Cloudflare did a lot of advertising, everything was dead including the people who made it that when Snapdragon team who design Kryo 820.
Nuvia is going to deliver a miracle as Server Ambitions is going to be decided once this product is out. Until then it's vaporware.
The thing I'll never quite understand about that is why Apple left the hulking husk of Imagination for someone like the Chinese to harvest of its IP. Was Apple feeling that Imagination was overplaying its hand and demanding too high an acquisition price, thus rejecting that path on principle? In practice, I think it would've been worth paying even too high a price just to swallow it and keep all that IP to themselves.
The saga of what has played out at Imagination, since then, is sad and hard to watch.
I dedicated 12 years of my wife to working at Qualcomm. Everybody there runs around thinking they have a target on my back. I never saw such paranoia. If I could do one thing for Qualcomm, it would be, give everyone a tranquilizer every day at work, there's some serious mental illness their about the company's always-impending destruction next month ...
Centriq and the custom CPU team were axed, in an effort to protect the company from a cheap takeover. They were sacrificed on the alter of corporate survival. Times have changed, and Qualcomm weathered the ruthless attacks by Apple/FTC and Broadcom, and they've shrugged off Apple's modem slot loss, by diversification. With the help if the Nuvia team, innovation will again be funded, and rule the day. Competition is good for everyone. Let's check back in a year, and see who's got what.
Big claims. Has CPU design become so easy that everybody is doing it? Doing it and then beating laptop class performance of Intel/AMD/Apple? And few years ago, people were like 'my ivybridge can do all my work, so I am off to buy Ipad'. And here we are in post-pc era with every company targeting laptop market.
I would have like to see or hear any comparison of NUVIA cores vs. ARM's "stock" X2. If QC can beat those AND be more efficient, they have a good chance. Otherwise, these SoCs will be niche players in the ultraportable segment at best. Apart from this, a lot depends on how much oomph Microsoft is willing to throw behind Windows on ARM. Native Win 11 plus native MS Office, Teams etc, all taking advantage of the SVEs NUVIA better has on offer, and it could work; maybe.
Mac marketshare is under 10%, their revenue cut towards over all Apple money stream is also 10%, and world runs on Windows not Mac. Mac is very expensive that price barrier is impossible for many people. Plus new OS and learning curve on top.
Games do not work on Mac as well, that is one major thing. More software restrictions on Mac like ARM only and no more 32Bit, all the old games are dead on Mac thanks to Apple.
I give current Microsoft leadership credit for being much smarter than you (you sound like a Ballmer-era guy -- how are you enjoying your Zune and Windows Mobile phone these days?)
Mac marketshare bottomed out at around 2% (maybe lower) at the turn of the century. It has been very gradually creeping higher since then. Today it's a bit lower than 10% globally, greater than 10% in the US. And almost all the share that Apple takes is at the high end -- by definition, since Apple doesn't compete much in the low-end.
Microsoft and Intel have clearly taken notice of this. It helps explain why Microsoft got into hardware and it explains Intel's recent ad campaign targeted at the Mac.
(anyone who remembers Zune has a fair chance of getting that reference.)
> Apple doesn't compete much in the low-end.
Apple TV. Clearly aimed at establishing their presence in the living room.
Apple is big and needs to keep growing. This will push them into new markets and drive them towards dominance. And they won't stop at games consoles. I'm predicting Apple Healthcare devices, before the decade is up.
Given Cortex-X2 devices will be out soon, beating those wouldn't be good enough in order to be competitive in 2023. Given NUVIA's experience it's likely they come up with a great design, but remember others are not standing still in the meantime.
I've said it many years ago (since 2014) that Apple will eventually produce superior chips and humiliate Intel and AMD..
ALSO, since first hearing about Nuvia acquisition by QC, I claim Qualcomm is the one to watch out for now. And it's because simple reason everyone keeps ignoring - people. Nuvia is a startup ran by THREE BEST CPU EXPERTS, the same one that created Apple's M architecture. Enough said.
Agreed. Weird that people make predictions based on Qualcomm’s past performance rather than the past performance of the guys who actually designed the darn chip. Companies are mostly just the people who work there (plus IP).
And who are those EXPERTS you talking about ? I wonder if it's going to shatter Redwood Cove and Zen 5, which are going to destroy every single CPU out there.
Apple M got nothing, all that fluff marketing got busted out once the benches were out, it's competitive but only with 5nm design vs Intel's old parts and Zen's inferior designs. In GPU it's a massacre by Nvidia and AMD.
Bonus is $2666 starting price tag with a BGA soldered SSD junk and Max SoC is $3300+. Qcomm talks a lot, their CPUs cannot even beat Apple and some magical Nuvia is going to deliver industry shattering performance ? I already saw what Apple had to do with 57bn transistors with massive GPU cores and still unable to beat the cheaper manufactured Silicon from AMD, Intel and Nvidia. This joke of ARM taking over the world has been there since 10 years still people beat the same old drum.
Wait for Alder Lake BGA processors wreck havoc on M1 Max. That's just Intel 10nm. Once Zen 4 launches with Intel's Meteor Lake it will be a big bloodbath.
This dude's obviously a bit off his rocker, but I do think Alder Lake has a legitimate chance in succeeding in mobile. It will be interesting to see how it fairs against the M1, new Snapdragon chips, and possibly these new Nuvia ones in 2023. I agree with Ian that 2022 is going to be a really fun year for CPUs.
> It will be interesting to see how it fairs against the M1, new Snapdragon chips, > and possibly these new Nuvia ones in 2023.
For sure. I'll have my popcorn ready, for when the reviews are up. I don't think Alder Lake will beat Apple or Nuvia, but the question is more one of whether it'll be competitive enough to stave off the threat of ARM eating into its Windows laptop marketshare.
I don't quite know what to make of the fact that Intel launched Alder Lake on desktop, first. Was that mostly a PR move? Or do they face some efficiency challenges in their new Intel 7 node?
I'm also intrigued by the lack of any messaging around Gracemont-based entry-level SoCs, like a successor to Jasper Lake. Again, this could speak to power-efficiency challenges they're facing, or simply (more likely) prioritizing their limited Intel 7 production capacity. However, failure to counter Mediatek's aggressive moves in the Chromebook market segment could cost them a substantial share of it.
Nuvia is a cool product addition but they were never gunning for low-power products. That means that, once again, we will see a garbage A55 or an even-more-garbage A510, barring a major surprise -- essentially, Qualcomm would have had to have lied about leaving the custom CPU space and have been working on a low-end CPU core design since about the time they announced they were leaving the space.
By contrast, Alder Lake and its Atom E-Cores (as well as a massive Atom performance improvement timed to land as Alder Lake landed) speak to responding to Apple as soon as possible. It's clear one of these two companies feared competition, and one didn't. Hopefully, the continued inadequacy of Qualcomm mobile SoCs leads to an eventual US antitrust lawsuit. It's disturbing that nothing has happened to force their hand.
Nuvia has the knowledge and experience of how Apple's low-power cores are designed. So, it definitely seems like the *could* make good low-power cores, if they wanted.
> Alder Lake and its Atom E-Cores ... speak to responding to Apple
It might be the overall ARM-based laptop threat, not specifically Apple. I don't expect Alder Lake to be very competitive against Apple, within a similar power envelope.
> It's clear one of these two companies feared competition, and one didn't.
I think Qualcomm's decision to stop making mobile CPU cores was based on the cost and trouble of trying to keep just marginally ahead of ARM's cores. They clearly decided to differentiate their SoCs in other areas. Maybe, if their homegrown cores had been more competitive, they'd have stuck with it. That's not to cast aspersions on the engineers, either. It could be that they were simply under-resourced for the task at hand.
whatever. end consumer product will cost more than an Intel and probably Apple product since they put a huge premium on cellular connectivity like on those Arm based MS laptops/tablets despite poor performance
If like me, you follow the on-going court case regarding Gerard Williams III’s theft of Apple’s trade secrets and his “plot” to spin-up Nuvia to compete directly against Apple, you will note that while a trial is still a ways off, Mr. Williams and Nuvia are not doing well trying to spin that he/they did nothing wrong, despite discovery already quite clearly painting a highly suspicious picture, and the judge getting quite upset numerous times with Mr. Williams gamesmanship and warning that he MUST unequivocally produce all of the trade secret documents that came out that he “mistakenly” had copied to his personal devices and directly testify to the court about numerous questions he is refusing to answer fully/honestly about. You can access all records by going to the Santa Clara court website and searching by party name first:Gerard last:Williams to see the progression of this legal action.
With only $57m of initial seed money for Nuvia, they don't have a single penny left over to work on a GPU (let alone any GPU software). Therefore, Nuvia is short for NoGpuvia.
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
50 Comments
Back to Article
Raqia - Thursday, November 18, 2021 - link
They had a nice preview of "next-gen" snapdragon performance against what is likely the A15's GPU on page 18 of Jim Thompson's slides:https://d1io3yog0oux5.cloudfront.net/_33a7aedec0c1...
The sustained performance is formidable this gen. despite what is still likely to be handicap of Samsung's 4nm process vs TSMC's 5nm.
Page 19's projections for Adreno GPUs performance in PC laptops is 15+ TFLOPs which greatly exceeds even the M1Max's throughput, so that number seems wildly ambitious for now, although I'm sure the M2Max will be in that neighborhood by 2023.
ikjadoon - Thursday, November 18, 2021 - link
Odd to see no power nor thermal data on the chart. Looking at last year, Qualcomm’s Adreno 660 (SD888) had very high power draw, so is this next-gen “Adreno 670” actually more efficient or are they letting it run hotter for longer?Cooling design is also key: identical test bench or two phones?
“The Snapdragon 888’s peak performance state is pretty absurd, as at its 840MHz GPU frequency I’ve measured average power of around 11W.”
11W pains me. I really hope Qualcomm can teach Apple, the king of GPU throttling, a lesson in sustained perf this year.
From: https://www.anandtech.com/show/16463/snapdragon-88...
Before I forget: And people complained Apple didn’t label their graphs... 😅
Raqia - Thursday, November 18, 2021 - link
Agree w/ your points, it's just a tease. We'll see in about 2 weeks what they have in store, and it's been rumored that this "Adreno 730" is an entirely new architecture after the relatively long in the tooth but prolific 6xx line all the way from the Snapdragon 845.Ppietra - Thursday, November 18, 2021 - link
Anything can scale up! Apple GPU is supposedly going to scale up to 40+ TFLOPs, and that is based on a GPU from 2020.dotjaz - Sunday, January 2, 2022 - link
You obviously don't know anything. Scaling up is hard. Interconnect and synchronisation is very hard, harder than the cores even. ARM even resorted to uncore boosting because they can't scale up their geometry and command pipeline.Anything can scale **DOWN**
darkich - Friday, November 19, 2021 - link
There's nothing wildly ambitious about that since Andreno in SD 888+ reaches almost 2TFLOPS at under 10W peak power.. and Qualcomm already reached 2.6TFLOPS in an OLD 8cx designed for 15W TDP.As Apple's M1 demonstrated, arm chips can scale pretty much linearly.
Raqia - Friday, November 19, 2021 - link
GPUs (not ARM) can scale very nicely with clock-speed or die area due to the embarrassingly parallel nature of rendering. However, pushing clock speed doesn't linearly scare with power consumption, adding GPU blocks increases power, and die area for their likely design is shared with multiple other blocks and caches:https://image.itmedia.co.jp/l/im/ee/articles/2012/...
If you want to do 15 TFLOPS in a reasonable die area and target ~15-30W for a laptop, it would be very ambitious compared to current offerings. The 6XX Adreno is already very efficient relative to other designs iso-process, but scaling the 6XX to 15 TFLOPS would put them in the 80W range (on Samsung's current leading edge process.) They'd likely need the whole 8cx die for just the GPU as well.
yeeeeman - Thursday, April 28, 2022 - link
Unfortunately this comment didn't age well. The new adreno GPU is indeed much much stronger than previous gen and fights shoulder to shoulder with the a15 GPU but power and thermals are so bad that Qualcomm is forced to switch back to tsmc (plus also yields are a big issue with Samsung 4lpe).Luis Alejandro Masanti - Thursday, November 18, 2021 - link
My big question is: December 2022 Qualcomm-Nuvia's chips will be better than Apple's… November 2020's M1… October 2021's M1 Pro/Max… 2022's M… (for MacPros)… or even 2022's M2 chips?Teckk - Thursday, November 18, 2021 - link
When they’re targeting a product release for 2023, why would they plan to target a product from 2021 let alone 2020? They need to be better at perf/watt against Apple, Intel and AMD. I’m sure those smart folks would know that and have a competitive silicon by end of next year.Ever though they come with great credentials, matching/beating Apple silicon for Nuvia with their v1 outing seems pretty monumental. Will be delighted to see more competition though.
ikjadoon - Thursday, November 18, 2021 - link
Because the major microarchitecture for Phoenix was finished in 2020 likely—their performance simulations were beating A13 in perf/watt. Just like Zen1 couldn’t beat Skylake. It’s not like AMD tried to fail: they just needed another few generations. And then Intel came back.I hope NUVIA does well, but let’s wait for benchmarks first. 🤝
High perf + perf/watt is the goal. Anyone can get higher perf/watt than AMD/Intel, while you can play with numbers to beat Apple’s perf/watt.
OreoCookie - Thursday, November 25, 2021 - link
IMHO you are looking at the wrong company, their competitor is mainly Intel, not Apple. If they soundly beat Intel cpus in performance-per-watt and are comparable or better in terms of absolute performance, they have a winner on their hands.Spleter - Monday, December 13, 2021 - link
what if Windows 11 Arm version is free from rumored Qualcomm contractual exclusivity by 2022 or 2023?CrystalCowboy - Thursday, November 18, 2021 - link
"but I needed better were functional support" - this clause needs something, but I can't figure out exactly what.CrystalCowboy - Thursday, November 18, 2021 - link
But will it Linux?vlad42 - Thursday, November 18, 2021 - link
Like essentially all other ARM chips, it will likely require custom images to boot. So, I would expect Linux support to take a while. Also, lets not forget Qualcomm's terrible support strategy. They will probably require companies to pay for driver support just like on Android. So, the Linux community will also have to reverse engineer all the drivers before it is usable.Diogene7 - Thursday, November 18, 2021 - link
Do we have any idea of what fab and node do they intend to use for the chip ?Will it be Samsung 3GAE (which could nice if Samsung GAA transistors implementation does really happen to be more energy efficient than 3nm FinFet) or will it be TSMC 3nm ? Any of those 2 nodes should provide a better energy efficiency versus Intel 7 or TSMC 5nm (AMD)…
I hope it won’t be Samsung 5nm, neither TSMC 5nm because by end of 2022, Apple will very likely begin to transition to TSMC 3nm node and it is likely the Nucia designed chip would be at a disavantage versus Apple…
NextGen_Gamer - Thursday, November 18, 2021 - link
Looking at their timeframe, of only delivering limited samples to customers in Q3 2022 and not starting mass production for consumers until 2023, I would assume they are targeting 3-nm. If they want to have any hope of matching Apple, they know 3-nm will be needed because Apple has jumped to each new process node as fast as it can for a number of years now. Qualcomm uses both Samsung and TSMC, so it is anyone's guess as to which they might use.Blastdoor - Friday, November 19, 2021 - link
Apple and Intel might have booked all 3nm capacity for a whileSilver5urfer - Thursday, November 18, 2021 - link
More Hot Air BS claims by Qcomm, lets recall how the Centriq was heralded as next generation and marketed as well. And where is that ? Dead and they even axed the whole thing down along with Staff, chief people.Nuvia is now the next miracle of the Computing era ? yeah sure Qcomm. Fool us once but fool us second time ? Apple M1 got destroyed by all Mobile laptops in the general workloads yeah it's fast on the Cinebench and when loading applications which are not made optimized for Apple silicon is not great. And gaming performance was slaughter fest, a 3060 destroyed it.
Qcomm cannot even beat Apple, all the CPUs are like a generation behind, yeah application performance is similar apart from SPEC leadership but still they are slow. Apple destroys them.
Alder Lake despite it's massive flaws like E cores being only relegated to SMT workloads and not gaming, and new DDR5 etc. The performance is great on P cores which is what really matters and they will slaughter Apple processors. Even at 57Billion transistors in SoC they are marginally at best, and this joke Qualcomm will beat all that ? And PC is Desktop LGA not some BGA crapware with limited power and limited everything.
Finally Software. Where is 8cX ? Dead, nowhere to be found and Win11 is now 64-bit x86 emulation, and with Intel Big little and Zen 4c it's not possible for this junk to exist., esp the whole nonsensical emulation part.
Wrs - Thursday, November 18, 2021 - link
GPUs are one of the easier things to scale up and down. If Qualcomm had some miracle GPU tech one would think their phone SoCs would get it first. That's their bread and butter! But here I have this suspicion they're just spamming silicon on a low volume part. Easy to win the benchmarks, hard to commercialize.darkich - Friday, November 19, 2021 - link
M1 got destroyed??!!xD
Your brain is destroyed
JeffreyHF - Thursday, November 25, 2021 - link
Your reference to Centriq is specious. Centriq was not abandoned because it was technologically untenable. To the contrary. It was cut in what became a successful effort to save the company as an independent entity.Apple had ordered the contract manufacturers to withhold all royalties from Qualcomm, at the same time that Apple was itself refusing to enter a PLA, and was instead instigating the FTC v Qualcomm case.
Broadcom tried to steal the company for $79/sh, bottom feeding the company that Apple was trying to destroy.
Centriq was reluctantly abandoned as part of the effort to withstand the existential threat, a major buyback of shares was implemented, CFIUS was lobbied, and the multi-frontal litigation onslaught was defended. Nuvia gets to fulfill their server ambitions, among other products.
mode_13h - Friday, November 26, 2021 - link
Thanks for sharing that - it's a perspective I've not heard.I thought the Centriq point was pretty far out there, and was just waiting for someone knowledgeable to step in and pour some cold water on it. I'm glad you did.
Silver5urfer - Tuesday, November 30, 2021 - link
Centriq was in 2017 and Broadcomm and Apple take over was in 2018, I know Apple wanted to steal Qcomm and eat their LTE patents it was out at EETimes for that, Apple is a scum garbage. That's not new. They robbed Imagination Tech, they bankrupted GT Advanced tech as well and ruined Dialog Semi as well.Centriq was heralded as next thing and Cloudflare did a lot of advertising, everything was dead including the people who made it that when Snapdragon team who design Kryo 820.
Nuvia is going to deliver a miracle as Server Ambitions is going to be decided once this product is out. Until then it's vaporware.
mode_13h - Wednesday, December 1, 2021 - link
> They robbed Imagination TechThe thing I'll never quite understand about that is why Apple left the hulking husk of Imagination for someone like the Chinese to harvest of its IP. Was Apple feeling that Imagination was overplaying its hand and demanding too high an acquisition price, thus rejecting that path on principle? In practice, I think it would've been worth paying even too high a price just to swallow it and keep all that IP to themselves.
The saga of what has played out at Imagination, since then, is sad and hard to watch.
systemBuilder33 - Saturday, June 11, 2022 - link
I dedicated 12 years of my wife to working at Qualcomm. Everybody there runs around thinking they have a target on my back. I never saw such paranoia. If I could do one thing for Qualcomm, it would be, give everyone a tranquilizer every day at work, there's some serious mental illness their about the company's always-impending destruction next month ...JeffreyHF - Sunday, December 12, 2021 - link
Centriq and the custom CPU team were axed, in an effort to protect the company from a cheap takeover. They were sacrificed on the alter of corporate survival. Times have changed, and Qualcomm weathered the ruthless attacks by Apple/FTC and Broadcom, and they've shrugged off Apple's modem slot loss, by diversification. With the help if the Nuvia team, innovation will again be funded, and rule the day. Competition is good for everyone. Let's check back in a year, and see who's got what.systemBuilder33 - Saturday, June 11, 2022 - link
The company really doesn't understand computers, or the internet. I worked there once. mark my words. they only understand wireless system design.sseemaku - Thursday, November 18, 2021 - link
Big claims. Has CPU design become so easy that everybody is doing it? Doing it and then beating laptop class performance of Intel/AMD/Apple?And few years ago, people were like 'my ivybridge can do all my work, so I am off to buy Ipad'. And here we are in post-pc era with every company targeting laptop market.
Blastdoor - Friday, November 19, 2021 - link
Maybe it’s easier when not tied to x86. Maybe RISC guys were right, and x86 was propped up by Intel’s process advantage (such an antiquated phrase!)eastcoast_pete - Thursday, November 18, 2021 - link
I would have like to see or hear any comparison of NUVIA cores vs. ARM's "stock" X2. If QC can beat those AND be more efficient, they have a good chance. Otherwise, these SoCs will be niche players in the ultraportable segment at best. Apart from this, a lot depends on how much oomph Microsoft is willing to throw behind Windows on ARM. Native Win 11 plus native MS Office, Teams etc, all taking advantage of the SVEs NUVIA better has on offer, and it could work; maybe.Blastdoor - Friday, November 19, 2021 - link
It wouldn’t have worked without Apple. But now MS has to fear losing share to Mac, esp at high end. So I bet MS throws full weight behind this.Silver5urfer - Friday, November 19, 2021 - link
Biggest joke of the comment.Mac marketshare is under 10%, their revenue cut towards over all Apple money stream is also 10%, and world runs on Windows not Mac. Mac is very expensive that price barrier is impossible for many people. Plus new OS and learning curve on top.
Games do not work on Mac as well, that is one major thing. More software restrictions on Mac like ARM only and no more 32Bit, all the old games are dead on Mac thanks to Apple.
MS has fear and that too at High End LMAO.
mode_13h - Saturday, November 20, 2021 - link
> world runs on Windows not Mac.The cloud runs on Linux, and kids are using Chromebooks in most schools. So, Windows doesn't have the inertia it once did.
Blastdoor - Monday, November 22, 2021 - link
I give current Microsoft leadership credit for being much smarter than you (you sound like a Ballmer-era guy -- how are you enjoying your Zune and Windows Mobile phone these days?)Mac marketshare bottomed out at around 2% (maybe lower) at the turn of the century. It has been very gradually creeping higher since then. Today it's a bit lower than 10% globally, greater than 10% in the US. And almost all the share that Apple takes is at the high end -- by definition, since Apple doesn't compete much in the low-end.
Microsoft and Intel have clearly taken notice of this. It helps explain why Microsoft got into hardware and it explains Intel's recent ad campaign targeted at the Mac.
mode_13h - Tuesday, November 23, 2021 - link
> how are you enjoying your ZuneIt's crack-a-lackin!
; )
(anyone who remembers Zune has a fair chance of getting that reference.)
> Apple doesn't compete much in the low-end.
Apple TV. Clearly aimed at establishing their presence in the living room.
Apple is big and needs to keep growing. This will push them into new markets and drive them towards dominance. And they won't stop at games consoles. I'm predicting Apple Healthcare devices, before the decade is up.
Wilco1 - Friday, November 19, 2021 - link
Given Cortex-X2 devices will be out soon, beating those wouldn't be good enough in order to be competitive in 2023. Given NUVIA's experience it's likely they come up with a great design, but remember others are not standing still in the meantime.darkich - Friday, November 19, 2021 - link
I've said it many years ago (since 2014) that Apple will eventually produce superior chips and humiliate Intel and AMD..ALSO, since first hearing about Nuvia acquisition by QC, I claim Qualcomm is the one to watch out for now.
And it's because simple reason everyone keeps ignoring - people.
Nuvia is a startup ran by THREE BEST CPU EXPERTS, the same one that created Apple's M architecture.
Enough said.
Blastdoor - Friday, November 19, 2021 - link
Agreed. Weird that people make predictions based on Qualcomm’s past performance rather than the past performance of the guys who actually designed the darn chip. Companies are mostly just the people who work there (plus IP).Silver5urfer - Friday, November 19, 2021 - link
And who are those EXPERTS you talking about ? I wonder if it's going to shatter Redwood Cove and Zen 5, which are going to destroy every single CPU out there.Apple M got nothing, all that fluff marketing got busted out once the benches were out, it's competitive but only with 5nm design vs Intel's old parts and Zen's inferior designs. In GPU it's a massacre by Nvidia and AMD.
Bonus is $2666 starting price tag with a BGA soldered SSD junk and Max SoC is $3300+. Qcomm talks a lot, their CPUs cannot even beat Apple and some magical Nuvia is going to deliver industry shattering performance ? I already saw what Apple had to do with 57bn transistors with massive GPU cores and still unable to beat the cheaper manufactured Silicon from AMD, Intel and Nvidia. This joke of ARM taking over the world has been there since 10 years still people beat the same old drum.
Wait for Alder Lake BGA processors wreck havoc on M1 Max. That's just Intel 10nm. Once Zen 4 launches with Intel's Meteor Lake it will be a big bloodbath.
mode_13h - Saturday, November 20, 2021 - link
> I wonder if it's going to shatter Redwood Cove and Zen 5,> which are going to destroy every single CPU out there.
Not in perf/W. Remember, Qualcomm is going for the laptop market.
Farfolomew - Tuesday, November 30, 2021 - link
This dude's obviously a bit off his rocker, but I do think Alder Lake has a legitimate chance in succeeding in mobile. It will be interesting to see how it fairs against the M1, new Snapdragon chips, and possibly these new Nuvia ones in 2023. I agree with Ian that 2022 is going to be a really fun year for CPUs.mode_13h - Wednesday, December 1, 2021 - link
> It will be interesting to see how it fairs against the M1, new Snapdragon chips,> and possibly these new Nuvia ones in 2023.
For sure. I'll have my popcorn ready, for when the reviews are up. I don't think Alder Lake will beat Apple or Nuvia, but the question is more one of whether it'll be competitive enough to stave off the threat of ARM eating into its Windows laptop marketshare.
I don't quite know what to make of the fact that Intel launched Alder Lake on desktop, first. Was that mostly a PR move? Or do they face some efficiency challenges in their new Intel 7 node?
I'm also intrigued by the lack of any messaging around Gracemont-based entry-level SoCs, like a successor to Jasper Lake. Again, this could speak to power-efficiency challenges they're facing, or simply (more likely) prioritizing their limited Intel 7 production capacity. However, failure to counter Mediatek's aggressive moves in the Chromebook market segment could cost them a substantial share of it.
lmcd - Saturday, November 20, 2021 - link
Nuvia is a cool product addition but they were never gunning for low-power products. That means that, once again, we will see a garbage A55 or an even-more-garbage A510, barring a major surprise -- essentially, Qualcomm would have had to have lied about leaving the custom CPU space and have been working on a low-end CPU core design since about the time they announced they were leaving the space.By contrast, Alder Lake and its Atom E-Cores (as well as a massive Atom performance improvement timed to land as Alder Lake landed) speak to responding to Apple as soon as possible. It's clear one of these two companies feared competition, and one didn't. Hopefully, the continued inadequacy of Qualcomm mobile SoCs leads to an eventual US antitrust lawsuit. It's disturbing that nothing has happened to force their hand.
mode_13h - Sunday, November 21, 2021 - link
> they were never gunning for low-power products.Nuvia has the knowledge and experience of how Apple's low-power cores are designed. So, it definitely seems like the *could* make good low-power cores, if they wanted.
> Alder Lake and its Atom E-Cores ... speak to responding to Apple
It might be the overall ARM-based laptop threat, not specifically Apple. I don't expect Alder Lake to be very competitive against Apple, within a similar power envelope.
> It's clear one of these two companies feared competition, and one didn't.
I think Qualcomm's decision to stop making mobile CPU cores was based on the cost and trouble of trying to keep just marginally ahead of ARM's cores. They clearly decided to differentiate their SoCs in other areas. Maybe, if their homegrown cores had been more competitive, they'd have stuck with it. That's not to cast aspersions on the engineers, either. It could be that they were simply under-resourced for the task at hand.
zodiacfml - Sunday, November 21, 2021 - link
whatever. end consumer product will cost more than an Intel and probably Apple product since they put a huge premium on cellular connectivity like on those Arm based MS laptops/tablets despite poor performanceCaptGingi - Tuesday, December 14, 2021 - link
If like me, you follow the on-going court case regarding Gerard Williams III’s theft of Apple’s trade secrets and his “plot” to spin-up Nuvia to compete directly against Apple, you will note that while a trial is still a ways off, Mr. Williams and Nuvia are not doing well trying to spin that he/they did nothing wrong, despite discovery already quite clearly painting a highly suspicious picture, and the judge getting quite upset numerous times with Mr. Williams gamesmanship and warning that he MUST unequivocally produce all of the trade secret documents that came out that he “mistakenly” had copied to his personal devices and directly testify to the court about numerous questions he is refusing to answer fully/honestly about. You can access all records by going to the Santa Clara court website and searching by party name first:Gerard last:Williams to see the progression of this legal action.systemBuilder33 - Saturday, June 11, 2022 - link
With only $57m of initial seed money for Nuvia, they don't have a single penny left over to work on a GPU (let alone any GPU software). Therefore, Nuvia is short for NoGpuvia.