Intel Innovation event
- Thread starter hist78
- Start date
P
Portland
Guest
He's probably making it up but you never know. There a lot of disruptions. I thought what quantum computers are doing now would be done in 2 decades. Technology is moving faster than the comprehension of the public.
> Pat Gelsinger stated Intel will match and even beat what Moore's law expected for the next ten years.
Yep, I watched it in disbelief. Pat sure does commit to his public persona. He did push ups and jumping jacks before his keynote. They even had a hipster DJ for a warm up act. Very polished event.
Yep, I watched it in disbelief. Pat sure does commit to his public persona. He did push ups and jumping jacks before his keynote. They even had a hipster DJ for a warm up act. Very polished event.
hist78
Well-known member
For whatever reason Intel deleted the original YouTube video and posted a new video at the link below. YouTube's policy is that you are not allowed to modify a video's content once it's uploaded and published. I don't know what's the critical issue to make Intel to do so.
hist78
Well-known member
Pat has an ability to seize a moment to turn an unhappy past to a beautiful future. Toward the end of the video, he's excited to announce the "Aurora" Exascale supercomputer will be delivered next year and it will double the performance from one to two exaFLOPS.
He did not mention the Aurora project has been delayed multiple times and they were all caused by Intel internal issues.
1. The $500 million Aurora super computer project was announced in 2015 and scheduled to go live in 2018 with 180 petaFLOPS capability at Argonne National Laboratory. The contract was awarded to Intel and Cray (now part of HPE(HP Enterprise)). But in 2018 the system was not delivered due to Intel internal product and manufacturing problems.
2. In 2019, Intel negotiated with Department of Energy (DOE) to modify the Aurora project to make it bigger and better. A new exaFLOPS Aurora supercomputer was schedule to be delivered in 2021.
3. In August 2020, DOE announced that Intel can't deliver Aurora in 2021. The new target date was unknown. According the news release:
"Argonne and Energy Department officials remain committed to the project and “are in discussions with Intel to update the delivery plan for Aurora,” the Argonne lab said in a statement. The partners are “actively working to mitigate any potential impacts to the schedule,” Intel said in a separate statement."
Source: https://www.nytimes.com/2020/08/27/technology/intel-aurora-supercomputer.html
4. On October 28, 2021, Intel CEO Pat Gelsinger announced the "Aurora" supercomputer will be delivered in 2022 and the performance will be increased to two exaFLOPS, double the original one exaFLOPS design.
5. It's reported that Intel will have a $300 million write-off in Q4 2021 due to Federal government contracts. It's possible caused by the Aurora project.
In the parallel development:
1. In March 2019, DOE and Oak Ridge National Laboratory awarded a $600 million, >1.5 exaFLOPS supercomputer "Frontier" contract to AMD/HPE-Cray. Currently it's on schedule to be delivered by the end of 2021. "Frontier" will become the first exascale supercomputer in US due to Aurora's delay.
2. In August 2019, DOE and Lawrence Livermore National Laboratory awarded a $600 million, >1.5 exaFLOPS supercomputer "El Capitan" contract to AMD/HPE-Cray. It's expected to be delivered in late 2022.
My thoughts:
1. HPE-Cray is the same contractor for all three projects. Intel (Aurora) or AMD (Frontier, El Capitan) provide the CPU and GPU processors but the on-time delivery seems not in Intel's favor.
2. It's great Intel can increase the performance to two exaFLOPS for Aurora. But I believe the government is one of few entities who can afford a four-year delay for projects in such large scale, as we saw in the Aurora project.
3. The Aurora system will use Intel new Xe-HP GPU. But it's reported on October 28, 2021 that Intel will drop the Xe-HP from their product offering before it can even make into the market. It seems too quick and too soon.
"We currently don’t intend to productize Xe HP commercially, it evolved into HPG and HPC that are on general market production path" ~ Raja Koduri, senior vice president and general manager of the Accelerated Computing Systems and Graphics (AXG) Group at Intel
Last edited:
Dr. Ian Cutress posted a video of Pat answering media questions at the end of the Innovation event.
Note I didn't watch the 98 minute video posted by Intel so some of the things mentioned in Cutress' video may no longer be a surprise to you. But here's what surprised me...
Super Moore's Law
Pat believes four things that will enable Intel to "go at Moore's or Super Moore's Law": (1) High-NA EUV (2) RibbonFET (3) PowerVia (4) Packaging Technology (Foveros, EMIB, Foveros Omni, etc.)
Further, Pat compared Intel vs the competition...
Intel's IBM relationship
Pat seems to suggest that Intel will be taking over Samsung's current role. That is IBM will assist Intel in advancing semiconductor and packaging research. Further, Intel will (or may) be assisting IBM on the manufacturing of their products. One such product might be IBM's Power10 processor. The processor was most recently manufactured by Samsung using their 7nm process node. It was noted that one of the Power10 processor cores is ALWAYS disabled due to yield issues. A early review of the processor indicated that the processor was somewhat held back by the limitations of Samsung's process node technology. With Pat seemingly claiming to be IBM's (preferred?) partner, I'm sure Samsung isn't too happy with this latest development.
Note I didn't watch the 98 minute video posted by Intel so some of the things mentioned in Cutress' video may no longer be a surprise to you. But here's what surprised me...
Super Moore's Law
Pat believes four things that will enable Intel to "go at Moore's or Super Moore's Law": (1) High-NA EUV (2) RibbonFET (3) PowerVia (4) Packaging Technology (Foveros, EMIB, Foveros Omni, etc.)
Further, Pat compared Intel vs the competition...
Obviously the competition feels differently here. Oh man...the DRAMA that Pat's statements have generated is going to make the Intel/TSMC/Samsung foundry war fun to watch!
Intel's IBM relationship
Pat seems to suggest that Intel will be taking over Samsung's current role. That is IBM will assist Intel in advancing semiconductor and packaging research. Further, Intel will (or may) be assisting IBM on the manufacturing of their products. One such product might be IBM's Power10 processor. The processor was most recently manufactured by Samsung using their 7nm process node. It was noted that one of the Power10 processor cores is ALWAYS disabled due to yield issues. A early review of the processor indicated that the processor was somewhat held back by the limitations of Samsung's process node technology. With Pat seemingly claiming to be IBM's (preferred?) partner, I'm sure Samsung isn't too happy with this latest development.
So, out of these three supercomputers, none has been delivered yet. Intel delays are a real fact but it might be too soon to draw the final conclusions. Aurora is not going to use Xe-HP. Xe-HP has been used for Aurora development (including software development). The final hardware will use Xe-HPC (Ponte Vecchio).
hist78
Well-known member
I believe Frontier at Oak Ridge National Laboratory will be delivered on time by the end of year. If not, DOE should have announced the delay long time ago. DOE announced in August 2020 when Intel won't meet the 2021 Aurora delivery deadline. I have no reason to think DOE will do differently and say nothing about the Frontier.
If Frontier can arrive on time, then I have no reason to think El Capitan will not arrive on time. Both projects are handled by AMD/HPE-Cray with similar >1.5 exaFLOPS performance targets.
hist78
Well-known member
@prime007, this is a great video link to let us understand Intel more.Dr. Ian Cutress posted a video of Pat answering media questions at the end of the Innovation event.
Note I didn't watch the 98 minute video posted by Intel so some of the things mentioned in Cutress' video may no longer be a surprise to you. But here's what surprised me...
Super Moore's Law
Pat believes four things that will enable Intel to "go at Moore's or Super Moore's Law": (1) High-NA EUV (2) RibbonFET (3) PowerVia (4) Packaging Technology (Foveros, EMIB, Foveros Omni, etc.)
Further, Pat compared Intel vs the competition...
Obviously the competition feels differently here. Oh man...the DRAMA that Pat's statements have generated is going to make the Intel/TSMC/Samsung foundry war fun to watch!
Intel's IBM relationship
Pat seems to suggest that Intel will be taking over Samsung's current role. That is IBM will assist Intel in advancing semiconductor and packaging research. Further, Intel will (or may) be assisting IBM on the manufacturing of their products. One such product might be IBM's Power10 processor. The processor was most recently manufactured by Samsung using their 7nm process node. It was noted that one of the Power10 processor cores is ALWAYS disabled due to yield issues. A early review of the processor indicated that the processor was somewhat held back by the limitations of Samsung's process node technology. With Pat seemingly claiming to be IBM's (preferred?) partner, I'm sure Samsung isn't too happy with this latest development.
Regarding the Aurora project debacle, I don't understand why Pat spent so many words circling around the question and try to pretend nothing had happened. See my brief Aurora's summary above in this thread.
Pat, as the CEO of Intel, has been frequently criticizing his Intel predecessors, important Intel customer (Apple) and important Intel supplier/partner (TSMC). I don't see there is any difficulty for him to admit that the Aurora is a failure (or called it a setback). When people on the Street already heard it from Intel about a $300 million write-off (out of a $500 million contract) in Q4 2021, why Pat bothers to waste many empty words to hide it?
Overall, I think all those questions are meaningful and excellent questions during the Q&A session. But the answers are not the ones reflecting the new Intel's determination, thoughtful planning, and hopeful future.
Last edited:
hist78
Well-known member
The following is Pat's explaination about the Aurora project. To me it's a four-year late and possible a money losing project. To Pat, it's another opportunity to demonstrate the power of the words.
"Timothy Prickett Morgan, The Next Platform: I'm trying to understand the new Aurora system: the original machine was supposed to be north of [an exaflop] and $500 million. Now it's two Exaflops, or in excess of two Exaflops, and you've got a $300 million write-off for federal systems coming in the fourth quarter. Is that a write-off of the original investment, or is Argonne getting the deal of the century on a two [exaflop] machine?
PG: [Since] the original concept of Aurora, we've had some redefinitions of the timelines and the specifications associated with the project efforts. Obviously some of those earlier dates when we first started talking about the Aurora project we've moved out and changed the timelines for a variety of reasons to get there. Some of those changes lead to the write-off that we're announcing right now. The way the contract is structured, part of it is that the moment that we deliver a certain thing, we will incur some of these write-offs simply from the accounting rules associated with it. As we start delivering it, some of those will likely get reversed next year as we start ramping up the yields of the products. So some of it just ends up being how we account for and how the contracts were structured.
On the two versus one Exaflop: largely it’s PVC, you know, Ponte Vecchio. The core of the machine is outperforming the original contractual milestones. So when we set it up to have a certain number of processors, and you can go do the math of what two [Exaflop] is, we essentially overbuilt the number of sockets required to comfortably exceed one Exaflop. Now that PVC is coming in well ahead of those performance objectives for some of the workloads that are in the contract, we're now comfortably over two Exaflop. That's pretty exciting at that point - that we will go from one to two pretty fast.
But to me, the other thing that's really exciting in this space is our Zetta Initiative. What we've said is that we're going to be the first to Zettascale by a wide margin. We're laying out as part of the Zetta initiative what we have to do in the processor, in the fabric, in the interconnect and the memory architecture, what we have to do for accelerators, and the software architecture to do it. So Zettascale in 2027 – it’s a huge internal initiative that's going to bring many of our technologies together for a 1000x gain in five years. That's pretty phenomenal."
Source: https://www.anandtech.com/show/17042/bringing-geek-back-qa-with-intel-ceo-pat-gelsinger
Last edited:
hist78
Well-known member
Dr. Ian Cutress posted a video of Pat answering media questions at the end of the Innovation event.
Note I didn't watch the 98 minute video posted by Intel so some of the things mentioned in Cutress' video may no longer be a surprise to you. But here's what surprised me...
Super Moore's Law
Pat believes four things that will enable Intel to "go at Moore's or Super Moore's Law": (1) High-NA EUV (2) RibbonFET (3) PowerVia (4) Packaging Technology (Foveros, EMIB, Foveros Omni, etc.)
Further, Pat compared Intel vs the competition...
Obviously the competition feels differently here. Oh man...the DRAMA that Pat's statements have generated is going to make the Intel/TSMC/Samsung foundry war fun to watch!
Intel's IBM relationship
Pat seems to suggest that Intel will be taking over Samsung's current role. That is IBM will assist Intel in advancing semiconductor and packaging research. Further, Intel will (or may) be assisting IBM on the manufacturing of their products. One such product might be IBM's Power10 processor. The processor was most recently manufactured by Samsung using their 7nm process node. It was noted that one of the Power10 processor cores is ALWAYS disabled due to yield issues. A early review of the processor indicated that the processor was somewhat held back by the limitations of Samsung's process node technology. With Pat seemingly claiming to be IBM's (preferred?) partner, I'm sure Samsung isn't too happy with this latest development.
The full session transcript can be found in the following link:
Bringing Geek Back: Q&A with Intel CEO Pat Gelsinger
hist78
Well-known member
@Daniel,
From your extensive knowledge of the EDA, IP and semiconductor ecosystem, is Intel's four years five nodes plan feasible and practical?
For those potential IFS clients, especially if their products are planned to rollout in next two to three years, I thought they have to have the important technical information or tools available today from IFS to allow them to evaluate, make decision and start working soon, if not already. Does Intel's timeline make sense?
Last edited:
I have asked multiple people this same question and there is serious doubt for different reasons. Intel just executing on plans is not going to do it. There are some serious technical challenges ahead such as HVM: EUV, GAA, and HNA EUV. Given Intel is still struggling with 10nm/7nm it is just not probable to get so many nodes out in such a short amount of time. Even TSMC is experiencing challenges with EUV at 3nm and Samsung, well don't even get me started about Samsung.
HVM, high volume manufacturing is the key. If Intel is going to trickle out chips and claim success then yes they may do it. Intel has done this since 14nm. We can all blame bad numbers on COVID and supply chain issues for another year or so but what then? Especially when the industry on a whole is exploding in a good way:
Semiconductor shipments swelled to a record high of $145 billion in Q3 2021
Nov. 01, 2021 9:12 AM
- Worldwide sales of semiconductors in Q3'21 were $144.8 billion, an increase of 27.6% over Q3'20 and 7.4% more than Q2'21, reported by the Semiconductor Industry Association.
- “Semiconductor shipments reached all-time highs in the third quarter of 2021, demonstrating both the ongoing high global demand for chips and the industry’s extraordinary efforts to ramp up production to meet that demand,” says John Neuffer, SIA president and CEO.
- Global sales for the month of September 2021 were $48.3 billion, an increase of 27.6% over the total from September 2020 and 2.2% more than sales from August 2021.
- Regionally, year-to-year sales increased in the Americas 33.5%, Europe 32.3%, Asia Pacific/All Other 27.2%, Japan 24.5%, and China 24.0%.
- Month-to-month sales increased in the Americas 3.9%, Europe 2.0%, Japan 2.0%, Asia Pacific/All Other 1.9%, and China 1.5%.
- The chart below shows the YTD price returns of INTC, ON, MU, TSM, NXPI, and TXN.
Semiconductor shipments grew to a record high of $145 billion in Q3 2021 (NASDAQ:SOXX)
Worldwide sales of semiconductors in Q3'21 were $144.8 billion, an increase of 27.6% over Q3'20 and 7.4% more than Q2'21, reported by the Semiconductor Industry...
seekingalpha.com
I think that Intel is now on a very dangerous path. I don't remember the exact name of the concept, but the idea is simple. Intel is and will be innovating in some ways, they will invest aggressively in new fabs and equipment + the company will go through some minor restructuring. This, to investors and normal folks, creates an illusion of a real change in the company, but in the end, they accelerate their demise. It is clear to me, they will be innovating and they will be investing heavily, but they are choosing a bad strategy which will lead them to failure faster...I have asked multiple people this same question and there is serious doubt for different reasons. Intel just executing on plans is not going to do it. There are some serious technical challenges ahead such as HVM: EUV, GAA, and HNA EUV. Given Intel is still struggling with 10nm/7nm it is just not probable to get so many nodes out in such a short amount of time. Even TSMC is experiencing challenges with EUV at 3nm and Samsung, well don't even get me started about Samsung.
HVM, high volume manufacturing is the key. If Intel is going to trickle out chips and claim success then yes they may do it. Intel has done this since 14nm. We can all blame bad numbers on COVID and supply chain issues for another year or so but what then? Especially when the industry on a whole is exploding in a good way:
Semiconductor shipments swelled to a record high of $145 billion in Q3 2021
Nov. 01, 2021 9:12 AM
- Worldwide sales of semiconductors in Q3'21 were $144.8 billion, an increase of 27.6% over Q3'20 and 7.4% more than Q2'21, reported by the Semiconductor Industry Association.
- “Semiconductor shipments reached all-time highs in the third quarter of 2021, demonstrating both the ongoing high global demand for chips and the industry’s extraordinary efforts to ramp up production to meet that demand,” says John Neuffer, SIA president and CEO.
- Global sales for the month of September 2021 were $48.3 billion, an increase of 27.6% over the total from September 2020 and 2.2% more than sales from August 2021.
- Regionally, year-to-year sales increased in the Americas 33.5%, Europe 32.3%, Asia Pacific/All Other 27.2%, Japan 24.5%, and China 24.0%.
- Month-to-month sales increased in the Americas 3.9%, Europe 2.0%, Japan 2.0%, Asia Pacific/All Other 1.9%, and China 1.5%.
- The chart below shows the YTD price returns of INTC, ON, MU, TSM, NXPI, and TXN.
Semiconductor shipments grew to a record high of $145 billion in Q3 2021 (NASDAQ:SOXX)
Worldwide sales of semiconductors in Q3'21 were $144.8 billion, an increase of 27.6% over Q3'20 and 7.4% more than Q2'21, reported by the Semiconductor Industry...
hist78
Well-known member
Is it called "Innovation Trap"? I'm not sure it match what you meant?
Innovation: The Classic Traps
Reprint: R0611C Never a fad, but always in or out of fashion, innovation gets rediscovered as a growth enabler every half dozen years. Too often, though, grand declarations about innovation are followed by mediocre execution that produces anemic results, and innovation groups are quietly...
hbr.org
Oh yes exactly. I believe intel suffers from the business-model trap. https://www.inc.com/samuel-bacharach/5-innovation-traps-to-avoid.htmlIs it called "Innovation Trap"? I'm not sure it match what you meant?
Innovation: The Classic Traps
Reprint: R0611C Never a fad, but always in or out of fashion, innovation gets rediscovered as a growth enabler every half dozen years. Too often, though, grand declarations about innovation are followed by mediocre execution that produces anemic results, and innovation groups are quietly...
"This happens when a company ventures into a product or sector which may seem ripe, but the firm doesn't readily possess the competencies or materials necessary to execute. While it may seem like a good idea to expand horizons and product lines, if a company is not equipped to take on the new task, then it can fall victim to the business-model trap."
Intel will basically exhaust themselves trying to catch up to TSMC while not having enough EUV and smart people to execute on their promises.
As we all know it already Moore's Law was already at a breakneck pace of doubling transistor's density every 18 months or so, but Pat now wants to break that barrier with 4 node advances in 5 years I'd say that's a lot of hot airs. It's one thing to build a 2nm chip in the labs and another animal to be in HVM. If it wasn't, IBM would be flooding the market with 2nm chips already by now based on their announcement about a year ago.