Intel's Future Chips: News, Rumours & Reviews

[goldstone77]

Intel's 28-Core 5GHz Processor And Test System Breaks Cover
by Paul Alcorn June 5, 2018 at 6:20 PM

But Intel's demo didn't tell the full story. Many in the press mistakenly assumed the new processor runs at 5.0 GHz at stock settings, but we carefully analyzed video from the event and spotted a few obvious signs that the processor was overclocked. Intel apparently was running some sort of closed-loop cooling that required insulating material around the tubing. This turns out to be a water chiller that was hidden under the table. We also spot more shielding over the long rectangular waterblock and what appears to be six sticks of RAM flanking the processor on each side. We theorized that this platform is based on the (until now) enterprise-class LGA3647 socket, and that has proven to be true. This means the processor could be a variant of the $8,700 Xeon Platinum Scalable processor we reviewed here, albeit with an unlocked multiplier.

https://www.tomshardware.com/news/intel-28-core-processor-5ghz-motherboard,37213.html

 

[aldaia]

Wondering about the TDP and power consumption.
The xeon platinum 8180 TDP is already more than 200W. Frequencies are 2.5GHz base and 3.8 Max turbo. However all core turbo is only 2.8 GHz (2.3 GHz if AVX512 is used)

The beast runs at 5GHz all cores, and as a rule of thumb power requiremets grow cubically with frequency. I expect no less than 1000W just for the CPU.

As a reference, it is said that a coffe lake overclocked to 5GHz requires 25-30W PER CORE. Add 2 AVX-512 pipes and multiply by 28 cores and we are in the ballpark of 1000W

Hence the exotic cooling system used. The water chiller has a cooling power of 1770W and requires 1000W for its operation (as a comparison a domestic refrigerator with integrated freezer requires around 200W).

Interesting demo but not expecting to see 28-core comercial products runing at such frequencies.

 

[goldstone77]

Computex 2018: Intel demos 28-core processor at 5 GHz
Subject: Processors | June 5, 2018 - 11:40 AM

UPDATE 6/6/18: It has now been confirmed by people on the ground in Taipei that the Intel 28-core demo was a complex feat. The motherboards were built by ASUS and Gigabyte, modifications of a server-class LGA3647 socket board that required a 32-phase power system, and a 1HP (horsepower) water chiller and refrigerant to drop the liquid to a cool 4 degrees Celsius. The processor is a single-die part, basically a Xeon Scalable Platinum 8180, that has a list price of $10,000.

Obviously this is not a configuration that any reasonable consumer, even the crazy ones really, would be willing to employ. It means motherboards with the X299 chipset will not be compatible with this part as it requires a new socket. It also means that clock speeds for real-world designs will be much lower, likely a Gigahertz or more.

There are a lot of questions to poke around about before the end of the year if we are truly going to understand Intel's plans for the enthusiast platform at the end of 2018.

https://www.pcper.com/news/Processors/Computex-2018-Intel-demos-28-core-processor-5-GHz


update:

 

[goldstone77]

Enter for a chance to win an 8th Gen Intel® Core™ i7-8086k Limited Edition processor Less than 7 minutes to go before you can enter!
When the countdown timer reaches zero you will have the opportunity to register for your chance to win one of 8,086 8th Gen Intel® Core™ i7-8086k Limited Edition processors.
Participants will only have 24 hours to enter so make sure you return!
https://game.intel.com/8086sweepstakes/

 

[jimmysmitty]

Enter for a chance to win an 8th Gen Intel® Core™ i7-8086k Limited Edition processor Less than 7 minutes to go before you can enter!
When the countdown timer reaches zero you will have the opportunity to register for your chance to win one of 8,086 8th Gen Intel® Core™ i7-8086k Limited Edition processors.
Participants will only have 24 hours to enter so make sure you return!
https://game.intel.com/8086sweepstakes/

This would be awesome. While I am not looking to upgrade a free top end 8700K would be fine with me. Could probably give my girl my 4670K setup.

If I don't win then I might just go down to their office and ask them nicely. Its only 10 minutes away.

 

[goldstone77]

Anything is worth a shot!

 

[aldaia]

Wondering if this news is more relevant for the AMD thread or for the Intel thread.
https://www.phoronix.com/scan.php?page=news_item&px=Hygon-Dhyana-AMD-China-CPUs

 

[jimmysmitty]

Wondering if this news is more relevant for the AMD thread or for the Intel thread.
https://www.phoronix.com/scan.php?page=news_item&px=Hygon-Dhyana-AMD-China-CPUs

That's actually interesting. I wonder if AMD is able to lease out an entire design since it still uses the x86 base design with their x64 extensions. Might make for an interesting debate.

 

[gamerk316]

Wondering if this news is more relevant for the AMD thread or for the Intel thread.
https://www.phoronix.com/scan.php?page=news_item&px=Hygon-Dhyana-AMD-China-CPUs

That's actually interesting. I wonder if AMD is able to lease out an entire design since it still uses the x86 base design with their x64 extensions. Might make for an interesting debate.

I put a more detailed post over in the AMD discussion. AMD can certainly license the Zen CPU architecture. AMD can *probably* license the x86-64 portion of the x86 ISA (honestly, you could argue x86-64 is an entirely separate ISA and not even an extension). AMD can *not* transfer the x86 ISA proper; that would have to be a clean-room re-implementation of the ISA.

 

[aldaia]

Ice Lake has vanished from most recent Intel roadmap. So far nothing is known, but usually when a product disappears from roadmap means it has been cancelled. Just a rumor.

Late 2018 Intel will launch: Whiskey Lake-U, Amber Lake-Y and Coffee Lake Refresh-S all in 14 nm

2019Q1 Coffee Lake Refresh-H
2019Q2 Comet Lake
Again all in 14 nm

Comet Lake seems to target same market as Canon Lake. Why would Intel market a 14 nm part in mid 2019? Unless High volume Canon lake is not expected until a year later.

 

[gamerk316]

At this point, Intel might just want to abandon 10nm and try moving right to 7nm. It's clear they've lost their fab edge.

Heck, at this point, you can argue they'd be better served outsourcing their fab production.

 

[YoAndy]

At this point, Intel might just want to abandon 10nm and try moving right to 7nm. It's clear they've lost their fab edge.

Heck, at this point, you can argue they'd be better served outsourcing their fab production.

Yes that's correct that finally other chip manufacturers are pretty close to intel, and finally AMD is challenging Intel, but for those who don’t know, Intel with the help of Ted Hoff introduced the first microprocessor, the Intel 4004 on November 15, 1971. The 4004 had 2,300 transistors, performed 60,000 operations per second (OPS), addressed 640 bytes of memory, and cost $200.00. the first multi-core processor was produced by IBM in 1996, Core architecture has been the basis of Intel chips since 2006 and the good old AMD(non existent) since 2005. According to a new job posting Intel ended up revealing its next-generation high-performance CPU architecture named “Ocean Cove.” It's difficult to speculate Intel's future plans right now, or what Ocean Cove could bring to the table but talking nonsense will make you look like a fool in the near future..

 

[goldstone77]

At this point, Intel might just want to abandon 10nm and try moving right to 7nm. It's clear they've lost their fab edge.

Heck, at this point, you can argue they'd be better served outsourcing their fab production.

Yes that's correct that finally other chip manufacturers are pretty close to intel, and finally AMD is challenging Intel, but for those who don’t know, Intel with the help of Ted Hoff introduced the first microprocessor, the Intel 4004 on November 15, 1971. The 4004 had 2,300 transistors, performed 60,000 operations per second (OPS), addressed 640 bytes of memory, and cost $200.00. the first multi-core processor was produced by IBM in 1996, Core architecture has been the basis of Intel chips since 2006 and the good old AMD(non existent) since 2005. According to a new job posting Intel ended up revealing its next-generation high-performance CPU architecture named “Ocean Cove.” It's difficult to speculate Intel's future plans right now, or what Ocean Cove could bring to the table but talking nonsense will make you look like a fool in the near future..

Intel eliminated 15,000 jobs in 2016 as part of the largest overhaul in the company's history as it worked to reduce its dependence on the fading PC market and pursue faster growing sectors, such as data centers, mobile communications and artificial intelligence.
The Oregonian/OregonLive reported the 2016 layoffs skewed older, with workers over 40 more than twice as likely to lose their jobs as those under 40. The disparities were greater among older groups of employees - workers over 60 were more than 8 times more likely to lose their jobs as those under 30.

There was another, smaller layoff in 2015 with a similar demographic profile.

https://www.oregonlive.com/silicon-forest/index.ssf/2018/05/federal_commission_investigati.html

And yet, the WSJ’s review of Intel’s own internal documents reveals that in one set of 2,300 layoffs, the median age was 49 years old, seven years older than the median age of the remaining staff.

https://www.theverge.com/2018/5/28/17401892/intel-age-discrimination-layoffs-investigation

They fire ~16,000 people, mostly older people, and hire young ones. Real classy move Intel!

 

[gamerk316]

They fire ~16,000 people, mostly older people, and hire young ones. Real classy move Intel!

Fact is, there's a ton of people in the over 55 age group that the only reason they are still working is they need the money and benefits to make it to SS/Medicare. That in turn means less jobs for people coming out of college, which is part of the reason why unemployment for younger adults is significantly higher then it should be.

And yes, I work at a company in the industry where the average age is 45, and that's counting the wave of young people who got hired in the last year. Two years ago, the average age was over 50. The company WANTS younger people who can be taught before everyone retires, but there isn't enough work the justify hiring them.

Welcome to economics.

 

[goldstone77]

They fire ~16,000 people, mostly older people, and hire young ones. Real classy move Intel!

Fact is, there's a ton of people in the over 55 age group that the only reason they are still working is they need the money and benefits to make it to SS/Medicare. That in turn means less jobs for people coming out of college, which is part of the reason why unemployment for younger adults is significantly higher then it should be.

And yes, I work at a company in the industry where the average age is 45, and that's counting the wave of young people who got hired in the last year. Two years ago, the average age was over 50. The company WANTS younger people who can be taught before everyone retires, but there isn't enough work the justify hiring them.

Welcome to economics.

Welcome to age discrimination law suits. Literally!

 

[gamerk316]

They fire ~16,000 people, mostly older people, and hire young ones. Real classy move Intel!

Fact is, there's a ton of people in the over 55 age group that the only reason they are still working is they need the money and benefits to make it to SS/Medicare. That in turn means less jobs for people coming out of college, which is part of the reason why unemployment for younger adults is significantly higher then it should be.

And yes, I work at a company in the industry where the average age is 45, and that's counting the wave of young people who got hired in the last year. Two years ago, the average age was over 50. The company WANTS younger people who can be taught before everyone retires, but there isn't enough work the justify hiring them.

Welcome to economics.

Welcome to age discrimination law suits. Literally!

Impossible to prove; you can just as easily argue they were let go because they make more money and thus letting them go leads to more savings.

The motivation here is 50% cost savings, 50% the need to hire younger workers.

 

[aldaia]

Intel 10 nm Logic Process Analysis
http://www.techinsights.com/technology-intelligence/overview/latest-reports/intel-10-nm-logic-process/

 

[goldstone77]

They fire ~16,000 people, mostly older people, and hire young ones. Real classy move Intel!

Fact is, there's a ton of people in the over 55 age group that the only reason they are still working is they need the money and benefits to make it to SS/Medicare. That in turn means less jobs for people coming out of college, which is part of the reason why unemployment for younger adults is significantly higher then it should be.

And yes, I work at a company in the industry where the average age is 45, and that's counting the wave of young people who got hired in the last year. Two years ago, the average age was over 50. The company WANTS younger people who can be taught before everyone retires, but there isn't enough work the justify hiring them.

Welcome to economics.

Welcome to age discrimination law suits. Literally!

Impossible to prove; you can just as easily argue they were let go because they make more money and thus letting them go leads to more savings.

The motivation here is 50% cost savings, 50% the need to hire younger workers.

Not impossible to prove, being 40 years or older with a good employment record, and being replaced by someone say 20. Significantly, easier to prove if it's a class action with thousands of workers being discriminated against.

The Age Discrimination in Employment Act of 1967 (ADEA) protects applicants and employees who are 40 years of age or older from employment discrimination based on age.

https://www.eeoc.gov/eeoc/publications/age.cfm

Intel facing possible class-action lawsuit over age discrimination
https://www.bizjournals.com/sanjose/news/2018/05/29/intel-age-discrimination-class-action-lawsuit-intc.html

 

[goldstone77]

Intel 10 nm Logic Process Analysis
http://www.techinsights.com/technology-intelligence/overview/latest-reports/intel-10-nm-logic-process/

I also, hear Intel tried to use Ruthenium(Ru) as well.

 

[jimmysmitty]

At this point, Intel might just want to abandon 10nm and try moving right to 7nm. It's clear they've lost their fab edge.

Heck, at this point, you can argue they'd be better served outsourcing their fab production.

At this point, Intel might just want to abandon 10nm and try moving right to 7nm. It's clear they've lost their fab edge.

Heck, at this point, you can argue they'd be better served outsourcing their fab production.

Yes that's correct that finally other chip manufacturers are pretty close to intel, and finally AMD is challenging Intel, but for those who don’t know, Intel with the help of Ted Hoff introduced the first microprocessor, the Intel 4004 on November 15, 1971. The 4004 had 2,300 transistors, performed 60,000 operations per second (OPS), addressed 640 bytes of memory, and cost $200.00. the first multi-core processor was produced by IBM in 1996, Core architecture has been the basis of Intel chips since 2006 and the good old AMD(non existent) since 2005. According to a new job posting Intel ended up revealing its next-generation high-performance CPU architecture named “Ocean Cove.” It's difficult to speculate Intel's future plans right now, or what Ocean Cove could bring to the table but talking nonsense will make you look like a fool in the near future..

Intel eliminated 15,000 jobs in 2016 as part of the largest overhaul in the company's history as it worked to reduce its dependence on the fading PC market and pursue faster growing sectors, such as data centers, mobile communications and artificial intelligence.
The Oregonian/OregonLive reported the 2016 layoffs skewed older, with workers over 40 more than twice as likely to lose their jobs as those under 40. The disparities were greater among older groups of employees - workers over 60 were more than 8 times more likely to lose their jobs as those under 30.

There was another, smaller layoff in 2015 with a similar demographic profile.

https://www.oregonlive.com/silicon-forest/index.ssf/2018/05/federal_commission_investigati.html

And yet, the WSJ’s review of Intel’s own internal documents reveals that in one set of 2,300 layoffs, the median age was 49 years old, seven years older than the median age of the remaining staff.

https://www.theverge.com/2018/5/28/17401892/intel-age-discrimination-layoffs-investigation

They fire ~16,000 people, mostly older people, and hire young ones. Real classy move Intel!

Well to be fair that is probably from a buildup of people who have worked at Intel over many years for a long time and it might also depend on the projects. They cut the jobs in some areas but are hiring for a different area, much like how they started to hire 3000 people in Chandler for FAB 42s move to 7nm.

Intel 10 nm Logic Process Analysis
http://www.techinsights.com/technology-intelligence/overview/latest-reports/intel-10-nm-logic-process/

That's pretty cool and honestly its impressive that it is 2x denser than their 14nm. I saw an article that GloFlos "7nm" is 2.8x as dense as their 14nm. I wonder what Intels 7nm would be like.

 

[goldstone77]

That's pretty cool and honestly its impressive that it is 2x denser than their 14nm. I saw an article that GloFlos "7nm" is 2.8x as dense as their 14nm. I wonder what Intels 7nm would be like.

Intel's 7nm is going to be 2.4x of 10nm. That is still a long ways into the future. Next year will still be 14nm.

 

[YoAndy]

CHeck this out...To put a quantum computer on your desk, Intel has a plan unlike any other..



The future of computing is quantum — or so headlines would have you believe. They’re not incorrect. Yet like the flying car, quantum computing is a technology that’s as elusive as it is enchanting. A computer that can go beyond the simple, binary 0s and 1s of today’s ‘classical’ designs opens a new world of possibility, but the technical hurdles are massive, and no one knows how long it’ll take to overcome them.

That’s not discouraging researchers, however. Quantum computers have reached an important milestone in recent years and piqued the interest of massive companies. Companies you’ve heard of. Companies like Intel, which has shipped Core processors in hundreds of millions of computers across the globe.

AN OLD DOG LEARNING SOME NEW TRICKS
Intel may seem an unlikely choice for innovation in quantum computing. Sure, it’s known for its powerful PC processors, but the company’s expertise is concentrated in classic computers built for the x86 instruction set. Intel’s 8086 chip, the first x86 processor, is nearing its 40th birthday. The fundamental underpinnings of Intel’s modern chips harken back to that now-ancient predecessor.

Jim Clarke, Intel’s Director of Quantum Hardware, explained to Digital Trends that the company will continue to lean on past expertise to drive future research, and quantum computing is no exception. During a visit to Intel’s campus, Clarke pulled out a slick, rainbow-like wafer. It looked a lot like those you’ve seen in the news or Intel’s own ads, but this one was different.

Intel Quantum Computing Holding Wafer
Jim Clarke, Intel’s Director of Quantum Hardware, showing off a wafer. Rich Shibley/Digital Trends
“What you see here is a wafer that comes from our 300-millimeter technology line just a few miles down the Ronler Acres campus,” Clarke told us. “We’re doing this wafer with the same technology that we’re using for our advanced chips. And what you see here are basically small spin qubit arrays.”

That’s the company’s quantum computing effort in a nutshell. It wants to advance the technology, just as does Google and IBM, as well as many universities. But Intel has a different approach. It wants to do it with silicon.

“We’re essentially using the same process lines, same tools, same design rules to do this, and that’s an advantage for Intel.”

“There are a total of, I think, roughly 50000 qubits on this wafer,” Clarke explained. “They’re not coupled together so we can’t use them together. But you see the power of using Intel’s advanced process lines. If we can get this technology working we’re going to be making chips that are the same from wafer to wafer, and we’re going to have lots and lots of qubit arrays on any given wafer.”

What Clarke’s talking about is not just production. He’s talking about mass production.

It’s easy to see why Intel would approach the technology with mass production in mind. It’s the only company among its competitors that makes a bulk of its money directly from processor sales. It’s a self-serving goal, to be sure, but one that gives Intel a unique incentive.


There’s many reasons why Google might want to build quantum computer, but selling quantum chips isn’t high on the list. IBM does sell chips, but only to enterprise customers. Intel’s the only horse in the race that might, one day, seek to sell you a quantum computer.

It’s not a crazy idea. As Clarke made clear, “this is running in the same fab that’s doing the cutting edge Core chips. […] With, I’ll say, some different challenges with making the wafers, we’re essentially using the same process lines, same tools, same design rules to do this, and that’s an advantage for Intel.”

Read more here:

https://www.digitaltrends.com/computing/intel-quantum-computing-research-lab/?utm_medium=push&utm_source=1sig&utm_campaign=One%20Signal

 

[jimmysmitty]

That's pretty cool and honestly its impressive that it is 2x denser than their 14nm. I saw an article that GloFlos "7nm" is 2.8x as dense as their 14nm. I wonder what Intels 7nm would be like.

Intel's 7nm is going to be 2.4x of 10nm. That is still a long ways into the future. Next year will still be 14nm.

Probably unless they are purposefully not talking about it to launch it earlier. I doubt that but considering that others are touting "7nm" I am not sure Intel would want to let it seem as if they have been surpassed.

On the other end they could also just change the name of their 10nm to 7nm since it should be just as dense as if not more dense than GloFlos 7nm. I would hope they don't as I feel like its false advertising but who knows.

 

[goldstone77]

That's pretty cool and honestly its impressive that it is 2x denser than their 14nm. I saw an article that GloFlos "7nm" is 2.8x as dense as their 14nm. I wonder what Intels 7nm would be like.

Intel's 7nm is going to be 2.4x of 10nm. That is still a long ways into the future. Next year will still be 14nm.

Probably unless they are purposefully not talking about it to launch it earlier. I doubt that but considering that others are touting "7nm" I am not sure Intel would want to let it seem as if they have been surpassed.

On the other end they could also just change the name of their 10nm to 7nm since it should be just as dense as if not more dense than GloFlos 7nm. I would hope they don't as I feel like its false advertising but who knows.

Meaning lost
At the 45 nm process, Intel reached a gate length of 25 nm on a traditional planar transistor. At that node the gate length scaling effectively stalled; any further scaling to the gate length would produce less desirable results. Following the 32 nm process node, while other aspects of the transistor shrunk, the gate length was actually increased.

With the introduction of FinFET by Intel in their 22 nm process, the transistor density continued to increase all while the gate length remained more or less a constant. This is due to the properties of FinFET; for example the effective channel length is a function of the new fins (Weff = 2 * Hfin + Wfin). Due to how the transistor changed dramatically from how it used to be, the current naming scheme lost any meaning.

https://en.wikichip.org/wiki/technology_node

Intel made a shift to measure transistor density, which they now use MTr/mm² (mega-transistor per squared millimeter) is a transistor density unit that attempts to quantify a process node.

Intel's 10 nm process consists of a minimum metal pitch of 36nm with 8 diffusion lines for a cell height of 272 nanometers. Additionally, the process has a poly pitch of 54 nm and cells on Intel's 10 nm use a single dummy gate. For 0.6 NAND2 + 0.4 SFF, Intel's 10nm has a density of 100.76 MTr/mm² along with a high-density 6T SRAM measuring 0.0312 µm². Note that Intel itself reported their 10nm at 100.8 MTr/mm².

https://en.wikichip.org/wiki/mtr-mm%C2%B2

Intel's 10nm and other foundries 7nm naming schemes are all close to 100 MTr/mm². So, just to recap, the node naming scheme hasn't really been an accurate measure looking at the Synopsys graph.

 

[jimmysmitty]

I always thought the node name was based on the smallest possible part of the node and not just the gate length but I guess not.

I guess nm means nothing then and so long as Intel keeps density up they remain ahead of the game. They really need to get 10nm or 7nm out though. I am honestly surprised they have had this much trouble. They had 10nm ready years ago but I guess yields and leakage must be the biggest problems.