Intel's Line of Notebook CPUs Gets More Confusing With 14nm Comet Lake (arstechnica.com) 62
Intel today launched a new series of 14nm notebook CPUs code-named Comet Lake. Going by Intel's numbers, Comet Lake looks like a competent upgrade to its predecessor Whiskey Lake. The interesting question -- and one largely left unanswered by Intel -- is why the company has decided to launch a new line of 14nm notebook CPUs less than a month after launching Ice Lake, its first 10nm notebook CPUs. From a report: Both the Comet Lake and Ice Lake notebook CPU lines this month consist of a full range of i3, i5, and i7 mobile CPUs in both high-power (U-series) and low-power (Y-series) variants. This adds up to a total of 19 Intel notebook CPU models released in August, and we expect to see a lot of follow-on confusion. During the briefing call, Intel executives did not want to respond to questions about differentiation between the Comet Lake and Ice Lake lines based on either performance or price, but the technical specs lead us to believe that Ice Lake is likely the far more attractive product line for most users.
Intel's U-series CPUs for both Comet Lake and Ice Lake operate at a nominal 15W TDP. Both lines also support a "Config Up" 25W TDP, which can be enabled by OEMs who choose to provide the cooling and battery resources necessary to support it. Things get more interesting for the lower-powered Y-series -- Ice Lake offers 9W/12W configurable TDP, but Comet Lake undercuts that to 7W/9W. This is already a significant drop in power budget, which Comet Lake takes even further by offering a new Config Down TDP, which is either 4.5W or 5.5W, depending on which model you're looking at. Comet Lake's biggest and meanest i7, the i7-10710U, sports 6 cores and 12 threads at a slightly higher boost clock rate than Ice Lake's 4C/8T i7-1068G7. However, the Comet Lake parts are still using the older UHD graphics chipset -- they don't get access to Ice Lake's shiny new Iris+, which offers up to triple the onboard graphics performance. This sharply limits the appeal of the Comet Lake i7 CPUs in any OEM design that doesn't include a separate Nvidia or Radeon GPU -- which would in turn bump the real-world power consumption and heat generation of such a system significantly.
Intel's U-series CPUs for both Comet Lake and Ice Lake operate at a nominal 15W TDP. Both lines also support a "Config Up" 25W TDP, which can be enabled by OEMs who choose to provide the cooling and battery resources necessary to support it. Things get more interesting for the lower-powered Y-series -- Ice Lake offers 9W/12W configurable TDP, but Comet Lake undercuts that to 7W/9W. This is already a significant drop in power budget, which Comet Lake takes even further by offering a new Config Down TDP, which is either 4.5W or 5.5W, depending on which model you're looking at. Comet Lake's biggest and meanest i7, the i7-10710U, sports 6 cores and 12 threads at a slightly higher boost clock rate than Ice Lake's 4C/8T i7-1068G7. However, the Comet Lake parts are still using the older UHD graphics chipset -- they don't get access to Ice Lake's shiny new Iris+, which offers up to triple the onboard graphics performance. This sharply limits the appeal of the Comet Lake i7 CPUs in any OEM design that doesn't include a separate Nvidia or Radeon GPU -- which would in turn bump the real-world power consumption and heat generation of such a system significantly.
Whiskey Lake (Score:4, Funny)
Re: (Score:3)
What AMD beatdown? Intel are completely uncontested in the mobile world. AMD may be slapping Intel sideways on the enthusiast / non-corporate desktop market but Intel's laptop division is just sitting back and printing cash.
Re: (Score:2)
What AMD beatdown?
They're thinking four-dimensionally; the "beatdown" (an exageration, surely) will be the inevitable long-term result of AMD's current success: everything in this industry follows the "enthusiast/non-corporate desktop market."
Re: (Score:1)
The effect is called "sphincter entanglement", when Intel's ass hole is hurt, so is yours.
Re: (Score:2)
Note that Zen2 (7nm DUV) beats greatest desktop that Intel has for desktop market at both instructions per clock and perf/watt, but loses at max clock speed.
At the moment, AMD's 3xxxx series mobile APUs are Zen+ cores on 12nm. They lose a bit on perf/watt, are roughly on par CPU perf wise and spank Intel in graphics department, already being the better buy if you do use GPU intensive apps, such as games. Even the said chips made it into premium lines of the OEMs such as Lenovo, HP, Asus.
7nm EUV process, whi
Competition (Score:2)
Intel is insufficiently managed, it seems to me. (Score:2)
Agreed. Human development depends partly on the microprocessor industry.
Intel CEO apparently has no technical knowledge. [slashdot.org]
Intel not successful with consumer products. (Score:2)
Re: (Score:2)
Well... about that MIPS stuff... "mips" (the most widespread of their archs) got dropped from Debian literally today. And no other major distribution had the resources to keep it even that long. Little-endian mipsen are still alive but I guess that this won't last long.
Re: (Score:2)
I suspect that desktop ventures outside X86 ecosystem are doomed to failure. Binary compatibility is very important.
If a system were developed that doubled X86 performance at the same price, it would stand a chance. I don't anticipate that happening.
No 6 core (Score:2)
Re: (Score:2)
Six cores at 1.1GHz base frequency [intel.com]. Make your conclusions. You cannot scale the 14nm node indefinitely even if you bin your CPUs 24/7.
Re: (Score:2)
Sorry, I meant 1.9GHz.
It would be great to get the edit button on /. at least for a couple of minutes. Would have saved a lot of pain and embarrassment.
Re: (Score:2)
Re: (Score:2)
Damn, I didn't make a mistake - it was someone at Intel who was messing with the website while I was writing the comment.
Base frequency is now back to 1.1GHz.
Re: (Score:2)
Anyway, I agree. Really tired of Intel taking older CPU's, marking them new with a slightly higher speed, but really it's the same 14 nm junk with Meltdown and all of their other issues still there because it's still just a stepping 10. I'm looking at you, 9750.
Re: No 6 core (Score:1)
Re: No 6 core (Score:4, Informative)
My web browser will use 10 cores on ad blocking alone.
Re: (Score:3)
Re: (Score:2)
Re: (Score:1)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:3)
Re: (Score:1)
Re: (Score:2)
Re: No 6 core (Score:2)
Comet Lake (Score:5, Insightful)
The sad truth about Comet Lake release is that Intel's 10nm node still does not work in any meaningful volumes, yields or performance.
It also demonstrates that monopolies are bad, very very bad especially in the semiconductor industry. It's year 2019 and now we're getting the fifth (!) iteration of the SkyLake uAach (which was first released in 2015, i.e. four years ago). It was unheard of in the past.
It also demonstrates that someone at Intel should have been fired a lot earlier that he actually was. It also shows how incompetent were his advisers, the board of directors and other senior staff.
Intel must be lucky they have billions of dollars saved - it might help them recover in 2021-2022 and become a CPU powerhouse again.
Re:Comet Lake (Score:4, Insightful)
I'm starting to wonder if the state of CPUs isn't just the long-term state of high technology more or less for the long haul.
The technology and engineering is pretty extreme, and the costs associated with it don't really support large markets with dozens of vendors. Plus once you get into the network effects of software support for various CPU instruction sets, it seems like the market will never support much more than a couple of vendors serving the mass market of computing.
I think you'll get some level of diversity and innovation, but only in niche products like smartphones that are kind of closed-loop and meant to support a very specific product.
Re: (Score:2)
The technology and engineering is pretty extreme, and the costs associated with it don't really support large markets with dozens of vendors. Plus once you get into the network effects of software support for various CPU instruction sets, it seems like the market will never support much more than a couple of vendors serving the mass market of computing.
In theory any other instruction set could take advantage of TSMC's 7 nm node, write the appropriate LLVM backend, a few kernel modules and a boot loader, and be up and running with any of the open source OSs. It's probably easier right now to launch a new ISA on a state-of-the-art node than it's been in 30 years. That's not to say it's easy. You do have to pay some expensive people who are competent in Verilog and Spice. Still, it's possible.
Re: (Score:2)
I'm just thinking that market dominance and network effects and duopolies are just part of the sad-sack dystopia we face.
Re: (Score:2)
I've done Verilog and SPICE work, and it's mostly grunt work, and the available improvements only nibble around the edges. If you want a big step forward in design (assuming it's possible) then what you need is a team of system architects: find out how to move the data faster, decide what to do with the data faster, and do it faster.
Here's a suggestion: design processors to explicitly be unsafe. Don't sell them for internet use or anywhere where hackers can be found. Place no limits on speculative execution
Re: (Score:2)
Re: (Score:3)
Even with Intel? You've got to be kidding. AMD signed on with a fab (TSMC) that has functional plans to move to 5nm soon, while Intel still can't get their 10nm from 2017 working properly. If they don't hit all their targets on 7nm EUV, they are toast as an IDM.
Re: (Score:2)
Re: (Score:2)
Low yields are probably more to do with cobalt and quad patterning. Intel is a huge organization with deep pockets. They can run both teams simultaneously (and they are). The 10nm node was just cursed from the start. Too many poor decisions.
Re: (Score:3, Insightful)
What's really hilarious... in 2014 Intel announced it was going full bore diversity because it makes for better decisions etc. It pulled at least 350 million USD (that we know of) out of research and put it into diversity programs inside and outside Intel. I doubt it was only 350 mil - likely it was much, much more in practise.
Basically a bunch of feminists got into Intel and extracted hundreds of millions out of it and funnelled it into their own garbage.
I said at the time... I give Intel 5 years before th
Re: (Score:2)
Nah, diversity isn't really the cause. It just didn't help them either. Their design shop is doing fine . . . it's the fab that's fubar. Meanwhile, a completely non-diverse TSMC is whipping them in the foundry race.
Diversity is orthogonal to departmental success. It won't hurt you, but it doesn't help you either.
Re: (Score:2)
Re: (Score:1)
No, it doesn't. Go read the corporate newspeak. Diversity is about bringing in new ideas to get people out of a rut. The notion is that, in order to do this, you get people of a different sex and/or ethnicity.
In theory it's an okay idea. In practice, you can get a guy with the same skin color and get some very different ideas if that's what you need. Technical skill is the bare minimum to get into the semicon game. You are not going to have clueless diversity hires except in some useless media office
Re: (Score:2)
Krzanich is ultimately responsible for that, and he is not a "dot Indian". Meanwhile, they hired a "dot Indian" to run their dGPU effort. Go Raja go!
Re: (Score:2)
Process size (Score:2)
If Intel can get better performance with a larger process size, it makes perfect sense to do it. This way, they have a good CPU to sell while they work on making the 10nm process perform better, and a free performance upgrade when they switch the new CPU to use it. I'm sure yields for the 14nm process are much better.
Comet Lake is contingency for Ice Lake Failure (Score:4, Interesting)
The most likely explanation is that Comet Lake was developed using existing processes and tools in case Ice Lake didn't come together on schedule.
It's still a sizable investment to create Comet Lake, but it's better than the situation where Intel didn't have anything to announce.
Manufacturing process change != upgrade (Score:4, Insightful)
How small the connection are made via the manufacturing of the chip, doesn’t really state if the CPU is faster, more power efficient, runs cooler or any benchmark. Just the fact it has bigger connections on the chip.
Is the smaller Nm the new MHz myth?
Re: (Score:2)
How small the connection are made via the manufacturing of the chip, doesn’t really state if the CPU is faster, more power efficient, runs cooler or any benchmark. Just the fact it has bigger connections on the chip.
Actually that is false. All else being equal the manufacturing change actually does make a design more power efficient and it will run cooler. The key point you're missing is that there's no such thing as a pure manufacturing change. Not currently, not in the tic-toc days. The manufacturing change has always been accompanied by a change in the underlying chip which the new manufacturing process enabled. For desktop chips that historically has been clock increases up until a practical limit was reached.
As fo
Re: (Score:2)
How small the connection are made via the manufacturing of the chip, doesnâ(TM)t really state if the CPU is faster, more power efficient, runs cooler
It greatly determines all of those things.
Is simple: Availability (Score:4, Interesting)
Ice-lake chips @ 10nm are very difficult to manufacture at the moment. Not only because the yields are at the moment quite likely abysmal, but also because intel does not have many fabs prepared to manufacture 10nm parts.
The Comet-Lake parts can be manufactured in really high numbers, with very high yiels, and be ultra-deep-binned to get all sorts of different SKU-s.
So, for high clockrate laptops in high quantities (with extra graphics chips) and very-low cost laptops with crap graphics manufacturers are supposed to pick the Comet-lake parts, reserving the Ice-Lake parts to low volume ultra-integrated solutions with decent graphics performance.
Re: (Score:2)
Re: (Score:1)
Not sure how debugging 10nm is going to help them on their vastly-different 7nm EUV node that is being handled by a different team. 10nm is a dumpster fire at this point.
Also, the fabs have been paid for (Score:5, Insightful)
For a recent fab like 14nm, if the processor guys can improve the processor design (which they did anyway for the 10nm fab), it makes sense to release a new CPU version on 14nm. It would be stupid to keep the 14nm lines producing last year's CPU design, when you have a perfectly viable new CPU design which can also be manufactured at 14nm.
Re: (Score:1)
14nm Comic Late (Score:1)