Laptops with Intel’s new 7th-gen Core “Kaby Lake” chips are now available, and desktops with the chips should be available in early 2017. Meanwhile, the first low-power notebooks, 2-in-1 tablets, and desktops with Celeron and Pentium chips based on Intel’s Apollo Lake designs should be available soon.
So what’s next? According to a leaked product roadmap, Cannon Lake, Coffee Lake, and Germini Lake. That’s a lot of lakes.
Keep in mind that this roadmap seems to have been drawn up in April, so there’s a chance some plans have changed. That said, let’s take things one at a time:
Gemini Lake
Now that Intel is positioning its Atom chips as products for IoT, robotics, drones, and other embedded systems, the company’s 4 watt to 10 watt Celeron and Pentium chips are its cheapest, lowest-performance processors for traditional PCs such as laptops, desktops, and 2-in-1 tablets.
Gemini Lake is the code-name for the chips that will eventually replace the Apollo Lake processors coming out this fall. But we’ll have to wait a while before they show up: the first Gemini Lake chips are slated to ship in the fourth quarter of 2017.
At this point there aren’t many details about the chips, other than their code name (which has been floating around for a while), but the product roadmap suggests they’ll be available in 4 Watt and 6 Watt varieties, much like earlier members of the Intel “N” series processor family.
Cannon Lake
The chips that will eventually replace Skylake and Kaby Lake processors in most portable notebooks are code-named “Cannon Lake,” and according to the roadmap, they’ll also ship in late 2017… possibly a little later than Gemini Lake.
The new chips are expected to introduce a new graphics architecture, and they’ll likely be the first Intel chips built using a 10nm process.
Like Broadwell, Skylake, and Kaby Lake before, the new Cannon Lake chips for mobile devices will be split into two product families: U series chips for mainstream notebooks and Y-series for ultra-low power systems.
What’s new is that it looks like the Y-series chips (which were mostly branded as Intel Core M processors, up until recently) seem to be 5.2 watt chips rather than 4.5 watt processors. Those numbers are a bit flexible though, since Intel allows device makers to adjust the TDP on its chips a bit, so some current Y-series chips might actually use up to 7 watts, for instance, or as little as 3.5 watts.
Something else that’s noteworthy is that the roadmap shows the most powerful Cannon Lake-U chips will be 15 watt processors, which means there won’t be a more powerful 28 watt variant.
Instead, Intel will brand its 28 watt Kaby Lake successor under a different name…
Coffee Lake
Coming in the second quarter of 2018, Intel’s Coffee Lake chips will likely include Core i3, Core i5, and Core i7 processors that use between 15 watts and 45 watts of power.
Unlike Cannon Lake, the Coffee Lake processors are expected to be 14nm chips.
The entry-level 15 watt Coffee Lake-U variant will likely be a dual-core laptop processor. There’s also a 28 watt Coffee Lake-U line of chips on the way, which will probably be a set of quad-core laptop chips.
The 45 watt Coffee Lake-H chips are higher-performance, 6-core processors. These chips are also designed for laptops, but since they generate a lot more heat than Cannon Lake or Gemini Lake chips, you probably won’t see the new H series processors in thin and light notebooks.
So NOTHING from Intel on the X series Desktop processors thru 2018? I couldn’t care less what happens on the mobile front … those devices are capped in performance by their limitations on available power (batteries in most cases or 50Watt when plugged in – just NOT going to hack it for any serious computing needs). For those of us that are more serious about computing, where are the X series desktop CPUs? I don’t need more “idle” cores, I need higher frequency, DDR5 RAM, better chipsets, and a USB 3.0 specification that can handle more than about 12 USB devices.
Wow.. I can’t tell if you are serious or not. It’s possible you’re trolling based on all your comments across many articles. But it’s clear you have no understanding of Thermodynamics, the difficulties of mainstreaming water cooling, and power requirement restrictions for chips. More Power = More Heat.
The reason Intel and the like are going towards more Cores is that they HAVE hit a threshold they haven’t been able to bust through. Maintaining any overclocked CPU above 4Ghz is going to ruin that chip. They don’t mass produce chips to be disposable every year.
The majority of processing ‘time’ is spent waiting on I/O bound events. Networks, Disks, Display, Onboard Cache, USB throughputs are some of the few things that are exponentially slower than a CPU. A 10Ghz CPU won’t help you computer a 3D image any faster if it’s being slowed down by having to read texture data, geo points, etc. off a disk.
Also, the difference between Intel Desktop and Intel Xeon chips is not about threads. Where did you even read that? What are you even talking about? Something about threads running in any order on the desktop but being linked on server chips? What? They both are based on x86 64-bit instruction sets. Xeon chips LAST longer. They have SLA’s to NOT crash. They can use ECC – Error Correcting – memory. Memory that will self correct if a bit is flipped via electrical chaos. Its all about stability and no downtime. Desktop CPU’s are about vast consumer consumption. It doesn’t really matter if your CPU segfaults due to being too hot while playing World of Warcraft. It really doesn’t. But it does matter if a CPU crashes during a stock market trade, or while performing life saving calculations, pharmaceutical work. Executables run the EXACT same way on Desktop and Server CPU’s. It’s more akin of taking a car and driving it really fast down a road tht has been cleaned, has bumbers on the sides, and ensures no other drives can crash into you versus you hoping on the interstate. The same car drives on both, but one is more stable and safe. Threads.. what article did you misinterpret for that argument? Are you talking about branch prediction being tweaked or turned off on server chips?
Being able to do more things, in an intelligent, efficient method is where the future is at. Smarter branch prediction, lower power consumption, faster wireless data transfer rates, etc. are all going to be the key to faster computing. A 10ghz single core chip will not help you nearly as much as you think. And your electrical usage will sky rocket. Sure, if you have solar power that’s great. But it means you’re literally wasting money as you could be selling it back to the Power Company. All for what? So that something can finish 5% faster?
The transistors are already so small there are already issues of electrons jumping when they shouldn’t be. A new way of computing is around the corner. Quantum computing will change everything forever. Chasing after an ever increasing GHz number is futile and wasteful.
We need smaller, faster, LESS power hungry computing machines. Less power = longer battery lives. Better batteries equals even longer computing times. We really don’t need anything more than what we have in the desktop arena. What can you possibly be doing that would benefit from a single core 10ghz chip? Nothing that I can think of that couldn’t be solved by using grid computing.
If you’re trolling then you’ve got me. If you aren’t I suggest picking up any Engineering book and educating yourself in order to not spread such terrible ideas.
Wow, can’t believe you responded with so much inaccurate and fundamentally wrong information? I’ve been running my 5960X at 4.6Ghz for almost 2 years now, not a single problem with it … but I do have it well water cooled CPU that rarely gets above 110 F (F not C). So real world evidence suggests your statement about running 4Ghz or higher is FALSE.
Agree with you that IO operations are slower, but textures are cached either in RAM or more likely in VRAM (most modern GPUs will cache textures). There is an initial one time loading of textures but after that point any reasonably efficient code will buffer the textures into RAM. The CPU speed/performance is key to moving textures to the GPU … and that IS the bottleneck directly associated with the CPU and RAM. Much larger CPU L3 cache could help performance considerably in this regard. The 8MB L3 cache found in Skylake CPUs IS NOT sufficient, the 20-25MB L3 Cache in the X series CPUs is certainly moving in the right direction, much larger would be better (think 64MB to 128MB). The reason L3 cache is small is due to the physical space it uses, Intel can’t get as many CPU yields out of a single wafer/disc which mean higher cost and potentially less profit.
As far as electrons “jumping” … I think you mean “electromigration”. This is a design choice made by Intel, “electromigration” can be prevented with thicker walls but once again, this means more cost and less yields from a single wafer/disc which again means less profit for Intel.
Mobile computing needs less power hungry, Desktop computing does NOT, Desktop has 1500+ Watts of typical home output to play with.
I didn’t suggest single core 10Ghz chip? I suggested “less cores” at 10Ghz, a 4 core 10Ghz chip with 128MB L3 cache taking up more wafer space and power would run circles around an 8 core 3Ghz chip with 25MB L3 cache using less wafer space and power.
It’s all about Profit, nothing to do with Thermal dynamics.
Quantum computing is still many many years away from hitting “retail” markets.
Anyway, it finally looks like AMD is understanding the need for higher performing desktop CPUs and are set to topple Intel’s lack of progress. Intel has left the door open for desktop market, and AMD are taking that opportunity to fill it with CPUs that perform much better than Intel’s X series.
BTW, your statement about “picking up any Engineering book” is beyond pathetic … sorta like telling a GP doctor to pickup any “doctor” book and go ahead and start doing brain surgery. No “any” Engineering book will not do … but not that it matters, what you are describing is “physics” of which there are MANY disciplines also … so please, if you’re going to dish out insults at the very least be accurate.
Rob, right from the start I seen Andrew was an uneducated idiot babbling along. At first I actually thought he was just trolling you. Then I seen his comment on thermodynamics, and I knew he just had no idea what he was talking about. Anyway, Bravo Rob! Bravo!
Correction: It is Coffee Lake that will be a holdover–the second “optimization” or third tock, if you will–of Intel’s 14-nanometer fabrication process, and it is the name for 14-nanometer Cannonlake processors. Look for yourself at your own sourced link from Fool.com. Cannonlake, which, by the way, is one word and not two, will be Intel’s first processor design to use 10-nanomater fabrication technology. What the linked Fool’s article reveals is that higher TDP processors (S- and H-class) will use Cannonlake’s architecture technology but on 14-nanometer fabrication technology instead. Only ULV (ultra-low power) designs of Cannonlake (U- and Y-class) will use 10-nanometer fabrication technology. This 14-nanometer offshoot, Coffee Lake, is in all likelihood a cost-cutting measure and a means for avoidance of any potential delays since 10-nanometer die yields will still be very low. In other words, Intel does not want to have a repeat of Broadwell on their hands, where they had to delay its release for many months, and ultimately skipping a widespread desktop release altogether, due to yields not meeting minimum percentages.
You’re right. Thanks!
Cool! No problem! 🙂
Just out of curiosity, are you in need of any junior columnists or staff members at the moment? I am a senior computer engineering undergrad and love researching about the latest and greatest in computer technology. I don’t have any professional experience in journalism, per se, but I have written much in forums and I have been a reviewer of several product samples (tablet accessories and audiophile headphones) on Amazon and Head-Fi.
I’m not currently hiring, but if that changes I’ll definitely keep you in mind.
I thought Cannon Lake would be 10nm, since Kaby Lake is the third and last iteration of the 14nm die shrink.