Intel will allegedly be releasing another 14nm processor following Kaby Lake (which is itself a 14nm successor to Skylake) in 2018. The new processors are code named "Coffee Lake" and will be released alongside low power runs of 10nm Cannon Lake chips.
Not much information is known about Coffee Lake outside of leaked slides and rumors, but the first processors slated to launch in 2018 will be mainstream mobile chips that will come in U and HQ mobile flavors which are 15W to 28W and 35W to 45W TDP chips respectively. Of course, these processors will be built on a very mature 14nm process with the usual small performance and efficiency gains beyond Skylake and Kaby Lake. The chips should have a better graphics unit, but perhaps more interesting is that the slides suggest that Coffee Lake will be the first architecture where Intel will bring "hexacore" (6 core) processors into mainstream consumer chips! The HQ-class Coffee Lake processors will reportedly come in two, four, and six core variants with Intel GT3e class GPUs. Meanwhile the lower power U-class chips top out at dual cores with GT3e class graphics. This is interesting because Intel has previous held back the six core CPUs for its more expensive and higher margin HEDT and Xeon platforms.
Of course 2018 is also the year for Cannon Lake which would have been the "tock" in Intel's old tick-tock schedule (which is no more) as the chips will move to a smaller process node and then Intel would improve on the 10nm process from there in future architectures. Cannon Lake is supposed to be built on the tiny 10nm node, and it appears that the first chips on this node will be ultra low power versions for laptops and tablets. Occupying the ULV platform's U-class (15W) and Y-class (4.5W), Cannon Lake CPUs will be dual cores with GT2 graphics. These chips should sip power while giving comparable performance to Kaby and Coffee Lake perhaps even matching the performance of the Coffee Lake U processors!
Stay tuned to PC Perspective for more information!
If Intel gives me half their
If Intel gives me half their marketing budget, I’ll come up with a better naming scheme in 10 minutes. Have them PM me here.
Those aren’t marketing names.
Those aren’t marketing names. They are internal code names.
Skylake is used in marketing
Skylake is used in marketing materials. It’s on the first page of the product brief. It is also used in ARK. If it were truly an internal name, you wouldn’t see it on external materials.
https://www-ssl.intel.com/content/dam/www/public/us/en/documents/product-briefs/6th-gen-core-family-desktop-brief.pdf
And even if you don’t consider Skylake to be an external name, “Intel 6th generation Core i7 and i5 processors” is not any better.
That’s a byproduct of 24-hour
That’s a byproduct of 24-hour news cycles and constant fixation on all things tech. They let the “insiders” know what they’re talking about without completely transitioning to the official name.
BTW, it clearly shows “codename” on that first page.
Still gonna milk 14nm through
Still gonna milk 14nm through 2018, huh, Intel?
Remind me how long AMD has
Remind me how long AMD has been selling their shitty 32nm CPUs with “8 cores” that get outperformed by i3s? Ok.
A slacking non-competitor
A slacking non-competitor isn’t something Intel should aim to compete or compare itself with.
They arent… theyre
They arent… theyre competing with Nvidia and ARM for datacenters and supercomputers.
Your assertion that theyre going to milk 14nm means nothing. Complain about TSMC, AMD and the rest of them for not having 10nm already?
Actually i need to elaborate
Actually i need to elaborate on what i just said.
Youre implying that shrinking features is going to increase performance. It hasnt for a long time, and it wont do much in the future.
The problem with CPUs and GPUs now is NOT the feature size needing to get smaller. The problem is twofold: memory and interconnects.
Memory has not changed significantly for most CPUs and GPUs in many years, and dual or quad channel DDR3 or 4 is simply insufficient to keep cores fed. GDDR5 is only now being replaced in GPUs by HBM2, and only in one GPU that is not used in any consumer parts.
Intel is using HMC in their Xeon Phis and Fujitsu has been using it in their PrimeHPC FX100 since 2014, but once again there is no consumer part using HMC.
The necessity for more memory bandwidth and capacity is driving HBM, HMC and XPoint adoption.
Intel and Fujitsu are both developing silicon photonic interconnects for their next generations of CPUs.
Currently, all PCs are using 2-4 channel DDR3 or DDR4 with less than 50GB/s per CHIP, NOT PER CORE, which seriously limits performance.
The CPU/GPU interface is still PCI-E. Unless youre using an NVMe SSD that Intels CPUs can directly talk to, your I/O subsystem is also ancient and slow.
If you improved the memory, interconnects and I/O subsystems, you could use a 22nm CPU and quadruple your actual performance.
I think the main point of
I think the main point of this is there will be no 10mm chips until at least 2019. That is a sad state of affairs for the entire industry, especially considering Intel has always lead fabrication shrinks.
Yes, but my point is that
Yes, but my point is that people are focused on something that is almost irrelevant, considering how little progress has been made in areas that create actual bottlenecks.
Its not new either. Node names dont even mean that much, and havent for years.
http://spectrum.ieee.org/semiconductors/devices/the-status-of-moores-law-its-complicated
A lot of the shrinks are more about marketing than actual performance improvements.
For desktop CPUs, the performance increase has been about 5% per generation for the last 8 years or so. Thats because people have been conditioned to think that smaller nodes are automatically better, and dont understand that the real bottlecks involve memory and cache.
Intel also focuses a huge, and rather pointless for many people, amount of effort on their integrated GPUs. If they used the die space for extra cores or cache on their small chips, who really cares if its 22nm or 10? Id rather have a 22nm chip with a better architecture, but people dont care about architecture. They care how many nanometers a node name is labeled, even if that name doesnt even indicate a meaningful feature size.
Wrong. For most tasks,
Wrong. For most tasks, memory isn’t a significant bottleneck, simply because both the CPU and GPU pre-caches a lot of the needed information. Why do you think we suddenly need 4GB VRAM, or have 128MB of L3 cache? It’s to get around memory latency.
Now yes, iGPUs/APUs get starved by latency related problems, but that’s the downside to their architecture. You can’t get high bandwidth/low latency RAM at an affordable cost, and CPUs care more about latency then bandwidth. So iGPUs/APUs get starved out. That’s why, for gamers, dedicated GPUs are not going away any time soon.
Secondly, PCI-Express isn’t a significant bottleneck. Bandwidth wise, even PCI-E 2.0 x16/PCI-E 3.0 x8 isn’t maxed out yet, let alone PCI-E 3.0 x16. Again, VRAM is being used to pre-cache a lot of the necessary information to get around any latency related issues.
You can prove this by doing simple CPU tests that don’t touch main memory; memory isn’t a significant bottleneck. The real problem is creating new process nodes is expensive, and there’s only so much performance you can gain by throwing more transistors at the problem.
It just shows some things:
1)
It just shows some things:
1) How difficult it is to actually go down to 10nm
2) How expensive it is
3) How much they are affected by stagnating PC sales (as more and more people are satisfied with smartphones or tablets)
4) That the higher end of their offerings don’t bring in as much profit as the lower end.
5) That they don’t see Zen as a short term threat.
It makes perfect sense to stay on 14nm as long as possible considering these points.
Had AMD actually been more of a competitor, maybe they’d accelerate going to 10nm for higher end as well.
Well Intel fans will finally
Well Intel fans will finally stop hitting their heads on the wall shouting “Why Intel, WHYYYYY?”. For (cpu) generations Intel was giving 4 cores +5% extra performance and a somewhat better iGPU while fans where asking for more cores. It seems that, with coffee lake, Intel will finally have to give to those still using even Ivy Bridge OCed models, a reason to upgrade. I am sure Zen is just a coincidence(and no reason for those hoping AMD to die to start using their brains).
Hell Sandy Bridge is still
Hell Sandy Bridge is still largely holding it’s own in many workloads even on the performance/enthusiast side of things. Especially of it has a decent OC.
They should call it Coffee
They should call it Coffee Break, because that’s what Intel’s consumer division has been on for years, since there is no need to work hard when your main competitor is AMD.
Coffee Lake…….sounds like
Coffee Lake…….sounds like somebody at intel admitting they are taking their sweet time, aka a coffee break for the last what…..5 years.
…admitting they’v had
…admitting they’v had enough coffee between now and then to fill a lake.