Intel debuts Comet Lake: Further complicates laptop CPU family with eight new 14nm chips

Shawn Knight

Posts: 15,256   +192
Staff member
What just happened? With 19 total mobile processors announced this month alone, there’s going to be an awful lot of confusion when it comes time to shop for a new Intel-powered laptop. And we also can’t turn a blind eye to the fact that Intel is revisiting 14nm. Why? Are there production issues with 10nm that forced Intel’s hand? What will pricing look like between the two lines?

Intel earlier this month announced nearly a dozen 10th-generation “Ice Lake” processors built on its 10nm manufacturing process. On Wednesday, the chipmaker rolled out eight additional 10th Gen Intel Core processors for laptops that, confusingly enough, utilize the 14nm process. Wait, what?

Intel’s 10th Gen “Comet Lake” U-Series mobile processors include the Core i7-10710U, the Core i7-10510U, the Core i5-10210U and the Core i3-10110U while the Y-Series line is comprised of the Core i7-10510Y, the Core i5-10310Y, the Core i5-10210Y and the Core i3-10110Y.

As you can see from the PowerPoint slide above, there’s quite the variety in terms of core / thread count, operating frequency, cache and memory support. The picture starts to become even clearer after examining the following slide from Intel’s press deck.

Despite being built on a 10nm process and having fewer cores / threads and a lower Max Turbo frequency, Ice Lake actually consumes more power than the 14nm Comet Lake.

In Ice Lake’s defense, some iterations do use Intel Iris Plus graphics while Comet Lake is limited to Intel UHD graphics.

According to Intel, Comet Lake delivers up to 16 percent better overall performance compared to previous U-Series CPUs and over 41 percent better productivity in Office 365 versus previous U-Series chips. Like Ice Lake, Comet Lake also gets Intel’s Wi-Fi 6 (Gig+) wireless technology and support for Thunderbolt 3.

With the introduction of Ice Lake earlier this month and now Comet Lake, Intel said the result is a “complete family of processors that complement each other, giving people choices for what usages matter most to them and re-imagining the possibilities of a modern laptop experience.”

Permalink to story.

 
It would appear that Intel's 14nm+++ process is better than Intel's new 10nm process.

Look at the max boost. Look at the core counts, 2 extra with the 14nm.

Do you think those old 14nm process CPUs are going to keep up with AMD's 7nm cores??

Does this mean they put the Ice Lake cores into Commet lake's die, notice the 16% reference for performance with no increase in boost or base clocks, and the same 14nm process.

Wow the base clock for the 6 core part is only 1.1 GHz, so after about a minute of using the computer, it will slow down to 1/4 the original speed.
 
Hahaha, this is a joke, right? So you can get a 15-W TDP hexa-core CPU... but at 1.1 GHz?! That makes it 300 MHz higher than throttling frequency. Marketing then proceeds to show turbos of ~4 GHz that will still exceed that TDP by at least 5 times.

My "45 W TDP" Kaby Lake reached ~115 W on turbo, according to my calculations (C * f * V^2) using CPUID reported values, and I did see a delta of 80 W at the power outlet in the same CPU stress workload just comparing Turbo Boost disabled and enabled (I read the core voltage under stress with Turbo disabled and then enabled, to factor it in the equation).
So let's say, hypothetically, that at 1.1 GHz the core voltage is 0.8 V and at 3.9 GHz it is 1.15 V. You would have a hypothetical TDP factor of 7.3 ((3.9 * 1.15^2) / (1.1 * 0.8^2)) a.k.a ~110 W.
 
In Ice Lake’s defense, some iterations do use Intel Iris Plus graphics while Comet Lake is limited to Intel UHD graphics.
Ice Lake's UHD and Iris Plus are both Gen11 GPUs, where as Comet Lake's UHD is Gen9.5 - there's quite a bit of difference between the two architectures and the former is larger in terms of transistor count, so it's possible that some of the power difference lies in this area (especially since the 'full' Iris Pro is 64 EUs in size, compared to Gen 9.5 UHD's 32 ).
 
It would appear that Intel's 14nm+++ process is better than Intel's new 10nm process.



Wow the base clock for the 6 core part is only 1.1 GHz, so after about a minute of using the computer, it will slow down to 1/4 the original speed.

That's not how Turbo works. If you have an Intel CPU in your PC, download Intel Power Gadget and do some tests for yourself.

Your PC will run at its all core Turbo speed permanently as long as it stays below the max Wattage set in BIOS and your cooler can keep the CPU temps below Tj (max). For these mobile parts from Intel, those numbers are the config Up TDP (25W) and 100°C.

Here's an example for the CPU in the NUC sitting on my desk using my own tests:

Core i5-7260U
2.2 GHz, 3.4 GHz turbo
15W, TDP-up @25W

Handbrake 1080p h.264 to h.265 video conversion: runs at 3.4 GHz for an hour long encode using~22W at 78°C constant. No throttle, no backing off to 2.2 GHz after a minute or anything silly like that.

Just for fun I re-ran the test using XTU to set TDP to 15W: runs @~3.1 GHz (bouncing from 3.0-3.2GHz) at 70°C constant. No big throttle, no backing off to 2.2 GHz after a minute or anything silly like that.

These new chips probably cannot do 4 or 6-core at max all-core turbo for more than a few seconds at 25W so they will throttle down to a 4-core speed in the low 3GHz range and the 6-core parts will run in the high 2GHz range with all 6 active and in the low 3GHz range like the others with 4 cores active.

The only condition you might see that 1.1 GHz speed is when running the 6-core part in 15W mode with an all-core AVX load. Even then, with that load I bet it will still run above that speed.
 
Last edited:
It appears to me that performance is going to largely vary based on the configuration the laptop is put into. More importantly, we likely won't see much if any improvement in regards to performance per watt. We'll see when benchmarks start going out as it's nearly impossible to gauge based on these slides alone.
 
I am completely confused, and I consider myself a techie

Ok I have unconfused myself by reading the following article:

https://www-pcworld-com.cdn.ampproject.org/v/s/www.pcworld.com/article/3431668/intel-announces-comet-lake-a-faster-10th-gen-whiskey-lake-chip-for-mainstream-laptops.amp.html?amp_js_v=a2&amp_gsa=1#referrer=https://www.google.com&amp_tf=From %1$s&ampshare=https://www.pcworld.com/article/3431668/intel-announces-comet-lake-a-faster-10th-gen-whiskey-lake-chip-for-mainstream-laptops.html

Basically, intel can't meet all the demand for its limited 10nm CPU production, so this 14nm+++ is meant to mitigate that. As to what's different, the 10nm (Ice Lake) has much better integrated graphics, whereas the 14nm+++ (Comet Lake) is faster than the 10nm in CPU performance (although Comet Lake is 41% faster in Office 365 than Whiskey Lake for some reason, strange as that is). Oddly enough, they're both the same in terms of battery life, owing to a very refined & mature 14nm node by Intel. You can differentiate between the 2 because Ice Lake CPUs will have a "G" in their name (i5-1035G7), and Comet Lake CPUs will have the traditional "U" at the end of their name (i7-10710U).

There, that's everything in a nutshell.
 
Last edited:
The 8th generation of Core(Coffee Lake) had a huge upgrade on the number of cores and single core IPC to the 7th Kaby Lake under the pressure of Ryzen. I don't understand why Comet Lake uses back the 6th generation of Sky Lake architecture.
Plus, the 10th generation of Core family now has both 10nm Ice Lake and 14nm+++ Sky Lake? What a mess of Intel!
 
With all those security holes, it also has WiFi integrated? Wow! That hole connects the chip directly to the internet??
 
The 8th generation of Core(Coffee Lake) had a huge upgrade on the number of cores and single core IPC to the 7th Kaby Lake under the pressure of Ryzen. I don't understand why Comet Lake uses back the 6th generation of Sky Lake architecture.
Plus, the 10th generation of Core family now has both 10nm Ice Lake and 14nm+++ Sky Lake? What a mess of Intel!
Kaby Lake uses the Skylake architecture.
Coffee Lake uses the Skylake architecture.
Comet Lake uses the Skylake architecture.

There are minor silicon differences addressing security vulnerabilities in each gen but the basic architecture, and most importantly the IPC, has remained unchanged since 2015.
 
Kaby Lake uses the Skylake architecture.
Coffee Lake uses the Skylake architecture.
Comet Lake uses the Skylake architecture.

There are minor silicon differences addressing security vulnerabilities in each gen but the basic architecture, and most importantly the IPC, has remained unchanged since 2015.
Thank you for your clarification for the architecture topic.
 
Intel's slides from the 2019 Investor Meeting paint a clear picture of their 10 nm struggles:

JvQtRAA.jpg

Hum8e3l.jpg

f4ZkQhk.jpg

9ZVzvqz.jpg

ybjWyZa.jpg


Interesting to see that the 14nm processes will still be running out products (new ones?) for another two years. Kinda makes sense, given that the Intel have spent 5 years constantly revising a process they only expected to run for 2 years.
 
LMFAO that efficiency graph 1272 -> 1274. Marketing scaling. So they basically got SFA gain on perf/W and graph the **** out of that thing. HAHAHAHAHA!!!
 
It's further evidence that Intel's 10nm woes are extremely serious. Their revised 10nm process appears unable to significantly outperform the 14++ process they have been using for a few years now. This news has been circulating for some time.

Their 10nm allows much greater transistor density and thus nice IPC gains of course. But they cannot clock the parts high enough to beat all these older, high 4GHz-5GHz+ 14nm parts they have had on the market for a while.

The 10nm yields are poor and the chips they get won't reach the speeds needed on the low voltage required to see an improvement in performance, or TDP reduction.

So they have poured billions of dollars into their 10nm process over a 5 year period and the best they can get out of it is performance parity with 14nm. Erk.
 
Last edited:
Back