[Ashraf] 10nm "Lakefield" SoC with Intel big + little cores

Page 17 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

coercitiv

Diamond Member
Jan 24, 2014
6,151
11,686
136
Significantly better battery life is one thing I counted for on Lakefield, but it doesn't seem to be enough, and Tigerlake with 10nm+ may do a fair bit better than Icelake.
Nah, Tiger Lake will have poor battery life, those big cores are hungry. Only 17+ hrs is with a huge 48Wh battery, and the device weighs almost 1kg!

EdD1OnvWsAQ32Or.jpg
 

Thala

Golden Member
Nov 12, 2014
1,355
653
136
I was kidding, the Tiger Lake implementation in X1 is looking very good in terms of size, weight, battery life - meaning Lakefield implementation in Galaxy S looks awkward with 10-12 hours on a 42Wh battery.

Looking at the video runtime is pretty much meaningless when comparing Tiger Lake with Lakefield. In such use-cases the main consumers are the screen, the video decoder, dram and other system components much more than the cores. So unless we are running something meaningful for the Tiger Lake CPUs any speculation about battery duration is moot.
 
  • Like
Reactions: moinmoin

coercitiv

Diamond Member
Jan 24, 2014
6,151
11,686
136
Looking at the video runtime is pretty much meaningless when comparing Tiger Lake with Lakefield.
The X1 is Project Athena certified, which according to Intel means 16 or more hours of local video playback and 9 or more hours of real wireless web browsing on WiFi and with 250 nits of brightness. The Galaxy S scored 10 hours of browsing in the Notebookcheck test with 150 nits of brightness and 12 hours in the video playback test. Given these figures, there is a good chance that the X1 Nano will match the Book S in normalized battery life for both video playback and web browsing. Therefore speculation about battery duration on the X1 Nano is anything but moot.

Now whether other products based on Lakefield will exhibit the same behavior as the Book S, that remains to be seen. It's obvious we need more devices to get a proper gauge of what Lakefield can offer.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
The X1 is Project Athena certified, which according to Intel means 16 or more hours of local video playback and 9 or more hours of real wireless web browsing on WiFi and with 250 nits of brightness.

We'll see how it really does. I was hyped because of the Project Athena certification but it came out to be nothing special in Icelake.

The XPS 13 7390 for example is a regression compared to Kabylake-R based 9370: https://www.notebookcheck.net/Dell-...Before-It.436573.0.html#toc-energy-management

The Yoga C940 doesn't do any better per WHr either: https://www.notebookcheck.net/Lenov...#toc-energy-management-of-the-lenovo-computer

You'll see the Galaxy Book S Intel is also rated at 17 hours in video browsing.

It gets 17 hours and 50 mins here: https://pcforalla.idg.se/2.1054/1.721751/samsung-galaxy-book-s

It actually beats the Surface Pro X implementation: https://pcforalla.idg.se/2.1054/1.724424/surface-pro-x

I am hopeful Tigerlake will do better because of the improved process, but we'll see.
 

coercitiv

Diamond Member
Jan 24, 2014
6,151
11,686
136
We'll see how it really does. I was hyped because of the Project Athena certification but it came out to be nothing special in Icelake.
I used the project Athena replacement for the rather vague 17+ hrs claim in the Lenovo marketing slide.

That being said, the 2019 X1 Carbon got 9h 22 min of browsing in Notebookcheck's battery test without Athena certification, running on Whiskey Lake and a 51WH battery.

As for Ice Lake examples, no offense but it's an unmitigated mess. For the first time ever I actually avoided buying a cheaper ICL laptop over Renoir not because of performance (plenty on both for the user in question), but rather to avoid a nasty surprise in terms of thermals/battery life. Even with all my rants on TGL projected performance I am reasonably convinced I wouldn't have had to make the same one-sided decision in a TGL vs. Renoir situation (would have gone with better deal instead).
 

Thala

Golden Member
Nov 12, 2014
1,355
653
136
The X1 is Project Athena certified, which according to Intel means 16 or more hours of local video playback and 9 or more hours of real wireless web browsing on WiFi and with 250 nits of brightness. The Galaxy S scored 10 hours of browsing in the Notebookcheck test with 150 nits of brightness and 12 hours in the video playback test. Given these figures, there is a good chance that the X1 Nano will match the Book S in normalized battery life for both video playback and web browsing. Therefore speculation about battery duration on the X1 Nano is anything but moot.

Interesting. You were going from "X1 is looking very good in terms of battery life and Lakefield looking awkward" to "there is a chance that X1 Nano beats Book S in normalized battery life". I'd say both devices are looking to provide pretty average normalized battery life, very comparable to Icelake devices.
 
Last edited:

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
I used the project Athena replacement for the rather vague 17+ hrs claim in the Lenovo marketing slide.

That being said, the 2019 X1 Carbon got 9h 22 min of browsing in Notebookcheck's battery test without Athena certification, running on Whiskey Lake and a 51WH battery.

As for Ice Lake examples, no offense but it's an unmitigated mess. For the first time ever I actually avoided buying a cheaper ICL laptop over Renoir not because of performance (plenty on both for the user in question), but rather to avoid a nasty surprise in terms of thermals/battery life. Even with all my rants on TGL projected performance I am reasonably convinced I wouldn't have had to make the same one-sided decision in a TGL vs. Renoir situation (would have gone with better deal instead).

The only place where I see Ice Lake doing well is on the bottom end of it's stack. The i3 1005G1 seems to be a clear value choice in the sub-$400 segment of the market at the moment. It is a noticeable improvement over the (admittedly outdated) 3200u based laptops and the previous gen 8000 series i3 laptops. It makes it absolutely pointless to get any of the celeron/pentium models that liter the bargain bins.
 

coercitiv

Diamond Member
Jan 24, 2014
6,151
11,686
136
Interesting. You were going from "X1 is looking very good in terms of battery life and Lakefield looking awkward" to "there is a chance that X1 Nano beats Book S in normalized battery life".
Do you acknowledge the topic of discussion here or are you just nitpicking to save face?

The subject at hand is Lakefield and it's supposedly better battery life over big core only chips. I'm not going anywhere from "X1 is looking very good in terms of battery life and Lakefield is looking awkward", you're the one who needs to change the narrative to still prove me wrong (because apparently my speculation wasn't moot). It doesn't matter if Lakefield is matched in battery life by Whiskey Lake, Ice Lake, Tiger Lake, the fact that any of these platforms manage to do it is a big question mark that makes Lakefield look awkward.

I'd say both devices are looking to provide pretty average normalized battery life, very comparable to Icelake devices.
Did you just speculate on a Tiger Lake machine battery life?!
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Even with all my rants on TGL projected performance I am reasonably convinced I wouldn't have had to make the same one-sided decision in a TGL vs. Renoir situation (would have gone with better deal instead).

I get you. Remember the review about Cannonlake? How it used more power at the same frequency level? It used 50% more power!

I think Icelake's 10nm still has lot of the same issues. It's just good enough to be better than the predecessor due to the improved uarch, and graphics.

I see lot of user reviews on reddit and elsewhere that Icelake-based systems aren't getting the battery life expected.

There is one benefit of getting such a low TDP chip though. The Galaxy Book S in a load scenario gets nearly 4 hours of battery life, and with AC connected uses only 15W. You'd be lucky to see 1.5 hours on a U system.

The only place where I see Ice Lake doing well is on the bottom end of it's stack. The i3 1005G1 seems to be a clear value choice in the sub-$400 segment of the market at the moment.

Take a look at this. It's from last year.

 

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
I'm actually currently digging around on the ~$300 market for a starter laptop for my daughter. My choices are largely 3200u laptops vs. i3-1005g1 laptops. When I compare benchmarks, the i3 is consistently winning, and sometimes by a lot, even in graphics.

It would be different if AMD had scored low cost wins for the 3300u, but those all wind up as pro SKUs and cost much more.

I wish there were 4300u products out there...
 

Thala

Golden Member
Nov 12, 2014
1,355
653
136
Do you acknowledge the topic of discussion here or are you just nitpicking to save face?

The subject at hand is Lakefield and it's supposedly better battery life over big core only chips.

I just explained that you are looking at the wrong power numbers. Both video and web-browsing are iso-work, why would Lakefield excel here at all? You might want to look at the load and maybe idle/standby power numbers to understand where Lakefield promises a longer battery runtime.
If you just want longer battery duration for web-browsing etc., you could just get an ARM device with much longer runtimes and call it a day.

ps. And yes i agree that Lakefield is mostly disappointing - and the disappointment comes when comparing to the ARM devices but not with its x86 siblings.
 
Last edited:
  • Like
Reactions: Tlh97

coercitiv

Diamond Member
Jan 24, 2014
6,151
11,686
136
I just explained that you are looking at the wrong power numbers. Both video and web-browsing are iso-work, why would Lakefield excel here at all? You might want to look at the load and maybe idle/standby power numbers to understand where Lakefield promises a longer battery runtime.
If you just want longer battery duration for web-browsing etc., you could just get an ARM device with much longer runtimes and call it a day.
Fair enough.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
I'm seeing a 15-20% gap between Lakefield's SNC and Icelake's SNC. Some may be due to TDP but now I'm starting to think it might be due to lack of AVX support on Lakefield. This trend holds true through R11.5, R15 and R20.

28% gap with SNC having a 7.4% clock advantage is extremely good for Tremont. That puts Tremont at Skylake level perf/clock, not Haswell at least for Cinebench.

Maybe it is just for Cinebench. If we assume Tremont is indeed running at 2.7GHz from Golem's test, then its not so impressive against Goldmont Plus, with only ~20% gains. It's extremely poor in R15, with Tremont underperforming Goldmont Plus per clock. And in Geekbench, its almost zero gain.

Then again, we're assuming it runs at a stable 2.7GHz. I can't see why a single Tremont core can't run at that frequency though. The low TDP and hybrid configuration really complicates comparisons.
 
Last edited:

Antey

Member
Jul 4, 2019
105
153
116
Sunny cove core 10nm INTEL: 4,46 mm²
Tremont core 10nm INTEL: 0,84 mm²

Sunny cove (icelake) 10nm INTEL: 6,91 mm² (wikichip).

Lakefield sunny cove is way smaller than it's icelake counterpart, or i wrongly read the total area or it is indeed smaller.

Edit:
Knowing that Lakefield was going to have to take the lowest common denominator from the two core designs, Intel physically removed the very bulky AVX-512 unit from the Sunny Cove core.
wow, it seems AVX takes a lot of space!


Image: die área is 82mm² (compute die). 1200px*825px or 10,92*7,511 mm

 

Attachments

  • LAKEFIELD.png
    LAKEFIELD.png
    2.7 MB · Views: 24
Last edited:

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
They actually produced a cut-down Sunny Cove core for Lakefield instead of just doing a direct optical copy? Interesting. Wonder how much that cost them.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Lakefield sunny cove is way smaller than it's icelake counterpart, or i wrongly read the total area or it is indeed smaller.

Edit:

wow, it seems AVX takes a lot of space!

No, and No.

You can see from the die shot its the exact same core as the one in Icelake. And you can see the AVX units.

Icelake's version is only larger because it includes the L3 cache and the FiVR.
 

Roland00Address

Platinum Member
Dec 17, 2008
2,196
260
126
Reading the 7nm news and Intel is going to use the Lakefield tech such as EMIB and Foveros with more chips.

So Lakefield LIVES, it is alive! ... In a kind of Frankenstein manner. I hope the mad scientists style of Intel will be far more successful than the 2015 to 2020 Intel.
 

Thala

Golden Member
Nov 12, 2014
1,355
653
136

I just read the golem.de article and i do not understand where the numbers in table are from. The only MT test they did was Geekbench and here the 8CX is about 50% faster then the Lakemont - as we already know from notebookcheck. The ST test consisting mostly of Office tests and Web. Office is emulated on the 8CX though.

ps. Seems they counting the Excel test as multithreaded - which indeed might be reasonable.
 
  • Like
Reactions: Tlh97

vigilant007

Junior Member
Dec 7, 2014
21
10
81
Lakefield just seems like such a bad idea. The Atom cores were dogs to begin with.

I’d wager that any modern ARM high efficiency core is probably better than a recently rehashed ARM chip. It seems pretty obvious to me that ARM High Performance cores would obliterate anything Atom.


Sent from my iPad using Tapatalk Pro
 

vigilant007

Junior Member
Dec 7, 2014
21
10
81
From a product positioning standpoint I agree, but as a test vessel for advanced packaging it's probably a great idea.

I agree, some of the concepts are interesting. But trying to bolt four trouts to a shark doesn’t mean a better killing machine.

For a thermal perspective, stacking CPUs like this, seems like it would lead to the potential of the thermals of the Atom core could indirectly impact the ability of the Core core to perform.

The other issues I potentially see is from a scheduling perspective. Please, correct me where I’m wrong, but largely speaking doesn’t most operating systems assume that all threads are the same? i know some adjustments were made to make “Hyperthreading” work, but having 2 grossly different architectures that can run only further complicates the matter.

Sent from my iPad using Tapatalk Pro
 
  • Haha
Reactions: clemsyn

Thala

Golden Member
Nov 12, 2014
1,355
653
136
The other issues I potentially see is from a scheduling perspective. Please, correct me where I’m wrong, but largely speaking doesn’t most operating systems assume that all threads are the same? i know some adjustments were made to make “Hyperthreading” work, but having 2 grossly different architectures that can run only further complicates the matter.

Sent from my iPad using Tapatalk Pro

As i already mentioned elsewhere, Windows on big.LITTLE ARM cores like on the Surface Pro X works really well. You see the small core utilized either on very low workloads like background processes or when the application requests 8 threads with full utilization, then all 8 cores are loaded 100%.
Of course the application can still make stupid decisions, like distributing the work equally to all 8 threads - but this is rarely the case in application where MT performance counts. For example Blender distributes one tile at a time to each core. Once a core is finished with a tile it gets the next tile assigned. You can literally see which tiles are assigned to the small cores :)

ps. I do still have the issue, that under Linux the CPU does not increase the clock when all cores are requested with full utilization on my Surface Pro X. As workaround i start a program on the Windows side which pulls the clock-speed up while running the Linux app under virtualization. Even then i am beating Lakefield in Blender by miles even with a relatively ancient Cortex A76 system - all while using less power from battery.
 
  • Like
Reactions: Tlh97