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

Page 10 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
If so, then a dual-core SNC may not look so bad in perf/watt compared to Lakefield, except maybe in idle power consumption which is arguably just as important for tablets and other small devices.

Dunno, the 2020 MBA i3 gets around 290-300 in sustained R15 MT (at around 10 package W) versus the fastest n5000 score they had, which was 280-290 @ 6 W.
 

insertcarehere

Senior member
Jan 17, 2013
639
607
136
But it doesn't have to take precedence. That's the point.

The whole Lakefield exercise is about pushing efficiency as far as possible without losing functionality.

Honestly, I don't understand why this is so underestimated by people on enthusiast forums.
Most of the software you use today on Windows won't work on that ARM-powered Samsung Book S. It will on the Lakefield one. Simple as that.

Most of the software that will actually net a substantial performance benefit from being natively on x86 simply won't work well on any Lakefield laptop anyway. The observation that stuff like PC gaming/Video editing/CAD will work better on Lakefield vs ARM is not particularly relevant when given power limits and form factors it's still going to be pretty unusable on the former.
 

Thala

Golden Member
Nov 12, 2014
1,355
653
136
Honestly, I don't understand why this is so underestimated by people on enthusiast forums.
Most of the software you use today on Windows won't work on that ARM-powered Samsung Book S. It will on the Lakefield one. Simple as that.

Most of the software i use today on a tablet form factor runs natively on my Surface Pro X (either under Windows or Linux). So do not make assumptions about SW usage.
Of course i also use software, which wont run on the ARM device - like AAA games - but i would not run these on a 7W tablet computer anyway - so it is a moot point, that this software would theoretically run.
And then there is SW, which runs perfectly fine under emulation. Like one of my favourite game series, the Trine Trilogie is running at max settings totally smooth 60fps under emulation.
 
Last edited:
  • Like
Reactions: Tlh97

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
Most of the software i use today on a tablet form factor runs natively on my Surface Pro X (either under Windows or Linux). So do not make assumptions about SW usage.
Of course i also use software, which wont run on the ARM device - like AAA games - but i would not run these on a 7W tablet computer anyway - so it is a moot point, that this software would theoretically run.
And then there is SW, which runs perfectly fine under emulation. Like one of my favourite game series, the Trine Trilogie is running at max settings totally smooth 60fps under emulation.

Interesting point. Your usage habits may not be indicative of the average lightweight notebook user, or maybe they are. But if the most-used software out there can run fine on multiple hardware platforms, then x86 is not really a major advantage merely as an ISA.
 

SAAA

Senior member
May 14, 2014
541
126
116
Very interesting technology soon available to everyone.

By the way no one pointed out how the naming for the two released chips, i5-L16G7 and i3-L13G4, hints at possibly more powerful variants: a 2+6 cores i7 would be a sizable jump, or 2+8 if they have to keep the ratio of big to small cores for some reason. Alder lake being 8+8 doesn't hint at that though.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
The biggest issue with Lakefield is that its coming so late. Had it be available Feb-Mar of this year it would have been a lot better.

Basically now the point of the product is to displace WoA products. That's fine, but it won't be financially compelling for Intel.

I can see some wanting the smallest, lightest, best battery life in an x86 device. But its not a huge market.

By the way no one pointed out how the naming for the two released chips, i5-L16G7 and i3-L13G4, hints at possibly more powerful variants: a 2+6 cores i7 would be a sizable jump, or 2+8 if they have to keep the ratio of big to small cores for some reason. Alder lake being 8+8 doesn't hint at that though.

I can see that happening with a 7nm successor but die size is too small to accommodate that many cores. If there's an i7 version it'll just be higher clocked. Also, based on boost clocks, there's no thermal/power room for more cores.

Ryefield with 2x Golden Cove and 8x Gracemont cores with Gen 13 graphics on 7nm!
 

Thala

Golden Member
Nov 12, 2014
1,355
653
136
Basically now the point of the product is to displace WoA products. That's fine, but it won't be financially compelling for Intel.

Thats an interesting point. But what in the world would make me "upgrade" from my SQ1 SoC to this piece of garbage? I shouldn't be so harsh, but i assume that the performance is just not there. Keep in mind that a Cortex A76 can be considered outdated already - its 2 years old by now.
 
  • Like
Reactions: Tlh97

podspi

Golden Member
Jan 11, 2011
1,965
71
91
I wonder how long it will be before this becomes standard. It opens a lot of possibilities, particularly when you (I don't think this is the case for Lakefield) don't have to fab all the cores on the same silicon. It should allow them to really tweak things for max perf or max perf/W, get the best of both worlds.

Thats an interesting point. But what in the world would make me "upgrade" from my SQ1 SoC to this piece of garbage? I shouldnt be so harsh, but i assume that the performance is just not there.

I think garbage is too harsh a description. Its performance is very close to the SQ1, beating it in ST and losing in MT. This isn't meant to be an upgrade to what you have, it really is an upgrade if you have an older WARM device (raises hand) or an core M or Atom-based device already.

Is this performance native or emulated? https://browser.geekbench.com/v5/cpu/2478358
If its native, not bad. If its emulated... that is a GREAT result.

The i5 doesn't do too poorly: https://browser.geekbench.com/v5/cpu/1899203

Multithread performance isn't as high, but I'm not sure how much that matters for the use cases you'd expect these tablets/laptops to be used for. If you're going to use this as a web browsing or emailing machine, I doubt you'd notice much of a difference between the two. If you have a piece of x86-64 software you need to run the Intel product is clearly better, but only for compatibility reasons. Performance is a wash.
 

Thala

Golden Member
Nov 12, 2014
1,355
653
136
I think garbage is too harsh a description. Its performance is very close to the SQ1, beating it in ST and losing in MT. This isn't meant to be an upgrade to what you have, it really is an upgrade if you have an older WARM device (raises hand) or an core M or Atom-based device already.

Is this performance native or emulated? https://browser.geekbench.com/v5/cpu/2478358
If its native, not bad. If its emulated... that is a GREAT result.

The i5 doesn't do too poorly: https://browser.geekbench.com/v5/cpu/1899203

Thats almost factor 2x performance delta. Thats what i call garbage. And even if you have an older WoA device, this looks to be a sidegrade at best.
But yes, there might be some people who can cope with the low performance if the devices are really cheap.
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Thats almost factor 2x performance delta. Thats what i call garbage. And even if you have an older WoA device, this looks to be a sidegrade at best.
But yes, there might be some people who can cope with the low performance.

I'd wait to see. Yes I'm not too hopeful but the MT is only running on Tremont. If they can get SNC in there we'll be at ~2500.

Tremont core ST is 200MHz lower than Sunny Cove ST.

If their claims of SNC having 50% better ST performance is in a shipping SKU then Tremont is at Ivy Bridge levels of perf/clock.
 
Last edited:

coercitiv

Diamond Member
Jan 24, 2014
6,151
11,685
136
Multithread performance isn't as high, but I'm not sure how much that matters for the use cases you'd expect these tablets/laptops to be used for. If you're going to use this as a web browsing or emailing machine, I doubt you'd notice much of a difference between the two.
If MT performance doesn't make much of a difference in these devices then how is this going to be superior to a dual-core Cove chip, say a Tigerlake Y at 7-10W TDP?
 

beginner99

Diamond Member
Jun 2, 2009
5,208
1,580
136
If MT performance doesn't make much of a difference in these devices then how is this going to be superior to a dual-core Cove chip, say a Tigerlake Y at 7-10W TDP?

Idle power usage being very close to 0. Albeit I would assume the could also make a cove-core based chip like that?
 

Tabalan

Member
Feb 23, 2020
41
25
91
The biggest issue with Lakefield is that its coming so late. Had it be available Feb-Mar of this year it would have been a lot better.

Basically now the point of the product is to displace WoA products. That's fine, but it won't be financially compelling for Intel.

I can see some wanting the smallest, lightest, best battery life in an x86 device. But its not a huge market.
I think Lakefield might be realease mainly to get experience with 3D chiplet stacking (afaik, noone did that before?), to iron out issues which for sure did/will appear during designing, manufacturing such SKU. But overall, I agree that Lakefield is super late. Ttemont core was unveiled 8 months ago, while Sunny cove (Ice Lake) went into production 1,5 year ago.
 
  • Like
Reactions: beginner99

SAAA

Senior member
May 14, 2014
541
126
116
I think Lakefield might be realease mainly to get experience with 3D chiplet stacking (afaik, noone did that before?), to iron out issues which for sure did/will appear during designing, manufacturing such SKU. But overall, I agree that Lakefield is super late. Ttemont core was unveiled 8 months ago, while Sunny cove (Ice Lake) went into production 1,5 year ago.

Indeed, that's my only complaint: tech wise it's the future, regardless of current implementation with low power hence low performance. I see a lot of criticism around Lakefield but performance isn't the point here, rather usable performance (core in a tablet, maybe even phablet sized device) at the lowest possible power.

But it's true for every category it looks like Intel's late, always too late to matter recently, so ARM cores have been catching up to the point that at 2-3 GHz they compare not too badly to something like a Sunny cove at 3GHz, who completely negates the point of the large core in there.

If that leak/rumor with Goldmont performing around 1.5x Skylake IPC and Gracemont being almost equal is true then it would be another matter, not to talk about Ocean and successors. Because then it'd make sense: still a single core but performing severely faster will speed up many applications, with the aid of 4-8 Skylake-like cores for all the other tasks.
 

Roland00Address

Platinum Member
Dec 17, 2008
2,196
260
126
I think Lakefield might be realease mainly to get experience with 3D chiplet stacking (afaik, noone did that before?), to iron out issues which for sure did/will appear during designing, manufacturing such SKU. But overall, I agree that Lakefield is super late. Ttemont core was unveiled 8 months ago, while Sunny cove (Ice Lake) went into production 1,5 year ago.

Yes this will be a rationalization that even if Lakefield fails and costs lots of money you can sell it to the board and everyone else that at least it was practice for the next things.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
I think Lakefield might be realease mainly to get experience with 3D chiplet stacking (afaik, noone did that before?), to iron out issues which for sure did/will appear during designing, manufacturing such SKU. But overall, I agree that Lakefield is super late. Ttemont core was unveiled 8 months ago, while Sunny cove (Ice Lake) went into production 1,5 year ago.

That's one of the reasons why I found it so interesting. What concerns me is that it took Intel a long time to get it out to market. It looked "ready to go" alongside IceLake-U last year . . . and it wasn't.
 

Tabalan

Member
Feb 23, 2020
41
25
91
Yes this will be a rationalization that even if Lakefield fails and costs lots of money you can sell it to the board and everyone else that at least it was practice for the next things.
I'm 99% sure Lakefield will "fails and costs lots of money", that's usually the case when you release bleeding edge product with hope that it's technology will be useful for later generations. After almost 1 year from being announced we get 2 SKUs and so far only 2 design wins. No way it can pay off.

That's one of the reasons why I found it so interesting. What concerns me is that it took Intel a long time to get it out to market. It looked "ready to go" alongside IceLake-U last year . . . and it wasn't.
I think Intel might have some problems with production/assembly of this chip. At this point, Lakefield will be available (I mean availability of product with Lakefield ) only 3-4 months before Tiger Lake. Maybe Intel will act faster with 2nd gen Lakefield, Gracemont + Golden Cove could be amazing.

Also, both dies are really small and very low voltage/power so this might be good product for early 7nm production (Intel estimates 2021 for 7 nm, this aligns with release date for Gracemont and Golden Cove).
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Citation?

Based on leaked results. The MT score is always in the range of approximately 2x the ST score in Geekbench, which is what Intel claimed in their presentation for Tremont 4x versus Sunny Cove 1x.

Also we know the maximum Turbo for Tremont is 1.8GHz, while Geminilake can do 2.6-2.7. The MT Geekbench scores are pretty close between Lakefield and Pentium Silver N5000, clock speed differences making up for uarch differences.

@Tabalan We know the successor is Lakefield-R, with the graphics refreshed to using Xe. Nothing about the cores being changed though, although if Lakefield is on 10nm, moving to 10nm+ with LKF-R should result in nice clock improvements.

There's almost no advantage in product availability for Lakefield. They should come within a month or two of each other.
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
Also we know the maximum Turbo for Tremont is 1.8GHz, while Geminilake can do 2.6-2.7. The MT Geekbench scores are pretty close between Lakefield and Pentium Silver N5000, clock speed differences making up for uarch differences.

Geekbench does appear to be using 5 threads. But yes it is not that much faster in MT than the 6W N5000.
 

mikk

Diamond Member
May 15, 2012
4,112
2,106
136
Geekbench does appear to be using 5 threads. But yes it is not that much faster in MT than the 6W N5000.

Because N5000 uses much more power if it's not much slower.

Gemini Lake

PL1= 6W
PL2= 15W
PL3= 30W

Lakefield

PL1= 7W
PL2= 9.5W

That's why there are such big fluctuations in the Geekbench N5000 scores, because the real consumption can differ quite a lot between a run or device. The very best results vary between 1300-1500 which I'm sure these are 15W level results. I'm not even sure if N5000 can reach 1000-1100 when running strictly at 6W, there are lots of results under 1000.
 

mikk

Diamond Member
May 15, 2012
4,112
2,106
136
I doubt it. The N5000 in the tiny Topjoy Falcon gets 1580 in Geekbench 5.


This is a small device but it needs active cooling! It's also much much thicker than the Samsung Galaxy Book S. Because of the active cooling you can be sure this score has been achieved running at 15W. If it would be at 6W the big majority of N5000 devices would be close to this score but they are not.

@jpiniero

The Gemini Lake datasheet says it, so I'm sure.