Monday, March 23rd 2020

Intel Core i5-L15G7 Lakefield Processor Spotted

Intel has been experimenting with a concept of mixing various types of cores in a single package with a design called Lakefield. With this processor, you would get a package of relatively small dimensions that are 12-by-12-by-1 millimeters withing very low TDP. Thanks to the Twitter user InstLatX64 (@InstLatX64) we have some GeekBench 5 results of the new Lakefield chip. The CPU in question is the Core i5-L15G7, a 5 core CPU without HyperThreading. The 5C/5T would be a weird configuration if only Lakefield wasn't meant for such configs. There are one "big" Sunny Cove core and four "small" Tremont cores built on the 10 nm manufacturing process. This is the so-called compute die, where only the CPU cores are present. The base dies containing other stuff like I/O controllers and PHYs, memory etc. is made on a low-cost node like 22 nm, where performance isn't the primary target. The whole chip is targeting the 5-7 W TDP range.

In the GeekBench 5 result we got, the Core i5-L15G7 is a processor that has a base frequency of 1.4 GHz, while in the test it reached as high as 2.95 GHz speeds. This is presumably for the big Sunny Cove cores, as Tremont cores are supposed to be slower. The cache configuration reportedly puts 1.5 MB of L2$ and 4 MB of L3$ for the CPUs. If we take a look at performance numbers, the chip scores 725 points in single-core tests, while the multi-core result is 1566 points. We don't know what is the targeted market and what it competes with, however, if compared to some offerings from Snapdragon, like the Snapdragon 835, it offers double the single-threaded performance with a similar multi-core score. If this is meant to compete with the more powerful Snapdragon offerings like the 8cx model, comparing the two results in Intel's fail. While the two have similar single-core performance, the Snapdragon 8cx leads by as much as 76.9% in a multi-core scenario, giving this chip a heavy blow.
Intel Core i5-L15G7 Intel Lakefield
Source: Tom's Hardware
Add your own comment

47 Comments on Intel Core i5-L15G7 Lakefield Processor Spotted

#26
Caring1
notb
I hope this was a weird joke. :)
Nope, I'll admit my math sucks, but it makes sense to me.
Posted on Reply
#27
ratirt
thepath
GeekBench 5 favor ARM heavily. It is not a good benchmark for comparing X86 to ARM



It is 82mm
https://www.tomshardware.com/news/intel-lakefield-die-shot-shows-82mm-square-chip

Even Ice lake is 122mm. There is no way lackefield is bigger than ice lake
https://www.anandtech.com/show/15380/i-ran-off-with-intels-tiger-lake-wafer-who-wants-a-die-shot
Yeah. But the territory this CPU is going to be used is mobile which is ARM's area and the latter one is big here and faster. (tablets, phones)
notb
Zen 2 alternative doesn't exist.
And AMD just launched a 6W Zen SoC. Not even Zen+.

We've seen some leaked benchmarks and Zen2 mobile APUs will probably match Ice Lake U in single-thread performance (give or take) and won't scale so well for low power.
If using words like "annihilation" makes you happy - fine. But you may be disappointed...
What you comparing here is Desktop APU with something designed for mobile. The ARM is beating it in single core by a double and it has same multi threaded performance according to what OP wrote and yet you think this is going to beat fully fledged desktop processor? This Intel's CPU is going to be used in the mobile but I doubt Laptops will do much outta it. Most likely tablets (mentioned by someone before).
Anyway lets hope for the best and see where this particular processor is going to end up. For what OP said, it is really hard to hold your breath. The numbers show something different than what you say.
Posted on Reply
#28
londiste
Slowest Ice Lake CPUs are capable of 4000-4500 single-core scores in Geekbench. Somewhere around or above 2000 seem to be the worst case scores (with how unstable Geekbench is, that these are probably not too representative). I would suspect there is something not quite right about this Lakefield test run. Either bench or Windows scheduler not using the right cores?

ratirt
What you comparing here is Desktop APU with something designed for mobile. The ARM is beating it in single core by a double and it has same multi threaded performance according to what OP wrote and yet you think this is going to beat fully fledged desktop processor?
It could very easily beat a fully fledged desktop processor. Remember that we are talking about very low power limits for desktop CPUs here. Mobile cores might do OK but I suspect these would post lower multi-core scores due to power limits. Zen2 in form of Matisse cannot compete at all. Renoir should be able to compete but we know next to nothing about Renoir so I do not even have an estimation of how it compares.
Posted on Reply
#29
Mats
Caring1
Nope, I'll admit my math sucks, but it makes sense to me.
Here you go.


Posted on Reply
#30
ratirt
londiste
Slowest Ice Lake CPUs are capable of 4000-4500 single-core scores in Geekbench. Somewhere around or above 2000 seem to be the worst case scores (with how unstable Geekbench is, that these are probably not too representative). I would suspect there is something not quite right about this Lakefield test run. Either bench or Windows scheduler not using the right cores?
This one we are talking about here scored 756 in single core in Geekbench. Maybe it is not so representative but if we take this approach then basically we can say that about any benchmark. It is slower than ARM here. Maybe it is utilizing not the right core but single core performance should be better since this is the new arch and should have had some juice right?

londiste
It could very easily beat a fully fledged desktop processor. Remember that we are talking about very low power limits for desktop CPUs here. Mobile cores might do OK but I suspect these would post lower multi-core scores due to power limits. Zen2 in form of Matisse cannot compete at all. Renoir should be able to compete but we know next to nothing about Renoir so I do not even have an estimation of how it compares.
It could, it would, it may have been. I'm talking about the number we see now. Sure there will be more benchmarks in the future when the cpu is out but this bench does say something. I understand we are talking about very low power limits. I'm sure Intel has a plan with all this Big.Little although I can't see it as of now and it would seem like it has given up on desktops and try it's luck with mobile where it still stands a fighting chance? Maybe things will change when this one is out as a final product.
Posted on Reply
#31
londiste
The number we see now is not logical. We know how these cores should perform and this is not at that level. Hell, last-gen Atoms score at over 1000p in single-core at <2GHz.
Given this is a leak, not a release CPU or system, it is quite likely that things are not going right.

If I had to guess, this single-core score is from one of the Atom cores.
Posted on Reply
#32
notb
ratirt
What you comparing here is Desktop APU with something designed for mobile.
I'm not sure what you mean. I haven't mentioned a single desktop chip in that post.
The ARM is beating it in single core by a double
?
No, it's not.
and yet you think this is going to beat fully fledged desktop processor?
I never said that either.
This Intel's CPU is going to be used in the mobile but I doubt Laptops will do much outta it. Most likely tablets (mentioned by someone before).
I'm not sure how you draw the line between laptops and tablets. Keyboard? Processing power?

The idea behind hybrid architecture is extremely simple.
If you're buying a high-end desktop CPU, you're probably getting the fastest single-thread performance an architecture offers. That's because high-end desktop users aren't really concerned with power consumption. So even if you're doing single-thread tasks most of the day (which is true for most people), you're not harmed in any way. You've paid for 8 cores, you use 1. It's still fast. You get fast browsing, fast software, responsive OS. Great.

If you're buying pretty much any other consumer PC, there's always a single-thread performance compromise. It doesn't matter if it's 35W or 15W or 6W.
A single core CPU makes no sense because of platform stability. You get at least 2. They have to split the die size and power budget.
On the other hand, you end up with large cores anyway. So even in idle, they'll draw more power than is needed.

The hybrid architecture fixes everything.
You get a single massive core that can use most of the power budget - so single-thread tasks run almost as fast as on a single-core chip OR a much larger chip with faster cores.
And you get a smaller core that is extremely frugal, yet powerful enough to support PC's basic functions.

For now we're only see these hybrid SoCs in mobile devices, because that's the priority market (and the implication is more substantial).
But there's no reason why they would not be used in desktops or even small servers (NAS etc) in the future.
A homogeneous architecture just isn't the optimal solution. It never was.
Posted on Reply
#33
ratirt
notb
?
No, it's not.
No according to OP.
notb
'm not sure how you draw the line between laptops and tablets. Keyboard? Processing power?

The idea behind hybrid architecture is extremely simple.
If you're buying a high-end desktop CPU, you're probably getting the fastest single-thread performance an architecture offers. That's because high-end desktop users aren't really concerned with power consumption. So even if you're doing single-thread tasks most of the day (which is true for most people), you're not harmed in any way. You've paid for 8 cores, you use 1. It's still fast. You get fast browsing, fast software, responsive OS. Great.

If you're buying pretty much any other consumer PC, there's always a single-thread performance compromise. It doesn't matter if it's 35W or 15W or 6W.
A single core CPU makes no sense because of platform stability. You get at least 2. They have to split the die size and power budget.
On the other hand, you end up with large cores anyway. So even in idle, they'll draw more power than is needed.

The hybrid architecture fixes everything.
You get a single massive core that can use most of the power budget - so single-thread tasks run almost as fast as on a single-core chip OR a much larger chip with faster cores.
And you get a smaller core that is extremely frugal, yet powerful enough to support PC's basic functions.

For now we're only see these hybrid SoCs in mobile devices, because that's the priority market (and the implication is more substantial).
But there's no reason why they would not be used in desktops or even small servers (NAS etc) in the future.
A homogeneous architecture just isn't the optimal solution. It never was.
There is variety of laptops but if you focus only on the low power, yeah, maybe but can't say the performance is outstanding in comparison to others' companies offerings. It is OK at most.
Posted on Reply
#34
notb
ratirt
There is variety of laptops but if you focus only on the low power, yeah, maybe but can't say the performance is outstanding in comparison to others' companies offerings. It is OK at most.
Well yes, I focus on the low power segment, because it's what Lakefield targets. Why would I compare it to other PCs? Or shoes?
:D

From the OP's text:
"if compared to some offerings from Snapdragon, like the Snapdragon 835, it offers double the single-threaded performance with a similar multi-core score. If this is meant to compete with the more powerful Snapdragon offerings like the 8cx model, comparing the two results in Intel's fail. While the two have similar single-core performance, the Snapdragon 8cx leads by as much as 76.9% in a multi-core scenario, giving this chip a heavy blow. "

This is EXACTLY what I'm saying. Different, much more useful scaling between single and multi thread. You get twice as much performance in single-thread.
And it's x86, so it can run way more software.

Yes, i5-L15G7 is not as fast as the halo Snapdragon. But you can't dis a product just because it isn't the best in the world.
i5 is the mainstream, money-making range of Intel's product.
Posted on Reply
#35
ratirt
notb
Well yes, I focus on the low power segment, because it's what Lakefield targets. Why would I compare it to other PCs? Or shoes?
:D

From the OP's text:
"if compared to some offerings from Snapdragon, like the Snapdragon 835, it offers double the single-threaded performance with a similar multi-core score. If this is meant to compete with the more powerful Snapdragon offerings like the 8cx model, comparing the two results in Intel's fail. While the two have similar single-core performance, the Snapdragon 8cx leads by as much as 76.9% in a multi-core scenario, giving this chip a heavy blow. "

This is EXACTLY what I'm saying. Different, much more useful scaling between single and multi thread. You get twice as much performance in single-thread.
And it's x86, so it can run way more software.

Yes, i5-L15G7 is not as fast as the halo Snapdragon. But you can't dis a product just because it isn't the best in the world.
i5 is the mainstream, money-making range of Intel's product.
Still the competition is tough so hope this one can somehow pull this off. In the other terms it is maybe an entry level of a product. Not saying it is bad just have the impression that it is outstanding which so far, in my eyes it is not spectacular. Besides the 8cx Snapdragon consumes 7watts of power. If power is your most concern and you want to compare these only with power to performance, the snapdragon is better. I'm not convinced that I5's are money making in this low power segment.
Posted on Reply
#36
londiste
Snapdragon 8cx is also 122 mm², 50% bigger. Closer to Renoir than this thing.
Posted on Reply
#37
notb
ratirt
Still the competition is tough so hope this one can somehow pull this off. In the other terms it is maybe an entry level of a product. Not saying it is bad just have the impression that it is outstanding which so far, in my eyes it is not spectacular.
Well, this comparison against ARM doesn't make as much sense and some think.
ARM and x86 are different platforms.
Comparing to high-end Snapdragons is fine for products that could use either.

Keep in mind that even today, with ARM chips being so obviously efficient and fast (Snapdragon 835 is almost 3 years old), a lot of devices stick to cheap x86 parts (e.g. NAS and cheap laptops - even Chromebooks!). It's a much more robust and easier to use platform.
Since most products will keep using x86 hearts, we should compare to the x86 chips that are used today. That's why I'm mentioning <10W Celerons all the time, not high-end Snapdragons (even if they cost roughly the same).
And the improvement here would be enormous.
Besides the 8cx Snapdragon consumes 7watts of power. If power is your most concern and you want to compare these only with power to performance, the snapdragon is better.
My main concern is single-thread performance in this power envelop. It's really game changing.
Because pushing fast low-power x86 CPUs would be a significant achievement. Intel tried a few times with mixed success. Lakefield is the most extreme and interesting approach to date.
Posted on Reply
#38
ratirt
londiste
Snapdragon 8cx is also 122 mm², 50% bigger. Closer to Renoir than this thing.
OK but it still consumes 7 watts of power and it's faster. That was the main idea and I don't think, if the processor is "bigger" means it is bad or less attractive. Not sure where you are going with this but if you can get a bigger chip, faster than any smaller and slower equivalent when both consume same 7 watt of power. Imagine if Intel's 9900 could have been as large as is now but consuming 10 watts and be twice as fast. Would that be a bad thing due it is bigger?

londiste
Snapdragon 8cx is also 122 mm², 50% bigger. Closer to Renoir than this thing.
Why? These are both processors even though they are different they are still doing the same thing (or consumers want them to do the same thing) when you say the comparison doesn't make much sense is because these are different or because Intel lacks a bit?
Imagine this. It is 3 years old CPU and still give Intel's brand new chip stimulus. Isn't that something worth mentioning? Intel is trying doing something and kudos for that but it would seem, Intel still has to catch up a bit or improve.

notb
My main concern is single-thread performance in this power envelop. It's really game changing.
Because pushing fast low-power x86 CPUs would be a significant achievement. Intel tried a few times with mixed success. Lakefield is the most extreme and interesting approach to date.
Sure it is but it has to be fast enough. You think this one is? I really wanna share your enthusiasm but I'd rather wait.
Posted on Reply
#39
Berfs1
Virtual machine... could be faster...
Posted on Reply
#40
Caring1
Mats
Here you go.



Very funny, but 12 X 12 =144 is not the same as 12 X 12 = 144mm²
Remember the squared bit? It's not hard.
Posted on Reply
#41
Mats
Caring1
Very funny, but 12 X 12 =144 is not the same as 12 X 12 = 144mm²
Remember the squared bit? It's not hard.
12 x 12 = 12² = 144
Or what are you talking about? :D

Edit: LMGTFY
Posted on Reply
#42
Caring1
Mats
12 x 12 = 12² = 144
Or what are you talking about? :D

Edit: LMGTFY

Congrats, you finally caught on.
Posted on Reply
#43
Mats
Caring1
Congrats, you finally caught on.
Nice trolling, you were wrong from the beginning. ;)
Posted on Reply
#44
Caring1
Mats
Nice trolling, you were wrong from the beginning. ;)
No, I was correct, you just confirmed it. :cool:
Posted on Reply
#45
Mats
Caring1
12mm X 12mm =144 mm² now?
My math puts it at 12mm²
Nope, this is still so wrong.
Posted on Reply
#46
Caring1
Mats
Nope, this is still so wrong.
OMG, do you even math?
You've proved me right, and you agree with my math, then tell me i'm wrong lol.
Posted on Reply
#47
Mats
Yup, that makes of of us. :)

I'm seriously curious how you're thinking.

Caring1
You've proved me right, and you agree with my math, then tell me i'm wrong lol.
You were the one who didn't agree that 12 mm x 12 mm = 144 mm².
I've never changed my point of view.
Posted on Reply
Add your own comment