Review 'The Microsoft Surface Laptop 3 Showdown: Amd Picasso vs Intel Ice Lake' - Anandtech

[UsandThem]

https://www.anandtech.com/show/1521...ptop-3-showdown-amd-picasso-vs-intel-ice-lake

AMD has made great improvements with their desktop Ryzen CPUs, and can go toe-to-toe with Intel's best.

However, their mobile Ryzen 'Picasso' CPUs have a ways to go. They are much closer in terms of performance of course, but they will have to do better to gain a significant share in that market segment. Honestly, since the specs of these two test laptops are almost identical, there really isn't a good reason to go with the AMD-based one unless it is significantly cheaper than the Intel model (IMO definitely not worth it when priced only $100 cheaper). The most striking difference (at least to me) was the difference in battery life.

Spoiler

 

[Mopetar]

Makes sense to see something like that considering the differences in idle power draw that AT shows in their review:

Spoiler

But Picasso is Zen+ on GF 12nm so it's not surprising that Intel's newest chips on 10nm would perform better. I think the bigger surprise is the graphics performance where Intel is competitive in almost all cases and even pulls ahead of Vega in a few of them, though it's hard to draw too many conclusions since some of that may come down to CPU differences.

But as badly as Intel is constrained and can't possibly get as many wafers as they want, AMD is even worse off since they've got to get all of their products (outside of their APUs and the IO dies for Zen 2) from TSMC. Even if they could be more competitive in this space, I'm not sure it's worth using wafers to make APUs when they could be selling more Epyc chips or Navi GPUs.

 

[Markfw]

Only one reason to get the Ryzen surface 3, if you are a gamer, the GPU is better. Hopefully, the laptop version of Zen2 will be out soon, then we might have a real competition.

 

[RetroZombie]

I don't understand a few things:

- So the Intel chip at WALL POWERED is using up to 44 Watts and 'goes down' to 25 Watts when it gets 'hot', so all the BENCHMARKS results are a little flawed since they are against the amd part with only 15 Watts?

- The BATTERY tests are done in a scenario where the Intel cpu when doing nothing uses less power and is paired with a special memory that only uses 0.6V, so obviously will last longer when doing 'standing still web browsing'?

- Based on the previously said, wouldn't make some logic to also do the benchmark testing on battery powered, and also get some battery life results from those tests since the performance and battery life is not the same with the Intel chip since supposedly it will be limited to 15 Watts instead of 44 Watts, and if it uses 44 Watts when battery powered the battery will run out much faster?

Many AVX2 favored tests also didn't help...

 

[Zucker2k]

Based on the previously said, wouldn't make some logic to also do the benchmark testing on battery powered, and also get some battery life results from those tests since the performance and battery life is not the same with the Intel chip since supposedly it will be limited to 15 Watts instead of 44 Watts, and if it uses 44 Watts when battery powered the battery will run out much faster?

The Intel power numbers are for the SoC package, but the AMD power numbers appear to be just the CPU cores, which is an unfortunate byproduct of testing two different platforms.

With all cores loaded the graph is considerably altered. Here the AMD processor is able to maintain a much higher frequency across its cores for much longer, while Intel's chip is only able to maintain 3.5 GHz for about 30 seconds before it runs out of headroom, dropping the cores down to around 2.6 GHz. But despite the lower frequency, the much higher IPC on Sunny Cove allows the Ice Lake platform to finish quite a bit sooner.

Makes sense to see something like that considering the differences in idle power draw that AT shows in their review:

Are battery life tests not the same as running continuous tests on battery until the system shuts down? What has idle power got to do with it? Besides, even if idle consumption is an advantage for Ice Lake, isn't that part of the package? The bolded quote above shows the AMD chip is aggressively maintaining boost clocks at 4GHz. Maybe that's where the energy went?

 

[arandomguy]

This is something I've wondered about Zen APUs going forward given the chiplet design especially with the upcoming Ryzen 4000 series is I wonder how much of a split in design there will be. The current chiplet setup while advantageous for Server, HEDT, and by extension enthusiast desktop market does in theory have certain drawbacks that would need to be overcome for mobile.

Both power draw and package size due are inherent weaknesses relative to a small single monolithic design. With Ryzen 2000 AMD has shown they are willing to design a separate Die for mobile inlcuding having the CPU design having differentiation. So I wonder if they will further split how APU and CPU lines are designed. What would be interesting with the above is if they did not use the large 12NM Die (due to package size and power concerns). If that were to by extention bring benefits such as lowered memory latency (eg. I/O moves back to on die) could that result in Ryzen 4000 APUs being faster (at least on a per clock/per thread basis) vs Ryzen 3000 CPUs (at least for certain tasks like gaming)?

Are battery life tests not the same as running continuous tests on battery until the system shuts down? What has idle power got to do with it? Besides, even if idle consumption is an advantage for Ice Lake, isn't that part of the package? The bolded quote above shows the AMD chip is aggressively maintaining boost clocks at 4GHz. Maybe that's where the energy went?

Typical mobile usage is very burst in nature, therefore how well the hardware can transition between lower and higher power states, and how low those states can go is very important to actual real battery life as well as user experience. The ARM based mobile SoCs are even better here due to the highly mobile oriented nature of their development.

A "representative" type test to mimic usage would therefore also test how well the system handles the above, which would mean it would cause power state changes (including going down to idle). For example in this case the test was "labled" web browsing, if you think about typical web browing it will be long periods of low power states (nncluding idle) while oyu are taking in the content punctuated by bursts to bring up the content.

Given the nature of the product type in this discussion the above is much more relevant than measuring full load battery life. The usage case and most important consideration isn't likely tasks such as gaming away from the wall, encoding/rendering to failure and etc.

A more applicable sustained usage case would be something like video playback battery life which is also a test often done for mobile devices but not in this particular article.

 

[Arkaign]

It's not terrible, but for the $ charged, it makes no sense to me in the context of the SP.

Branding-wise, I *hate* that they call non-Zen2 SKUs 3000 series such as the 3200G and this mobile series. It makes things murky and is a bite off Intel's playbook. Should have stuck with :

All 1000 series = Zen derived
All 2000 series = Zen+ derived
All 3000 series = Zen2 derived

Beh. Agree with Mark, things will get interesting with 7nm Zen2 mobile products. They're really giving Intel some breathing room here (in mobile) with this half baked last gen offering.

 

[Zucker2k]

For example in this case the test was "labled" web browsing, if you think about typical web browing it will be long periods of low power states (nncluding idle) while oyu are taking in the content punctuated by bursts to bring up the content.

Are the web browsing tests not scripted, though? Otherwise there'll likely be discrepancies in the tests.

 

[DrMrLordX]

However, their mobile Ryzen 'Picasso' CPUs have a ways to go.

No surprises here. AMD doesn't prioritize APU development anymore. They haven't since 2016. Intel will continue to have the best mobile offerings once TigerLake comes out, which hopefully will launch in greater volume and at higher clocks than IceLake. AMD's big chance to pick up sales will come from 14nm shortages and poor 10nm yields at Intel's fabs.

 

[exquisitechar]

Hard to care about either Ice Lake or Picasso when Renoir’s coming very soon.

This is something I've wondered about Zen APUs going forward given the chiplet design especially with the upcoming Ryzen 4000 series is I wonder how much of a split in design there will be. The current chiplet setup while advantageous for Server, HEDT, and by extension enthusiast desktop market does in theory have certain drawbacks that would need to be overcome for mobile.

Both power draw and package size due are inherent weaknesses relative to a small single monolithic design. With Ryzen 2000 AMD has shown they are willing to design a separate Die for mobile inlcuding having the CPU design having differentiation. So I wonder if they will further split how APU and CPU lines are designed. What would be interesting with the above is if they did not use the large 12NM Die (due to package size and power concerns). If that were to by extention bring benefits such as lowered memory latency (eg. I/O moves back to on die) could that result in Ryzen 4000 APUs being faster (at least on a per clock/per thread basis) vs Ryzen 3000 CPUs (at least for certain tasks like gaming)?



Typical mobile usage is very burst in nature, therefore how well the hardware can transition between lower and higher power states, and how low those states can go is very important to actual real battery life as well as user experience. The ARM based mobile SoCs are even better here due to the highly mobile oriented nature of their development.

A "representative" type test to mimic usage would therefore also test how well the system handles the above, which would mean it would cause power state changes (including going down to idle). For example in this case the test was "labled" web browsing, if you think about typical web browing it will be long periods of low power states (nncluding idle) while oyu are taking in the content punctuated by bursts to bring up the content.

Given the nature of the product type in this discussion the above is much more relevant than measuring full load battery life. The usage case and most important consideration isn't likely tasks such as gaming away from the wall, encoding/rendering to failure and etc.

A more applicable sustained usage case would be something like video playback battery life which is also a test often done for mobile devices but not in this particular article.

Yeah, Renoir is monolithic and will likely have lower memory latency than Matisse. However, the L3 cache will likely be smaller too.

No surprises here. AMD doesn't prioritize APU development anymore. They haven't since 2016. Intel will continue to have the best mobile offerings once TigerLake comes out, which hopefully will launch in greater volume and at higher clocks than IceLake. AMD's big chance to pick up sales will come from 14nm shortages and poor 10nm yields at Intel's fabs.

Tiger will have better single/low threaded and iGPU performance, probably, but Renoir will still crush it in multithreaded performance when it’s only going to have 4 cores. In terms of volume, Intel’s 14nm CPUs will still make up the majority of it.

 

[DrMrLordX]

Tiger will have better single/low threaded and iGPU performance, probably, but Renoir will still crush it in multithreaded performance when it’s only going to have 4 cores.

It'll be an interesting matchup. There' s still a dearth of information on core counts for Renoir.

In terms of volume, Intel’s 14nm CPUs will still make up the majority of it.

Well yes and . . . yes, with caveats. 10nm will continue with yield problems that presumably keep TigerLake volumes low by historical standards. It should do better than Cannonlake or IceLake. But Intel's 14nm shortage is shaping up to be very interesting. Unless the Samsung rumours are true, there's a very real chance that Intel may be forced to reduce availability of comet/rocket mobile parts depending on how they prioritize wafer usage.

 

[arandomguy]

Are the web browsing tests not scripted, though? Otherwise there'll likely be discrepancies in the tests.

I'm not sure the exact criteria/methodology of Anandtech's specific test but typically these would be scripted but the scripts will attempt to simulate similar human behavior. I believe in AT's cause they use a script to load a series of offline replica's of major websites and then simulate reading through it (including scrolling) before moving onto the next one.

You would not just have a script that just continuously loads webpages one right after another with no wait inbetween as no one would actually use a device like this in that scenario. As such the CPUs (and other hardware) will transition between power states through the test.

 

[RetroZombie]

Are battery life tests not the same as running continuous tests on battery until the system shuts down? What has idle power got to do with it? Besides, even if idle consumption is an advantage for Ice Lake, isn't that part of the package? The bolded quote above shows the AMD chip is aggressively maintaining boost clocks at 4GHz. Maybe that's where the energy went?

Idle is an advantage for intel for sure, and for the 'idle tests' it really shows that. But my point was not that, is how some 45 Watts mobile cpu is benchmarked against one at 15 Watts and then the 45 Watt cpu does the battery tests at the 15 Watts, wouldn't that change all the performance results and also the battery life results, from techspot:

"So on a 15W laptop, that would typically be a boost up to 45W for a few seconds with higher clock speeds, then a drop straight down to 15W for the rest of the workload."

My doubts are that the battery will last that long if the cpu is allowed to go for 45 Watts, that will drain the battery in no time, 45 Watts is 3X higher than 15 Watts, that's a lot!

So my all point is, for battery test one suite is used, and for benchmarks another suite is used and then all the results are all screwed because the cpus behave differently in each of the suites scenarios!

 

[Thunder 57]

It'll be an interesting matchup. There' s still a dearth of information on core counts for Renoir...

I expect Renoir to be quad core. Until the CCX grows to 8 in Zen 3, I don't see AMD going past 4 cores. That said, Renoir should get a nice boost in performance from Zen 2 and LPDDRX4, both of which will also help with power consumption. That should firmly hand the GPU side to AMD, and come rather close CPU wise. I still expect Intel will have some advantage in battery life, if for no other reason than not having a separate Wifi controller.

 

[DrMrLordX]

I expect Renoir to be quad core. Until the CCX grows to 8 in Zen 3, I don't see AMD going past 4 cores. That said, Renoir should get a nice boost in performance from Zen 2 and LPDDRX4, both of which will also help with power consumption. That should firmly hand the GPU side to AMD, and come rather close CPU wise. I still expect Intel will have some advantage in battery life, if for no other reason than not having a separate Wifi controller.

That was my thinking. I don't automatically expect that Renoir will wind up winning MT workloads vs IceLake-U/Y.

 

[jpiniero]

I expect the Renior die to have 8 cores... if nothing else but for marketing reasons to push moar corez. What models have more than 4 is an open question.

 

[arandomguy]

Wasn't there rumors that Renoir would have 8 cores?

What would be wild is if they debut and move to the 8 core CCX with a unified 8/16MB L3 with Renoir. Didn't Raven Ridge have some Zen+ improvements even though it was officially classified as Zen?

 

[LikeLinus]

Idle is an advantage for intel for sure, and for the 'idle tests' it really shows that. But my point was not that, is how some 45 Watts mobile cpu is benchmarked against one at 15 Watts and then the 45 Watt cpu does the battery tests at the 15 Watts, wouldn't that change all the performance results and also the battery life results, from techspot:

The CPU is not 45w. It's a 25w CPU on 10nm. The Surface Laptop 3 uses a i7-1065G7.

I have no clue why, but you've linked an article with an i5 CPU that has nothing to do with the CPU in the Surface Laptop 3.

All of your questions make no sense because you're not even talking about the correct hardware.

"So the Intel chip at WALL POWERED is using up to 44 Watts and 'goes down' to 25 Watts when it gets 'hot', so all the BENCHMARKS results are a little flawed since they are against the amd part with only 15 Watts? "

The AMD on the Surface Pro 3 charges at virtual the same rate as the Intel. They were both in the 40-45w range on wall power during charging. Intel has lower power hardware and is better optimized for power usage. There are several reasons why the 25w cpu is more power efficient than the 15w AMD.

 

[DrMrLordX]

I expect the Renior die to have 8 cores... if nothing else but for marketing reasons to push moar corez. What models have more than 4 is an open question.

Matisse shares a chiplet design with Rome, but Renoir does not. AMD can select the number of cores per die to meet the needs of the market without having to share package topology with Rome. Personally I would like to see Renoir have up to 8 cores per die, but I do not automatically expect that.

 

[Thunder 57]

What would be wild is if they debut and move to the 8 core CCX with a unified 8/16MB L3 with Renoir. Didn't Raven Ridge have some Zen+ improvements even though it was officially classified as Zen?

Pretty sure that's accurate regarding RR. I don't see AMD surprising us with an 8 core CCX with Renoir though, mobile just hasn't been a priority for awhile. That would certainly shake things up if they did though. That would be a lot of Zen 3 features vs Ice Lake, and I imagine that would beat the snot out of ICL. I really do not expect that, though.

Then again, AMD probably wants to address the fact that they are limited to quad cores on mobile while Intel can offer more. I've heard rumors of a hex core Zen 2 laptop chip. Maybe it's Zen 2 and requires a dGPU, and that launches along with Renoir. Or maybe, and this would be interesting, they have a Zen 2 chiplet and some sort of Vega chiplet. They could launch that to cover the high end with 6-8 cores and an iGPU, and have Renoir for the quad core market. I do not expect that at all though. My best guess is Renoir is quad core, and if there is anything with more cores, it will be Matisse with a dGPU.

 

[jpiniero]

AMD has said no GPU chiplet with Renoir. It is almost certainly monolithic.

 

[Thunder 57]

AMD has said no GPU chiplet with Renoir. It is almost certainly monolithic.

I get that, and that is why I am saying it is probably a quad core only design. I was just speculating as to the rumors of higher core counts on laptops next year. We could see Matisse with a dGPU to fill that role (plausible) or Matisse with a GPU chiplet (laughably unlikely, but hey, people go on about SMT4 or FDSOI, so why not?). I think Renoir will be quad core and that is all. There is a spot for AMD to try to push high end gaming laptops though. For that, Matisse + mobile Vega may be a thing. Power usage would be terrible though.

 

[jpiniero]

I get that, and that is why I am saying it is probably a quad core only design.

Two CCX would be pretty small if you cut it to 4 MB L3 each. Even if you end up using 6+ only for H parts it might be worth it.

 

[RetroZombie]

The CPU is not 45w. It's a 25w CPU on 10nm. The Surface Laptop 3 uses a i7-1065G7.

Did you really interpreted what I said?
And its you who is doing mistakes not me, according to Intel it's an 15 Watts cpu not 25 Watts like you said

I have no clue why, but you've linked an article with an i5 CPU that has nothing to do with the CPU in the Surface Laptop 3.
All of your questions make no sense because you're not even talking about the correct hardware.

I'm talking about the correct hardware, Intel mobile cpus overall not this one specifically which only makes things even worst

The AMD on the Surface Pro 3 charges at virtual the same rate as the Intel.
They were both in the 40-45w range on wall power during charging. Intel has lower power hardware and is better optimized for power usage.

And yet from the review "We weren’t able to accurately measure the power draw for the AMD platform"

There are several reasons why the 25w cpu is more power efficient than the 15w AMD.

If the efficiency was measured with 45 Watts for the performance tests, and the battery efficiency with other group of tests sorry but I cant correlate with that, because the data that is given to me is crossfired

Go out into the comments section of the article, there are many 'feeling the review' mistakes, but can't point them out like I did.

 

[LikeLinus]

Did you really interpreted what I said?
And its you who is doing mistakes not me, according to Intel it's an 15 Watts cpu not 25 Watts like you said

I'm talking about the correct hardware, Intel mobile cpus overall not this one specifically which only makes things even worst


And yet from the review "We weren’t able to accurately measure the power draw for the AMD platform"


If the efficiency was measured with 45 Watts for the performance tests, and the battery efficiency with other group of tests sorry but I cant correlate with that, because the data that is given to me is crossfired

Go out into the comments section of the article, there are many 'feeling the review' mistakes, but can't point them out like I did.

It is 15w, but even the Intel link says "Configurable TDP-up 25W" and down to 12w. Depends on how it's configured.

"Intel’s chip here is allowed to draw up to 44W for short durations before it falls down to a sustained 25W. "

Intel Core i7-1065G7 Benchmarked: Ice Lake with Iris Plus Graphics

This has been a long time coming, as we're finally getting around to testing Intel's Ice Lake architecture. This is Intel's first real attempt at a 10nm...

www.techspot.com

Appears the Surface Laptop 3 Intel CPU can range from 15w to 25w depending on how it's configured or boosting the speed.

I do know that the AMD version's battery life is shorter than the Intel from personal experience. But the Intel has hardware advantages and lower power consumption.