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

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

LikeLinus

Lifer
Jul 25, 2001
11,518
670
126
All the time I mentioned TDP, power consumption, ... not performance. But they some how are related:
CPU
TDP
Power consuption peak
I9-9900K​
95W​
210W​
I9-10900K​
125W​
250W​
I7-1065G7​
15W​
45W​
i5-1035G4​
15W​
45W​
R9-3950X​
105W​
?​
R5-3550H​
35W​
35W​
R7-3780U​
15W​
?​

Feel free to say what is the value at ?
AMD with renoir could say it's 15W TDP and feed the 8 core monster with 45W and I want to see if you will complain or not.
Last time I'm responding to you because you apparently do not understand the thermal dynamics. You're only providing an absolute peak consumption number. Which is just one data point in these test. No one sits around running their chips at 45w turbo boosted power at all times. The chips and OS themselves dynamically adjust the power consumption based on current usage. That peak is only used on-demand. Why aren't you posting numbers where those same chips take under 1w while idle? Or even paying attention to the test that during web and simple usage, that the CPUs battery life are vastly different. It's a known fact and many others in this thread have pointed this out, some CPUs just aren't as power efficient, despite using the same power consumption range. There are a number of factors for this, but it's completely lost on you. You're even posting desktop chips? This thread is about the Surface Laptops. Get over it.

Enjoy your day.

Not responding to any of this off-topic stuff anymore. It's a waste of everyone's time and not productive.
 

RetroZombie

Senior member
Nov 5, 2019
464
386
96
You're even posting desktop chips? This thread is about the Surface Laptops. Get over it.
I had too, you are the one going off topic all the time, I had to relate the more known desktop behavior that most people is aware to the laptop that you are not aware and because intel start using it some time ago, I posted a link you didn't read it did you?
You seam to devalue absolute peak consumption , if you where building an itx with some of the intel stuff would you use some 180W psu knowing the intel power numbers don't comply? That's how important peak numbers are.

Why aren't you posting numbers where those same chips take under 1w while idle?
Those numbers are very important for the people that use their laptop like a tablet, doing nothing, all the time standing still, which is good and important, very important! You have those in the anandtech review.
But where are the battery life numbers with the laptop doing work like some people use them, for working, productivity, maybe even gaming?
 

krumme

Diamond Member
Oct 9, 2009
5,952
1,585
136
No, I prefer Renoir to Picasso. If AMD has to keep taking wafers from GF (which I doubt), then of course they can use Picasso to unseat Bristol Ridge. But Renoir should be the product in the Microsoft Surface 3+ or Surface 4 or whatever it is that MS releases next. Not Picasso.

My expectation is that AMD will move to Renoir/Dali and cover their entire lineup with that. Van Gogh for a very specialized market segment. Picasso, Bristol Ridge, and Stoney Ridge should all be tossed at once (and my expectation is that Bristol/Stoney are already gone).

Its unfortunately not about our personal preferences for Ferrari to a Fiat but what other people buy. Its not about the most beautiful girl in the bar, but who you are able to take home and tell stories.
The wsa is terminated in 2024.
The current agreement runs until 2020 and will be renogatiated then and presumably cover the last part until 2024, - if its not already set in stone.

Surely renoir is the apu for Surface 3+ and is very much needed, but that segment is only a minor part of the market. Unless some miracle happens at GF, Picasso will even outlive Renoir.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
All the time I mentioned TDP, power consumption, ... not performance. But they some how are related:
CPU
TDP
Power consuption peak
I9-9900K​
95W​
210W​
I9-10900K​
125W​
250W​
I7-1065G7​
15W​
45W​
i5-1035G4​
15W​
45W​
R9-3950X​
105W​
?​
R5-3550H​
35W​
35W​
R7-3780U​
15W​
?​

While the desktop Core chips can run PL2 values indefinitely as long as you have it set on the BIOS and with a good cooling solution, laptops only reach those figures in AC mode and only for a short time, because cooling and current requirements.

XPS 13 2-in-1 is a prime example of that as its the fastest when on AC mode and in the beginning, but falls behind Microsoft Surface Laptop 3 and Lenovo C940(in intelligent cooling mode).

On NBC reviews they talk about whether performance is limited or not in battery mode. Also shows sustained performance levels on Cinebench R15, which is very useful.

The XPS 13 isn't particularly a good implementation as it also has DPC latency issues, but its prioritizing short term boost performance, while the SL3 is for sustained and performance under battery. SL3 also beats the XPS quite easily in gaming, but as resolutions and settings get lower it starts having advantage. So XPS 13 might also be putting more power to use on the CPU than its on the SL3.
 
Last edited:

DrMrLordX

Lifer
Apr 27, 2000
21,610
10,804
136
I found this stating Raven Ridge based:
AMD's Dali APU Could Be Based on Raven Ridge

If they want GF 12nm keep going on, not surprised.


No reason why that has to be on 12nm . . .

Its unfortunately not about our personal preferences for Ferrari to a Fiat

Renoir is not a Ferrari. Dali is definitely not a Ferrari. Bad analogy. Dali, in particular, would be tiny and cheap, if it is in fact Raven Ridge shrunk to 7nm.

Just because the WSA extends to 2024 doesn't mean that AMD has spare wafers to burn and that they want to continue to poison their lineup with old products. Now I'll admit, AMD doesn't take the mobile market as seriously as server, workstation, or desktop; but they still have a brand image to manage here. Selling 2018's product in 2020 does not help them at all. It also gives Intel a pass in what is currently Intel's strongest market segment.
 

krumme

Diamond Member
Oct 9, 2009
5,952
1,585
136

No reason why that has to be on 12nm . . .



Renoir is not a Ferrari. Dali is definitely not a Ferrari. Bad analogy. Dali, in particular, would be tiny and cheap, if it is in fact Raven Ridge shrunk to 7nm.

Just because the WSA extends to 2024 doesn't mean that AMD has spare wafers to burn and that they want to continue to poison their lineup with old products. Now I'll admit, AMD doesn't take the mobile market as seriously as server, workstation, or desktop; but they still have a brand image to manage here. Selling 2018's product in 2020 does not help them at all. It also gives Intel a pass in what is currently Intel's strongest market segment.
Picasso is not an old product, and who cares if its old. My kid can play overwatch at 1080 low 75% res at more or less consistently 60 fps, while i get half fps on my quad kbl.
I dont know where you guys get the idea its an absolete product. Its performing perfectly fine as a low end product and is lightyears above any bd derivative and even above the old quad skl++ products in real life consumer usage.
I can only think you get the impression from reading stupid reviews like that surface review on AT. That review is imo some biased stuff intended to please partners. What is the next skewed review, comparing some top end renoir to a old kbl product? Meaningless.
I am pretty sure amd knows what is best for themselves.
I have owned litteraly every Intel mobile cpu sans broadwell since core duo in all the expensive lenovo and dell stuff, and the 3500u is a huge, huge step forward for the low end.
I will surely get a lenovo renoir this spring, but it doestnt take away what a leap this cheap apu is. Surely its dependant on forced wsa agreement that skews the entire picture. But i dont care. Consumers can now get cheap -and very balanced- apu with loads of gpu and cpu power.
 

DrMrLordX

Lifer
Apr 27, 2000
21,610
10,804
136
I dont know where you guys get the idea its an absolete product.

How could it not be obsolete? Renoir is about to obliterate it completely in every metric. It's like arguing that AMD should still be making Pinnacle Ridge products for some market segment (other than as replacements for long-term support clients who bought Pro products). Sure, it works, but why? There's absolutely no reason for AMD to continue producing or selling the CPU when they have a full raft of products produced at TSMC to fill their lineup from top to bottom.

AMD has cut off Pinnacle Ridge completely. They aren't using it anymore. Picasso is based on the same core family (Zen+). AMD needs to put it out to pasture as well, especially in its 4c form as we saw in the Surface 3. I'm also assuming that AMD will have a 7nm Ryzen 400u sooner or later, which would knock off the lowest-end Picasso. There's simply no reason to keep producing the things. Definitely not in a product like the Surface where it will be compared unfavorably to an Intel mobile CPU sold in the same chassis under the same product family.
 

Gideon

Golden Member
Nov 27, 2007
1,619
3,643
136
I doubt Picasso is going anywhere soon, in the ultra-low-end craptops.

We are still getting Bulldozer-era chromebooks, desipte 2-core Raven Ridge die being out for a while (and should be dirt-cheap). 7nm will be expensive for a while, just look at AMD's low-end GPU pricing), and even when we get a cut-down Renoir die, it will still be noticeably more expensive to produce than GF's 12nm or 12nm+ (AMD might refresh these chips, if it requires no new masks, etc).

I agree that Picasso should have no place in > $500 laptops, but below it actually makes sense. I'd much rather have a 3500U with a bit better chassis/display/SSD and at least 8GB of dual-channel memory than the cheapest Renoir coupled to total junk.
 
  • Like
Reactions: RetroZombie

DrMrLordX

Lifer
Apr 27, 2000
21,610
10,804
136
We are still getting Bulldozer-era chromebooks

28nm is cheap. Cheaper than 12nm or 12nm+. Probably the only reason why those keep getting made. Or it might be zombie overstock of the chips.

desipte 2-core Raven Ridge die being out for a while

They weren't over-produced, and AMD stopped making them. Which is what I think they should do with Picasso. That's what I expect will happen.
 

krumme

Diamond Member
Oct 9, 2009
5,952
1,585
136
28nm is cheap. Cheaper than 12nm or 12nm+. Probably the only reason why those keep getting made. Or it might be zombie overstock of the chips.



They weren't over-produced, and AMD stopped making them. Which is what I think they should do with Picasso. That's what I expect will happen.
We dont know the details of the wsa, so we dont know the economic consequence of keeping it on market.

But what options are there for amd? How should they fullfill the wsa? Is io dies enough?
I doubt it and besides i can also see its important for amd to keep gf strong. And of all their product on 14/12nm picasso is the most compettitive and balanced imo.
It also makes sense to make a 8c renoir and sell some as 6c. The rest is fine for 4c picasso where yield will be fine for more or less 100% 4c at that time. Seems like fine segmentation to me. There is actually a real product difference then unlike 3500u vs 3700u where the 3500u eg have a battery advantage and performs more or less the same. Or some old kbl i7 i5 where you couldnt tell the difference.
 

moinmoin

Diamond Member
Jun 1, 2017
4,944
7,656
136
How could it not be obsolete?
Your are going to give the OEM/ODM industry with their super long TTM a hearth attack. :tearsofjoy:

I don't think anybody is disagreeing with you on a technical level. It's just that big parts of the industry moves slow as molasses. E.g. embedded products having a guaranteed availability of 10 years should tell you all you need to know.
 

RetroZombie

Senior member
Nov 5, 2019
464
386
96
On NBC reviews they talk about whether performance is limited or not in battery mode. Also shows sustained performance levels on Cinebench R15, which is very useful.
Very cool, didn't know that, thanks. Just curious do you know anyone that test battery life on some scenario like gaming for example, with game x with settings x will gives x battery play time? I know it's not an very easy test to do but maybe someone does it.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Very cool, didn't know that, thanks. Just curious do you know anyone that test battery life on some scenario like gaming for example, with game x with settings x will gives x battery play time? I know it's not an very easy test to do but maybe someone does it.

Yes, NBC tests it with Witcher 3 sometimes. But not always. Their load tests are roughly representative of a gaming scenario.

It's kinda meaningless to test when the U laptops will last 2 hours at best. You have just the CPU using 15-20W, and rest of the platform to consider, plus the display. With a typical 50WHr battery its simple arithmetic. -Y systems are between 20-25W, and the -U systems are 25-35W.

Actually, the 13-inch Y systems can go as low as 10W. Leather-bound HP Spectre Folio uses Amberlake-Y to do just that. Probably the most efficient 13-incher out there. The Swift 7 is pretty good there too. Better efficiency under load and uses slightly over 10W.

Gaming is one of the worst scenarios for battery life. There's no way to power down in between. dGPUs are even worse. Not only you lose 30-50% of the performance, you'll get about 50 mins battery when gaming even with a large 80WHr battery. Of course the battery can only supply so much current, so the system has to throttle, because at peak it might briefly reach 150W or more.

Well, one way to make it last longer is if you cap the performance, so 60 fps rather than 100, 30 fps rather than 60.
 
Last edited:
  • Like
Reactions: RetroZombie

Gideon

Golden Member
Nov 27, 2007
1,619
3,643
136
An interesting take on the Surface laptops from notebookcheck:

The really interesting chart is the last one, Battery life:
Battery RuntimeMicrosoft Surface Laptop 3 15 i7-1065G7
1065G7, Iris Plus Graphics G7 (Ice Lake 64 EU), 45 Wh
Microsoft Surface Laptop 3 15 Ryzen 5 3580U
3580U, Vega 9, 45 Wh
Reader / Idle21001121
WiFi v1.3490546
Load112102

The Idle/reader difference is huge, nearly 2x in Intel's favor, but AMD's version actually wins in the WiFi test, and by an hour!

Initially I thought that their v1.3 WiFi test doesn't idle that much, at least not nearly as much as say Anandtech's web test

But looking at their test criteria link that looks to not be the case:
Wi-Fi mode: the possible battery life while surfing the Internet via Wi-Fi with medium brightness (~150 cd/m²) and power-saving options ("balanced" mode) switched on. We measure the runtime by letting the device run an automatic script (HTML 5, JavaScript, no Flash - update 03.05.2015 v1.3), which picks a mix of websites and switches between them every 30 seconds.
Maximum runtime: the "Reader's" test of the Battery Eater tool is used to measure the maximum runtime of the test model. The brightness is set to minimum and all power-saving options are turned on. The Windows power plan is set to "Power Saver" and WLAN and Bluetooth are switched off.
The individual (more in depth) reviews seem to confirm this as well
I really hope they dig into this further, as Anandtech's result is quite different.

And anyway, can't wait to see the Ryzen 4xxxU series Surfaces. I really do hope they were smart enough to design the chassis similarily in advance for it to be a drop-in replacement (therefore coming soon, rather than Q4)
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
The Idle/reader difference is huge, nearly 2x in Intel's favor, but AMD's version actually wins in the WiFi test, and by an hour!

I want to play with the latest Ryzen laptop or the Intel laptop. My XPS 12 is too old. Also needs a new battery and keyboard.

There's a huge amount of power-based customizations that can be done at least for the Intel side. But there may be some to do on the AMD side as well. Even new, out of the box configurations aren't the most optimal.

By the way, that SL3 Intel is actually quite well balanced. The gaming performance is comfortably on the high end for the Iris Plus graphics. Also it seems MS optimized it for higher settings and resolutions as the advantage over other Iris competition and Picasso increases. It doesn't throttle the CPU or the GPU under battery either.

The Dell XPS 13 is not a particularly well optimized one. It has DPC latency issues which will cause problems with audio. If its too high even talking over skype or discord can cause skipping.

Based on the sustained Cinebench numbers, the XPS 13 outperforms the rest greatly when on AC and for a short time. But sustained and on battery, it underperforms a lot of them. The SL3 is very good on that regard, and does it without going above 25W PL1 even on AC.
 

krumme

Diamond Member
Oct 9, 2009
5,952
1,585
136
An interesting take on the Surface laptops from notebookcheck:

The really interesting chart is the last one, Battery life:
Battery RuntimeMicrosoft Surface Laptop 3 15 i7-1065G7
1065G7, Iris Plus Graphics G7 (Ice Lake 64 EU), 45 Wh
Microsoft Surface Laptop 3 15 Ryzen 5 3580U
3580U, Vega 9, 45 Wh
Reader / Idle21001121
WiFi v1.3490546
Load112102

The Idle/reader difference is huge, nearly 2x in Intel's favor, but AMD's version actually wins in the WiFi test, and by an hour!

Initially I thought that their v1.3 WiFi test doesn't idle that much, at least not nearly as much as say Anandtech's web test

But looking at their test criteria link that looks to not be the case:

The individual (more in depth) reviews seem to confirm this as well
I really hope they dig into this further, as Anandtech's result is quite different.

And anyway, can't wait to see the Ryzen 4xxxU series Surfaces. I really do hope they were smart enough to design the chassis similarily in advance for it to be a drop-in replacement (therefore coming soon, rather than Q4)

I have been trying for ages to get that message into your heads :)
For normal office use or just surfing picasso is fine. What most use a laptop for.
There is not such a thing as total idle in typical normalnusage. Its small usage all the time. Typically 1 to 10 percent load or so.

Note though:
1.
The ryzen uses 8gb vs 16g for the Intel. I presume that means 1 stick vs 2 (?) so explains some of the ryzen advantage for battery. And btw also explains why its mostly slower in games.
2.
The picasso is a ryzen 5. For some reasons as shown by notebookcheck earlier that processor is a good deal more efficient than r7. By all accounts never buy a 3700u at extra cost.
3.
I dont know; Icelake migt go into deep low hipernation bear winter mode, but 10nm is just junk it seems. Junk. No need to frame it otherwise. Imo just get a quad kbl derivative. Just as good bar some few differences. Heck a 3500u is more or less the same.
 

uzzi38

Platinum Member
Oct 16, 2019
2,608
5,824
146
Note though:
1.
The ryzen uses 8gb vs 16g for the Intel. I presume that means 1 stick vs 2 (?) so explains some of the ryzen advantage for battery. And btw also explains why its mostly slower in games.
Soldered kinda-dual-channel for both. Intel gets quad-channel LPDDR4X (each channel is half the width).

That gives the Intel system the advantage, not the Ryzen one.
 

LikeLinus

Lifer
Jul 25, 2001
11,518
670
126
An interesting take on the Surface laptops from notebookcheck

Not really. Two different range of CPU (lower end AMD vs. higher Intel). They also use two different WIFI chips. The Intel uses the latest and fastest chip. Also amount of RAM. Intel is using more which could account for more power usage if the browser/OS is using it for cache and such.

Comparing apples to oranges.