Asus Zenbook S 16 laptop review - The first Copilot+ laptop with AMD Zen 5 inside a 1.3-cm-thick case

[Redaktion]

The Zenbook S 16 marks Asus' first Copilot+ laptop to be fitted with AMD's new Zen 5 mobile processors. Adding to this is a 1.3-cm-flat case as well as a brilliant 120-Hz OLED display. Has this already made the Snapdragon laptops obsolete after only one month?

https://www.notebookcheck.net/Asus-Zenbook-S-16-laptop-review-The-first-Copilot-laptop-with-AMD-Zen-5-inside-a-1-3-cm-thick-case.868219.0.html

[Nedim Tabakovic]

How is it possible that this performs worse in games then Z1 extreme?

[ArsLoginName]

Thank you for a great write-up and analysis of this brand new platform. All of the information is there and this does look to be a step forward for AMD in some capacities. However, there are a few tidbits one should consider when looking in comparison.

1 - Processor runs at 33/28 W as noted in article. When comparing this to the 7840U in the HP Elitebook 845 G10 in which the 8-core processor runs at 30/25 W, things don't look as rosy as one expects. In particular, the new 12-core Ryzen 9 HX 370 scores 2407 vs 1925 in CBR15. Thus the new R9 HX 370 consumes 10% more power and has 50% more cores for a 25% increase in score. Not the best. This is further backed up by the CBR23 scores of 16522 (Ryzen 9 HX 370) vs 12385 for the 7840U. Better with a 33% gain for the increase in cores and nominal 10% increase in power consumption.

2 - There are several older 8 core Ryzen 7X40/7X45/8X40/8X45 (e.g., ROG G14 2622 CBR15 = +10%, 17079 CBR23 = +3.4%) series laptops that outperform this new Asus S16 in terms of total performance albeit at a higher power consumption (65/45 W sustained) vs 33/28 W for the new HX 370. So almost the same performance at 62% the power consumption (almost 40% reduction in power consumption). Platform and laptop looks waaaay better. U series power consumption for almost H series performance in terms of *relative* performance. But total performance stays the same.

3 - However, the one caveat that makes me think this Asus still needs some driver or low level updates and will improve in time is the battery consumption tests. The HP Elitebook 845 G10 (or the Lenovo T14s G4 has very similar performance) achieves 760 min of H.264 video playback and 779 min of wi-fi time on a very small 51 Whr battery compared to 1204 min of H.264 playback and ONLY 640 min of wi-fi time from a 78 Whr battery. So for slightly over a 50% increase in battery capacity, the H.264 time increased by 58% (so there is a system level improvement) but a substantial decrease of over 2 hours in wi-fi time despite a much larger battery!! Something is totally wrong there. My thoughts are this severe decrease in wi-fi time is due to their power control features. The core clocks and package power graphs remind me of the old power consumption graphs of several older Dell laptop models you reviewed. They should be constant and not fluctuating between max and 0.

Thank you again for all of your detailed information and reviews. Looking forward to the Asus ProArt 16 series review.

[ArsLoginName]

How is it possible that this performs worse in games then Z1 extreme?

Asus doesn't have the power control/profiles worked out yet properly. See my other comments regarding total performance and especially those when looking at battery lifetimes.

[Konstantinos]

There is something wrong with your multi-core Cinebench R15, R20 and R23 scores.

All three show around 40% less performance than the ProArt PX13 when the multi-core Cinebench 2024, Geekbench 6 and Geekbench 5 show only 12-15% better performance for the ProArt PX13 which is what you expect due to the difference in TDP between the two laptops.

Please check you power settings (e.g. 17 watt or 28 watt?)

[LXR]

There's something seriously wrong with this laptop. That other Proart HX 370 is 41 % faster in Cinebench R23.

[vladteapa]

Typical Asus fumble...

By the looks of it, soldered RAM is the only available future and other brands have to wait until the fall to get their "hands" on this chip.

This chip is another reason to hate the AI bandwagon.

[NikoB]

For example, the X1E-80-100 in its direct competitor, the Samsung Galaxy Book4 Edge 16, was slightly slower during the multi-core tests, but the X1E-78-100 inside the Asus Vivobook S 15 was slightly faster in standard mode.

No, X1E-78-100 inside the Asus Vivobook S 15, 15-20% slower in CBR15 and let's not forget about Arm's poor compatibility with the x86 code base.
Also consider that the X1E-78-100 consumes more than 35W in PL1 mode, instead of 28W for the AI 9 HX 370.

comfortable keyboard

It's a shame for NB to list the cut-down keyboard as a plus for the laptop. Obviously, for a 16" model, this is a shameful keyboard without a classical numpad, which excludes effective work in the office and at home. Well, except for those who just write texts...

brilliant 120-Hz OLED

For IPS, even if it is semi-matte, yes. But for AMOLED, where "super" contrast of 1M:1+ is advertised, this is a shameful panel.

Note that the laptop screen does not even have a True Black HDR certificate - precisely because of the shameful contrast of 20,000:1, instead of 1M:1+ minimum.

I wonder how the manufacturers managed to raise the PWM frequency on AMOLED to 480Hz+ (240x2, which immediately leads to some bad thoughts), instead of 200-240Hz? What new achievements in AMOLED panels allowed to improve the frequency exactly 2 (two) times? The authors of NB need to explain this change in the last year in the reviews.

Although, as I already wrote, the minimum PWM frequency, which is considered safe, is 1.2 kHz according to the international standard.

CPU part:
Compared to 7945HX with Zen4 cores "5/6nm" (4900-5000 points in CBR15 at 85W in PL1 from www.notebookcheck.net/Minisforum-AtomMan-G7-PT-review-Compact-gaming-mini-PC-with-AMD-Ryzen-9-7945HX-and-Radeon-RX-7600M-XT.859693.0.html) the real increase in energy efficiency is approximately +40-45% for Zen5 cores (~2380@28W in PL1), which is quite good. It is clear that this is a thin and extremely light laptop (for 16") with a weak cooling system, but to get the performance of Zen4 7940HS again, albeit with minus 7W of consumption in PL1 (2330@35W from www.notebookcheck.net/Geekom-A7-with-0-5l-case-in-review-Premium-mini-PC-with-AMD-Ryzen-9-7940HS-32-GB-DDR5-RAM-and-2-TB-SSD.811438.0.html) - this is just boring and not impressive...

The real leap (for AMD) is finally a much more efficient memory controller with the same boring LDDR5 7500, which was finally fixed (as they previously claimed - supposedly there were problems with Zen4 with 7500, but in reality there were models in which it worked, it's just that other manufacturers, like Lenovo, were not qualified enough at the level of development departments to make motherboards with working LDDR5 7500 + Zen4 Phoenix). Finally, we see numbers close to 100 GB / s. Although from the point of view of the efficient operation of DP2.1 (UHBR20) and igpu, this bandwidth is obviously not enough, you need 3 times more, at least. That is why this junior line of Zen5 Strix does not support DP2.1 with UHBR20, because there is not at least a 256-bit memory controller with twice the bandwidth. We are waiting for Zen5 Strix Halo, the release of which was postponed, as usual - another "paper" AMD family, to its shame. Although it should be the most popular in business laptops without dGPU, but with real support for 8k monitors and igpu at least at the level of GTX1650.

I still naively expect that they will finally unsolder the HBR3 in SoC with a 512-1024 data bus and 8-32GB as VRAM. This will be a real breakthrough for AMD igpu...

But the memory latency is again depressing - more than 100ns even for the HX series 9

I don't like that the hot air exhaust on the screen exceeds 47C - this can have a bad effect on the AMOLED panel, although I did not look at the critical temperatures for it, but most likely they also do not exceed 50C.

Battery life with Wi-Fi (and this is an openly fake test, the time in which for real surfing should be immediately divided by 1.5-2 times) is not impressive. There are no stable 8 hours+ with real work and load.

Considering the stated price, 32 GB of soldered RAM is too little - there should be at least 64 GB (and they are 100% supported in Zen4/Zen5, as we have already found out earlier).

Because even taking into account the novelty, but taking into account all the shortcomings and slightly high performance of last year's 7940HS with slightly lower consumption at -7 W, not a very good offer for more than $2200. What are the extra 800$-900$ for, Asus?

If there were 64GB/4TB here, it could still be justified somehow, but as it is... until there is a 30-40% discount, it is unlikely that there will be many people willing to buy the already flawed series, both in terms of the keyboard and the glossy flickering screen without real support for HDR10 and 8k monitors on USB-c ports (where we were waiting for USB40 V2 at 80Gbps)...

Thank you for your attention.

[troy]

No touchscreen? The ASUS website says it does ! It even comes with a stylus. You sure you have the right one?

[NikoB]

unsolder the HBR3 in SoC

Sorry, typo - HBM3

[lolryzen]

How can they call it ryzen NINE if it gets completely raped by passively cooled M4 tablet in geekbench.

[lolryzen]

Asus Vivobook S 15, 15-20% slower in CBR15

CBR 15 is a 2013 test with zero support for 2024 platforms, welcome to reality

[vladteapa]

Was hoping these would be priced more competitively being soldered RAM machines...

Also what's the point of cramming a non low power chip inside a chassis that clearly can't take the heat? And since 90% of reviewers got this laptop only it doesn't look good.

Pretty sure it's another Asus fumble but clearly AMD were aware...

[NikoB]

CBR 15 is a 2013 test with zero support for 2024 platforms, welcome to reality

Welcome to reality, IT amateur - the code used in CBR15 is the code of 95% of modern software. All the "new" extensions are used in 5% of software and less than 5% of cases in practice among most buyers.

About the same as the 99% useless NPU block in 99% of everyday tasks, as well as another scam with "AI" and "Co-pilot+".

It is software like CBR15 that shows the real multi-threaded performance of hardware.

And in CBR2024 there are few changes in this regard. As Asus S15 was weaker with Qualcomm against Zen5, so it is in any x86 code, because the second-level x86 emulator will never be faster than the first-level emulator from Intel/AMD, because RISC cores have long been used there, which emulate a significant part of the x86 instruction set at the microcode level.

[NikoB]

if it gets completely raped by passively cooled M4 tablet in geekbench.

This is absolutely not a relevant test for x86 code. It is more tailored for web surfing than for heavy real-world floating-point loads. Arm has always lost, even without emulation, because x86 processors scale much better with increasing TDP in desktops and laptops. The tablet's destiny is banal content consumption, not serious workloads and tasks.