The mobile GeForce RTX 3000s are finally out! While the availability of the Desktop versions is still problematic, this is not (yet?) The case for the mobile versions. We were able to get our hands on an MSI GS66 Stealth, a machine that features the fastest mobile GPU around since it is an RTX 3080.
After presenting its new generation of graphics chips in a mobile version during CES 2021 , Nvidia kicked off hostilities on January 26 so that the Ampere architecture and its cards 2nd generation GeForce RTX are coming to laptops. With more than 70 portable PCs announced, for the players as for the creators, it is also necessary to count, initially, on the arrival of 28 variants for these GeForce RTX 3060, GeForce RTX 3070 and GeForce RTX 3080. An important nuance to understand that we explain below.
The mobile RTX 3080 which equips our test machine incorporates a GA104-775-A1 chip, engraved in 8 nm by Samsung. It takes advantage of a maximum clock speed of 1,245 MHz, a total of 6,144 CUDA cores, and 16 GB of GDDR6 memory on a 256-bit bus with a bandwidth of 384 GB / s. The basic TGP is 80 W on this chip, but rises to 95 W on our machine thanks to Dynamic Boost 2.0 which we will discuss next.
MSI GS66 Stealth 10UH: our test configuration
Regarding the machine itself, this MSI GS66 Stealth includes an Intel Core i7-10870H processor with 8 cores and 16 threads for a base frequency of 2.2 GHz and 5 GHz in turbo. On the RAM side, we have the right to 2 x 8 GB of DDR4 in Dual-Channel, clocked at 3200 MHz.
The MSI GS66 Stealth 10UH also has a 15.6 ″ IPS panel, 240 Hz in QHD (2560 x 1440), definition with which we carried out our tests. Finally, Windows 10 is here in Professional edition (version 20H2), everything takes place on a 1TB Western Digital SN730 NVMe SSD. The Nvidia Game Ready drivers have been updated to version 461.40 before starting the test.
RTX 3000: a broken strategy
Before getting to the heart of the matter, it seems wise to come back to this new generation of mobile GPUs. While it implies a definite generational leap thanks to the Ampere architecture, it also marks a departure from the previous generation, which is sure to cause confusion among consumers.
First of all, and contrary to what we have read here and there, Nvidia has not abandoned the Max-Q denomination for its mobile GPUs. In the past, GeForce cards for laptops were segmented according to Max-Q and Max-P configurations, the former referred to a reduced TGP in order to be suitable for thin and light chassis, while the latter involved more demanding GPUs and therefore more efficient, ultimately a story of energy efficiency and temperature management.
However, the energy efficiency of a laptop is mainly based on the design of its chassis, for this reason Nvidia lets the OEM determine itself the TGP of each reference, as we have seen, from 80 to 150 W for the RTX 3080, 80 to 125 W for the RTX 3070, and finally from 60 to 115 W for the RTX 3060. To see these variations, as well as the data sheets of the computers unveiled recently, the risk for the consumer is not to understand the respective performance of each machine, firstly because it is possible that a laptop equipped with an RTX 3060 with a TGP of 115W offers better performance than a model equipped with an RTX 3070 at 80 W, but also, because it seems difficult to know precisely whether this or that low-power GPU will or will not benefit from the improvements of the 3rd generation Max-Q technology and in particular of the Dynamic Boost 2.0 which directly influences performances.
Ultimately, with the 3rd generation Max-Q, Nvidia evokes a “broader” approach than in the past, with “a holistic set of platform technologies”, for the design of powerful and powerful laptops. purposes. We thus find the features Dynamic Boost 2.0, Whisper Mode 2.0, Advanced Optimus, but also the famous Resizable BAR and DLSS, all things to which we will come back later. Where things get complicated is that PC manufacturers may choose to integrate only some of these features.
To summarize, you will have to be careful and vigilant when buying a laptop computer equipped with a GeForce RTX series 30. Like RTX series 20 cards, performance is naturally not really comparable. homonymous cards intended for desktop computers. However, here we can expect the performance gap to widen more between a mobile and desktop version of an RTX 3000 compared to what was done on the 20 series.
Without further ado, let’s take a look at what the MSI GS66 Stealth has in store for performance.
3DMark gives us a good idea of the power of a given map for a relatively precise comparative approach. During a first test with Time Spy, we obtained a graphic score of 10,126 points, for an overall score of 9,706 points, all with an average of 67.66 fps and 56.83 fps on the first and second part of the benchmark and a temperature that does not exceed 61 ° C.
Time Spy Extreme, a similar test but carried out in 4K with DirectX 12, gives us a graphic score of 4,887 points, for an overall score of 4,492 points. Our copy averaged 31.75 fps in the first part of the test and 28.11 fps in the second, with a maximum temperature of 64 ° C for the GPU.
If we are to believe the averages observed here and there, our test copy shows an increase in performance of between 18 and 19% on these two tests compared to an RTX 2080 Max-Q. However, even if the comparison is irrelevant, here we see a performance difference of around 40% compared to the desktop version of an RTX 3080.
3DMark DirectX Ray Tracing
To assess the Ray Tracing performance of our mobile RTX 3080, we first used 3DMark’s DirectX Ray Tracing test, then also launched Port Royal in order to gauge the Ray Tracing performance in real time.
In the first test, our GPU managed to get an average of 26.91 fps, a better result than the mobile versions of the RTX 2080, which even came close to an RTX 3070 desktop. The second test shows what to expect regarding real-time performance since we get a score of 6,142 points, which represents a performance gain of about 25% compared to the MSI GS66 Stealth equipped with an RTX 2080. Max-Q. Finally, during these two tests, the GPU reached a temperature peak of 71 ° C.
Benchmarks in games without ray tracing
Since the arrival of the Turing architecture, and now Ampere, ray tracing has become a reality, even on portable machines. Nevertheless, there is still only a handful of games to really enjoy it, which is why we chose Assassin’s Creed Odyssey and Red Dead Redemption 2, two titles demanding enough to determine the performance of our machine excluding the [19459059 ] ray tracing of the equation.
Assassin’s Creed Odyssey
Using the benchmark tool built into Assassin’s Creed Odyssey, we raised the average frame rate by selecting the highest graphics settings (Ultra High), and turning off vertical sync, as well as the adaptive definition, all in QHD (2560 x 1440).
With these parameters, we note an average of 55 fps, a sign that the game runs almost perfectly at this definition in its maximum quality. For comparison, we get 59 fps by setting the quality slider to “very high” and 62 fps to “high”.
Red Dead Redemption 2
If there is one game that remains particularly demanding on our graphics cards, it’s Red Dead Redemption 2! The title of Rockstar Games also allows us to evaluate the performance of our RTX 3080 mobile through an integrated benchmark. To do this, we set the graphics quality sliders to the maximum, while taking care to turn off vertical sync. Like our other tests, this one still takes place in QHD, the native screen definition of the MSI GS66 Stealth.
As a result of the races, the various scenes broadcast during the benchmark show us that the game runs at 58 fps on average in this case, so we can easily enjoy the superb open world offered here. The game runs at an average of 63 fps when the quality slider is set to “balanced”.
Benchmarks in games with Ray Tracing
What about Ray Tracing with this second generation of mobile GPUs under Ampere? We tried to gauge the performance of our copy by analyzing several games and comparing the readings without and with Ray Tracing, then finally with Ray Tracing supported by DLSS.
First scenario with the title of Remedy Entertainment. No benchmark built into the game here, but scenes that allow us to measure performance fairly precisely from the start of the game. With our mobile RTX 3080, we manage without problem to hold the bar in QHD and without Ray Tracing, with 61 fps on average. Activating Ray Tracing is of course a whole different story, with 36 fps.
DLSS however saves the day by keeping the high quality of Ray Tracing with an average of 65 fps, but a resolution rendering that goes to 1707 x 960 in our case. By setting Ray Tracing to “medium” we manage to get 84 fps.
Ha, Cyberpunk! Mocked upon its release due to significant optimization issues, CD Projekt Red’s long-awaited game has however enjoyed a few welcome patches in recent weeks. We wanted to see what it is like with Ray Tracing on the very qualitative panel of our GS66, but also to measure its performance.
The results are obviously less attractive than with other titles, but we managed to run it without too much hassle, but without ever exceeding 60 fps, even though DLSS is activated in performance mode, in which case we get an average of 58 fps. With Ultra Ray Tracing, the game crumbles at 22 fps on average, selecting “quality” mode for DLSS we measure 41 fps. Hopefully future fixes will allow us to enjoy it more peacefully on a laptop!
Watch Dogs Legion
Watch Dogs Legion is also an excellent game for judging the performance of our GPU, especially thanks to its open world and its many effects in Ray Tracing. Its built-in benchmark includes a long stage in the middle of London, which allows you to get a good idea of performance depending on the graphics settings.
The game runs relatively well in Ultra, without Ray Tracing, with 54 fps on average. By lowering the graphics quality option to “very high” we achieve 62 fps. By activating Ray Tracing, without DLSS, the game has a hard time running correctly with an average of 33 fps in Ultra and 40 fps with Ray Tracing on “high”. Once again, DLSS saves the day to take advantage of Ray Tracing with 57 fps when this is configured on “performance”.
Max-Q 3 th generation: update on new features
As said above, not all laptops equipped with RTX 3000 cards will benefit from it, but Nvidia has lent us here an MSI GS66 Stealth integrating all the Max-Q technologies of 3 e [19459078 ] generation.
Dynamic Boost 2.0
With the Resizable Bar, Dynamic Boost in its version 2.0 is the great novelty of this 3 th generation of Max-Q. New, because although this feature was deployed with the previous generation, it only worked in one direction, allocating power from the CPU to the GPU, not the other way around. It now allows the operation to be done in both directions, adding graphics memory to the equation. Nvidia especially evokes an important development in terms of the management of Dynamic Boost 2.0, now run by several AI networks.
On paper, Dynamic Boost 2.0 is therefore supposed to offer us a performance gain of up to 16% according to Nvidia. On our test copy, it actually makes it possible to increase the TGP from 80 to 95 W, but the manufacturer, MSI in our case, can quite choose to unblock only 5 or 10 W with this technology. Unfortunately, unlike the first iteration, it is no longer possible to enable or disable Dynamic Boost 2.0, which is why we were unable to test what Nvidia said about it. Either way, the performance gains are quite evident compared to the previous generation of mobile GPUs.
Another expected new feature is the Resizable Bar, a feature that allows the processor to access all of the VRAM in the form of a single addressable block instead of 256MB blocks. Already deployed by AMD as Smart Access Memory (SAM) for its Radeon RX 6000 graphics cards, this functionality is actually based on the Base Address Register (BAR) specification of the PCI-Express interface.
In practice, Resizable Bar technology not only accesses all of the video memory, but also performs transfers simultaneously rather than one after the other. This would result in a significant performance gain, up to 10% on games whose engine is optimized for, but also in certain 3D rendering applications.
Here, too, we were unfortunately unable to put the technology to the test. Indeed, if the Resizable Bar can be activated and deactivated at will via the BIOS on some motherboards, this is not the case on our MSI GS66 Stealth.
Whisper Mode 2.0
For this version 2.0 of Whisper Mode, Nvidia suggests that it has completely redesigned its acoustic control system. Using algorithms powered by artificial intelligence, WhisperMode 2.0 dynamically manages CPU and GPU usage, in order to control and act on machine temperatures and fan speed.
As the screenshot below shows, GeForce Experience lets us choose between three levels of volume, or rather noise pollution, as well as 7 levels of framerate, between 30 and 60 frames per second. Fortunately, the option can be activated and deactivated at will via the chameleon software. Whisper Mode 2.0 then takes care of reducing noise pollution while trying to maintain the indicated framerate.
Examining this feature is not really easy, for the good reason that this feature adapts different parameters dynamically, so volume changes are not necessarily noticeable upon activation. In its press release, Nvidia mentions for example a reduction of about 6 dB on Shadow Of The Tomb Raider in QHD without going below an average of 60 fps, while selecting the “balanced” parameter and aiming for 60 fps.
In our case, we tried the experiment on Control with Ray Tracing (high) and DLSS enabled. Our first measurement without WhisperMode 2.0 tells us a sound volume of 36.5 dB while the game is running at an average of 65 fps. Once activated in balanced mode (60 fps), we find an average of 59 fps and a slightly lighter nuisance at 34.2 dB. Keeping our target of 60 fps, but with the “quieter” mode this time around, the volume drops to 31.2 dB for an average framerate of 55 fps.
Concretely, although we have left the game volume at 0 for our measurements, this slight reduction in nuisance is almost imperceptible in normal conditions where very generally we use headphones, or the integrated speakers. However, it may make the difference to avoid annoying those around us in the middle of the game.
However, this mode may be more relevant for titles with a much higher frame rate. Only at the cost of a few decibels you will limit your gaming experience to 60 fps, which is certainly not what we want in competitive games like Apex Legends, Fortnite or even Overwatch to name a few.
Already known for many years (2010 to be precise), Optimus aims to save battery without sacrificing the graphics performance of the dedicated GPU, by automatically juggling between the latter and the iGPU (the economical graphics chip of the processor). We are not going to go into detail on its operating principle, but Optimus had a large drawback in the past since it did not allow you to take advantage of the display features that you would expect to find when you connected an external screen to a laptop PC equipped with an Nvidia GPU. We think, for example, of the variable refreshing technology G-Sync, the possibility of displaying 4K at 120 Hz, or a QHD definition at 240 Hz.
With Advanced Optimus, Nvidia has integrated a simple hardware switch to its new mobile cards, allowing the dedicated GPU to fully manage the screens connected to the machine while keeping Optimus activated. This means that when this is the case, the functionality of HDMI 2.1 is directly available to the external monitor or TV, in which case the PC screen will be managed by the iGPU. Finally, provided the laptop has other outputs, via USB-C or HDMI, it is now possible to manage up to 3 external screens with the Advanced Optimus.
As a conclusion, let’s start by pointing out that the generational leap with the Ampere architecture and Samsung’s 8nm chips seems to bear all its promises here. The performance gains are much more significant than upgrading to the GeForce RTX 20 Super last year.
However, it should be remembered that our test was structured around an RTX 3080 with 95 W TGP, the results of our benchmarks are exciting and show that it is quite possible to take advantage of Ray-Tracing in QHD in good conditions. These results are not, however, to be taken at face value for all the laptops that will be displayed on paper with the same GPU and we should probably expect more or less large variations depending on the choices made. by the OEM, regarding the TGP and functionality related to the 3 th generation Max-Q technology.
In our opinion, Nvidia has taken up the gamble perfectly with Max-Q. Confusion still reigns, which is not good news for consumers, and some may find the chameleon’s communication strategy haphazard. Despite everything, the overall increase in performance is accompanied here by essential technologies for designing ever more powerful, thin and light laptops, which are ultimately intended for gamers and designers alike. We think of course of the Dynamic Boost 2.0, the Advenced Optimus, but also of the Resizable Bar, with which Nvidia is catching up on its wagon over AMD. We’re less excited about the WhisperMode 2.0, however, but it depends a lot on the design of the machine itself and its cooling system, so it will be necessary to build a more in-depth opinion by looking at other machines. .
There remains the price of machines powered by RTX 3000 mobile GPUs. It is certain that at nearly € 3,700, our test copy will gently make some of our readers smile. It should be remembered, however, that this is a high-end, particularly compact kid’s laptop which includes all of the Max-Q features. The above reflection is therefore valid once again here: it will be necessary to compare what is comparable, at the same price, by testing other machines and also to see what Nvidia has prepared for us with its RTX 3060 and 3070 mobiles. .