Parallel to Intel's presentation of new high-end notebook CPUs, Nvidia is introducing the GeForce RTX 3050 (Ti) laptop GPU, two new entry-level GPUs for mobile gaming PCs. They follow GeForce RTX 3080 and 3070 laptop GPU (test) as well as GeForce RTX 3060 laptop GPU (test) and rely on the GA107 chip, which has RT and tensor cores.
The mobile 50s get RT and tensor cores
Based on Turing, Nvidia had the smallest graphics card from the 60 series (GeForce RTX 1660 Ti (Max-Q) with TU116) and the 50 series (GeForce GTX 1650 (Ti) (Max-Q) with TU117) as in the desktop the RT and Tensor cores have been deleted, the mobile GeForce RTX 3050 laptop GPU and RTX 3050 Ti laptop GPU now offer these functional units. Real-time ray tracing according to Microsoft DXR (Nvidia RTX) and AI functions such as DLSS or application scenarios accelerated by AI (e.g. denoising in the render preview) are possible.
GeForce RTX 3050 Ti and 3050 Laptop GPU
While the mobile GeForce GTX 1650 Ti and GeForce 1650 differ only in the maximum power dissipation and thus the clock rate (Ti ultimately called itself the fastest configuration), both variants of the GeForce RTX 3050 are based on two different configurations of the new GA107 chip the amp generation. The GeForce RTX 3050 Ti laptop GPU should be fully equipped with 2,560 amp shaders, the GeForce RTX 3050 laptop GPU was cut by 512 shaders and thus exactly 20 percent. This also applies to the RT and Tensor cores.
There is no difference in the memory expansion (4 GB GDDR6), its connection via a 128-bit wide interface and its clock, which is connected with 12 Gbps regardless of the variant and whether it is a Max-Q version.
New laptop GPUs from 35 to 80 watts
As with the larger mobile GeForce RTX 3000, OEMs can choose the maximum permissible power loss for the GeForce RTX 3050 (Ti) laptop GPUs from a wide range: Configurations from 35 to 80 watts are possible, with the versions with 35 to a maximum of 50 watts as “Max-Q models” apply.
As known from the GeForce RTX 3080, 3070 and 3060 laptop GPU, this is no longer included in the product name of the GeForce RTX 3050 (Ti) laptop GPU, but can only be seen via the driver. Nvidia sells Max-Q in this generation specifically as an efficiency feature and no longer wants to focus on the direct reference to particularly low power dissipation and thus also computing power. In response to the massive criticism at the beginning of the year, Nvidia has now imposed on partners to openly communicate maximum power loss and/or clock speed.
8 GB GDDR6 8 GB GDDR6 6 GB GDDR6 4 GB GDDR6 4 GB GDDR5 memory throughput 12/14 Gbps ** 12 Gbps 8 Gbps memory bus 256 bit 192 bit 128 bit 128 bit TGP (watt) *** 80 to 150 80 to 125 80 to 115 60 to 115 60 to 90 60 to 80 35 to 80 35 to 55 * Analogous to the possible TGP range,
** 12 Gbps (6,000 MHz) for Max-Q variants
** OEMs can provide the GPU with RTX 3000 via “Dynamic Boost 2.0” an additional 5, 10 or 15 watts, which the CPU would otherwise be entitled to. The maximum available electrical power for the 3080, 3070, 3060 and 350 in the notebook is 165, 140, 130 and 95 watts.
Significantly more raw performance than its predecessor
Compared to the GeForce GTX 1650 (Ti), the GeForce RTX 3050 promises to deliver significantly more performance at the same TGP level, although in this case too The rule is that the massive increase in FP32 shader performance, as long as there are no integer arithmetic tasks in between, cannot be used equally efficiently from Ampere to Turing in every application. Nvidia promises up to 100 percent more performance.
It remains with 4 GB of memory
Nvidia has also increased the memory bandwidth by 50 percent. On the other hand, a maximum of 4 GB GDDR6 no longer allows the maximum texture details in many games and can become a bottleneck under Full HD even with less high-resolution textures when using ray tracing. Obviously, Nvidia does not see any need to forego 4 GB in the notebook, especially not if the smallest expansion stage with 8 GB would have the same amount of memory as the top models and thus more than the GeForce RTX 3060 laptop GPU. On the desktop, however, Nvidia dared to take this step with the GeForce RTX 3060 12 GB to avoid only 6 GB on a GeForce RTX 3000.
Until 60 percent faster in games than its predecessor
As always, manufacturer benchmarks should be treated with great caution. They are by no means wrong, but extremely well selected and often only show the very best. Most recently, the game selection in particular had shown that you can turn a result completely in favor of a manufacturer with just a few titles, so a double-digit lead can turn into a double-digit deficit if desired. This of course makes comparative tests all the more difficult, because the accusation can always be made that manufacturers A or B are preferred. A test sample was not available to the editorial team at the start.
With the benchmarks presented, Nvidia promises 50 to 60 percent more FPS in pure rasterizer titles, but with three of the five titles mentioned, it focuses on ray tracing benchmarks that were not yet possible with the small Turing chips. With the help of DLSS (“Quality”) with medium graphic details, the GeForce RTX 3050 Ti laptop GPU should achieve over 60 FPS in Full HD in Control, Watch Dogs: Legion and Minecraft RTX, without DLSS this should at least be the case in Watch Dogs .
RT and tensor cores accelerate apps
While the concrete benefits of the RT and Tensor cores in games in the new 50s class still have to be proven in more comprehensive benchmarks, the advantages in applications are obvious: The fact that a GeForce RTX 3050 (Ti) laptop GPU will also be in the future compatible apps can accelerate RT and AI workloads, it sets it apart significantly from the GeForce RTX 1650 (Ti) or the GeForce GTX 1660 Ti. Players can now also let Nvidia's broadcast suite disappear in high quality without a green screen.
-
Image 1 of 2
In. Notebooks from 799 US -Dollar
Nvidia promises notebooks with GeForce RTX 3050 laptop GPU from 799 US dollars before taxes. That's $ 200 less than the cheapest GeForce RTX 3060 laptop GPU configurations. The first notebooks are theoretically available immediately, but given the expected time until Intel's new Tiger Lake H45 processors are on the market, it should take a few more weeks for many models.
ComputerBase has received information about this article from Nvidia under NDA. The only requirement was the earliest possible publication time.