Search the Community

For some generations now, NVIDIA uses the same desktop GPUs on laptops, although with some differences to reduce its consumption. In essence they are the same GPU, but NVIDIA drivers continue to differentiate those dedicated to dedicated graphics on PCs and laptops. Why do they do it and what are the differences? In this article we explain everything. What is the difference between a laptop and desktop GPU? First of all, it should be clarified that, some time ago, the GPUs of the laptop and those of the dedicated graphics of NVIDIA were completely different, and in fact those of the laptops had the distinctive "M" at the end of their name for better differentiation. However, for a long time the GPU is essentially the same, but with some reduced attributes in order to reduce its consumption. Thus, for example, an RTX 2080 in its reference model for desktop, has 2,944 shader units, with 8 GB of GDDR6 memory and a 256-bit interface that makes it work at 14 GHz effective, delivering a bandwidth of 448 GB / s. The RTX 2080 notebook is exactly the same, with the same parameters, but nevertheless its speed is slightly lower, around 9% slower in its base speed and 7% in its Boost speed. The performance is also slightly lower on the notebook model, but instead of consuming 215 watts, it consumes only 150W. We do not speak, however, of the Max-Q variants, which are graphics with a much lower consumption (90 watts in this case) and with much lower operating speeds, so the performance also plummets. If the GPU is the same, why are the NVIDIA drivers different? When we go to the NVIDIA driver download website, when choosing the graphics model for which we want to download the drivers, the normal models are clearly different from those of the notebook (Notebooks). But there is a difference that we can clearly see: size. The desktop drivers, to the right of the image, occupy 570.94 MB compared to 523.27 MB of the desktop graphics version. This difference in size is because the laptop drivers have a lower level of customization, although the firmware and the driver itself is exactly the same, as well as the same technologies because, remember, the GPU is the same. Why if the driver is the same occupy different size? If you have read this far, surely you are asking yourself precisely this question: why do they have a different level of personalization? This is because, with the "NVIDIA Notebook Driver" program, the company ensures the correct functioning of the reference model - we speak in this case only of laptops. However, most laptop manufacturers modify these GPUs to offer higher performance, lower consumption or lower temperature, and therefore the operation may be different and not work perfectly with the NVIDIA driver.

To date, the NVIDIA ‘SUPER’ designation was reserved for GTX 20 Series graphics cards. It seems that the company will finally extend this denomination to the GTX 16 Series. The company's simplest graphics with RT Cores and Tensor Cores will receive at least one version. During the last weeks there has been a lot of talk about a GTX 1660 SUPER, which has just been made official. This new NVIDIA GTX 1660 SUPER graphics card will be the most powerful in the company's mid-range. There will be no reference model, the models customized by the manufacturers will be seen directly. Possibly they come with lower prices even to the model they arrive to replace and combat the AMD RX 5500. NVIDIA GTX 1660 SUPER is confirmed Zotac has been the first of the manufacturers to have these charts ready. The company will have two models, the simple model and an AMP model. This second model, with a better cooling system will have factory overclocking and therefore, better performance. Another of the outstanding novelties of this graphic is that it will not arrive with GDDR5 memories, but will arrive with GDDR6 memories. The change of memories does not seem important, but it gives us extra performance. Obtaining more bandwidth improves the final performance of the graph. The frequencies have not yet been confirmed, which will be revealed shortly before launch or when NVIDIA officially presents them. We see how this graphic has an 8-pin PCIe connector and completely lacks an NVLink port. On its performance, we must expect 10-15% more than the GTX 1660 with a fairly similar consumption. We assume that in the coming days NVIDIA will reveal the launch date and Zotac the market launch price of its two models.

Reuters, via financial news outlet Calcalist, reports that Nvidia has outbid Intel in an attempt to buy the networking behemoth Mellanox. Previous unconfirmed reports claimed that Intel had offered $6 billion for Mellanox, but Nvidia has reportedly outbid the company by more than 10 percent. These offers come after Mellanox essentially offered itself for sale in October 2018, which reportedly also garnered attention from Microsoft and Xilinx. Mellanox specializes in Ethernet and InfiniBand networking products for the data center. The company currently has a market cap of $5.93 billion and a commanding market presence, but Intel's purported acquisition attempt could run afoul of regulators. Intel already has a heavy presence in the InfiniBand market through its line of products that come as the fruit of its $125 million acquisition of QLogic's IP back in 2012, while Mellanox is the only large-scale InfiniBand competitor. Nvidia's play at Mellanox would help the company diversify into networking, thus reducing its heavy reliance upon graphics cards sales that has seen the company suffer at the whims of the cryptocurrency market. The Mellanox IP could also tie in well with Nvidia's strategy to increase its penetration into the data center, and Reuters predicts the company wouldn't face the same regulatory concerns as Intel. Mellanox would also fold into Intel's portfolio nicely. Intel has spent several years transitioning away from being a PC-centric business to other profitable climes, such as data storage, memory, IoT, and 5G. Intel's overall goal is to leverage its commanding presence in the data center, estimated at ~95% of the worlds server sockets, to expand into these new adjacencies quickly. Networking is the glue that ties many of these critical components together, making the Mellanox portfolio a natural fit. InfinBand networking is a critical component in the high performance computing (HPC) space that could afford Intel an advantage with tightly-integrated systems-level products as it fends off the rising po[CENSORED]rity of AMD's Zen-based data center processors. Only time will tell if Intel will dig into its significantly deeper pockets to push Nvidia out of the running, or if another large player, like Microsoft, can steal the networking show. As expected, Nvidia, Intel, and Mellanox have not commented on the bids. Article created by „tomshardware”.