NVIDIA uses GA106 for RTX 3050, here is a possible explanation
NVIDIA uses GA106 for RTX 3050
Controlling 79% of the market for dedicated desktop graphics cards has its advantages. For example, NVIDIA can maximize production and profits thanks to the different dies it makes. The company recently unveiled its new GeForce RTX 3050, which is offered at the formal recommended retail price of $ 249. Many thought the company was using its relatively inexpensive and low-power GA107 graphics processing unit (GPU). Unfortunately, this is not the case. According to Inno3D, NVIDIA instead uses the much more expensive GA106 die for this entry-level product. However, this choice could be a good sign.Photo Credit: Inno3D Inno3D has released a high resolution render of its GeForce RTX 3050 without a cooling system as part of a press release, revealing that the PCB uses the GA106 GPU. Unfortunately, we can't say anything about the other GeForce RTX 3050 cards for obvious reasons. Still, they will most likely all use the same GPU at launch, but that may change over time. However, the GA106 and GA107 are completely different.
The GA106 is a reasonably large chip (272mm²) with up to 3,584 CUDA enabled cores and a TDP of 170W (the desktop version). The scaled-down edition has 2,560 CUDA cores, overlapping the fully enabled GA107 GPU with 2,560 CUDA cores used for laptops, and has a TDP of between 35 and 80 W.
Chip Chip Configuration (CUDA) Memory Interface Die Size MSRP GeForce RTX 3050 desktop GA106 2560 128-bit 276mm² $ 249 GeForce RTX 3060 desktop GA106 3584 192-bit 276mm² $ 329 GeForce RTX 3050 notebook GA107 2048 128-bit Smaller than 276 mm²? GeForce RTX 3060 Ti notebook GA107 2560 192-bit Smaller than 276mm²? if (jQuery ("# crm_srl-th_hardware_d_mh2_1"). is (": visible")) {console.log ("Edinet ADV adding zone: tag crm_srl-th_hardware_d_mh2_1 slot id: th_hardware_d_mh2"); } Usually, large manufacturers want a software driver, and once a company wins a contract, it lasts for a cycle, which happens to be a year. So we can assume that, for this reason, NVIDIA has opted to have two GPUs overlapping rather than having no GPUs to ship. The good news for PC gamers is that the company has two entry-level GPUs in production. The bad news is that there are miners, scalpers and, more generally, groups of other people who will put a spoke in the wheel of anyone looking to buy a GeForce RTX 3050 at its MSRP.
Nvidia GeForce RTX 3050 could launch with 2 models for desktops
It was already rumored that Nvidia is working on a desktop variant of its GeForce RTX 3050 graphics card. According to a new leak, however, the company is preparing to launch the GPU with two models.
Leaker Kopite7kimi (spotted by VideoCardz) disclosed that Nvidia is reportedly opting to integrate the GA106 GPU for the RTX 3050 desktop lineup (4GB and 8GB configurations) as opposed to the GA107 GPU that was originally reported.
Initial rumors detailing the card’s specifications indicated 3,072 CUDA cores. According to Kopite7kimi, that configuration has been downgraded by Nvidia. The GA106-150 GPU will now feature 2,560 CUDA cores alongside 8GB of memory, while the GA106-140 GPU that sports 2,304 CUDA cores comes with 4GB of memory.
It’s unclear why Nvidia chose to change the GPU architecture, but VideoCardz notes the decision could be associated with plans to potentially release an RTX 3050 Ti desktop variant that offers more cores, as well as the same GPU.
As for the GA106 GPU itself, it’s based on Nvidia’s Ampere architecture. With support for DirectX 12 Ultimate, the chip is capable of both hardware ray tracing and variable-rate shading, which are features largely used for video games. It’s also compatible with machine-learning applications through the use of 120 tensor cores.
Currently, the GeForce RTX 3050 graphics card is found on laptops. The card’s desktop series, meanwhile, seems to be Nvidia’s answer to Intel’s upcoming Arc Alchemist lineup, as well as AMD’s RDNA 2-based Navi 24 GPUs. The first video cards from these two series produced by Team Blue and Team Red are expected to be released during the first quarter of 2022.
With this in mind, the RTX 3050 desktop GPUs are most likely set for an unveiling at CES 2022 on January 4, followed by an official release a few weeks later at the end of the month. Of course, this is not the only graphics card Nvidia plans to release during the coming months. Its flagship GeForce RTX 3090 Ti is also due for a January launch, while the RTX 3080 12GB model could go head-to-head with Arc Alchemist’s high-end segment in the second quarter.
Component shortages, however, could cause one particular GPU, the rumored GeForce RTX 3070 Ti 16GB, to miss its January launch.
Editors' Recommendations
