TECH NEWS – Nvidia has launched a new Ada Lovelace architecture graphics card, but not under the GeForce branding!
Under the Turing (GeForce RTX 2000 and GTX 1600 series), Nvidia renamed its Quadro line of professional cards to RTX, and it can still be misleading to use the name RTX 2000, for example. Only here, the name doesn’t refer to the architecture from a few years ago, but to the latest one! According to Nvidia, the RTX 2000 is the entry level for “prosumer” users.
With its compact design, the RTX 2000 with 16 GB of VRAM and 2816 CUDA cores on a 128-bit bus can handle design and visualization processes accelerated by artificial intelligence. Its 22 third-generation RT cores deliver up to 1.7x ray tracing performance (27.7 TFLOPs), while its 88 fourth-generation Tensor cores deliver up to 1.8x AI performance (191.9 TFLOPs). The FP32 CUDA cores (12 TFLOPs) have increased their performance by up to 1.5 times, while their power consumption remains amazingly low (70W!). It’s so power efficient that it doesn’t need a separate PCIe power connector, so it gets enough power through the PCIe slot (max. 75W) and its performance is about twice as high as the previous generation!
The RTX 2000 is also equipped with DLSS 3, so the frame-generating technology results in a more stable frame rate when you’re using the card for gaming. The eighth-generation Nvidia Encoder (NVENC) now supports AV1, which is 40% more efficient than H264, so you’ll get a much more stable broadcast when streaming on Twitch, for example. The dual-slot card, which features Video Super Resolution (video upscaling during playback) and TrueHDR, has 4 MiniDisplayPort 1.4a. With TensorRT-LLM support, you get a rich AI experience in many large language models (LLM).
The reference model of the Nvidia RTX 2000 will be available for $625 (expected), which is a really good price for a workstation graphics card. The RTX 4000 SFF Ada Generation, the world’s fastest low-profile graphics card, is twice as expensive at $1250!
Leave a Reply