NVIDIA Launching RTX A400 and RTX A1000: Single-Slot 50W Workstation GPUs
New Cards, Familiar GPU
NVIDIA has announced a pair of workstation graphics cards – the RTX A400 and RTX A1000 – and while the cards are new they are based on the previous-gen Ampere architecture. And while Ampere isn’t as exciting as Ada, these are single-slot – and slot-powered – cards which consume 50 watts, max.
When the RTX 2000 Ada Generation launched I commented on how confusing it can be to have concurrent Ada and Ampere product lines with similar product names, but we all know that the only “A” that matters this year is the one in AI (and of course these are being marketed as AI integration products).
Features include (via NVIDIA):
Second-generation RT Cores: Real-time ray tracing, photorealistic, physically based rendering and visualization for all professional workflows, including architectural drafting, 3D design and content creation, where accurate lighting and shadow simulations can greatly enhance the quality of work.
Third-generation Tensor Cores: Accelerates AI-augmented tools and applications such as generative AI, image rendering denoising and deep learning super sampling to improve image generation speed and quality.
Ampere architecture-based CUDA cores: Up to 2x the single-precision floating point throughput of the previous generation for significant speedups in graphics and compute workloads.
4GB or 8GB of GPU memory: 4GB of GPU memory with the A400 GPU and 8GB with the A1000 GPU accommodate a range of professional needs, from basic graphic design and photo editing to more demanding 3D modeling with textures or high-resolution editing and data analyses. The GPUs also feature increased memory bandwidth over the previous generation for quicker data processing and smoother handling of larger datasets and scenes.
Encode and decode engines: With seventh-generation encode (NVENC) and fifth-generation decode (NVDEC) engines, the GPUs offer efficient video processing to support high-resolution video editing, streaming and playback with ultra-low latency. Inclusion of AV1 decode enables higher efficiency and smoother playback of more video formats.
Both of these new cards are based on GA107, which is an 8nm chip from 2021. The configuration in the A1000 contains 2304 CUDA cores, with a dramatic cut to just 768 CUDA cores in the A400. Memory bandwidth will also be a major differentiator, as the RTX A400 has a narrow 64-bit width for its GDDR6, and the RTX A1000 offers a 128-bit interface.
We should also mention display support, as the RTX A400 and RTX A1000 boast four DisplayPort outputs apiece – though if monitor support was the primary objective then a product like the Matrox LUMA A380 is quite competitive (the Matrox card offers 4x full-size DisplayPort 2.1, while the NVIDIA cards are equipped with 4x mDP 1.4a outputs, instead).
NVIDIA partner PNY has product pages for the RTX A400 and RTX A1000 up already, and channel availability is set for May. The full NVIDIA blog post is here.
