NVIDIA Launches 5 RTX Ada Lovelace Workstation Mobile GPUs Alongside the RTX 4000 SFF, Promised to Offer 2x Efficiency Over Last-Gen

Jensen Huang at the recently held GTC Keynote unveiled the Ada Lovelace mobile workstation series. We have already seen a few glimpses of these GPUs beforehand, but NVIDIA has finally made the official announcement. Better yet, team green has also unveiled something special for the desktop market, the RTX 4000 SFF.
Meet the RTX Ada Lovelace Mobile Workstation Lineup
The RTX Ada Lovelace workstation (mobile) lineup consists of the RTX 5000, RTX 4000, RTX 3000 and the RTX 2000. These GPUs are tipped to be 2x more efficient than last-gen which is a really impressive feat. Ada Lovelace incorporates special hardware-accelerated features to provide the necessary computational power for powering collaborative content creation, design and AI workflows.
“Running data-intensive applications like generative AI and real-time digital twins in the metaverse requires advanced computing power,” said Bob Pette, vice president of professional visualization at NVIDIA. “These new NVIDIA RTX GPUs provide the horsepower needed for creators, designers and engineers to accomplish their work from wherever they’re needed.”
NVIDIA
GPU Name | Chip | FP32 Count | Memory | Memory Spec | Bus Width |
RTX 5000 Ada | AD103 | 9728 | 16GB | GDDR6 | 256-bit |
RTX 4000 Ada | AD104 | 7424 | 12GB | GDDR6 | 192-bit |
RTX 3500 Ada | AD104 | 5120 | 12GB | GDDR6 | 192-bit |
RTX 3000 Ada | AD106 | 4608 | 8GB | GDDR6 | 128-bit |
RTX 2000 Ada | AD107 | 3072 | 4/8GB | GDDR6 | 128-bit |
RTX 4000 SFF
Next up, we have, not the RTX 4070, but the RTX 4000 SFF (Small Form Factor) for the desktop market. Let us make it clear that this GPU is not meant for gaming. In fact, this GPU is designed for mini-computers allowing compact and small yet capable workstations.
In terms of specifications, this GPU has 20GB of GDDR6 memory over a 160-bit memory bus translating to 320GB/s of bandwidth. The RTX 4000 SFF is probably based on the AD104 GPU given the CUDA count. The single-precision compute performance stands at 19.2 TFLOPS, similar to the RTX 3070. The interesting part is that all this performance comes in at just a 70W TBP.
As for the price, NVIDIA is asking consumers upwards of $1250 for the RTX 4000 SFF.

Source: NVIDIA