At the GPU Technology Conference (GTC) 2023, NVIDIA has updated its line of workstation GPUs, including the new RTX 4000 SFF workstation GPU as well as five new mobile workstation GPUs, covering the entire market segment. The new RTX 4000 SFF is a small form-factor professional GPU packed in a compact design, while the new mobile workstation GPU lineup will include a total of five new RTX Ada Generation GPUs.
In addition to the previously available NVIDIA RTX 6000 Ada Generation graphics solution, which NVIDIA targets at the "design and simulation" market, the lineup now includes the NVIDIA RTX 4000 SFF, in addition to other, previously available NVIDIA RTX A-series GPUs.
In case you missed it earlier, the RTX 6000 Ada Generation is the current flagship, packing 18,176 CUDA cores, 142 3rd generation RT cores, and 568 4th generation Tensor Cores. It packs 48 GB of ECC GDDR6 memory on a 384-bit memory interface. This leaves it with 960 GB/s of memory bandwidth and the TDP is set at 300 W. All of this is enough for 91.1 TFLOPs of single-precision and 1457.0 TFLOPs of Tensor-based compute performance. It has four DisplayPort 1.4a connectors, three encode/decode engines, as well as brings support for NVIDIA GPUDirect for Video support, GPUDirect Remote Direct Memory Access (RDMA) support, virtual GPU (vGPU) software support, Quadro Sync II compatibility, RTX Experience, RTX Desktop Manager software, RTX IO support, and NVIDIA Mosaic technology.
The new RTX 4000 SFF Ada Generation is obviously a cut-down version, and, as spotted by Videocardz.com, appears to be based on an AD104 GPU. It features 6,144 CUDA cores, 48 3rd generation RT cores, and 192 4th generation Tensor Cores. It comes with 20 GB of GDDR6 memory with error-correction code (ECC), offers four mini DisplayPort 1.4a outputs, and comes with two out of three encode/decode engines. This is a significant cut-down but not surprising considering its low-profile form-factor and 70 W total board power (TBP). It still offers up to 19.2 TFLOPs of single-precision and 306.8 TFLOPs of Tensor-based compute performance. It also features most, if not all NVIDIA workstation-oriented features.
In addition to the new RTX 4000 SFF Ada Generation workstation desktop GPU, NVIDIA has also updated its entire line of mobile workstation GPUs, introducing the RTX 5000, RTX 4000, RTX 3500, RTX 3000 and the RTX 2000 Ada Generation. These pretty much line up with all the consumer desktop graphics cards, but with a few exceptions, including the RTX 3500 with 5,120 CUDA cores.
The TDP range from 35 W to 140 W on entry-level, 60 W to 140 W on mid-range, and 80 W to 175 W for the RTX 5000 Ada Generation. All feature ECC memory, except for the RTX 2000 Ada Generation, and will offer a significant increase in both single-precision and Tensor-based compute performance compared to the previously available NVIDIA RTX A-series workstation mobile GPUs.
According to NVIDIA, the RTX 4000 SFF Ada Generation workstation desktop GPU should be available soon with price set at $1250, while there is no word on when we will see workstation laptops with those new RTX Ada Generation mobile GPUs.
View at TechPowerUp Main Site | Source
In addition to the previously available NVIDIA RTX 6000 Ada Generation graphics solution, which NVIDIA targets at the "design and simulation" market, the lineup now includes the NVIDIA RTX 4000 SFF, in addition to other, previously available NVIDIA RTX A-series GPUs.
In case you missed it earlier, the RTX 6000 Ada Generation is the current flagship, packing 18,176 CUDA cores, 142 3rd generation RT cores, and 568 4th generation Tensor Cores. It packs 48 GB of ECC GDDR6 memory on a 384-bit memory interface. This leaves it with 960 GB/s of memory bandwidth and the TDP is set at 300 W. All of this is enough for 91.1 TFLOPs of single-precision and 1457.0 TFLOPs of Tensor-based compute performance. It has four DisplayPort 1.4a connectors, three encode/decode engines, as well as brings support for NVIDIA GPUDirect for Video support, GPUDirect Remote Direct Memory Access (RDMA) support, virtual GPU (vGPU) software support, Quadro Sync II compatibility, RTX Experience, RTX Desktop Manager software, RTX IO support, and NVIDIA Mosaic technology.
The new RTX 4000 SFF Ada Generation is obviously a cut-down version, and, as spotted by Videocardz.com, appears to be based on an AD104 GPU. It features 6,144 CUDA cores, 48 3rd generation RT cores, and 192 4th generation Tensor Cores. It comes with 20 GB of GDDR6 memory with error-correction code (ECC), offers four mini DisplayPort 1.4a outputs, and comes with two out of three encode/decode engines. This is a significant cut-down but not surprising considering its low-profile form-factor and 70 W total board power (TBP). It still offers up to 19.2 TFLOPs of single-precision and 306.8 TFLOPs of Tensor-based compute performance. It also features most, if not all NVIDIA workstation-oriented features.
In addition to the new RTX 4000 SFF Ada Generation workstation desktop GPU, NVIDIA has also updated its entire line of mobile workstation GPUs, introducing the RTX 5000, RTX 4000, RTX 3500, RTX 3000 and the RTX 2000 Ada Generation. These pretty much line up with all the consumer desktop graphics cards, but with a few exceptions, including the RTX 3500 with 5,120 CUDA cores.
The TDP range from 35 W to 140 W on entry-level, 60 W to 140 W on mid-range, and 80 W to 175 W for the RTX 5000 Ada Generation. All feature ECC memory, except for the RTX 2000 Ada Generation, and will offer a significant increase in both single-precision and Tensor-based compute performance compared to the previously available NVIDIA RTX A-series workstation mobile GPUs.
According to NVIDIA, the RTX 4000 SFF Ada Generation workstation desktop GPU should be available soon with price set at $1250, while there is no word on when we will see workstation laptops with those new RTX Ada Generation mobile GPUs.
View at TechPowerUp Main Site | Source
