At Siggraph 2023, Nvidia introduced a range of workstation graphics cards and a server-grade board, all powered by the Ada Lovelace GPU. The lineup includes RTX 4000 20GB, RTX 4500 24GB, and RTX 5000 32GB cards, catering to demanding tasks like CAD and content creation. Nvidia’s L40S Ada data center card, featuring the AD102 GPU with 18,176 CUDA cores and 91.6 FP32 TFLOPS, is aimed at graphics AI and video processing. These solutions will be incorporated into workstations by companies such as Boxx, Dell, HP, Lenovo, and Lambda. The OVX servers with L40S GPUs are designed to accelerate AI, graphics, and video processing workloads.
What do we think?
Nvidia has had a bit of an identity crisis in its professional graphics AIB line. We can’t forget the Quadro brand, and the Pro designator, but in late 2020, the company decided it would drop the Quadro brand it had invested in for over 10 years and brand its newest WS AIBs as RTX followed by the GPU series, such as an RTX A6000 with the A to identify it as an Ada GPU. But with the latest WS AIBs, the GPU letter got dropped, too. But don’t be too sad because if you go to Nvidia’s Web page you can still find Quadro AIBs along with RTX and now the new RTX 4000, 4500, and 5000.
Nvidia RTX 4000 Ada Generation graphics card
During Siggraph 2023, in beautiful downtown LA, Nvidia unveiled a trio of workstation AIBs along with a server-grade graphics board, all based on the Ada Lovelace GPU. The lineup included the RTX 4000 20GB, RTX 4500 24GB, and RTX 5000 32GB graphics cards, aimed at demanding applications like CAD and content creation, as well as the L40S data center AIB with 48GB of memory. Several workstation suppliers announced rack-mount workstations with the L40S.
The RTX 4000 20GB uses the AD104 GPU with 6,144 CUDA cores, delivering up to 26.7 FP32 TFLOPS. The RTX 4500 24GB, powered by the AD104 GPU with 7,680 CUDA cores, provides up to 39.6 FP32 TFLOPS and is set to launch in October. The RTX 5000, with 32GB, utilizes the AD102 GPU with 12,800 CUDA cores, offering 65.3 FP32 TFLOPS.
These graphics solutions will be featured in upcoming workstations from companies like Boxx, Dell, HP, Lenovo, and Lambda.
Nvidia’s L40S Ada data center card, equipped with the AD102 GPU featuring 18,176 CUDA cores and 91.6 FP32 TFLOPS, will be initially used in Nvidia’s OVX servers designed for graphics AI and video processing. One of the target applications for the rack-mount data center workstations is remote work requiring this level of processing power (see “Boxx racks up AMD and Nvidia in a P1G”).
“As generative AI transforms every industry, enterprises are increasingly seeking large-scale compute resources in the data center,” said Bob Pette, vice president of professional visualization at Nvidia. “[Nvidia server] OVX systems with Nvidia L40S GPUs accelerate AI, graphics, and video processing workloads, and meet the demanding performance requirements of an ever-increasing set of complex and diverse applications.”
| RTX 6000 | RTX 5000 NEW | RTX 4500 NEW | RTX 4000 NEW | RTX 4000 SFF | |
| GPU Memory | 48GB GDDR6 with ECC | 32GB GDDR6 with ECC | 24GB GDDR6 with ECC | 20GB GDDR6 with ECC | 20GB GDDR6 with ECC |
| Display Ports | 4× DisplayPort 1.4a | 4× DisplayPort 1.4a | 4× DisplayPort 1.4a | 4× DisplayPort 1.4a | 4× Mini DisplayPort 1.4a |
| Max Power Consumption | 300W | 250W | 210W | 130W | 70W |
| Form Factor | 4.4” (H) × 10.5” (L) dual slot | 4.4” (H) × 10.5” (L) dual slot | 4.4” (H) × 10.5” (L) dual slot | 4.4” (H) × 9.5” (L) single slot | 2.7” (H) × 6.6” (L) dual slot |
| Thermal | Active | Active | Active | Active | Active |