Consequently, the Tensor core count has also decreased from 568 to 440. Officially, the card delivers 69.3 TFLOPS of single-precision compute performance, which is 24% slower than the RTX 6000 ADA. Despite the lower specifications, the RTX 5880 ADA maintains a board power of 285W, which is 15W less than its counterpart, the 6000 ADA. It continues to offer video encoding and decoding support and retains the dual-slot design with a blower-style cooler. The memory configuration remains consistent, boasting 48 GB of GDDR6 memory across a 384-bit interface, delivering a maximum bandwidth of 960 GB/s.
NVIDIA has officially introduced the RTX 5880 ADA on its website, indicating that orders are now being accepted, and the deployment of this GPU can be expected in the near future. It's important to note that NVIDIA's launch is not exclusive to the Chinese market; the product is featured on the global website, making it available to the worldwide professional market.
Here is a technical comparison of NVIDIA's RTX ADA workstation GPUs:
RTX 6000 ADA:
- GPU: AD102
- CUDA Cores: 18,176
- Tensor Cores: 568
- Boost Clock: 2.5 GHz
- Memory: 48GB G6 ECC
- Memory Bus: 384-bit
- Max Bandwidth: 960 GB/s
- FP32 Compute: 91.1 TF
- TDP: 300W
- MSRP: $6,800
RTX 5880 ADA:
- GPU: AD102
- CUDA Cores: 14,080
- Tensor Cores: 440
- Boost Clock: 2.5 GHz
- Memory: 48GB G6 ECC
- Memory Bus: 384-bit
- Max Bandwidth: 960 GB/s
- FP32 Compute: 69.3 TF
- TDP: 285W
- MSRP: To Be Confirmed