NVIDIA announced the introduction of cards RTX A5000 (desktop/laptop), A4000 (desktop/laptop), A3000 (laptop) and A2000 (laptop) for stationary and mobile workstations. The new layouts are aimed at professionals, content creators, and anyone looking for top performance at work, at home and on the go. NVIDIA RTX A5000 is equipped with a partially locked GA102 GPU with 8192 CUDA cores and 64 Tensor cores. The GA102 is available here in both a stationary and mobile version. The chip is equipped with 24 GB of GDDR6 memory with a 384-bit bus interface. The card is actively cooled with 230W TDP and offers 27.8 TFLOP FP32, 54.2 TFLOP RT and 222.2 TFLOP of Tensor Core performance. NVIDIA RTX A4000 is a full GA106 chip with 6144 CUDA cores, 48 Tensor cores, 16 GB of GDDR6 memory with 256 bit bus and 140W TDP. This model is also actively cooled. The card offers 19.2 TFLOP FP32, 37.4 TFLOP RT and 153.4 TFLOP performance of Tensor cores. Cards are equipped with four DP1.4 outputs, Quadro Sync technology.
NVIDIA has announced the introduction of RTX A5000, A4000, A3000 and A2000 cards for desktop and mobile workstations.
The previously presented system is the most powerful of the family NVIDIA RTX A6000 which uses the full GA102 GPU with 10752 CUDA cores, 336 Tensor cores and 84 RT units for hardware ray tracing acceleration. On the other hand, systems intended for mobile devices RTX A3000 and RTX A2000 are quite limited GA104 and GA106 cores. The new NVIDIA RTX laptop GPUs will be available on mobile workstations in the second quarter of this year. And the new NVIDIA RTX desktop GPUs and NVIDIA data center GPUs will be available from global distribution partners and OEMs this month.
NVIDIA RTX Ampere Mobile Graphics Processors for Workstations
|Graphics Card||RTX A2000||RTX A3000||RTX A4000||RTX A5000|
|Transistors||12 billion||17.4 billion||17.4 billion||28 billion|
|Process||8 nm||8 nm||8 nm||8 nm|
|Size||276 mm²||392,5 mm²||392,5 mm²||628,4 mm²|
|ALU FP32 units||2560||4096||5120||6144|
|ALU INT32 units||1280||2048||2560||3072|
|FP32 performance||9.3 TFLOPS||12,8 TFLOPS||17,8 TFLOPS||21,7 TFLOPS|
|RT performance||18.2 TFLOPS||25,0 TFLOPS||34,8 TFLOPS||75,6 TFLOPS|
|Tensor performance||74,7 TFLOPS||102,2 TFLOPS||142,5 TFLOPS||174,0 TFLOPS|
|Memory capacity||4 GB||6 GB||8 GB||16 giga bytes|
|Memory interface||128 bits||192 bits||256 bits||256 bits|
|Memory bandwidth||192 GB/s||264 GB/s||384 GB/s||448 GB/s|
|TDP||35 - 95 W.||60 - 130 W.||80 - 140 W.||80-165 W.|
All NVIDIA RTX GPUs support the latest technologies in the NVIDIA Ampere Architecture:
- Second Generation RT Cores: Up to 2x the bandwidth of the previous generation, with the ability to perform ray tracing, shading and noise reduction tasks simultaneously.
- 3rd Generation Tensor Cores: Up to 2x the bandwidth of the previous generation, up to 10x on a rarity, with support for the new TF32 and BFloat16 data formats.
- CUDA Cores: Up to 2.5 times the FP32 throughput compared to the previous generation, resulting in a significant increase in workloads.
Laptop GPU features and technology include:
- 3rd Generation Max-Q Technology: For thin and light laptops that run quieter and more efficiently with Dynamic Boost 2.0, WhisperMode 2.0, Resizable BAR and NVIDIA DLSS technology.
- Up to 16 GB GPU memory: for the largest models, scenes, assemblies, and advanced multi-application workflows.
NVIDIA Ampere Desktop Graphics Cards for Workstations:
|Graphics Card||NVIDIA RTX A4000||NVIDIA RTX A5000||NVIDIA RTX A6000|
|GPU||Ampere GA106||Ampere GA102||Ampere GA102|
|Process GPU||Samsung 8 nm||Samsung 8 nm||Samsung 8 nm|
|Matrix size||276 mm²||628mm²||628mm²|
|Single precision||TBC||TBC||40 TFLOP|
|VRAM||16 GB of GDDR6 memory||24 GB GDDR6||48 GB GDDR6|
|NVLINK VRAM||TBC||TBC||96 GB GDDR6|
|Memory bandwidth||448 GB/s||768 GB/s||768 GB/s|
Source: ITHardware.pl by ithardware.pl.
The article has been translated based on the content of ITHardware.pl by ithardware.pl.
We stand by the rights of the original author of the post, no matter what. We always respect and prioritize the copyright of the content and always include the original link of the source article. If the author of the original article has any issue with it, just leave a report below, we’ll edit it or delete it. Whatever it takes. We will make it right as quickly as possible to protect the rights of the author.
Thank you very much! Best regards!