Nvidia has kicked off its latest GPU Technology Conference, with a keynote from Jensen Huang's kitchen, again. There are lots of announcements today with the Nvidia news blog bursting with 21 news releases at the time of writing. As we have grown accustomed, lots of the news is about AI, cloud, robotics and supercomputing. However there is some hot news too about the latest pro visualization graphics cards from the green team, aimed at artists, designers, and engineers.
The latest pro visualization graphics cards from Nvidia are the Nvidia RTX A6000 and Nvidia A40. Yes, the Quadro branding appears to have been consigned to history with these first pro cards furnished with Ampere GPUs. Remember, we don't have a Titan branded Ampere card on the consumer side either.
Before going further, I might as well make clear that the Nvidia RTX A6000 and A40 have basically the same spec but the A40 is a headless (3x DP 1.4 option) two slot passive version of the product for high density server use and it has a slightly slower memory clock. Check out the specs summary below:
- GPU: Ampere GA102, Samsung 8nm
- CUDA cores: 10,752
- Tensor cores: 336
- RT Cores: 84
- Memory: 48GB
- Memory clock/bus: 16Gbps GDDR6/384-bit (A40 is 14.5Gbps GDDR6/384-bit)
- TDP: 300W
- Available: mid-December (A40 arrives Q1 2021)
The key attractions of the new Nvidia RTX A6000 and A40 are the performance benefits to professionals. Nvidia's news blog contains performance testimonials from visual effects companies and games developers, as well as the likes of Groupe Renault. In brief, Nvidia boasts that the second gen RT cores are up to twice as fast, third gen Tensor cores are up to five times faster, and the new CUDA cores deliver up to 2x the FP32 throughput compared to the previous generation. Add to that the 48GB of memory available (or 96GB using two GPUs and NVLink), Nvidia Virtual Workstation, and PCIe Gen 4 and Nvidia thinks it has a compelling offering.
Nvidia RTX A6000 based workstations from the likes of Dell, HP and Lenovo will be available early next year but channel partners like PNY, Leadtek, and Ingram Micro will be selling the cards in December (as well as at Nvidia.com). Those interested in the Nvidia A40 wil have to wait for servers from the likes of Cisco, Dell, Fujitsu, Hewlett Packard Enterprise and Lenovo to arrive early next year.
Nvidia and Arm
Like I mentioned in the intro, Nvidia has made a plethora of announcements today. Huang's keynote mentioned some major initiatives to advance Arm computing. In particular he highlighted that some currently x86 exclusive capabilities will become available to Arm users:
- First, Nvidia will complement Arm partners with GPU, networking, storage and security technologies to create complete accelerated platforms.
- Second, Nvidia is working with Arm partners to create platforms for HPC, cloud, edge and PC — this requires chips, systems and system software.
- And third, Nvidia is porting the Nvidia AI and Nvidia RTX engines to Arm.
