facebook rss twitter

Nvidia launches the RTX A2000 low-profile pro graphics card

by Mark Tyson on 11 August 2021, 13:11


Quick Link: HEXUS.net/qaeqx6

Add to My Vault: x

Nvidia has launched the compact, power-efficient RTX A2000 professional graphics card at SIGGRAPH 2021. In the main photo below you can get an idea of how compact the new design is, on a table next to the likes of the Nvidia RTX A6000, A5000, and A4000 (hmmm, there's a gap left for an A3000).

The Nvidia RTX A2000 is based upon an Ampere GA106 GPU with a few SMs disabled, and a relatively low clock speed helping this card come in at 70W, just under the PCIe power limit meaning you don't need a seperate power connector. Being a smaller GPU running at this pace also facilitates the dual-slot low profile form factor that has been used (measurements are 68.6mm tall and 167.6mm long, by 2 slots wide).

Versions of Nvidia's GA106 GPU are also used by the GeForce RTX 3060/Ti/LHR and Mobile models, and the RTX 3050Ti. In this professional example we see 3,328 CUs enabled, with 26 RT cores, and 104 Tensor cores. Nvidia claims it is capable of 8.0 TFLOPS of single-precision performance, 15.6 TFLOPS of RT Core performance, and offers 63.9 TFLOPS of Tensor performance. You can see how this stacks up against its Nvidia brethren here, but for quick reference the next model up, the A4000, is capable of 19.2 TFLOPS of single-precision performance – leaving quite a gap – which in turn is half as fast as the RTX A6000 by this metric (gap filled by A5000).

For memory, Nvidia has paired the GA106 with 6GB of GDDR6 (EEC) on a 192-bit interface, offering 288GB/s of memory bandwidth. I mentioned the board dimensions above (68.6mm tall and 167.6mm long, by 2 slots wide) but some other physical qualities you need to know are the ports (4x mini DP 1.4a), and you can see the shroud has a single active blower fan for cooling.

Remember, Nvidia RTX AXXXX solutions are certified with a broad range of professional applications, tested by leading independent software vendors (ISVs) and workstation manufacturers, and backed by global support specialists. This particular solution is built for everyday workflows, "so professionals can develop photorealistic renderings, build physically accurate simulations and use AI-accelerated tools".

Nvidia says that the A2000 pro graphics card will be available on its own for US$450, and in systems from the likes of Asus, Lenovo, HP and Dell, starting from October.

HEXUS Forums :: 5 Comments

Login with Forum Account

Don't have an account? Register today!
That would look great for my Velka 5!
Not now, Nvidia!

We'd just convinced the beancounters we all needed A4000's for Solidworks 2021.
Our current 5-year-old Dells don't have extra power cables inside, so we were waiting on all new desktops for the 3DCAD dept.
Not now, Nvidia!

We'd just convinced the beancounters we all needed A4000's for Solidworks 2021.

Fill the finance department fridges with lots of beer. You'll find the beans being counted can go awry quite nicely!
They don't even bother listing double precision performance now.
Well, While I'm sure it will be very good for CAD Drawings, and AI software. I'm also certain it's NOT meant for heavy photorealitic rendering workloads… As a 3D Artist that works mainly developing Architectural visualisations, I know that raytracing software like V-Ray can easily use upwards of 4GB of VRAM with only a Small bathroom scene, cinsidering the 3D Software viewport by itself can use easily over 1GB of VRAM, its easy to see how 6GB of VRAM is extremely limiting for actual Photorealistic rendering. Unless you're just rendering a simgle product, like a Watch, or a Phone, then it will be a great card.