facebook rss twitter

AMD launches FirePro W9100 32GB workstation graphics card

by Mark Tyson on 15 April 2016, 09:31

Tags: AMD (NYSE:AMD)

Quick Link: HEXUS.net/qacz7b

Add to My Vault: x

AMD has announced an updated FirePro W9100 workstation graphics card at the 2016 National Association of Broadcasters (NAB) Show in Las Vegas. With the new FirePro W9100 32GB, AMD recaptures the 'industry leading memory support' crown. Less than a month ago Nvidia unleashed its Quadro M6000, updated with 24GB GDDR5, beating the FirePro W9100 16GB workstation graphics card in support for large asset workflows.

Raja Koduri, senior vice-president and chief architect, Radeon Technologies Group (RTG), AMD talked about the new FirePro fitting in with AMD's twin pronged professional graphics card strategy, alongside the Radeon Pro Duo. "Some professional creative workflows demand very high bandwidth memory subsystems while others demand high memory size," said Koduri, referring to the two graphics cards at the head of AMD's workstation stable.

At NAB, AMD is showcasing the FirePro W9100 32GB running industry-leading applications from the likes of Adobe, Apple, Autodesk, Avid, Blackmagic Design, and Rhino, in the StudioXperience Technology Zones.

As a reminder of the capabilities of the AMD FirePro W9100, it offers 5.24 TFLOPS of peak single-precision floating-point performance. Now buyers have a choice of 16GB or 32GB GDDR5 configurations. The card sports six mini DisplayPort outputs and support for DisplayPort 1.2a, and is capable of driving up to six 4K displays. AMD's FirePro W9100 supports OpenCL 2.0, DirectGMA and SDI, Framelock and Genlock. Bullet pointed hardware specs are as below:

  • Memory: 32GB or 16GB GDDR5 GPU memory, 512-bit memory interface, 320GB/s memory bandwidth
  • Compute Performance: 2,816 stream processors (44 compute units), 5.24 TFLOPS peak single-precision floating-point performance, 2.62 TFLOPS peak dual-precision floating-point performance
  • Display Outputs: Six mini DisplayPort 1.2a outputs, Maximum DisplayPort 1.2a resolution 4096x2160, Maximum DisplayPort 1.1 resolution 2560x1600
  • API/Features/OS Support: DirectX 11.2/12, OpenGL 4.4, OpenCL 2.0, Shader model: 5.0, AMD Eyefinity multidisplay technology support, AMD HD3D Pro support via stereoscopic 3-pin mini DIN3, DirectGMA support, OS Support: Microsoft Windows 10, Windows 8.1, Windows 7 and Linux (32- or 64-bit)
  • Cooling/Power/Form Factor: 275W maximum power consumption, Discreet active-cooling solution, Full-height dual-slot form factor, PCIe 3.0 compliant, x16 bus interface
  • System Requirements: Available PCIe x16 (dual-slot), 3.0 for optimal performance, A power supply, one PCIe AUX power connector (8-pin) and one PCIe AUX power connector (six-pin), 16GB system memory

AMD has yet to inform us about pricing and availability of its new FirePro W9100 32GB workstation graphics card. At the time of writing the 'Shop Now' link on the product page appears to show two FirePro W9100 16GB SKUs priced at $2999 and $4399, but no 32GB model.



HEXUS Forums :: 21 Comments

Login with Forum Account

Don't have an account? Register today!
I am shocked….. AMD re-releases a top spec workstation GPU with double the ram just after Nvidia does. Nah who am I kidding, I was expecting this the moment I read about the quadro update lol
Grrrr! Compute Performance: 2,816 stream processors- the professional range cards don't bring much profit but at-least AMD should have removed the 4gb HD memory from the Fury x and some 32GB of EEC GDDR5 memory on the board, NOW that could be awesome.
lumireleon
… at-least AMD should have removed the 4gb HD memory from the Fury x and some 32GB of EEC GDDR5 memory on the board, NOW that could be awesome.

Impossible - the memory architecture is completely different. You can't just swap stacked HBM on an interposer for a direct connection to GDDR5. And to get anything near the bandwidth they wanted you'd have to add more channels to the memorty controller, which means more motherboard traces, more phsyical memory chips … potentially you'd actually go beyond what's phsyically possible on a sensibly-sized PCIe card. AMD need to make what's both possible and practical. If you need a card with a huge memory buffer for working with large datasets, this is the one for you.

Once cards with HBM2 go into mass production we'll start seeing 32GB frame-buffers with HBM2 for the professional market. You just need to patiently wait for the technology to catch up with your desires ;)
lumireleon
Grrrr! Compute Performance: 2,816 stream processors.

That's still comfortably the highest DP FLOPs we've ever seen isn't it?
kalniel
lumireleon
Grrrr! Compute Performance: 2,816 stream processors.

That's still comfortably the highest DP FLOPs we've ever seen isn't it?

Pretty sure the Tesla P100 beats it
21.2 TeraFLOPS half-precision performance
10.6 TeraFLOPS single-precision performance
5.3 TeraFLOPS double-precision performance

Feel free to correct me if I'm wrong though.