In a somewhat tricksy move today, AMD hosted a press conference a couple of miles from nVidia’s active GTC event going on down the road. In yesterday’s keynote by nVidia CEO Jen-Hsun Huang, we saw the introduction of the new Titan Z video card, Pascal architecture, machine learning, and other upcoming GPU technologies. Now, less than 24 hours later, AMD has invited us by to look at their new high-end workstation solution – the W9100 FirePro GPU.
The presentation was pretty quick compared to what we got with nVidia, but the primary focus was on computationally-intensive OpenCL tasks, real-time color correction and editing playback in full 4K resolution, and “enabling content creation.”
Let’s start with the obvious.
AMD W9100 FirePro GPU Specs
Many marketing adjectives and movies aside (“incredible” seems to be the trend this year, if we learned from ASUS at CES), it all boils-down to specifications.
- Hawaii-based.
- 16GB GDDR5 RAM.
- 5TFLOPs single-precision performance.
- 2TFLOPs double-precision performance.
- 320GB/s memory bandwidth.
- DP1.2 support up to 6x4K UltraHD displays.
Unfortunately, we don’t presently have precise GPU specifications or MSRP at this time, but that should get you started; I’m told we’ll have more information in the first week of April.
The FirePro card was branded as “the first GPU to exceed 2TFLOPs of double-precision compute performance.” To reiterate, this isn’t a gaming-targeted card. The W9100 is most at home when used with Adobe software, like the video editing Premiere program, 3D modeling and rendering solutions (Maya), and CAE tools. Putting all of this into perspective, ideal use-case scenarios for the new FirePro card (as presented by AMD) would include fast physics solvers in Maya’s Bullet software, reduced render time for simulations, reduced encoding time for video rendering, and improved multi-GPU performance.
AMD sees a shift from HPC toward workstation use going forward and is adjusting its business model appropriately. The company defined its “growth opportunities” as existing in professional graphics (that’d be the FirePro), ultra-low power client, and embedded and dense server solutions.
General Education: We’ve Moved to GPU-Accelerated Workflows
Regardless of what AMD is doing, the industry has moved wholeheartedly toward GPU-accelerated pipelines in all aspects of encoding, simulation, and rendering software. Adobe’s new Creative suite leverages OpenCL compute and GPUs to significantly decrease time spent encoding, Autodesk has done the same to reduce render time in Maya and other tools, just as two examples. These eventually feed into the media we consume – it means we get movies and games with higher visual fidelity, so although gamers aren’t sticking a FirePro W9100 in their systems any time soon, they will benefit from the advancements. It is also worth noting that our leading CPU and GPU companies produce for high-end consumption first, often Enterprise, and then use Enterprise as a means to reduce cost of production for slimmed-down consumer versions of the products.
AMD Presentation Key Focuses & W9100 Benchmark vs. K6000, K5000
Again, the big drive here was enabling OpenCL tasks. Here are some internal benchmarks AMD ran between the W9100 and existing nVidia Quadro FX K6000 and K5000 solutions. Grain of salt recommended:
In perhaps more immediately relevant and less adjective-saturated news, we’ve got a large feature on nVidia SDKs publishing later today that will interest a lot of the PC gaming enthusiasts reading.
- Steve "Lelldorianx" Burke.