Value and Conclusion

When the Radeon Pro Duo goes on sale by April 26, we expect it to be standard equipment for some of the most expensively branded gaming desktops, and for certain performance leaderboard climbers to pick some of the AIB-branded cards up to get their fix. Beyond that, other consumers, gamers, including those vying for 4K gaming nirvana, have much to think about.

Our standing complaint with AMD CrossFire still applies with the Radeon Pro Duo. It's a dual-GPU card each application sees as two GPUs. If it's not optimized for CrossFire, or lacks support for multi-GPU in general (eg: "JustCause 3"), you will have paid close to three times as much as for your R9 Fury X. What's more, being a multi-GPU card, most of your applications see 4 GB of usable video-memory. That's a bit too little for a $1500 card. AMD has not done anything substantive to change its reputation of being slower to the table with multi-GPU game optimizations released through driver updates.

There's a reason NVIDIA didn't bother with a dual-GPU graphics card based on its GM200 silicon - the next-generation is on the horizon. NVIDIA's "Maxwell" architecture, driving the GeForce GTX 900 series, lived out 18 months on the market, and the company is reportedly making efforts to launch its first next-generation GPUs a little over a month from now. Even though the rumored GP104 silicon doesn't succeed the GM200, and more logically, the GM204, keeping up with past trends, it could outperform current GM200-based single GPU cards. This could greatly diminish the value of a dual-GM200 product. The forgettable Kepler-based GTX TITAN Z quietly depreciated 50% with the onset of "Maxwell," for instance.

The Radeon Pro Duo is threatened with a similar marketing problem. It's really cool today, but the lure of next-generation "Pascal" and "Polaris" GPUs being literally around the corner could keep buyers away. Dual-GPU cards, if launched early enough in the life cycle of a GPU's architecture, could present buyers with genuinely good future-proofing options. One can still find R9 295X2 users getting lucky with playable 4K Ultra HD performance. As if "Fiji" and the R9 Fury X weren't late enough, the Pro Duo is grossly mistimed. AMD's 2015 reveal of the "Fiji" GPU by company CEO Lisa Su could have been a better time to come up with this card. Even a month after the R9 Fury X's launch could have been fine.

The other dimension of the marketing problem, and one AMD learned absolutely nothing from with its Radeon R9 Nano launch drama, is putting the card in the hands of reviewers with hundreds of data points. There's a reason why our test-bench consists of such a vast diversity of games and game-engines. It gives you an idea of what kind of games aren't ideal for the graphics card being tested.

We don't have a card yet, and we aren't getting one. Almost as if a repeat of the R9 Nano PR trainwreck, there is a surreal celebration of cluelessness among AMD and its marketing partners responsible for Germany. Last week, we asked AMD's German PR agency on what the plan was regarding sampling and scheduling and received a "no idea on timing." We also checked with AMD, and they didn't know either, giving us a "will keep you updated" response. We've been reviewing Radeons for over a decade now, and if AMD does not want something to be reviewed, a "no" goes over better than flat-out denying the very existence of an upcoming product's launch (one that has already seen a paper-launch).

At its press-briefing for the Radeon Pro Duo, and even briefly at its "Capsaicin" event, AMD saw the need to tease its future "Polaris" and "Vega" architectures and upcoming GPU product stacks. AMD itself wants you to hope for the next-generation. This is hype more than a tease on AMD's part.