AMD Radeon R9 Fury X Reviews

I'm getting vibes of a "save our CEO from being a total liar to investors" product, in some ways reminiscent of Nvidia's spinning Shield as the current-gen console their tech went into.

I don't know if AMD's marketing is to blame on this one. The final spec of this product didn't match some of the earlier rumors. I think there's a possibility that marketing really was told to sell a less overbuilt Gemini board that wasn't going to be this late or with this TDP, and the hardware couldn't be delivered.

And Polaris might factor into this. A pair of ~40 CU Polaris boards in their rumored power and performance range would have left no room for a dual Fury solution that wasn't jacked beyond viability as a consumer solution. It doesn't sound like this is available in quantities commensurate with the VR market it was supposedly synced to.
 
Charlie's piece is surprisingly forgiving.
http://semiaccurate.com/2016/04/26/amd-finally-really-honestly-launches-the-dual-fury/
Small excerpt:
„Other than that it is just two Fijis plus 2x 4GB HBM stacks on a black PCB. It still looks good but unlike its predecessors it is aimed at a very different market, content creation. AMD is all hot on VR with new acronyms and initiatives every release, we can’t honestly tell you why any of it matters though. This card is aimed at VR creators, not users, hence the prosumer label and the Pro in the Radeon Pro Duo. AMD was up front about this and in a very refreshing change also notified press that there would be no sampling, we consider this to be a good step forward for the company.“

How exactly it is aimed at VR Developers (iow: what makes it aimed at them, apart from the marketing lines), I'm missing to read from this otherwise soinvestigative website. ;)
 
Those VR developers could have just bought 2 Fury X, of course. But the same argument can be used for all dual GPU cards...
OTOH, the timing of this is just completely off. Dual GPUs are supposed to be a filler when new silicon isn't quite ready yet. This thing could have had a useful life when introduced right along Fury X.
 
Anandtech and Techreport noted that FirePro drivers can be installed for the Pro Duo, and there is coverage for a subset of the applications the fully professional lines have support for.
 
Those VR developers could have just bought 2 Fury X, of course. But the same argument can be used for all dual GPU cards...
OTOH, the timing of this is just completely off. Dual GPUs are supposed to be a filler when new silicon isn't quite ready yet. This thing could have had a useful life when introduced right along Fury X.

If you are developping for VR, you dont want 1 gpu, who can be replaced by 2.. you want 4 gpu's who cant be replaced 8.. Same goes for raytracing. ( where i can have some mood as 4-8GB of HBM is a bit low on "home system" when you dont have a network compute available )

The Dual pro / or FirePro duo, have pass extensive test, the hardware is not completey the same of a standard gaming part.. It have need to been compliant with server, pro workstations test, even the watercooling system is compliant with professional tests ( the 50/50/50 )..

The Power consumption of this gpu, is extremely low, really low for a dual core gpu's ... it cant pass 300W and strangely it is able to maintain is core speed in much situations. ( funny to see the TDP report where it consume just a few watts more than a 980TI or a single FuryX (around 10W more )

Each parts are binned extensively for be compliant on the pro market.. because there's absolutely no difference between the dual FirePro and the Radeon dual Pro.

This said, for the gaming market, this gpu is too late and not much indicated: for the Professional market ( who dont need ECC and FP64, as Raytracing and VR developpement ), this gpu is good. No, in this case, it is just incredibly good. 16.Tflops on a single gpu, we have right now, AIB who provide a 5+1 systems with them, 5x 16 Tflops of FP 32.. All of this on a really compact system.

Without modding, you can use both drivers, FirePro and Catalyst one. So drivers are compliant with professional softwares and need pass the QA tests for them.

I think this article of VR World resume it all in some lines: http://vrworld.com/2016/03/22/how-amd-radeon-pro-duo-gpu-went-prosumer/

Due to the public market dominant, i think this card will surely quickly be forgetten.. but honestly this is just a pure monster..
 
Last edited:
I'm was able to find the compute-only Fiji-based FirePro.
The Pro Duo also gets the added bonus of AMD's Typical Board Power, so the 3 8-pin connectors make more sense in that regard.
 
Anandtech and Techreport noted that FirePro drivers can be installed for the Pro Duo, and there is coverage for a subset of the applications the fully professional lines have support for.
And that's really the clencher here. You can use it as a gaming card - far from stopping you, AMD will support it - but AMD is going to flog it as a semi-pro/prosumer card, as they know it wouldn't do well as a gaming card. Instead you have the FirePro driver set for development purposes, and then the Radeon driver set for testing purposes.

IMO it's the best thing they could do with this card given the poor state of AFR. More importantly if it works even half as well as AMD is hoping, then this will be a great tool for getting developers to include Affinity Multi-GPU support (and VR SLI) in their VR games, since developers still need to spend a bit of time enabling that.
 
I'm calling it now: the Polaris introduction will be absolutely flawless. After so many failed attempts, statistics simply dictate that AMD marketing will get things right one way or the other.

Unfortunately, if you cast a fair die a thousand times without getting a six, and you cast it once more, the odds of getting a six are still 1/6.

But one may hope that AMD's marketing people are better learners than dice.
 
And that's really the clencher here. You can use it as a gaming card - far from stopping you, AMD will support it - but AMD is going to flog it as a semi-pro/prosumer card, as they know it wouldn't do well as a gaming card. Instead you have the FirePro driver set for development purposes, and then the Radeon driver set for testing purposes.
How do the FirePro drivers help with development? Last I checked, their main difference was a nicely reddish Catalyst Control Center, 10 bpp color option and optimizations for CAD/CAM/CAE programs. None of this would really help in VR development outside of the professional space (where the 4 GiB limit instead will make matters worse compared to the 32 GiB Hawaii-based Fire Pro).
 
How do the FirePro drivers help with development? Last I checked, their main difference was a nicely reddish Catalyst Control Center, 10 bpp color option and optimizations for CAD/CAM/CAE programs. None of this would really help in VR development outside of the professional space (where the 4 GiB limit instead will make matters worse compared to the 32 GiB Hawaii-based Fire Pro).
Certification. When something goes wrong with your pro software, the developer of the software will actually support you, as opposed to telling you that they can't help you since you're using an unsupported card/driver combination.

People like you and I are used to doing everything on our own. But the real graphics pros are users, not techies, so they want proper technical support.
 
True, but that rather falls under the category CAD/CAM/CAE, not developing general VR applications.
But then, maybe it's yet to be determined, what "the general" VR application will be :)
 
True, but that rather falls under the category CAD/CAM/CAE, not developing general VR applications.
But then, maybe it's yet to be determined, what "the general" VR application will be :)
That's my reaction as well. For typical professional solutions, the amount of memory is on the small side, and I don't think crossfire is commonly used in those applications?

So what does it mean "VR application development"? What kind of applications do they run that makes the Radeon Pro so uniquely qualified? Just the Liquid VR driver?
 
I'm getting vibes of a "save our CEO from being a total liar to investors" product, in some ways reminiscent of Nvidia's spinning Shield as the current-gen console their tech went into.

I don't know if AMD's marketing is to blame on this one. The final spec of this product didn't match some of the earlier rumors. I think there's a possibility that marketing really was told to sell a less overbuilt Gemini board that wasn't going to be this late or with this TDP, and the hardware couldn't be delivered.

And Polaris might factor into this. A pair of ~40 CU Polaris boards in their rumored power and performance range would have left no room for a dual Fury solution that wasn't jacked beyond viability as a consumer solution. It doesn't sound like this is available in quantities commensurate with the VR market it was supposedly synced to.

Well, sure. It was delayed for too long (there were problems with the water cooling months ago) and now it's so late to market that new architecture is going to be available imminently, even if the highest end stuff for it won't be widely available. Bad luck, bad timing, bad watercooling, etc. etc.
 
Unfortunately, if you cast a fair die a thousand times without getting a six, and you cast it once more, the odds of getting a six are still 1/6.

But one may hope that AMD's marketing people are better learners than dice.

This is more like golf though; if you go out there everyday rain or shine and try to hit a hole in one, you will eventually get struck by lightning.
 
I'm calling it now: the Polaris introduction will be absolutely flawless. After so many failed attempts, statistics simply dictate that AMD marketing will get things right one way or the other.
Are you willing to put money on that? If so I'm gonna be rich :D

AMD and Success are oil and water.
 
Are you willing to put money on that? If so I'm gonna be rich :D

AMD and Success are oil and water.

He might as well also include major success for Nintendo with that bet too, since its just as likely to happen too.
 
Back
Top