AMD's current dual GPU card, 295X2, was released at $1500. Because of the catastrophic failure that was the Titan Z, AMD owned the dual GPU market this generation. Despite that, less than a year later it was selling for $600 after $30 rebate:
http://forums.anandtech.com/showthread.php?t=2422332&highlight=295x2
And as you can see, no one cared. 60% off a card that is still the fastest single card in some games today, and no one cared.
That has a lot to do with the perception of AMD's entire R9 200 series. Even if R9 295X2 cost $299 when GTX980 was $550, it still wouldn't have outsold the 980.
Generally speaking if you are going to say that R9 295X2 was a card no one cared about, might as well draw the line and say that all dual-chip cards starting from NV 295X2 and AMD's HD3870X2 are a waste of time and money. As you said even when R9 295X2 was $600, it was hardly popular. I am of the view that dual-chip cards hardly ever made sense. HD5970 was CRAM crippled and had no tessellation performance worth talking about. GTX590 had a horribly underpowered/cheap VRM system that resulted in those cards blowing up. GTX690 was VRAM gimped making it DOA long-term and Kepler driver crippled over time. HD6990 was unbearably loud. HD7990 had CF frame issues and cost more than buying 2xHD7970Ghz cards stand-alone. Titan Z was a thermal throttling mess and cost 2X more than R9 295X2 that was faster! The only card in recent times that was actually any good was the R9 295X2 but its $1500 price made it still a bad value.
The other issue with the Fiji X2 is that for 4K it likely won't support HDMI 2.0 and 4GB of VRAM makes it a poor buy long-term. Also, the timing doesn't work since we are near the end of 2015 and 2016 brings 16nm HBM2 cards.
I don't care what the rumors will be for the next generation cards, if a 980Ti can be had for $260 next March, the internet will explode.
Current rumors have Pascal possibly launching in April 2016 at the earliest. So how in the world would a 980Ti cost $260 by March 2016?
It doesn't matter how little you think this cost AMD and will cost them in support, they wasted money developing it and should focus their efforts on main stream products.
It could be much more complex than that:
1) The foundation of a dual chip 596mm2 HBM1 on 1 board will give AMD the experience they need for their next generation HBM2 dual-chip cards in 2016-2018. There for sure will be some learning here.
2) AMD has the WSA with GloFo which means they do have to buy a minimum number of wafers of they face major penalties. Not only is Fiji the largest chip they make, Fiji likely has the highest profit margins out of any card they make. The types of customers who pay > $600 for GPUs tend to be price inelastic. That would allow AMD to sell more Fiji die in pairs. You say that it's a waste of $ which could be true but since AMD has been selling HD3870X2->R9 295X2, that means these cards must be making some money for them to keep using this strategy?
3) AMD wants to move away from the "budget" brand image in graphics. To do that, the halo flagship card such as the Fiji X2 will strengthen their brand image. Even if the R9 295X2 didn't sell in many units, there is no doubt that it provided a positive brand image for AMD - 1st ever reference GPU with AIO CLC, the most powerful card of last generation, cool, quiet. IMO, the R9 295X2 marks the turning point in AMD's brand image even though the market tends to be 2-3 generations behind in perceptions, much like "poor quality of American cars" still persists for many consumers.
4) With the market for miniITX cases gaining in popularity, even though I still think dual-chip flagship cards are a niche, I think they are starting to make more and more sense in 2015 than they did in say 2010-2013.
Before, we didn't even have
700W small form-factor PSUs. Who knows how quickly the miniITX market will gain in popularity over the next 5 years but if you don't take risks, you miss the growth.
5) Playing the devil's advocate: if AMD is wasting $ making dual-chip flagships cards, why has NV been making those cards as well since 295X2? Since we have sufficient data to prove your point correct, I am going to say that overall those cards must make sense which is why both AMD and NV keep making them almost every generation.
Look, I agree with you that dual-chip flagship cards don't make much sense but I am still open to the idea that a small market exists for ultra enthusiast games who buy these things. Similar to how I think all Titan cards are a giant waste of $, the market keeps defying logic which means that for some customers knowing they have the best of the best is all that matters, within a 'reasonable' $1000-1500 price range. If Fiji X2 is a $1000-1500 card, it'll find its buyers. I mean think about it in March of 2015 we had $1000 Titan X but if AMD launches a $1000-1500 Fiji X2, it's already better than the Titan X and not even a year has passed. It's already progress as far as comparing the maximum performance on a single card -- the same for dual GM200.
@RussianSensation
I think vram usage has peaked, at least for awhile. Games are now developed with console specs as the driver and these consoles will remain the target for several more years at least. For the PC, we're seeing the optional 4K textures and 4GB GPUs handle SoM, Ryse, Evolve just fine, even at 4K resolutions.
The only setting that pushes vram beyond 4GB at this point is 4K + 4x MSAA or supersampling 4K+, at which point the GPU grunt has run out long before vram.
To an extent I agree with you. It seems neither the Fury X nor the 980Ti have the GPU power required to fully benefit from more than 4GB of RAM. Over the next 2 years the amount of VRAM will explode as we move into 8-16GB HBM2 cards so really in the context of Fury X and 980Ti, I presume people who buy them are almost always upgrading often (i.e., cutting edge top 5% of PC gamers).
There's a place for dual-GPU cards, as long as its cheaper than 2 separate GPUs, ie. $1200 or less for Fury X2. If its priced more than 2 separate cards, the use scenario shrinks to be really niche.
Ya, that's the part where I agree with you that Fiji priced at $1100-1300 makes sense but if they charge a premium to $1500+, it starts to make less and less sense. Pricing is what made R9 295X2/Titan Z less than stellar.
Still, I am not against a Fiji X2 but I think AMD is prioritizing the extreme high-end while ignoring:
1) No R9 370X outside of China
2) No R9 380X
3) No new cards below $150 now that R7 265/260X/270/270X are EOL
4) No binned 3584 Fury for laptops in the sub-150W space
5) Very few AIBs are even selling the standard Fury non-X -- just XFX, Asus and Sapphire. Why?
6) AMD stopped bundling games after Dirt Rally with their cards while some NV's AIBs continue to
use this strategy.
I just think AMD is prioritizing the wrong products but then again I don't know what their WSA speculates which might require them to buy a lot of wafers and then they are basically stuck.
I mean ya you summed up why this card needs to come out then. Great for you guys who want 2 cards.
I want 1 card, 2 gpus, Wc aio, single slot and amd is continuously to deliver and refine so good for them. Arctic islands is where I'll buy in although I'm tempted to buy the dual fiji card when it's on sale for less and I start playing games from 2010-2013.
What exactly is the point of that? What kind of a modern case and PC customer willing to buy a $1000+ dual-chip flagship card will have issues with a card taking up 2 slots?
The entire single slot thing mattered 10 years ago when we had sound-cards in PCI slots that blocked airflow for the GPU. With the AIO CLC, the dual slot makes no difference since you can put a 2nd dual-slot GPU or a sound-card into the 3rd PCIe slot right below.
I am legitimately interested how there is a benefit to having a single-slot AIO CLC card? With AIO CLC, you can squeeze 3-4 cards as tight as possible since GPU airflow doesn't matter anymore. All that matters if finding space to fit the AIO CLCs.
