Fury XT and Pro prices

[Dribble]

The big problem is it's not really under AMD's control, it's the devs that decide what goes in the gpu's memory. How many do you think are going to optimize memory usage just for Fury, how hard do you think AMD will really try to help them? It's not like lack of memory is the way things are going, it's just a particular problem for one card that will be resolved as soon as HBM 2 arrives.

It will never be worth it for a dev to optimize just for fury - how many extra sales will that get them that they wouldn't have got anyway? Perhaps you'll get one or two games "optimized for fury" stamped on them where AMD has paid the dev to put in some special path for Fury, but that will be it. Not that 4GB is the end of the world - it's plenty most of the time, you just might have to turn down the odd memory intensive setting a notch now and again because of it.

 

[exar333]

Well there we have it.

Since GCN has no issue with tile resources or async compute & rendering, definitely drivers can stream in assets as required. I like the emphasis on async compute/rendering to prevent texture pop-ins as it reduces frame latency.

That's awesome. Hopefully we start seeing DX11.2 and DX12 focused titles soon. We have all these great DX capabilities, but its a total waste if they are rarely used outside of demos and so forth.

 

[Enigmoid]

In games like Witcher 3 and GTA V, when you are moving in the open world, it's seamless and its due to asset streaming. Stutter free. That's coming from system ram. Definitely possible with optimizations or as they call it: heuristics.

Sure you can stream it in. But then you have to drop something else to make space.

And you have a couple bottlenecks with streaming. First is the 16 GB/sec PCIe bus. Then is the actual ~25-30 GB/sec DDR3 bandwidth which has to be shared with the CPU.

 

[Enigmoid]

I agree that current games are extremely wasteful with VRAM. Lots of improvement there. However, devs have no current reason to optimize for vram.

 

[xthetenth]

That MaximumPC interview, he mentions using compression specifically in the context of 4 GB being enough, so that might take the sting out of games using uncompressed assets.

 

[n0x1ous]

I agree that current games are extremely wasteful with VRAM. Lots of improvement there. However, devs have no current reason to optimize for vram.

Whats really interesting is looking at Witcher 3 vs GTA V - both huge open world games. Witcher 3 uses minimal VRAM at all resolutions whereas GTA V seems to be a hog.

Witcher 3 certainly lends some credibility to what AMD has been saying about VRAM being wasted, but this appears to be down to the developer not a driver thing. I don't think AMD want's to count on lazy dev's to optimize properly. Witcher 3 seems to be the outcast in the latest games with most being more and more VRAM hungry though Witcher 3 proves thats probably not necessary if optimized properly.

 

[tviceman]

Let's not even forget the insane BS made up on this very forum about 2x2048 shader Tonga with HBM or Hawaii with HBM and no other improvements. Now AMD has released a card with 1.5X the perf/watt increase over 290X in Fury X and supposedly 2X the improvement in the Fury Nano

First of all, the card isn't released yet. No one can buy it. Second, lets wait until non-biased reviews to see where the performance falls and actual power draw ends up. Companies, every single one of them, always tout absolute best case scenarios and often fudge / round up to make it look better. Nvidia did it with GM107 and GM204. I'm absoultely sure AMD is not impervious to this type of marketing either.

Until I see otherwise, I still think aftermarket $670-680 open-air cooled 980 TI's will perform better than Fury X when both are overclocked. The differences in performance will likely be negligible, but it should help to continue fighting and crying for the next 12-18 months.

 

[AnandThenMan]

Second, lets wait until non-biased reviews to see where the performance falls and actual power draw ends up.

Agreed. We need to see professional reviews before making any declarations of performance, power draw, if 4GB is a limiting factor etc.

 

[IEC]

Agreed. We need to see unbiased professional reviews before making any declarations of performance, power draw, if 4GB is a limiting factor etc.

FTFY. I'll be waiting for benches from my friends across the Atlantic (particularly ze Germans) considering how flamingly bad most US-based sites are.

 

[Creig]

I wonder how angry Fury X owners will be if they try to run games like GTA V with VSR or 4K + AA and find it hitting their tiny 4GB limit? Hmmm..

Probably not nearly as angry as Titan X owners were after they spent $1,000 per card at launch only find people were getting the same performance for $650 per card with the 980 Ti a mere 2½ months later.

 

[Enigmoid]

Probably not nearly as angry as Titan X owners were after they spent $1,000 per card at launch only find people were getting the same performance for $650 per card with the 980 Ti a mere 2½ months later.

Anyone who bought a titan x

1) didn't care about cost and wanted it now

2) Want/need the 12 GB for something else

3) are a complete idiot and have no right to be upset, especially when the majority of the internet told them not to and to wait because prices would drop like with the 780/780ti/titan.

I really don't see very many upset titan x owners either.

 

[ChuckFx]

I hope you're joking. Most of the cards are basically rebrands with more memory (how many people actually need 8GB?); that won't sell, and the 390/X pricing is RIDICULOUS considering you can get R9 290/X for $240-270 after rebate RIGHT NOW. $430 for a warmed over 290X with 8GB won't cut it.

Even with Fury I don't think anything changes. I'm expecting Fury X to be about the same performance level as 980 Ti and when both are max OC'd the 980 Ti may be faster.

You are lucky to have all the benchmarks that nobody has seen yet, perhaps you could share them with us?

If you don't have a confirmed benchs, there is nothing to discuss yet in terms of perf. This is not just ''rebrand'' it's a remodel by the way.

 

[Creig]

I really don't see very many upset titan x owners either.

Of course you don't see many. They'll never admit to wasting $350 per card for 2½ months of gaming. They'd rather defend their purchases to the death rather than appear foolish.

 

[96Firebird]

This is not just ''rebrand'' it's a remodel by the way.

New hardwoods and granite countertops?

 

[Enigmoid]

Of course you don't see many. They'll never admit to wasting $350 per card for 2½ months of gaming. They'd rather defend their purchases to the death rather than appear foolish.

Perhaps they don't care?

Either way if you are upset you are an idiot.

 

[Jacky60]

When the Hell is dual Fury coming, I'm already getting 5fps out of VRAM slow downs in Arma 3 and I'm not waiting 6 months as the next generation will then be just around the corner. I don't want to go 980Ti but need more VRAM today not 'sometime soonish'. I don't see why they have to launch dual GPU card so much later, it can't be that difficult to design can it?

 

[exar333]

With just a $100 difference (~18%) and the difference in cooling and performance (not cut-down), the X really seems to be the card to get. If the pro was maybe $499 it would be more attractive. The X seems better, even if I toss the cooler out.

Edit: For example the 5870 and 5850 was 50% more to get the 5870. That was a LOT different! $379 vs $259 then and $649 and $549 now.

 

[monstercameron]

So the vram marketing is already in full swing. So they are shifting from perf/watt, noise, aesthetics and api compliance to vram.

 

[n0x1ous]

When the Hell is dual Fury coming, I'm already getting 5fps out of VRAM slow downs in Arma 3 and I'm not waiting 6 months as the next generation will then be just around the corner. I don't want to go 980Ti but need more VRAM today not 'sometime soonish'. I don't see why they have to launch dual GPU card so much later, it can't be that difficult to design can it?

so you have a VRAM limit problem and your solution is to buy another dual gpu card with the same amount of VRAM?

 

[chimaxi83]

so you have a VRAM limit problem and your solution is to buy another dual gpu card with the same amount of VRAM?

6990 has 2GB per GPU, Fury has 4GB. It's double the VRAM.

 

[n0x1ous]

6990 has 2GB per GPU, Fury has 4GB. It's double the VRAM.

His signature says 295x2 first. I would assume he would be looking at fury to pair with the 4770k + 4k monitor system and not the old 920

ps - congrats on Lord Stanley's Cup

 

[Subyman]

Let's keep this thread about Fury as the OP intended.
-Moderator Subyman

 

[bryanW1995]

So, are the 390 and 390X significantly faster than the 290 & 290X? If not, I don't see how the 970 & 980 are anything but helped as the cheaper-per-FPS cards phase into the SOS with exciting new price tag cards, and someone correct me if I'm wrong but isn't the >$500 graphics card market pretty much one or two percent of sales at best?

I've got to buy something this fall and frankly, I'm not seeing anything nearly as compelling as the 970 - which kinda sucks for me as I've better luck with AMD cards. (Got two defective 275s in a row, and in dozens of cards I've only gotten one other defective graphics card, a $1,500 Wildcat I smoked in the late 80s.) I was hoping for a compelling AMD card under $350 street or, failing that, at least something that would drive down the price of the 970. Am I just missing it?

We'll have to see how things shake out, but AMD doesn't look to have anything super compelling below Fury Nano. Likely your choice will be between 970 and 980.

 

[RussianSensation]

We'll have to see how things shake out, but AMD doesn't look to have anything super compelling below Fury Nano. Likely your choice will be between 970 and 980.

I don't know about the 980. $549 Fury PRO is going to be a better buy than a 980 if it has 3584 shaders, 224 TMUs and 64 ROPs and 4096-bit HBM. People on AT recommended a 6-15% faster $510-550 980 over a $280-300 290X for the last 8 months. I expect every single one of those people to recommend a card just $50-100 more expensive than a 980 if it's 6-15% faster, which is about where Fury PRO should fall in.

 

[exar333]

I don't know about the 980. $549 Fury PRO is going to be a better buy than a 980 if it has 3584 shaders, 224 TMUs and 64 ROPs and 4096-bit HBM. People on AT recommended a 6-15% faster $510-550 980 over a $280-300 290X for the last 8 months. I expect every single one of those people to recommend a card just $50-100 more expensive than a 980 if it's 6-15% faster, which is about where Fury PRO should fall in.

100% agree.

The 980 also doesn't have a 6GB to 4GB (although this remains to be seen what it means, but still reflects the specs) 'advantage' the 980Ti has over the X.

Unless the Pro is a very poor overlocker, the advantage of 4GB HBM will really help.

Plus, even if the cards use the same power, perform the same and cost about the same, its cooler to have some new, shiny HBM.

The exception might be picking-up some cheap 980s used? Those could get to be a decent deal as people dump those for 980Tis or Furys...

Edit: IMHO the 980 (new) would only be interesting at $429 or less.