Radeon Fury X Beats GeForce GTX Titan X and Fury to GTX 980 Ti: 3DMark Bench

Fury will be the best price/performance ratio card because of it's price and due to the fact that it has 980 Ti, tx performance.

 

If it is the same price as the 980Ti and trades blows with it how is it a better deal?

 

If fury won't bring at least 20% more performance on at least same price, amd achieved nothing.

NVIDIA just dominating with maxwell, and even if FIJI gonna be that awesome it will barely scarth nvidia, something tells me nvidia has a little secret to share soon.(or another Kepler downgrade )

 

^AMD has already achieved something with the fury and nano, but we'll see how benchmarks fare etc. Prices are already good.

We're talking about the normal fury not fury x.

 

And what would you have them say?

We had to jump on new, bloody expensive and capacity-limited technology, because that was our only shot at becoming competitive.
Just look at the rest of our furnace technology spewing nearly twice the heat of the competition
Unless we get our *** together, we're so gonna get it when 16nm/HBM2 becomes available.

 

Haha exactly rich, quote and unquote...

 

ya know amd could easily snap nvidia market share on two major occasion.
the 970 memory issue era, if AMD would had nano back then on some sort of fury , then nvidia would lose many customers.

The second one is the Kepler downgrade era which last now, they had good two opportunities, and what amd does? REBRAND!!!

 

If all things equal, but higher power draw, it will be a less appealing card than 980ti. AMD may have to reduce price to keep it in the game.

 

People don't care about power draw bro, end of story.

Why would it be less appealing if it's cheaper? I fail to see your point.

 

Higher power draw means more heat. I care about heat.

 

This is really misleading though...because that's not "12k"
12k has a resolution of 12288x6480 or 80 megapixels.
This is 3x UHD or 11520x2160 or 25 megapixels.

I mean it's still impressive, but "12k" it is not.

 

Doesn't necessarily mean more heat. Depends on the cooler.

I guess it may matter more with the Fury as it's an aircooled card, but we'll see how it performs. I'm sure it'll be better and won't be as bad as the 200 series cards.

 

Not GPU, but GPUs. For every frame each GPU had to fetch those data via PCIe which is cut in half in case you have 2 cards and not very expensive MB+CPU with 2x PCIe 3.0 x16 available at once.

So lets take it simply. You have gtx690 with 2x 2GB vram. And you have card with exactly same rendering power as that gtx 690 which sports just one GPU and has same 2GB of vram.

For each frame you need 2.5GB of data which equals 0.5GB needed from system ram.
PCIe 3.0 x16 can pull 16GB/s of data.
So, gtx 690 has to load for each GPU 0.5GB = 1GB in total. Frame time is therefore increased by 1GB/(16GB/s) = 1/16s = 0.0625s.
That another card which has same rendering power needs only 0.5GB as it has only one GPU, increase in frame time is 0.5GB/(16GB/s) = 1/32s = 0.03125s.

So now you can make any scenario and get how much more 2 GPUs sharing same bandwidth with system memory get affected in comparison to Single GPU.

In case GPU renders 30 fps before it hits this wall by jumping from 2GB utilization to 2.5GB:
- on 30 fps one frame is rendered in 0.0333s
- Dual GPU adds to frame time 0.0625s + 0.0333s = 0.0958s = 10.4 fps
- Single GPU adds to frame time 0.03125s + 0.0333s = 0.06455s = 15.5fps

In case GPU renders 60 fps before it hits this wall by jumping from 2GB utilization to 2.5GB:
- on 60 fps one frame is rendered in 0.01667s
- Dual GPU adds to frame time 0.0625s + 0.01667s = 0.07917s = 12.6 fps
- Single GPU adds to frame time 0.03125s + 0.01667s = 0.04792s = 20.9fps

In case GPU renders infinite fps before it hits this wall by jumping from 2GB utilization to 2.5GB (just in case someone likes limits):
- on infinite fps one frame is rendered in 0s
- Dual GPU adds to frame time 0.0625s + 0s = 0.0625s = 16 fps
- Single GPU adds to frame time 0.03125s + 0s = 0.03125s = 32fps

This is just to show that there is quite bigger impact when you have to get same data twice via shared slow bus.

And that if gtx 970/980 run ever out of vram in SLI or single, you would have known.

 

Getting some good reports from Gibbo on OcUK

 

Nice, when was the official NDA gonna lift, 24th?

 

what are the reports

 

http://forums.overclockers.co.uk/showpost.php?p=28200197&postcount=40

 

Nothing particularly wrong with that though.

The low stock pushing the price over £600 isn't so good though, so no price drop's for 980ti.

 

So when is the 980Ti 3072 core coming?