>AMD’s fastest Radeon yet, the fabled “Fury X2” dual Fiji graphics board has just been teased on twitter by the company’s resident frame pacing guru.
>The card will feature two full fledged Fiji XT GPUs with a total of 8192 GCN cores, 8GB of stacked High Bandwidth Memory and a total memory bandwidth of one Terabyte per second. That’s a whole lot of performance. Which if we’re talking about R9 Nano clock speeds, and we are, is equivalent to 16 Teraflops/s of single precision compute. That’s equivalent to four and a half R9 380 graphics cards, in a board that’s shorter than a reference design GTX 960.
>8192 Stream Processors
>128 GCN Compute Units and Render Output Units
>512 Texture Mapping Units
>Up to 1000 Mhz in GPU Frequency
>8GB of HBM Memory at a Interface of 8192 bits and frequency of 500Mhz
>Bandwidth of 1 Terabyte
Set to launch sometime this year
>All four Nvidia graphics boards in question are described as “COMPUTER GRAPHICS CARDS”. However, all four carry very similar per unit values and as such we could be looking only at the circuit boards and not necessarily graphics cards but there’s really no way of telling for sure.
>All four boards start with the same 699 serial number and the earliest record of a board carrying that serial number appears in December. So we know that we’re looking at Nvidia graphics boards that are new and did not exist at any point before December. This could potentially explain Pascal’s absence from CES and why Nvidia chose to showcase the Pascal Drive PX2 module with Maxwell GPUs instead.
>Pascal graphics architecture.
>2x performance per watt estimated improvement over Maxwell.
>DirectX 12 feature level 12_1 or higher.
>Built on the 16nm FinFET manufacturing process from TSMC.
>Allegedly has a total of 17 billion transistors, more than twice that of GM200.
>Will feature four 4-Hi HBM2 stacks, for a total of 16GB of VRAM and 8-Hi stacks for up to 32GB for the professional compute SKUs.
>Features a 4096-bit memory bus interface, same as AMD’s Fiji GPU power the Fury series.
It started with a thread using jenny from a few days ago and it resulted in a massive western vs eastern animation turf war. It seems to have calm down since that thread isn't being shitposted in like other threads were from yesterday and the day before. I guess as long as you post something /g/ related like a robot, nobody will have a problem unlike posting gumball or regular show
They used to, but the Titan Z flopped so hard that it's not likely they'll go that route again.
Quite frankly, I don't know what the point of dual-GPU cards is... it's usually worse than simply having SLI/CF, but on the flip-side, few ITX cases could actually provide adequate cooling for such a monster or support a radiator.
>more power efficient GPU like nVidia does
>like nVidia does
So, by making a GPU that is completely useless for compute tasks and double-precision?
Nvidia didn't make an outright efficient architecture, you fucking retard. They were stuck with 28 nm so they repurposed Maxwell to be filler product for their consumer (GeForce) line. They made it more efficient by stripping it of compute-oriented components and focusing strictly on gaming only, which made a decent gaming GPU but with awful compute performance. Which is why the current Titan is a joke (1/32 DP performance, compared to 1/3 DP performance for the original Titan), why Tesla didn't get updated to Maxwell, and why Quadro only recently got updated to Maxwell but people still recommend the Kepler models.
Pascal won't be like that, because Nvidia now needs a GPU that isn't shit for compute, so they can make proper Tesla and Quadro updates. So just like with Kepler, Nvidia's efficiency advantage will be gone.
>>52954967# block image MD5s for dumb drakeposters
They're not reports dipshit, I'm blocking a post from ever showing up on my screen and wasting pixel real estate. What part of blocking made you think report? Are you fucking retarded?
>Not loving the classics.
>>Allegedly has a total of 17 billion transistors, more than twice that of GM200.
>The number of transistors is a metric we should care about.
It isn't. If Maxwell had twice as many transistors as Kepler, it still wouldn't be impressive. The only metrics worthy of note are raw performance, and efficiency. I'd be seriously worried if Nvidia weren't doubling their transistor count with a die shrink of this magnitude.
Now if only we could get an actual date or price, which is all anyone cares about.
Old black lady yelled at me for riding by slowly so I put this on her door
Hover over the picture, it fills the screen.
Wow that was hard!
why the fuck would they still try to sell this with Polaris being supposedly <6 months out?
I just hope to fuck the AMD made the interposers for Fiji and big Polaris pin compatible so they can roll out the Fury 2x2 or whatever without a stupidly long delay.
not him, but even though CF now scales better than SLI, AMD still seems to be slower about rolling out profiles for games than Nvidia.
this shouldn't matter In Theory(tm) for DX12/Vulkan, but we'll see what really happens.