Monday, November 30th 2015

AMD Prepares December Launch of Dual-GPU "Fiji" Graphics Card

AMD could launch its next-generation dual-GPU graphics card based on the "Fiji" silicon, some time in December. Codenamed "Fury-Gemini," the card features a pair of "Fiji" GPUs in an internal multi-GPU setup; and is cooled by an AIO liquid-cooling solution, much like the R9 Fury X. Prototype boards of this card were shown by AMD top-brass at some of the chip's earliest reveals. It's expected that the PCB (pictured below), will be mated with a liquid-cooling solution; and unless NVIDIA releases its dual-GPU GM200 graphics card any sooner, is on course to becoming the fastest graphics card you can buy. It remains to be seen if AMD can cash in on the Holiday shopping season.
Sources: VideoCardz, Legit Reviews
Add your own comment

53 Comments on AMD Prepares December Launch of Dual-GPU "Fiji" Graphics Card

#1
Frick
Fishfaced Nincompoop
To put that in perspective, here's the 295x2.



I honestly can't wait for reviews. I love Fury and Nano.
Posted on Reply
#2
RCoon
Gaming Moderator
Guess it'll still be stuck with 4GB of HBM
Posted on Reply
#3
RejZoR
Basically this dual Fiji solution is as long as the usual 2x GPU solutions :) Will this one be using dual Fury X or a dual R9 Nano GPU? Dual Nano would be nice in regards to heat and power consumption. You can't overclock Fury X much anyway so Nano with slightly more relaxed TDP limits again makes more sense.
Posted on Reply
#4
Brusfantomet
RejZoR
Basically this dual Fiji solution is as long as the usual 2x GPU solutions :) Will this one be using dual Fury X or a dual R9 Nano GPU? Dual Nano would be nice in regards to heat and power consumption. You can't overclock Fury X much anyway so Nano with slightly more relaxed TDP limits again makes more sense.
The Fury x and nano has the same GPU, but difrent clocks and power delivery.
If we compare the number of phases to a nano: (from TPUs review if the nano)

and then a fury X: (from TPUs review if the fury X)

i would say it looks like 2 * nano, picture from article to compare:


That card for the X2 fury is shorter than the stock 290X. wonder how face melting the power requirements will be, 375 W for the connectors (PCI-E + 8 pin + 8 pin) but that did not stop the 295X2 from drawing 430W under normal loads:

then again, a 290X draws more than the fury X here, so if its a full fledged fury x * 2 it will probably be in the same ballpark as the 295X2, if tis 2 * nano then i guess it will be more in the 300 W area.
Posted on Reply
#5
TheinsanegamerN
Brusfantomet
The Fury x and nano has the same GPU, but difrent clocks and power delivery.
If we compare the number of phases to a nano: (from TPUs review if the nano)

and then a fury X: (from TPUs review if the fury X)

i would say it looks like 2 * nano, picture from article to compare:


That card for the X2 fury is shorter than the stock 290X. wonder how face melting the power requirements will be, 375 W for the connectors (PCI-E + 8 pin + 8 pin) but that did not stop the 295X2 from drawing 430W under normal loads:

then again, a 290X draws more than the fury X here, so if its a full fledged fury x * 2 it will probably be in the same ballpark as the 295X2, if tis 2 * nano then i guess it will be more in the 300 W area.
Nah, that's 2x fury x. There appear to be 12 VRMs close to the GPUs, 6 apiece. Nano only has 4.
Posted on Reply
#6
ArdWar
TheinsanegamerN
Nah, that's 2x fury x. There appear to be 12 VRMs close to the GPUs, 6 apiece. Nano only has 4.
If you look closely, there're actually only 8 VRM using DirectFET. I presume that's the main VRM going to GPU. Plus one VRM using DirectFET that I think supplying the PLX chip. The other 4 VRM are using SuperSO8. That's much like Nano, using 4 DirectFET + 2 SuperSO8 for each die. Fury X actually using 8 DirectFET VRM (and one unpopulated VRM).

Nevertheless, the main difference between Nano and Fury X is about silicon binning. There's really no technical constraints about VRM, and I believe that a properly designed 4 phase VRM can do as good as 6 phase VRM. Even on some voltage ratio, 4 phase will beat 6 phase on output ripple.
Posted on Reply
#7
awesomesauce
i love it, hope the price will be good :toast:
Posted on Reply
#8
r.h.p
To be honest and I'm no expert , I think the Radeon HBM powered cards are gonna kick ass with directx 12 . At the moment the majority all pc game reviews are run on DX11 ??

Also its impressive to see a Brand like AMD still actively adding to the Fight the machine war with CPU and GPU products not just GPU like NVidia in my opinion








at the moment
Posted on Reply
#9
Estaric
Sounds pretty dang powerful, hope Nvidia can come back with something just as good.
Posted on Reply
#11
the54thvoid
It's a shame if they don't strap on all cylinders and make it that little bit faster than a dual Nano config. It's not got size restraints and if under water, heats not an issue. The phases don't mean much (they're very much overstated by peeps but an adequate amount for a volt fixed card is fine) so fingers crossed.
Posted on Reply
#12
TheDeeGee
All they need now is a Driver which doesn't Toast their hardware.
Posted on Reply
#13
bogami
However, this would i love to have :) . I want to test 4 pieces :) 32k.core in multicore ..:):D
Posted on Reply
#14
trog100
the price will be based on whatever the competition offers.. so far the competition isnt offering anything.. he he

i think it will have to be less than the cost of a couple of 980 ti cards thats about all.. :)

trog
Posted on Reply
#15
natr0n
New smaller cards are nice.

I'm sure cards would have gone to a meter in length in the future.
Posted on Reply
#16
Bytales
Just give me
(16gb HBM gen 2, 8192 cores) times 2 on 14 nm,, 4x8pin power connectors, 4 mini display port and 2 mini hdmi 2.0 port - single slot capability, and ill take 2 of them.

>This one, doesnt really worth it. Perhaps next year well get that.
Posted on Reply
#17
siluro818
RCoon
Guess it'll still be stuck with 4GB of HBM
2x4=8 ^^
Posted on Reply
#18
Fabio Bologna
So if rumors are correct the Fury X2 (or whatever...) will have 2 highly binned Fiji XT chips which, as you may notice with the Nano, is a pretty "cool" running chip (cool in relation to what amd chips used to run at in terms of °C). It will be watercooled from the get-go by AMD (usual asetek setup with a fan in the middle for VRM, looking at you 295 x2...). Now NVidia is going to make something like a GTX TITAN X2-Z-Ti, probably, and as we know GM200 is a damn hot running chip (again compared to the likes of 980 970 & co. ). So if the TITAN Z was a three slot design... if this "thing", "monstrosity", whatever, is gonna be air cooled... mmmh... leaf-blower? turbine? maybe? Or we'll see the first departure of NVidia from that annoying magnesium shroud they've been using since the 690?
My opinion on this is that NVidia should just release something like a dual gtx 980... easy to manage, less power required, still air-coolable and it could support up to 16GB (8GB per GPU as we've seen on the new laptop model) and give us back the glorious "GTX x90" mark that we've all known and loved.
For me GTX 990 it is! :)

GTX 990 Specs:

2x GM204 binned
2x 8GB GDDR5 7010MHz (hinex possibly, better overclocks)
2x 8pin pci-e plugs
stock clock of 1275MHz + boost 2.0
Posted on Reply
#19
Musaab
I think nVidia is not interested in making dual GPU board at least not this generation because if you have mATX and up you can buy two cards for the same price and make your computer look great and you don't need to deal with the big power consumption and heat generation in small space. and for ITX cases you either deal with huge power consumption and heat generation in ITX small space and bad ventilation or deal with water cool or get a crippled card to keep power consumption and heat generation under control.
Posted on Reply
#20
Hit4
AMD keep sleeping cause your dream never come true
Posted on Reply
#21
the54thvoid
Hit4
AMD keep sleeping cause your dream never come true
What dream is that?

That under a deluge of pessimism and huge financial debt they manage to manufacture a gfx card that rivals/beats a 980ti at 4k with 2/3 of the Vram and a similar energy footprint. The only dream AMD missed on was the overclocking part - which should never have been blurted out by the PR launch. Aside from that, Fiji is a very capable design and moreover, going forward, the underlying GCN architecture is exceptionally well placed for DX12.

Perhaps instead of trollish comments you should educate yourself more on how good AMD are at making very capable graphics hardware?
Posted on Reply
#22
tyrone747
I imagine this is going to be another 'paper launch' and the cards won't be widely available until some time next year.
Posted on Reply
#23
erocker
*
So... Retail about March or so?
Posted on Reply
#24
Musaab
the54thvoid
What dream is that?

That under a deluge of pessimism and huge financial debt they manage to manufacture a gfx card that rivals/beats a 980ti at 4k with 2/3 of the Vram and a similar energy footprint. The only dream AMD missed on was the overclocking part - which should never have been blurted out by the PR launch. Aside from that, Fiji is a very capable design and moreover, going forward, the underlying GCN architecture is exceptionally well placed for DX12.

Perhaps instead of trollish comments you should educate yourself more on how good AMD are at making very capable graphics hardware?
In fact while AMD forced to squeeze every single technology just to design the Fury family, nVidia had the 980Ti on shelf for months and they lunch it in less than two months, AMD was forced to use untested and expensive technology like memory on die AKA HBM and a base circuit AKA Interposer.
you said that Fury X rivals/beats a 980ti at 4k but only reference cards and no one buy them to keep them reference they mostly used in open loop water cooled systems with hugr overclocking.
You said with 2/3 of the Vram and a similar energy footprint and the 4GB HBM cost much moor than 6GB GDDR5 and the reduce in energy footprint came from the memory change not the GPU design itself.
About DX12 it's still early because the only test tell us that R9 390X, Fury x, GTX 980 and 980Ti perform the same.
Posted on Reply
#25
Musaab
erocker
So... Retail about March or so?
That is the time frame for nVidia's Pascal lunch. Too late.
Posted on Reply
Add your own comment