Wednesday, December 23rd 2015

AMD "Fiji" Dual-GPU Graphics Card Delayed

by
btarunr
 Discuss (74 Comments)
Originally expected to unveil/launch some time in December-January, AMD's upcoming flagship dual-GPU graphics card based on its "Fiji" silicon could now be scheduled for a Q2-2016 (April-June) launch. In its E3 livecast, AMD CEO Lisa Su stated that the dual-GPU "Fiji" product could launch as early as Christmas 2015. We now know that it isn't happening.

Responding to a question by Hardware.fr, AMD stated that it's pegging the launch of the dual-GPU "Fiji" card to commercial availability of HMDs (head-mounted displays), and a general sense of maturity in the VR ecosystem. AMD is now expecting HMDs to be well proliferated no sooner than Q2-2016, and is hence "adjusting the Fiji Gemini launch schedule to better align with the market," to "ensure the optimal VR experience." AMD did state that samples of the card have been shipped to some of its B2B partners for internal testing within Q4-2015, and their response have been "positive." Could this be AMD buying time to re-engineer a non-Cooler Master cooling solution?
Source: Hardware.fr

Related News

Add your own comment

74 Comments on AMD "Fiji" Dual-GPU Graphics Card Delayed

#1
xkm1948
This should not be more power hungry than 295X2.
Posted on Reply
#2
btarunr
Editor & Senior Moderator
xkm1948This should not be more power hungry than 295X2.
R9 Nano turned out to be OK. Two chips with the same power tuning shouldn't make this card as bad as the 295X2.
Posted on Reply
#3
arbiter
AMD's upcoming flagship dual-GPU graphics card based on its "Fiji" silicon could now be scheduled for a Q2-2016 (April-June) launch.
Around same time frame as Pascal? O boy.
Posted on Reply
#4
the54thvoid
Super Intoxicated Moderator
I doubt they're being honest here. Pegging it to a VR maturity stage is unlikely given the known variables of VR development. They would have known the general timeframe involved. I more think it's pure sales based such that it would potentially cannibalise Nano sales, or even Fury X sales. Given the already noted Asetek legal case and its outcomes, it may well be a cooler issue.
Dual GPU cards are niche designs so the VR argument is null on that front, given dual (single) Fiji availability. Almost all (non bargain basement) mobos support crossfire and what mobo can't support the size of two Nano's?
Reference above to Pascal is irrelevant as its more pressing that Greenland could be out Q3 2016 if we're lucky. Unless Arctic Islands aren't on track (unlikely) then a dual Fiji in Summer might offset that?
Posted on Reply
#5
HumanSmoke
the54thvoidGiven the already noted Asetek legal case and its outcomes, it may well be a cooler issue....Reference above to Pascal is irrelevant as its more pressing that Greenland could be out Q3 2016 if we're lucky. Unless Arctic Islands aren't on track (unlikely) then a dual Fiji in Summer might offset that?
It actually looks disturbingly like the HD 7990 Redux. Dual GPU touted for the best part of year ( lest we forget Fury X2 was announced in June, so a year between announcement and launch is on the cards(!)), and by the time it released its appeal was diluted thanks to the imminent arrival of Hawaii - revealed 5 months later. Even if Arctic Islands isn't imminent, there should be enough buzz and benchmarks (some might even be legit) to turn what should be a GPU highlight into a damp squib.
Might be the cooler - but given that Fury X2 will be a limited edition part, and if push came to shove (assuming it is Nano X 2) air cooling could work, that seems a stretch. Sometimes there just isn't much rhyme or reason to graphics cards launches- it still beats me why AMD released the 380X a week before a driver launched that not only introduced new features, but added a performance uplift for Tonga. Wouldn't a combined launch have had more impact?


Posted on Reply
#6
qubit
Overclocked quantum bit
More AMD stalling tacktics.
Posted on Reply
#7
The Quim Reaper
In other words, we could launch it now but we need something available to counter Nvidia's Pascal launch.
Posted on Reply
#8
Ferrum Master
HumanSmokeIt actually looks disturbingly like the HD 7990 Redux.
Hey look at the fingerprints and marker symbol on the first die... it is actually the same board that Lisa holds and in this picture in the article...
Posted on Reply
#9
RejZoR
btarunrR9 Nano turned out to be OK. Two chips with the same power tuning shouldn't make this card as bad as the 295X2.
295X2 and "as bad as" don't go together. 295X2 is ridiculously cheap if you can get it and sweeps the floor with ALL competition. Even GTX 980Ti. And since it's R9 290 based, it's also fully DX12 and AsyncShader compatible. Meaning it's still pretty damn capable card. I was actually thinking about it before going for GTX 980. Didn't go with it because of case limitations at that time (miniATX with existing AiO in it). Otherwise I might have gone with that one...
Posted on Reply
#10
EarthDog
@RejZoR - R9 295x2 is R9 290X based. A full R9 290x implementation x2 (your underlying point..DX12.1...is correct though.

I asked AMD PR a few days ago about this and received the same exact answer. :)
Posted on Reply
#11
Dave65
AMD should not waste resources on that.
Posted on Reply
#12
GhostRyder
Putting my money on this being related to the cooler...
Posted on Reply
#13
ZoneDymo
qubitMore AMD stalling tacktics.
tactics? not much tactical about this, holidays are now and it was slated to come out now, perfect timing with Christmas etc.
Something hit a snag unfortunately.
Posted on Reply
#14
RejZoR
Yeah, like R9 300 series that were said for spring 2015 and then ended up being released in late summer 2015. With half of it rebranded old series.

I don't understand what's so complicated about dual GPU Fiji. Just smack two of them on a PCB and release the damn thing. It's not like they need to make GPU from ground up this time around. And they've done dual GPU in the past so PCB design isn't an issue either. Power delivery with dual R9 Nano cores shouldn't be an issue either. Damn AMD.
Posted on Reply
#15
awesomesauce
RejZoRYeah, like R9 300 series that were said for spring 2015 and then ended up being released in late summer 2015. With half of it rebranded old series.

I don't understand what's so complicated about dual GPU Fiji. Just smack two of them on a PCB and release the damn thing. It's not like they need to make GPU from ground up this time around. And they've done dual GPU in the past so PCB design isn't an issue either. Power delivery with dual R9 Nano cores shouldn't be an issue either. Damn AMD.
i remember a shortage of plx chip in the past

www.techpowerup.com/167683/shortage-of-pex8747-bridge-chip-disturbs-various-launch-schedules.html

And i think the cooler is the issue now.
But yeah AMD have alot of delay lately

keep on going AMD :toast:
Posted on Reply
#16
EarthDog
RejZoRYeah, like R9 300 series that were said for spring 2015 and then ended up being released in late summer 2015. With half of it rebranded old series.
Well, they came out June 18th, which is still spring. Summer starts June 20th. The Fury parts came out later. ;)
I don't understand what's so complicated about dual GPU Fiji. Just smack two of them on a PCB and release the damn thing. It's not like they need to make GPU from ground up this time around. And they've done dual GPU in the past so PCB design isn't an issue either. Power delivery with dual R9 Nano cores shouldn't be an issue either. Damn AMD.
If it was only that easy...

They didn't need to make the GPU from the ground up in any of their dual cards. They were based off of the existing generation.

PCB design from generation to generation will be different. Its not like they can take what the 295x2 has and use it. Ground up there for the most part. HBM in this one, ETC.
Posted on Reply
#17
PP Mguire
RejZoR295X2 and "as bad as" don't go together. 295X2 is ridiculously cheap if you can get it and sweeps the floor with ALL competition. Even GTX 980Ti. And since it's R9 290 based, it's also fully DX12 and AsyncShader compatible. Meaning it's still pretty damn capable card. I was actually thinking about it before going for GTX 980. Didn't go with it because of case limitations at that time (miniATX with existing AiO in it). Otherwise I might have gone with that one...
Was recently playing with a 295x2 and even though it's not as good as dual Maxwell chips for how cheap it is it performs DAMN good. I like it.

I was hoping FuryX2 would come out sooner so I could get my hands on one but alas when this card is released I'll be knee deep in Pascal and finances tied in that.
Posted on Reply
#18
Slizzo
Welp, this card just became irrelevant.

And does anyone honestly believe that stalling this to coincide with the launch of nVidia's Pascal is a good idea? That's a terrible idea. Release it now (if they could) and make the money now, not bank on the "feeling" that this may be able to perform on par with Pascal.
Posted on Reply
#19
TheGuruStud
SlizzoWelp, this card just became irrelevant.

And does anyone honestly believe that stalling this to coincide with the launch of nVidia's Pascal is a good idea? That's a terrible idea. Release it now (if they could) and make the money now, not bank on the "feeling" that this may be able to perform on par with Pascal.
And no one will buy it, anyway. Dual GPU has never been competition for single GPU cards. Anyone with a brain doesn't want to deal with shit games/drivers when buying an old gpu vs a new one. They would just buy two new ones, period, if they want dual gpu.
Posted on Reply
#20
Prima.Vera
SlizzoWelp, this card just became irrelevant.

And does anyone honestly believe that stalling this to coincide with the launch of nVidia's Pascal is a good idea? That's a terrible idea. Release it now (if they could) and make the money now, not bank on the "feeling" that this may be able to perform on par with Pascal.
Yeah. Is the 3dfx story with the Voodoo 3 and 4 all over again. To keep up with competition 3dfx launched dual, triple and even some quadruple GPU video cards. They only last as a company less than an year afterwards....
Posted on Reply
#21
ZoneDymo
TheGuruStudAnd no one will buy it, anyway. Dual GPU has never been competition for single GPU cards. Anyone with a brain doesn't want to deal with shit games/drivers when buying an old gpu vs a new one. They would just buy two new ones, period, if they want dual gpu.
Might as well skip on those and wait for even newer ones....actually those are probably bad compared to the next one so best wait for the one after that....
Posted on Reply
#22
the54thvoid
Super Intoxicated Moderator
ZoneDymoMight as well skip on those and wait for even newer ones....actually those are probably bad compared to the next one so best wait for the one after that....
You're being obtuse. You know fine well that many people won't consider dual GPU's because of the inherent per game/dev driver issues. His point being Greenland or Big Pascal will pack enough power for VR etc so why waste money on a dual GPU with expected issues.
Posted on Reply
#23
Finners
the54thvoidYou're being obtuse. You know fine well that many people won't consider dual GPU's because of the inherent per game/dev driver issues. His point being Greenland or Big Pascal will pack enough power for VR etc so why waste money on a dual GPU with expected issues.



Sorry!, Couldn't resist
Posted on Reply
#24
geon2k2
Unless Lisa Su is holding a cardboard card in her hands, they have a product ready and they should launch it.
I have doubts that this VR manjo-jumbo will fly anyway.
Look at 3D which doesn't require a helmet, and it is pretty much dead on PC.
Posted on Reply
#25
DeathtoGnomes
Ship the damn card to EK ( :rockout:) and have them make a water block for it. Then attach a radiator with a built-in pump, add some decent length tubing, and.... :banghead:
Posted on Reply
Add your own comment
Dec 22nd, 2024 02:13 EST change timezone

Latest GPU Drivers

New Forum Posts

Popular Reviews

Controversial News Posts