AMD Radeon R9 Fury X 4 GB

[Absolution]

GCN is actually showing more and more to be a jack of all trades and master of none.

I think that is the real culprit here. It is not efficient, and that limits the whole potential of GPU offerings with AMD. They are still just scaling up the die further and further like they did since the 7970, and Tonga was not enough to limit the inefficiency of this arch. With Tonga XL (Fiji) they are hitting limits, shader count brings no 1:1 performance increase anymore, which is also why overclocking yields virtually nothing.

This spells doom for any future AMD gpu unless they start fresh or totally revamp GCN. I even think a node shrink won't help them enough.

The difference in overclockability of Maxwell and Tonga/GCN is staggering. Maxwell is much more versatile, I think the comparison between a motorcycle and a fast car is the best one: Maxwell being the motorcycle that can accelerate faster than any car possibly could, even if they put a 4096 HP engine in that car.

Well it's good for LuxMark at least lol, but then again 4GB is a problem for 3d rendering

 

[_BARON_]

This could should've been @ MSRP $500 or $550 then it would've made sense. It costs like GTX980Ti, offers less performance than stock GTX980Ti, just terrible.

From what I've seen overclocks are shit.

Best card on the market now seems to be EVGA GTX980Ti SC+ 6gb @ 650$

 

[INSTG8R]

I guess what I take away from this is that AMD has now at least cracked the HBM implementation so it should put them one up on NV when it comes to the Arctic Islands vs Pascal. For AMD it will be 2nd gen and for NV 1st gen(okay they will both most likely be using HMB2 but you get my point)

 

[Kyuuba]

For the first time I am feeling worried about AMD...

 

[Ikaruga]

They also "forgot" the cool the VRMs, and since it's a closed enclosure, it heats up quite nicely:

(source)

I wonder how fast these "top of the line" cards will die in the hands of enthusiasts

 

[Sihastru]

They didn't, there's a copper pipe dedicated to that very task.

 

[Mistral]

Not quite what we hoped for, but not too shabby either. Will wait and see where the Nano ends up.

 

[nickbaldwin86]

LOL AMD fans bois crying at AMDs fail right now.... sorry had too.

I really REALLY wanted this to be a WIN for AMD... because I really wanted the prices to drop on the 980Ti's...

NVidia could raise the prices and still get sales now... thanks AMD you have had epic amounts of fails in the past week/month/year

I guess I will just sit back and be grateful with what I got.

 

[SirEpicWin]

I'm utterly disappointed I shouldn't have waited this long ........sigh
but can someone explain these numbers? just how in the world it's slower then 980Ti even by a small percentage!
Fury x
Shading Units: 4096
TMUs: 256
ROPs: 64
Compute Units: 64
Pixel Rate: 67.2 GPixel/s
Texture Rate: 269 GTexel/s
Floating-point performance: 8,602 GFLOPS

980Ti
Shading Units: 2816
TMUs: 176
ROPs: 96
SMM Count: 22
Pixel Rate: 96.0 GPixel/s
Texture Rate: 176 GTexel/s
Floating-point performance: 5,632 GFLOPS

 

[TheGuruStud]

Better be a secret 96 rop version... What a bunch of losers.

 

[Ikaruga]

They didn't, there's a copper pipe dedicated to that very task.

I used quote marks with the "forgot" word, but seriously: it's 104C under heavy load with the dedicated copper pipe?

 

[buildzoid]

They also "forgot" the cool the VRMs, and since it's a closed enclosure, it heats up quite nicely:

(source)

I wonder how fast these "top of the line" cards will die in the hands of enthusiasts

104C is fine. The MOSFETs are rated at 75A continuous drain current at 125C and there's 6 of them. The caps are all tantalums so those should be OK with it too.

@W1zzard where are you thermal camera images? I remember you had them for some GPU reviews right?

 

[L'Eliminateur]

what a monumental failure.
I too wanted this to be the next "core 2 duo", but nope, AMD has been doing nothing but failures for the past YEARS.

This is the faildozer of GPU releases for AMD, a ton of hype and shit for result.

why in gods name they released this in 28nm?, this should be 16nm!.

And if these are the results for the X then the fury and the nano will be complete dissapointment and wont be competitive with the 970 at all.

Nvidia's HQ must be dancing in joy right now, i guess i'll get a 970, no point in waiting, AMD won't have anything for more than a year(maybe when they rebrandeon the fury and do a die shrink of it) and then we will have Pascal which -so far- looks like an interesting gpu with a paradigm shift PLUS HBM and you know it will be crazy fast

 

[ensabrenoir]

Its great to finally see a review. Puts all those conspiracies to rest. Thanks wiz.

......no.......im sure Nvidia is responsible for this.........we go now to start the under ground resistance........
Seriously though if Amd was able to get this out a year or two earlier it would have really mattered. Now when Nvdia goes hbm2 next year......

 

[slick530]

The 980Ti pretty much destroys the Fury X on every single benchmark. It's misleading when Richard Huddy and Lisa Su claims that the Fury X is the fastest card in the world. The Titan X still holds the crown. The only advantage I see with this card is the small form factor and it's ability to fit in a HTPC. But then again, without HDMI 2.0, it just doesn't make sense and not an option. AMD still haven't nailed this as much as the hype surrounds it make people believe so. It's a decent card but honestly, doesn't worth $650 given how it performs. $599 would make it a more appealing card to new buyers.

 

[lilhasselhoffer]

...
True, sad but true as this round is not really much to speak of overall.

Either way, card does exactly what its supposed to. Its got great performance, lower power consumption (it basically if very close to Titan X and 980ti) and comes with an already great stock cooler. What do they need to wrap the card in solid gold to make it appealing at this point?

Yes and no.

I'm neither and AMD lover, nor an Nvidia lover. What I need both of these companies to do is actively smack their fabs around, and get a new node up and running. I'm not buying twice reheated leftovers, and especially not at the price of a fresh gourmet meal. It doesn't honestly matter what minor incremental improvements Fury and the 980 Ti brought to the table, when I don't have to spend one additional cent for the minor performance penalty my 7970 already offers.


I, like most people, am a cheap ba####d. If it costs 200% more than I've already invested, for a 40% increase in performance, my happy backside is punching out. If AMD wants to sell the Fury to me they'll have to run a goods air cooler, allow me to trade my 7970 in, and only charge $200. Given, that proposition is insane; likewise, the proposition of spending enough to make a new system on a single card of dubious improvement is equally insane.

 

[haswrong]

so tomshardware says the average idle power consumption is 4.8W. so why tpu says its 20W? *confused look*

 

[KainXS]

Its looking pretty bad for amd, ontop of pretty much rebranding nearly every card in the 300 series from cards in previous series the Fury X is not performing as well as they thought it seems and thats kind of to be expected since this is the first release of HBM. When Lisa said this was the fastest card in the world boy did she get that wrong and almost no overclocking potential vs what they said at the event, very very disappointing and they should drop the price as they're asking too much right now, I still want a nano though to be honest.

 

[LAN_deRf_HA]

You know it's really similar to politics. No matter how times a side gets burned they're always ready to get their hopes up again next election cycle.

 

[Ikaruga]

I'm utterly disappointed I shouldn't have waited this long ........sigh
but can someone explain these numbers? just how in the world it's slower then 980Ti even by a small percentage!

If you want to compare numbers, compare the Pixel rate. Maxwell is simply better. (note: not many talk about it here, but AMD's GCN will perform better under DX12, the higher number of units will help a lot there)

104C is fine. The MOSFETs are rated at 75A continuous drain current at 125C and there's 6 of them. The caps are all tantalums so those should be OK with it too.

Yea, they are fine alone, but look how large area of the PCB is dark redish, add the extra heat when the backplate is back in place and and tell me that's healthy on the long run if you push the card a lot (let alone if you want to OC it heavily).

 

[HD64G]

I'm utterly disappointed I shouldn't have waited this long ........sigh
but can someone explain these numbers? just how in the world it's slower then 980Ti even by a small percentage!
Fury x
Shading Units: 4096
TMUs: 256
ROPs: 64
Compute Units: 64
Pixel Rate: 67.2 GPixel/s
Texture Rate: 269 GTexel/s
Floating-point performance: 8,602 GFLOPS

980Ti
Shading Units: 2816
TMUs: 176
ROPs: 96
SMM Count: 22
Pixel Rate: 96.0 GPixel/s
Texture Rate: 176 GTexel/s
Floating-point performance: 5,632 GFLOPS

Most possible factors are ROPS and even for a few % the immature drivers.

 

[the54thvoid]

Will the trolls please stop calling it a fail.

It's far superior to the 290X in every way. Yes AMD hyped it too much at their PR call but it's still a damn good card.

Just not what we had all hoped for.

 

[steen]

Does anyone know if this card is 1/4 DP?

DP is 1/16.

 

[revin]

These Pre-Release card's just could be a sandbag..............just saying.
It's very possable AMD could release a very fine tuned unit to market................
Or pull off a price drop since the reviewer's all have found similar faults

Please Don't bother just flaming my post

 

[TheGuruStud]

These Pre-Release card's just could be a sandbag..............just saying.
It's very possable AMD could release a very fine tuned unit to market................
Or pull off a price drop since the reviewer's all have found similar faults

Please Don't bother just flaming my post

Too late. Your flame suit can't handle what's coming.