AMD Radeon R9 Fury X 4K Video Card Review @ [H]

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,629
AMD Radeon R9 Fury X 4K Video Card Review - We take the new AMD Radeon R9 Fury X and evaluate the 4K gaming experience. We will also compare against the price competitive GeForce GTX 980 Ti as well as a GeForce GTX TITAN X. Which video card provides the best experience and performance when gaming at glorious 4K resolution. Can the new AMD Radeon R9 Fury X rise up to the occasion?
 
AMD really shot themselves in the foot by insisting on HBM for their top end card on such an early go. They should have reserved it for upper midrange cards like their upcoming Nano and just shoved more cores and 12gb of GDDR5 ram as Nvidia wisely chose to do.
 
Dainas said:
AMD really shot themselves in the foot by insisting on HBM for their top end card on such an early go. They should have reserved it for upper midrange cards like their upcoming Nano and just shoved more cores and 12gb of GDDR5 ram as Nvidia wisely chose to do.
Click to expand...

I tend to agree, but I'[m not sure it's the HMB alone, but the HBM AND the GPU itself that's holding this thing back. I mean, I get the marketing reasons for putting it on the Fury: visibility and mindshare, but it's not going to translate into sales, and that means they aren't winning any new customers. At least on the consumer level. Maybe this is part pitch to Nintendo, MS and Sony for the next gen consoles, to lock in another generation. Unless this GPU architecture scales DOWN really well, into the inevitable lower tier cards, I don't see AMD doing anything but relying heavily on those console GPU sales for the foreseeable future.

I am really interested to see the Fury-gen cards that compete with the 970... whenever they drop.
 
So if I'm reading this right I can max out Dying Light and still get a minimum of 30fps. That's great because I can't notice a difference going above 30fps. Too bad about the FuryX still being so behind but it is what it is and I'm glad I went 980ti (couldn't have gone FuryX anyway since there weren't any and I had to replace a shorted out card). Also I still can't get over the prejudice of not wanting water any where near my computer.

Edit:
BTW nice review very useful info. Can't wait to see if you can manage to get two of all these cards for some SLI/xFire comparisons.
 
Last edited:
"Unfortunately, HBM is not the saving grace of the AMD Radeon R9 Fury X that propels it forward in 4K gaming currently. It is held back by capacity and performance."

Where's your evidence of this "capacity" limitations thus far?

It appears you enter with the preconceived notion that 4GB is holding back the Fury, but never demonstrate so. BF4 shows your preconceptions.

"Lowering settings by disabling MSAA, trying to give the Fury X the best chance it can get"

Why would disabling MSAA give the Fury X the best chance? If anything, more MSAA would help the Fury X as it tends to do better at higher resolutions due to AMD drivers and its memory bandwidth advantage. And, sure enough, the gap between the Fury X and 980 Ti is smaller with MSAA than without. Yet you felt no MSAA gave it the best chance...? I'm guessing this is because you think it is running out of memory.
 
Modred189 said:
I tend to agree, but I'[m not sure it's the HMB alone, but the HBM AND the GPU itself that's holding this thing back. I mean, I get the marketing reasons for putting it on the Fury: visibility and mindshare, but it's not going to translate into sales, and that means they aren't winning any new customers. At least on the consumer level. Maybe this is part pitch to Nintendo, MS and Sony for the next gen consoles, to lock in another generation. Unless this GPU architecture scales DOWN really well, into the inevitable lower tier cards, I don't see AMD doing anything but relying heavily on those console GPU sales for the foreseeable future.

I am really interested to see the Fury-gen cards that compete with the 970... whenever they drop.
Click to expand...

Yeah, the big issue with AMD cards as we saw with the 290x, is we often do not know what the card is truly capable until months after and everyone had lost interest. The 290x released to scarcely beat a Titan; began to out muscle the 780ti after a few months of drivers and later still now nips on the 980 hard. Nvidia takes months to get their optimisations in order, AMD takes years.

Fury X may have a rubbish core, or could have a good deal more potential. Sadly we sure as heck do not know now because of its 4GB restriction and raw drivers.
 
10% faster
24% faster

The 980 Ti is romping home to victory. FuryX just isn't there I am afraid. Completely gets stomped by the Ti at 4K where VRAM is really important.

So much for the HBM pundits saying that somehow HBM makes 4GB feel a lot smoother. Bear in mind that HBM is only 30% faster than the 7Ghz 980 Ti's vram. So not much to say once you overclock the Ti VRAM and being things very close.

Very nicely concluded there:

Let's be honest, the AMD Radeon R9 Fury X is getting its butt kicked at 4K. We don't think these are the results AMD wanted, especially marketing this video card as a video card developed for 4K gaming. However, these are the real-world results we have found between the AMD Radeon R9 Fury X, NVIDIA GeForce GTX 980 Ti and NVIDIA GeForce GTX TITAN X.
Click to expand...
 
So there's no reason to buy a single card Fury X over NV's offerings. Gets walloped at 1440p and below, and even at 4K it still falls behind slightly.

From what I've seen, the only saving grace for the Fury is Crossfire, it scales well and can match/slightly exceed the 980 Ti in SLI. But then, if you OC the 980 Ti it comes out ahead again.

It's been awhile since NV has had such a runaway victory at the high end, 290X was way more competitive.
 
Crisium said:
"Unfortunately, HBM is not the saving grace of the AMD Radeon R9 Fury X that propels it forward in 4K gaming currently. It is held back by capacity and performance."

Where's your evidence of this "capacity" limitations thus far?

It appears you enter with the preconceived notion that 4GB is holding back the Fury, but never demonstrate so. BF4 shows your preconceptions.
Click to expand...

It is HardOCP's stance that 4GB holds back 4K performance on any video card:

"At 4K though 4GB of VRAM is clearly not enough. At 4K you want at a MINIMUM 6GB. It is possible though that more may actually help as you start increasing the number of video cards in SLI. 6GB might actually not be enough for some games in 4K when SLI is involved, we will see."

http://www.hardocp.com/article/2015/06/15/nvidia_geforce_gtx_980_ti_video_card_gpu_review/12
 
AMD needs to switch to multiples of 3.

1GB/3GB/6GB/9GB for GPUs

3Core/6Core/9Core/12Core for CPUs.

Be a nice differentiator too.
 
madgun said:
10% faster
24% faster

The 980 Ti is romping home to victory. FuryX just isn't there I am afraid. Completely gets stomped by the Ti at 4K where VRAM is really important.
Click to expand...
Solid win for Nvidia but it seems that single GPU power is still not ready for 4K yet anyway so it's kind of academic (the 980ti is a beast @1440p though).
Frankly put, single-GPU video cards do not perform well enough to enjoy immersive gaming at 4K.
Click to expand...
Hope there is a SLI vs CF review in the works? Should exposed the 4GB handicap (and maybe even 6GB) more markedly than single cards.
 
madgun said:
10% faster
24% faster

The 980 Ti is romping home to victory. FuryX just isn't there I am afraid. Completely gets stomped by the Ti at 4K where VRAM is really important.

So much for the HBM pundits saying that somehow HBM makes 4GB feel a lot smoother. Bear in mind that HBM is only 30% faster than the 7Ghz 980 Ti's vram. So not much to say once you overclock the Ti VRAM and being things very close.

Very nicely concluded there:
Click to expand...

fury x is equal in the gaming experience.
20% at 40fps is 8fps and seriously who plays games at 40fps?
Hardocp review guys
 
Thanks guys, glad to see the review up! It's too bad the 4GB HBM can't handle 4K like you said... Looking forward to the crossfire/SLI scaling review to give us the full picture. Should drive home your point the 4GB HBM cards can't keep up with 6GB/12GB ones...or will it? Nvidia seems to own the single cards, but it's looking like if your building a monster rig, Fury scales much better. Looking forward to the crossfire review fellas, good job on this one and thanks again!
 
Dainas said:
AMD really shot themselves in the foot by insisting on HBM for their top end card on such an early go. They should have reserved it for upper midrange cards like their upcoming Nano and just shoved more cores and 12gb of GDDR5 ram as Nvidia wisely chose to do.
Click to expand...

more money more dev time bigger card. fury x is a the 1st fruit of amd and sk hynix's 7 years of work to create a new graphics memory standard. If tsmc/world did not screw up the 20nm node so epically willing to bet this would of been a vastly different 20nm part. However we are still stuck at 28m till next year.

It will be interesting to see if the rumor that amd has hmb2 priority over nvidia pans out if it does sometimes it is good to be 1st.

http://www.guru3d.com/news-story/rumor-amd-gets-priority-advantage-to-hynix-hbm2.html
 
Why am I not surprised about the review conclusion. hardocp is going to keep bashing Fury and Fury X because they have decided to for this generation. No point at all reading their reviews on Fury cards for this generation. Most other sites have a much more balanced view than such a extremely prejudiced view. Whats even better is Fury X CF is competing very well with GTX 980 Ti SLI. This blows the hardocp claims of 4K performance being held back by 4GB VRAM.

http://www.computerbase.de/2015-07/radeon-fury-x-geforce-980-ti-crossfire-sli-ultra-hd-12k/2/
http://www.hardware.fr/focus/111/crossfire-radeon-r9-fury-x-fiji-vs-gm200-round-2.html
http://nl.hardware.info/reviews/615...marks-hardwareinfo-gpu-prestatiescore-q2-2015
http://ht4u.net/reviews/2015/amds_furie_radeon_r9_fury_x_im_test/index45.php

I believe the conclusion is decided first or rather pre-decided and then they go about cooking some graphs. Then they go about writing rubbish to top it off. :p
 
I tend to look at the apples to apples results more myself so seeing AVG FPS within 5 FPS on most games not named Dying Light I don't know if I would agree that the Fury X got it's butt kicked. I would agree the 980 Ti is the better card and if priced the same I would recommend it to people.
 
it would be interesting to see some vram charts. i know 4gb must be getting maxed out, but just to be sure.

it would also be interesting to see a game or two benchmarked at 4k and the lowest settings. this way we could see if it is really the lack of vram holding the card back or just a lack of shader power. to be honest, i think it is a combination of the two.
 
Wow, it really gets handed its ***.

Flopper said:
fury x is equal in the gaming experience.
20% at 40fps is 8fps and seriously who plays games at 40fps?
Hardocp review guys
Click to expand...
You did read the review?
 
Simmonz said:
I tend to look at the apples to apples results more myself so seeing AVG FPS within 5 FPS on most games not named Dying Light I don't know if I would agree that the Fury X got it's butt kicked. I would agree the 980 Ti is the better card and if priced the same I would recommend it to people.
Click to expand...

and that's why you should read the entire review to understand the entire landscape but more important read the author's opinion and editor's opinion as first hand experience through the review.. just reading the apple to apple results, partially tend to create a false conception of results.
 
I really hope the new CEO gets their shit together. I'm making the assumption that this product was designed and scheduled before she took over and while normally i'd be like SCRAP IT, i dont think AMD can afford to. Any improvement is better than none. The next iteration for AMD i think will make or break them though and god that has me scared. I have used intel/nvidia since going quad core and its still the right choice to make. But i know what happens in a market with zero competition and i'm very worried.
 
This card is too slow for 4K gaming, but I fail to see how this review demonstrates the 4 GB RAM being limiting. It was only in one game and a known exceptionally GPU RAM heavy one. None of the other games showed 4 GB having any negative influence. I was fully expecting Fury cards to have problems when it was announced that they would only have 4 GB, but the fact is that in current games they are way too slow before getting capacity limited.

So, please, if you're going to complain about Fury, do so on its processing abilities, not RAM capacity as far as currently observable performance goes. For future proofing? Maybe. Oh, and modded Skyrim possibly.
 
Crisium said:
"

Why would disabling MSAA give the Fury X the best chance? If anything, more MSAA would help the Fury X as it tends to do better at higher resolutions due to AMD drivers and its memory bandwidth advantage. And, sure enough, the gap between the Fury X and 980 Ti is smaller with MSAA than without. Yet you felt no MSAA gave it the best chance...? I'm guessing this is because you think it is running out of memory.
Click to expand...

ROPs and Memory
 
Flopper said:
fury x is equal in the gaming experience.
20% at 40fps is 8fps and seriously who plays games at 40fps?
Hardocp review guys
Click to expand...

1437535126iwTl74Zfm5_3_1.gif

1437535126iwTl74Zfm5_4_1.gif

I'm not going to copy the other charts, but they are essentially the same: The 980 Ti has higher settings and largely higher fps even with the higher settings.
How can you possibly say, "fury x is equal in the gaming experience"?
 
+3.8 FPS

1437535126iwTl74Zfm5_3_5.gif


+2.8 FPS

1437535126iwTl74Zfm5_4_4.gif


+ 3.3 FPS

1437535126iwTl74Zfm5_5_4.gif


Wooeee.

And this is based on this huge collection of titles, that all happen to be nvidia sponsored.
 
Last edited:
Flopper said:
fury x is equal in the gaming experience.
20% at 40fps is 8fps and seriously who plays games at 40fps?
Hardocp review guys
Click to expand...

It matters at 4k when you are doing everything you can to get to 60fps and the highest settings you can manage.
 
So none of these cards are playable at 4K ultra settings on a single card...its what you guys had eluded to speaking to the results on other threads. But from what ive read elsewhere, 2-4 of these cards will beat the 980ti /TitanX, so the RAM limitation cant be that limiting if its scaling better than a card with 3X the RAM at ultra settings right? Im sticking with my current setup till next gen but things are looking promising for HBM and CF setups.
 
Yeah I'd really like to see this same evaluation but with 2-way or 3-way multiGPU configurations. The problem is availability, where can you even get another Fury X right now?

Do you guys at the [H] have more than one?
 
Meeho said:
This card is too slow for 4K gaming, but I fail to see how this review demonstrates the 4 GB RAM being limiting. It was only in one game and a known exceptionally GPU RAM heavy one. None of the other games showed 4 GB having any negative influence. I was fully expecting Fury cards to have problems when it was announced that they would only have 4 GB, but the fact is that in current games they are way too slow before getting capacity limited.

So, please, if you're going to complain about Fury, do so on its processing abilities, not RAM capacity as far as currently observable performance goes. For future proofing? Maybe. Oh, and modded Skyrim possibly.
Click to expand...
Why would you risk buying a card for 4k if it's already being vram limited by games right out of the gate? How is this card even going to handle games in the coming year if it's already having problems?
 
Only games that mattered in the review were FC4 and BF4, I'd say it's doing quite well, smacked the 980ti in FC4, then lost in BF4, two of the most important game engines IMO.

Not sure why GTA V is being used it doesn't show anything besides how poor their engine is, patch 1.28 was just released to fix FPS issues they created. Dying Light is also only used in 2 games :\

Modred189 said:
I'm not going to copy the other charts, but they are essentially the same: The 980 Ti has higher settings and largely higher fps even with the higher settings.
How can you possibly say, "fury x is equal in the gaming experience"?
Click to expand...
Cherry picking 2 games does not an argument win make. Like I said above, I'd look more at what engines are being used as it tells a more likely tale of future game performance.

Also GTA V is a garbage game, people bitched about the Order 1886 being short, GTA V is repetitive after about 2 hours. The Assassin's creed of open world
 
Last edited:
shad0w4life said:
Only games that mattered in the review were FC4 and BF4, I'd say it's doing quite well, smacked the 980ti in FC4, then lost in BF4, two of the most important game engines IMO.

Not sure why GTA V is being used it doesn't show anything besides how poor their engine is, patch 1.28 was just released to fix FPS issues they created. Dying Light is also only used in 2 games :\


Cherry picking 2 games does not an argument win make. Like I said above, I'd look more at what engines are being used as it tells a more likely tale of future game performance.

Also GTA V is a garbage game, people bitched about the Order 1886 being short, GTA V is repetitive after about 2 hours. The Assassin's creed of open world
Click to expand...

The quality of the gameplay is entirely irrelevant to this conversation. But go back and look at the other games' performance and they tell the same story, higher performance with more options enabled.
That said, you could not be more wrong regarding the FarCry 4 game engine. Dunia has been around since Far Cry 3/2012 and has not be used outside of the Far Cry games afaik. That said, Frostbite is being used in quite a few games (thanks to EA of course).
SO let's look at that...
1437535126iwTl74Zfm5_7_1.gif

Oh look, higher in game settings and higher average fps. From the review's BF4 page:
Once again, following the other games, we had to make the most sacrifices in image quality on the AMD Radeon R9 Fury X in order to make it playable at 4K. We had to completely turn off all forms of AA. We could not run MSAA or even Post AA FXAA option.
Click to expand...
The GeForce GTX 980 Ti allowed us to improve image quality settings by turning everything to "Ultra" except for Effects and Post Processing, we had to leave those on "High."
Click to expand...
So, again, as I said, there's no way to say that the "fury x is equal in the gaming experience."

More eye-candy features+higher fps =/= equal gaming experience
 
yourgrandma said:
Why would you risk buying a card for 4k if it's already being vram limited by games right out of the gate?
Click to expand...

I wouldn't buy any of these cards for 4K. They are barely capable of proper 1440p.

How is this card even going to handle games in the coming year if it's already having problems?
Click to expand...

I don't know, I can only go by today's testing. Since it is not RAM limited for 4K today, I don't think it will soon be limited for 1080/1440p, and by the time it is, it will probably be slow enough to warrant an upgrade.

Maybe we will be flooded with shitty ports that will require 8 GB RAM for 1080p. Maybe common resolutions will never be RAM, but GPU limited with new processing features. Buy what works for you today.
 
as i said before the pcie bus IS STILL the bottleneck the card can have all the bandwidth in the world

its like hooking a fire hose kitchen sink
 
Elios said:
as i said before the pcie bus IS STILL the bottleneck the card can have all the bandwidth in the world

its like hooking a fire hose kitchen sink
Click to expand...

I'm curious to see why you think so. I've been wondering if the bus will ever really be a limitation on cards, as it seems the PCI-E spec maturation largely outpaces the cards' demands.
And ever since this article, I didn't see it changing.
http://www.hardocp.com/article/2012...gpu_gaming_performance_review/14#.VbZNLaRVhBc
 
Remon said:
+3.8 FPS



+2.8 FPS



+ 3.3 FPS



Wooeee.

And this is based on this huge collection of titles...
Click to expand...
Those frame times, though... Average framerate doesn't tell the whole story when the graph shows what could be interpreted as a stuttering and laggy mess.
 
psyclist said:
So none of these cards are playable at 4K ultra settings on a single card...its what you guys had eluded to speaking to the results on other threads. But from what ive read elsewhere, 2-4 of these cards will beat the 980ti /TitanX, so the RAM limitation cant be that limiting if its scaling better than a card with 3X the RAM at ultra settings right? Im sticking with my current setup till next gen but things are looking promising for HBM and CF setups.
Click to expand...

VRAM capacity limitations are only apparent when you look at the continuous FPS graph, not the avg FPS figures other site does. It causes textures swapping, which is where the problem is.

Elios said:
as i said before the pcie bus IS STILL the bottleneck the card can have all the bandwidth in the world

its like hooking a fire hose kitchen sink
Click to expand...
Which is why you would want larger VRAM, so you could fit all textures into the VRAM and not swap them while the GPU is rendering. It has always been that way and it's no different now. The aim is to fit all required texture on the card before rendering. PCI-E bus will never be fast enough for you to move textures in and out without resulting in the GPU having to wait for data.

As for all other data exchange during rendering, they are no where bandwidth extensive enough for current bus' bandwidth to be a limiting factor.
 
Meeho said:
I wouldn't buy any of these cards for 4K. They are barely capable of proper 1440p.
Click to expand...
What would you buy exactly then? Do you mean strictly the Fury X? The 980 Ti has been quite capable for me at 1440p, replacing my 290X CF setup at 1440p and it's nearly the same performance in a single card solution.
 
Elios said:
assuming you can load everything in system ram to start with...

ram ~23GB/s > cpu > pcie ~16GB/s > vram ~512GB/s

see the problem now?
Click to expand...

Yeah, we're not testing basic SLI/CFX scaling here using fixed benchmarks, we're testing high-resolution interactive gaming (where any delays loading assets can be noticed by the human playing the game). Those charts only capture the average framerates so you have something to compare against. The actual frame times can affect how playable a video card is. And Tech Report found that it was about as playable as the GTX 980 when worst-case frame times were taken into account (click link, press 99th percentile button):

http://techreport.com/review/28513/amd-radeon-r9-fury-x-graphics-card-reviewed/14

These guys at [H] are picky-as-hell when it comes to smooth framerates, but then again they test the highest-end cards for us, so they SHOULD BE! Things like high frame times can't be quantified in simple benchmarks reporting min/max/average fps, and you can't extrapolate the playability difference between two cards SOLELY FROM THOSE NUMBERS.

If you are convinced that average fps from fixed benchmarks are all you need to know, I suggest you get your complaining ass over to Anandtech. Oh wait, they didn't find much to like either:

http://www.anandtech.com/show/9390/the-amd-radeon-r9-fury-x-review/27
 
Last edited:
You must log in or register to reply here.
Back
Top