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

but its possible in fact, nvidia are making that move with any card with asynchronous memory controller and asymmetrical memory configurations.. if nvidia can do that, i don't see a reason why AMD can't do it to keep fresh the memory pool.. bigger bus and faster bandwidth allow faster swap of textures which certainly can help, so basically is possible.. also I think not all games are so sensitive to 4GB vRAM so the work in a basis game to game can also work.. but we know slow are AMD in their driver development, don't work as good as nvidia in that aspect..

NVidia cant do it.
The drivers were updated so games end up limited to 3.5GB unless pushed very very hard, and then performance tanks.
And they have a local 0.5GB cache that is used in conjunction with system memory to give a slightly faster cache speed.
Fury has no local cache, it is solely dependent on the speed of system ram once it runs out of its 4GB.
 
NVidia cant do it.
The drivers were updated so games end up limited to 3.5GB unless pushed very very hard, and then performance tanks.
And they have a local 0.5GB cache that is used in conjunction with system memory to give a slightly faster cache speed.
Fury has no local cache, it is solely dependent on the speed of system ram once it runs out of its 4GB.

they can, the memory pool refreshing time can be really adjusted/modified per game, if they decide to work in a game to game basis as Rizen said, they can work to brute force the textures to out faster of the memory pool when not used, but it require to deeply work with the games to make the card decide really which kind of textures should be stored longer than others.. some games are just coded to use as much vRAM as available even if don't need it, Bioshock infinite is a good example, they can just work to know what is needed what is not..
 
they can, the memory pool refreshing time can be really adjusted/modified per game, if they decide to work in a game to game basis as Rizen said, they can work to brute force the textures to out faster of the memory pool when not used, but it require to deeply work with the games to make the card decide really which kind of textures should be stored longer than others.. some games are just coded to use as much vRAM as available even if don't need it, Bioshock infinite is a good example, they can just work to know what is needed what is not..

It will require the driver to be specifically tuned for every game to make use of the feature.
Otherwise it wont know which assets are expendable. This assumes it is possible as well, some games might not be easy targets for this tech. It will require very good prediction.
It will also require setting aside part of the 4GB as a cache which could make the problem worse for those games it isnt optimised for.

Once the next gen cards come out, this card will end up on the back burner and may be left to rot a bit early.
 
I really appreciate Brent Justice and Kyle Bennett review on the radeon fury X.

I have to say I'm disappointed coming from EVGA gtx 680 classified. I recently upgraded everything except video card awaiting arrival of the fury x or possible new nvidia card.

For me at this time based on everything I have read and seen recently GTX 980ti.

I really wanna thank everyone here for various opinions as well.

I'm committed to performance and right now today gtx 980ti just works well in what it does and that is performs.

I would welcome a AMD hardcore siege on the market gpus in recent time especially the highest end have risen to insane levels.

Again thank you all definitely looking forward to more opinions and further reviews from these guys!!!
 
Last edited:
Too bad you only tested 4K apples-to-apples. Would have loved to see a real test on how 980ti and Fury X performs at 4K with playable settings. I'm gonna run my 4960x1600p PLP setup with single Fury X for now and that's almost 4K in pixels, so it would have been nice to see how much eyecandy you actually have to disable to get things playable.
 
Last edited:
Too bad you only tested 4K apples-to-apples. Would have loved to see a real test on how 980ti and Fury X performs at 4K with playable settings. I'm gonna run my 4960x1600p PLP setup with single Fury X for now and that's almost 4K in pixels, so it would have been nice to see how much eyecandy you actually have to disable to get things playable.


Yeah, its a goddamn shame.
 
Too bad you only tested 4K apples-to-apples. Would have loved to see a real test on how 980ti and Fury X performs at 4K with playable settings. I'm gonna run my 4960x1600p PLP setup with single Fury X for now and that's almost 4K in pixels, so it would have been nice to see how much eyecandy you actually have to disable to get things playable.

So.. testing based on subjective criteria? A little unrealistic. You understand that would take 100 years and still no one would be happy.

The performance delta between "apples to apples" and what you personally consider "playable" is going to pretty much remain the same percentage anyway, so run calc.exe and extrapolate and you'll be in the ballpark.
 
Fury could turn out to be even better with mature drivers and amd commitment to it. I'm just saying today many of us especially those like myself if you have been on the sideline for a bit waiting until everything drops you just kinda like meh but still optimistic as well.
 
Fury could turn out to be even better with mature drivers and amd commitment to it. I'm just saying today many of us especially those like myself if you have been on the sideline for a bit waiting until everything drops you just kinda like meh but still optimistic as well.

to the moment when Fury Make any big advance in drivers, Nvidia's Pascal with HBM2 will be already in the market crushing the Fury. =). Remember Fury is GCN, GCN have already very very matured drivers. it's no a new architecture to xploit via drivers. :eek:
 
True I'm really starting to feel AMD might not be able to catch up and if that happens it will be bad for us all.
 
to the moment when Fury Make any big advance in drivers, Nvidia's Pascal with HBM2 will be already in the market crushing the Fury. =). Remember Fury is GCN, GCN have already very very matured drivers. it's no a new architecture to xploit via drivers. :eek:


Your right I forget GCN has been around for sometime already.
 
Too bad you only tested 4K apples-to-apples. Would have loved to see a real test on how 980ti and Fury X performs at 4K with playable settings. I'm gonna run my 4960x1600p PLP setup with single Fury X for now and that's almost 4K in pixels, so it would have been nice to see how much eyecandy you actually have to disable to get things playable.

Problem with "playable settings" is what is playable on 1 card might not be playeable one the other.

to the moment when Fury Make any big advance in drivers, Nvidia's Pascal with HBM2 will be already in the market crushing the Fury. =). Remember Fury is GCN, GCN have already very very matured drivers. it's no a new architecture to xploit via drivers. :eek:

um that is pretty stupid logic, gcn 1.0 to 1.1 or even 1.2 1.3 etc are not same. If you say that then maxwell and kepler and even fermi were same since they are all based on using cuda cores. they maybe all called GCN but there are changes in arc that makes what might work for 1.0 not work for 1.3.

Pretty weird how big off AMD's PR slide was from independent reviews. Seems like what AMD claimed and what most reviews see is a good 15-20%. This i would guess goes back to first person i quoted about playable settings, maybe AMD ran it that way not using same settings for both?
http://cdn.vrworld.com/wp-content/uploads/2015/06/AMD_Fiji_GPU_RadFuryXPerf.jpg
 
Last edited:
Problem with "playable settings" is what is playable on 1 card might not be playeable one the other.



um that is pretty stupid logic, gcn 1.0 to 1.1 or even 1.2 1.3 etc are not same. If you say that then maxwell and kepler and even fermi were same since they are all based on using cuda cores. they maybe all called GCN but there are changes in arc that makes what might work for 1.0 not work for 1.3.

Pretty weird how big off AMD's PR slide was from independent reviews. Seems like what AMD claimed and what most reviews see is a good 15-20%. This i would guess goes back to first person i quoted about playable settings, maybe AMD ran it that way not using same settings for both?
http://cdn.vrworld.com/wp-content/uploads/2015/06/AMD_Fiji_GPU_RadFuryXPerf.jpg

Good points didn't look at things in that way.
 
AMD's benches were super sketchy. They ran 0xAF in most of the games tested. Who runs without AF?
 
After a couple of days of reading reviews, it's really not a bad card at all, it's not even off by $100, more like $50, @ $600 this would still be a good card, considering the cooling solution + performance.
IF this card had come out before the 980TI, I think it would have been much better received, Kudos to NV for raining on AMDs parade.
 
AMD's benches were super sketchy. They ran 0xAF in most of the games tested. Who runs without AF?

Makes you wonder but im sure nvidia has ran some weird test to boost things. Just right now today I feel bad for AMD they go down man that would just suck,

I'm hopeful that next year both companies will expand on this technology that AMD has brought forth.
 
Problem with "playable settings" is what is playable on 1 card might not be playeable one the other.

So.. testing based on subjective criteria? A little unrealistic. You understand that would take 100 years and still no one would be happy.

But [H]'s reviews are heavily based upon what is considered "playable." I don't think its too unrealistic to put the settings on the lower card's "playable" and run the test, then the higher card's "playable" and test again.

It is more work for [H] and if the site decides the benefit isn't worth the cost, then whatever, I'll take what I can get. But this site does try to define what "playable" is.
 
Ah, fair enough.
I didnt use Furmark on my 290x because it ages cards too quick.

It doesnt change that AMD probably locked the voltage because the VRMs will quickly end up being the limiting factor.
Why the hell havent they given them any cooling, it might well have been an overclocking demon.

Perhaps there are other concerns with the memory interface if the GPU voltage is raised.
There may be an incompatibility or leakage that upsets the memory, not helped by the direct connection.

I imagine some home brewed cooling solutions will emerge and some manual voltage mods so we should find out its real capabilites in a few weeks/months.

Yeah I actually think the cooling system is one of the biggest weaknesses of Fury X. It looks like many of the VRMs are only passively cooled by the backplate, there's no actual water running over them, so you'd expect those VRMs to get extremely hot. Probably why it's such a poor overclocker.

The normal fury might prove to be a lot better for those reasons as well because they would probably have some better cooling on the VRMs..
 
Doesn't beginning of article say there will be follow-up 4k articles?

Also:

> We only got our Fury X card Saturday morning. We would have had it Friday, but AMD sent the card out signature required, and did not give us notice of shipment or tracking number. Brent was burning both ends of the candle to get out what we did.

> We are going to finish up a 390X overclocking article then we will move back to Fury X coverage that is a bit more granular at the 4K resolution, which we think its deserves.
 
Ah so there is going to be 4K article coming up at some point :)

And Brent, I didn't mean to be rude or overly pushy when I said that I would have loved to see "proper" 4K tests as well. I know it takes time and dedication to get the reviews out and it seems you didn't really have that much time before the release day.
 
But [H]'s reviews are heavily based upon what is considered "playable." I don't think its too unrealistic to put the settings on the lower card's "playable" and run the test, then the higher card's "playable" and test again.

It is more work for [H] and if the site decides the benefit isn't worth the cost, then whatever, I'll take what I can get. But this site does try to define what "playable" is.

They also give you the hard numbers as well so you can make the decision yourself.
 
Ah so there is going to be 4K article coming up at some point :)

And Brent, I didn't mean to be rude or overly pushy when I said that I would have loved to see "proper" 4K tests as well. I know it takes time and dedication to get the reviews out and it seems you didn't really have that much time before the release day.

Our focus also was not entirely spent at 4K either.

I for one will not ignore 1440p gamers, that is the resolution where you are able to mostly maximize settings in most games with new cards, and pretty much the resolution to use if you want great looking PC gaming with maximum game settings at acceptable performance. 4K, you cannot max out games, the GPUs aren't powerful enough yet. 1080p is mostly a given, every game will perform great at 1080p on high-end cards, but 1440p presents a challange still for some, and is IMO the best test of a graphics card. 4K gaming as I said is growing, but it is no where near the saturation levels of 1440p and 1080p PC gaming.
 
Problem with "playable settings" is what is playable on 1 card might not be playeable one the other.
Thats why each card uses different settings when "what is playable on 1 card might not be playeable one the other".
That is the WHOLE point of the playable settings.
 
Thats why each card uses different settings when "what is playable on 1 card might not be playeable one the other".
That is the WHOLE point of the playable settings.

Sort of like when people come out and say, hey, you guys got different results than all the other (canned benchmark) sites! No shit Sherlock, if I could get real world gaming results data using a canned benchmark, I would save the time and money and do it that way.
 
What's worst is that people fail to see that the FuryX trails the 980 Ti considerably in titles such as BF4 and GTA 5. Those are not gamework titles and hence no associations with nVidia.

This. The delta between Fury and 980Ti is fairly similar game to game. If you all of a sudden saw the fury going from neck and neck in all titles then drop off in Gameworks titles, there would be something too that.

But the Ti is leading by those types of numbers fairly across the board.
 
It looks like many of the VRMs are only passively cooled by the backplate...

Cooled? More like insulated. Look at the disassembly pics. There's no thermal pads between the two and I'm pretty sure there's no contact. Otherwise, the thermal images of it on and off wouldn't be so night and day. Plus, I've seen people refer to it as "rubberized" and even "plastic", making it even less likely that it's a cooling component. Just a really dumb design decision by AMD.

Problem with "playable settings" is what is playable on 1 card might not be playeable one the other.

That's kind of the whole point. Have you read an [H] video card review?

Seems like what AMD claimed and what most reviews see is a good 15-20%.

It surprises you that AMD cherry picked benchmarks and settings that showed them in a favorable light? Anybody who believes any company's claim about their own product's performance is delusional.

Ah so there is going to be 4K article coming up at some point :)

And Brent, I didn't mean to be rude or overly pushy when I said that I would have loved to see "proper" 4K tests as well. I know it takes time and dedication to get the reviews out and it seems you didn't really have that much time before the release day.

Given the time constraints and the inability of really any single card to perform well at 4K with anything resembling max settings, I think they made the right decision. Well, that, and I've got a 1440p monitor. ;)
 
Given the time constraints and the inability of really any single card to perform well at 4K with anything resembling max settings, I think they made the right decision. Well, that, and I've got a 1440p monitor. ;)

Running 1600p all this time, for new graphically intense games, even a single card for that to push max settings is not going to happen most of the time, and that hasn't changed the near 10 years now I've had this monitor.

I just don't see 4k being feasible single-card outside of some quantum leap in GPU power; otherwise it will always be a multi-GPU setup.
 
I just don't see 4k being feasible single-card outside of some quantum leap in GPU power; otherwise it will always be a multi-GPU setup.

I don't know about always. 2016 and 16nm should bring some big improvements to the table.
 
You just need to wait for Pascal HBM2 for that. Nvidia know what they are doing. None of this AMD Fiji HBM1 rubbish.
 
https://youtu.be/8hnuj1OZAJs?t=91
https://youtu.be/8hnuj1OZAJs?t=136

I was just watching some gameplay footage with this card at 4K. Look at what happens at the 1:39 and 2:20 mark, massive stutter. It's like the card is running out of VRAM because the usage goes down a lot when this happens, seems like it's swapping new assets in/out.

Massive.. massive.. stuttering at 4K.

Now you cannot call that a driver problem. End users will prove [H] to be correct.
 
What's worst is that people fail to see that the FuryX trails the 980 Ti considerably in titles such as BF4 and GTA 5. Those are not gamework titles and hence no associations with nVidia.

BF4 is shocking. I've read too many reviews, so I can't recall where this reported, but in some reviews it's just an ass kicking. I don't understand how this didn't raise a red flag.
 
Our focus also was not entirely spent at 4K either.

I for one will not ignore 1440p gamers, that is the resolution where you are able to mostly maximize settings in most games with new cards, and pretty much the resolution to use if you want great looking PC gaming with maximum game settings at acceptable performance. 4K, you cannot max out games, the GPUs aren't powerful enough yet. 1080p is mostly a given, every game will perform great at 1080p on high-end cards, but 1440p presents a challange still for some, and is IMO the best test of a graphics card. 4K gaming as I said is growing, but it is no where near the saturation levels of 1440p and 1080p PC gaming.

I, for one, appreciate the 1440p testing. 4k is still grasping a bit (IMO). The work needed for a good 4k experience is beyond most, and properly a part of those who frequent [H]. ;) But, for reviews, having all 3 (1080p, 1440p, and 4k) is still important.

Thanks.
 
ARGH!! The only Kung Fury happening here is TO AMD's Fury x card!

Bow down to the mighty 980 Ti, credit where credit is due

I still have "some" hope for the cutdown's of Fiji (Fiji Pro, Fiji Nano) with respect to the partners getting invovled and perhaps having the ability to tweak those cards. The cost reduction from an air cooler, etc may help these cards achieve better bang for the buck. Remains to be seen how much freedom they will have to customize there designs, in addition can a Fiji core with some 500 stream procs removed be able to run with a 980? I am not so sure... I am not so sure even if they unlocked the voltage and let people go ham it would matter as those same people already have a plethora of 980s to go ham on.

Thank you [H]ardocp for the review and honest opinions! Kung Fury indeed lol.
 
ARGH!! The only Kung Fury happening here is TO AMD's Fury x card!

Bow down to the mighty 980 Ti, credit where credit is due

I still have "some" hope for the cutdown's of Fiji (Fiji Pro, Fiji Nano) with respect to the partners getting invovled and perhaps having the ability to tweak those cards. The cost reduction from an air cooler, etc may help these cards achieve better bang for the buck. Remains to be seen how much freedom they will have to customize there designs, in addition can a Fiji core with some 500 stream procs removed be able to run with a 980? I am not so sure... I am not so sure even if they unlocked the voltage and let people go ham it would matter as those same people already have a plethora of 980s to go ham on.

Thank you [H]ardocp for the review and honest opinions! Kung Fury indeed lol.

Fury is balls

https://youtu.be/bS5P_LAqiVg?t=25m46s
 
https://youtu.be/8hnuj1OZAJs?t=91
https://youtu.be/8hnuj1OZAJs?t=136

I was just watching some gameplay footage with this card at 4K. Look at what happens at the 1:39 and 2:20 mark, massive stutter. It's like the card is running out of VRAM because the usage goes down a lot when this happens, seems like it's swapping new assets in/out.

Massive.. massive.. stuttering at 4K.

Now you cannot call that a driver problem. End users will prove [H] to be correct.

Its like watching the 770 reviews of Watch Dogs....
 
Back
Top