Dual-GPU face off: Gainward GeForce GTX 980 Ti SLI vs. Radeon R9 Fury X Crossfire

Steve

Posts: 3,035   +3,142
Staff member

We got our first look at Nvidia's GTX 980 Ti reference card almost two months ago and up until now we had yet to check out a board partner's production model. Compared to reference designs by Nvidia and AMD, their partners usually come up with creations that run cooler and quieter.

A perfect example of this is Gainward's GTX 980 Ti "Golden Sample", which we have on hand today, featuring a 15% factory overclock that Gainward says provides 11% more performance on average. The GTX 980 Ti GS has a "Zero RPM fan design," meaning all three fans can be stopped until the GPU temperature rises to 60°C or above. Gainward also says its card operates 10°C cooler and is 6dB quieter under factory overclocked conditions.

Not wanting to simply just revisit the GTX 980 Ti's performance, we'll be hooking up two of Gainward's cards for some SLI testing and to make the numbers even more interesting, we'll add a pair of AMD's Radeon R9 Fury X cards in Crossfire.

Read the complete review.

 
People buying a pair of these are certainly enthusiasts. Therefore, 980 Tis are a better option than FuryX's, especially since the Furys don't seem to overclock at all.

At least, 980 Tis seem worth over Fury Xs, but only if you overclock. Otherwise, at stock speeds, the FuryXs provide better performance overall.

PS: Maybe change to Skylake for CPU limited games ?
 
People buying a pair of these are certainly enthusiasts. Therefore, 980 Tis are a better option than FuryX's, especially since the Furys don't seem to overclock at all.

At least, 980 Tis seem worth over Fury Xs, but only if you overclock. Otherwise, at stock speeds, the FuryXs provide better performance overall.

PS: Maybe change to Skylake for CPU limited games ?

The testing was done 3 weeks before Skylake's release and I don't think Skylake would help much unless heavily overclocked. Just one game was CPU limited of the bunch anyway.
 
People buying a pair of these are certainly enthusiasts. Therefore, 980 Tis are a better option than FuryX's, especially since the Furys don't seem to overclock at all.

At least, 980 Tis seem worth over Fury Xs, but only if you overclock. Otherwise, at stock speeds, the FuryXs provide better performance overall.

PS: Maybe change to Skylake for CPU limited games ?

The testing was done 3 weeks before Skylake's release and I don't think Skylake would help much unless heavily overclocked. Just one game was CPU limited of the bunch anyway.

Steve, is Techspot looking at doing a follow up review for Skylake? Of all the reviews I've read so far, Techspot's was the harshest by some margin. It could be worth looking at again in the near future once more boards have come out along with the odd firmware/driver update.
 
From what I've seen skylake won't change a think, not unless it can give a 20%+ boost in performance vs 5960X which lets be honest, it wont.
 
Steve, is Techspot looking at doing a follow up review for Skylake? Of all the reviews I've read so far, Techspot's was the harshest by some margin. It could be worth looking at again in the near future once more boards have come out along with the odd firmware/driver update.

I am doing some clock-for-clock testing now. That said I have read all the other reviews and seen their results. Taking that into account I don't see any reason to change my conclusion, the things that I mentioned to be disappointing are and I feel I got the positive stuff right as well.

It looks like my 4790K was running the Turbo clocks too aggressively, though the 6700K was spot on. Anandtech pointed out that clock-for-clock they found the 6700K slower in games so I am interested to see what I find. Of course memory speed plays a big roll and you really want to go with a minimum of DDR4-3000 it seems.

The 6700K hasn't made its way onto newegg.com yet but here in Australia it costs around $50 more than the 4790K which is nuts. At 10% more expensive you don't get 10% more performance and then of course you have to factor in the motherboard and memory price.

Overall I think Skylake and its supporting platform is solid. Pricing just needs to become a little more aggressive which I am sure it will with time and DDR4 needs to get a bit cheaper as well. Still anyone with a Haswell or Ivy Bridge processor will be wasting their time/money with Skylake which is disappointing.
 
Nice review, interesting to see just how close Nvidia and AMD are with their Top tier cards, minus the Titan of course.

Now we just need the Fury X2 to come out and have our selves a battle of the $1K cards. And then crossfire those of course.
 
People buying a pair of these are certainly enthusiasts. Therefore, 980 Tis are a better option than FuryX's, especially since the Furys don't seem to overclock at all.

At least, 980 Tis seem worth over Fury Xs, but only if you overclock. Otherwise, at stock speeds, the FuryXs provide better performance overall.

PS: Maybe change to Skylake for CPU limited games ?

The testing was done 3 weeks before Skylake's release and I don't think Skylake would help much unless heavily overclocked. Just one game was CPU limited of the bunch anyway.
Maybe I missed it but did you test this on Windows 10? and would skylake on W10 make any difference?
 
People buying a pair of these are certainly enthusiasts. Therefore, 980 Tis are a better option than FuryX's, especially since the Furys don't seem to overclock at all.

At least, 980 Tis seem worth over Fury Xs, but only if you overclock. Otherwise, at stock speeds, the FuryXs provide better performance overall.

PS: Maybe change to Skylake for CPU limited games ?

The testing was done 3 weeks before Skylake's release and I don't think Skylake would help much unless heavily overclocked. Just one game was CPU limited of the bunch anyway.
Maybe I missed it but did you test this on Windows 10? and would skylake on W10 make any difference?

The test was done on Windows 8.1 its listed on the spec page.
 
This should have come out a few months later, when the Fury X Drivers are more mature and so that you could actually compare non-reference to non-reference. As it stands right now the verdict claims the obvious, a non-reference card beats a reference one with newer drivers in overclocking.

Given that HBM is a brand new tech, I would not be surprised to see huge driver gains to be coming.
 
This should have come out a few months later, when the Fury X Drivers are more mature and so that you could actually compare non-reference to non-reference. As it stands right now the verdict claims the obvious, a non-reference card beats a reference one with newer drivers in overclocking.

Given that HBM is a brand new tech, I would not be surprised to see huge driver gains to be coming.
Just like how DX12 is supossed to make fury x better than 980ti, just like how HBM was going to give amd a huge advantage, just like how fiji was an "overclockers dream"?

The world is not going to sit idly, twiddling their thumbs and wait for AMD to pull their head out of their crevasse. If AMD can't straighten out their drivers, and deliver what they promise, that's their fault. By the time the drivers are finally finished, pascal will be nearing release.

If someone is building a new pc, they are not going to think to themselves "golly gee, im going to wait for AMD to properly support their card, and then purchase a graphics card based on which model is superior" they will think"ok, what card is the best I can get RIGHT NOW, given the budget?" People who buy $650 cards expect them to work when they are purchased, not at some random point.
 
This should have come out a few months later, when the Fury X Drivers are more mature
If AMD were that concerned they would have launched the card when they were satisfied with the drivers. Are you seriously expecting a moratorium on reviews until the drivers are fully optimized? What is the point of launching hardware but requiring a blackout on comparative reviews?
and so that you could actually compare non-reference to non-reference.
AMD have themselves said that the Fury X will not be offered as anything other than reference design. The amount of lockdown AMD have exercised over Fiji is unprecedented - especially by the company itself- so I wouldn't hold your breath waiting for custom designs deviating from AMD's specification.
 
If AMD were that concerned they would have launched the card when they were satisfied with the drivers. Are you seriously expecting a moratorium on reviews until the drivers are fully optimized? What is the point of launching hardware but requiring a blackout on comparative reviews?

AMD have themselves said that the Fury X will not be offered as anything other than reference design. The amount of lockdown AMD have exercised over Fiji is unprecedented - especially by the company itself- so I wouldn't hold your breath waiting for custom designs deviating from AMD's specification.

I never said everyone should hold reviews, and this isn't a review. As I said in my post, this article doesn't say anything we didn't already know. It's funny how you sensationalize everything to be so dramatic, somehow you gathered that I wanted all reviews embargoed, when I said nothing akin.

"If AMD were that concerned they would have launched the card when they were satisfied with the drivers."

It's funny that you mention that, because I remember Nvidia having driver issues with many graphics cards when they were first released. Stop trying to apply a double standard to AMD.

Just like how DX12 is supossed to make fury x better than 980ti, just like how HBM was going to give amd a huge advantage, just like how fiji was an "overclockers dream"?

The world is not going to sit idly, twiddling their thumbs and wait for AMD to pull their head out of their crevasse. If AMD can't straighten out their drivers, and deliver what they promise, that's their fault. By the time the drivers are finally finished, pascal will be nearing release.

If someone is building a new pc, they are not going to think to themselves "golly gee, im going to wait for AMD to properly support their card, and then purchase a graphics card based on which model is superior" they will think"ok, what card is the best I can get RIGHT NOW, given the budget?" People who buy $650 cards expect them to work when they are purchased, not at some random point.

I believe AMD has already delivered with Fiji. It's the first generation of HBM to market. It's unreasonable to think such a dramatic change in memory architecture wouldn't take time to fully optimize for. You can try and paint AMD as lazy all you want but I highly doubt you would be involved in the implications of driver design. Even more, I believe Nvidia has many such examples of similar scenarios in the past, would you still call them lazy?

"they will think"ok, what card is the best I can get RIGHT NOW, given the budget?" People who buy $650 cards expect them to work when they are purchased, not at some random point."

So what your saying is that people don't care about their rig's performance in the future? Cuz everyone has $600 to drop on a top end card every year. Fact is, expected future performance does play a factor in buying a card. Any notion otherwise is just shortsighted stupidity. I'm guessing that you don't speak for every enthusiast.
 
It's funny that you mention that, because I remember Nvidia having driver issues with many graphics cards when they were first released. Stop trying to apply a double standard to AMD.
Newsflash guy. Most graphics hardware has driver issues at launch, or with new games/game engines/features, or a combination of both - the difference is most people don't resort to saddo straw man arguments.

Show me an instance where I have applied a different metric regarding Nvidia graphics cards.

When you don't find me making excuses and calling unfair because of drivers, I hope you'll have the good grace to offer an apology as quickly as throw around accusations.

/wont be holding my breath.

I'm guessing that you don't speak for every enthusiast.
I'm guessing you don't either.
Moreover, there is more than enough anecdotal evidence to show that enthusiasts place a higher value on overclocking and performance in the days and weeks following launch. Any perusal of enthusiast forums and benchmark result ladders would tend to reinforce that point. Check out any enthusiast card thread and see what kind of activity is evident 4, 5, or 6 months after it was initiated.
 
As I said in my post, this article doesn't say anything we didn't already know.

You know more than me then, but I am sure that is hardly surprising ;)

This should have come out a few months later, when the Fury X Drivers are more mature and so that you could actually compare non-reference to non-reference. As it stands right now the verdict claims the obvious, a non-reference card beats a reference one with newer drivers in overclocking.

Given that HBM is a brand new tech, I would not be surprised to see huge driver gains to be coming.

If I had a dollar for every time a fan of the red team cried immature drivers, I would have many more dollars than I do now. The Fury X was released 6 weeks ago now and since it doesn’t use a new architecture how long do we have to wait for mature drivers? Since you are telling us to wait please put a number on it. Also how do you suppose HBM and drivers are in any way related?

As others have already pointed out there isn’t going to be a non-reference Fury X, at least not according to AMD so how long do we need to wait for that?

Obviously if something drastic changes we will re-evaluate but it is crazy to suggest we hold off telling our readers which the better buy is right now on the off chance that something changes in the future, something unlikely I might add.
 
You know more than me then, but I am sure that is hardly surprising ;)



If I had a dollar for every time a fan of the red team cried immature drivers, I would have many more dollars than I do now. The Fury X was released 6 weeks ago now and since it doesn’t use a new architecture how long do we have to wait for mature drivers? Since you are telling us to wait please put a number on it. Also how do you suppose HBM and drivers are in any way related?

As others have already pointed out there isn’t going to be a non-reference Fury X, at least not according to AMD so how long do we need to wait for that?

Obviously if something drastic changes we will re-evaluate but it is crazy to suggest we hold off telling our readers which the better buy is right now on the off chance that something changes in the future, something unlikely I might add.

"Also how do you suppose HBM and drivers are in any way related?"

Entirely new memory architecture is going to be much harder that simply changing the timings for a GDDR5 interface. I don't really get how it's hard to grasp that new technologies like this have taken awhile to optimize for. Remember when multi-core processors first came out? Hyper-Threading? Tessellation Units on GPUs?

I hate sounding like a broken record but I did NOT say this article should be held and no amount of "fanboy this or fanboy that" is going to change it. It's shameful to play the fanboy card. In general, the driver card shouldn't work. In this case, when introducing a new technology like HBM, AMD deserves some slack.

Newsflash guy. Most graphics hardware has driver issues at launch, or with new games/game engines/features, or a combination of both - the difference is most people don't resort to saddo straw man arguments.

Show me an instance where I have applied a different metric regarding Nvidia graphics cards.

When you don't find me making excuses and calling unfair because of drivers, I hope you'll have the good grace to offer an apology as quickly as throw around accusations.

/wont be holding my breath.


I'm guessing you don't either.
Moreover, there is more than enough anecdotal evidence to show that enthusiasts place a higher value on overclocking and performance in the days and weeks following launch. Any perusal of enthusiast forums and benchmark result ladders would tend to reinforce that point. Check out any enthusiast card thread and see what kind of activity is evident 4, 5, or 6 months after it was initiated.

So based on your first paragraph, why are you even complaining? If AMD drivers are having issues and you expect it, what are you doing singling them out for it.

"the difference is most people don't resort to saddo straw man arguments."

I've "conversed" with you many times on this forum and you always to result to petty generalization when you can't tactfully provide a cogent reply. This adds nothing to what has been said.

"Show me an instance where I have applied a different metric regarding Nvidia graphics cards."

This conversation isn't about you, nor would me doing so add anything to the conversation. Chance to derail denied.

"When you don't find me making excuses and calling unfair because of drivers, I hope you'll have the good grace to offer an apology as quickly as throw around accusations."

You're a rather conceited person, aren't you?

Your last paragraph completely avoid the topic. You said

"I'm guessing you don't either.
Moreover, there is more than enough anecdotal evidence to show that enthusiasts place a higher value on overclocking and performance in the days and weeks following launch. Any perusal of enthusiast forums and benchmark result ladders would tend to reinforce that point. Check out any enthusiast card thread and see what kind of activity is evident 4, 5, or 6 months after it was initiated."

but it doesn't make any sense given the conversation was this

----
"they will think"ok, what card is the best I can get RIGHT NOW, given the budget?" People who buy $650 cards expect them to work when they are purchased, not at some random point."

So what your saying is that people don't care about their rig's performance in the future? Cuz everyone has $600 to drop on a top end card every year. Fact is, expected future performance does play a factor in buying a card. Any notion otherwise is just shortsighted stupidity. I'm guessing that you don't speak for every enthusiast.
----

You complete avoid the great blunder you made and refuse to even try talk about the fact that you said enthusiasts don't care about future performance. As if the longevity doesn't matter on a card as expensive as that.

"Moreover, there is more than enough anecdotal evidence to show that enthusiasts place a higher value on overclocking and performance in the days and weeks following launch. "

Look who's speaker for others again. Of course, you made sure to avoid the issue I pointed out above and add "in the days and weeks following launch" to avoid this point from being subject to the same problem.
 
I'm noticing that the games that AMD wins by "lots" are still unplayable for both cards... I believe Total War: Attila didn't even get to 30FPS... who cares which card is faster if they can't play it... I'd be throwing out any results where the framerates are below the 30FPS threshold...

Although it's not part of the test, would like to see how 2 Titan Xs in SLI match up here....
 
So based on your first paragraph, why are you even complaining?
I obviously wasn't - I merely pointed out that your driver excuses were lame pseudo-PR...
This should have come out a few months later, when the Fury X Drivers are more mature
You must really enjoy trolling.
If AMD drivers are having issues and you expect it, what are you doing singling them out for it.
I obviously wasn't - note my post - a direct response to yours-that you actually quoted:
Newsflash guy. Most graphics hardware has driver issues at launch...
...it obviously isn't vendor specific...whereas you're bleating on about how unfair it is that a review featuring released hardware and released drivers is being released before AMD's drivers are at the pinnacle of their effectiveness.
"Show me an instance where I have applied a different metric regarding Nvidia graphics cards." This conversation isn't about you, nor would me doing so add anything to the conversation.
If you're accusing someone of double standards, then you need to show now they are doing so. You aren't - and thus, you're trolling.
ThwHNjW.jpg
You complete avoid the great blunder you made and refuse to even try talk about the fact that you said enthusiasts don't care about future performance. As if the longevity doesn't matter on a card as expensive as that.
I never said any such thing - as you've just quoted:
enthusiasts place a higher value on overclocking and performance in the days and weeks following launch.
Higher value during the initial phase of ownership disinterest in future performance.
Comprehension fail on your part.....again.

Feel free to continue bleating, but I'm done. Your incessant hyperbole, trolling, fanboyism, and innate lack of comprehension make any dialogue an exercise in nothing more than reinforcing your ignorance, and a diversion from the review this thread is supposed to be concerned with.
I'm noticing that the games that AMD wins by "lots" are still unplayable for both cards... I believe Total War: Attila didn't even get to 30FPS... who cares which card is faster if they can't play it... I'd be throwing out any results where the framerates are below the 30FPS threshold...
The settings can be toned down for actual gameplay. The review is designed to remove the CPU and memory subsystem as the limiting factor. It is a fine line juggling between CPU/RAM saturation (which would present a false picture of the graphics ability - especially if an f.p.s. cap makes an appearance at the top end), and VRAM saturation - which doesn't so much highlight a large framebuffer, than cripple a smaller one.
Although it's not part of the test, would like to see how 2 Titan Xs in SLI match up here....
I'm sure Steve would gratefully accept the donation of a second Titan X. The only real difference would be in a very limited number of games - many of which would be artificially weighted (by texture packs to overwhelm the 4 or 6GB of the other cards) to accentuate the Titan X's 12GB of GDDR5. In the majority of games I've seen tested, the Titan X's extra shaders and VRAM is offset by the 980 Ti's higher (and more consistent) boost clocks (Here, for example)
 
Last edited:
Anyone who blames driver issues at this point is a fool - the only reason drivers should be an issue is for a brand new game - that is the only time you can allow for Nvidia/AMD to have a few days to come up with optimized drivers...

When a company releases a major piece of hardware for more than $600, there is no reason why their drivers shouldn't work... The Fury has now been around for awhile, and blaming drivers for their lack of performance is just being a naive jerk...

Yes, I expect driver updates to improve performance in the future.... but that's no reason not to hold AMD accountable for their performance RIGHT NOW.
 
Is it me I feel that $1600 for video cards and power supply should be better than this. Add the rest of the build and a $3500.00 plus build barely does 4k at playable rates?
 
Yeah nothing new to see here. If you want something small or quiet, and you don't overclock - get the Fury X. Otherwise once the dust has settled the 980 Ti is ~10-15% faster.

It just depends on what your needs/wants are.
 
Back