AMD Radeon R9 Fury X 4GB Review

[Elvis_Marmaduke]

Why are people comparing a stock graphics card with unreleased drivers to another's aftermarket and overclocked, and seemingly making a decision based on that? Especially when, in like the 5th post it's stated the test wasn't even using the latest drivers? I think that's silly, and shows poor decision management. I, for one, will be waiting a couple months before making a decision.

 

[tomc100]

I can't help but feel a little disappointed. Well at least tomshardware paints a brighter picture than some of the other sites like the techreport which is just doom and gloom. Now I don't feel so bad for buying that EVA GTX 970 OC for $280 and abandoning AMD after decades. My last Nvidia gpu was the first TNT and have always stuck with ATI but they just can't seem to compete with constant delays and lack of support for the newer games. Meanwhile it seems Nvidia is bribing every developer to use their gameworks technology. If the non-water cooled Fury can best the GTX 970 then I might switch back but right now I'll have to stick with Nvidia even though I detest that company.

 

[jimmysmitty]

Why are people comparing a stock graphics card with unreleased drivers to another's aftermarket and overclocked, and seemingly making a decision based on that? Especially when, in like the 5th post it's stated the test wasn't even using the latest drivers? I think that's silly, and shows poor decision management. I, for one, will be waiting a couple months before making a decision.

If you read the article, especially the test setup page, Chris clearly states which drivers he used for what.

GTA V: 15.4 Beta
Witcher 3: 15.5 Beta
Fury X: 15.15 Beta (specific for the R300 series and Fury series)

Using the Beta 15.6 wont work for the Fury X and would not benefit the R9 290X as the betas have been performance fixes for specific games. If they used the 15.6 Beta for the R9 290X it would still have the same optimizations for all the games so in the end the results would be the same.

If they decided to benchmark Batman Arkham Knight they would have had to use the 15.6 Betas to get decent performance.

 

[spentshells]

My 2 cents. It's a good performer @ 4k. It keeps up and beats NV. Lower resolution loses are quite dissapointing, but this is a resource issue as mentioned in the article. It is important to note the performance increases made with every driver release as has been the pattern so far with AMD cards since the release of GCN. The 7950 was being beat by the 660ti on it's release, yet now in a lot of scenarios it pushes past the GTX770. I'll place a 5 dollar "payable" bet that the fury pulls away from the 980ti by the end of the year..... at least @ 4k resolutions.

While I think AMD did well, all they did was tack HBM to a fatter Hawaii GPU. If NVIDIA were to do this same thing, let's say with the 980ti there would clearly be a no contest and high frame rate 4k gaming would already be at hand. Pascal is on the way, you would be a fool to buy a pair of 980 ti's, or really awesome, however you can look at it financially.

 

[sp00]

There seems to be a large fluctuation in performance from review sites. They are probably using different drivers and not the latest one. These benchmarks are already a week old. Depending on when they got their copy. Lisa Su did mention that in the beginning the Fury X might not beat the Titan X. I'm assuming that is due to enhancing drivers for this new architecture.

 

[Pei-chen]

No. 15.15

Would it be possible to get updated benchmarks with the latest drivers?
Would it make a big difference?

All the sites already used the 15.15 drivers. They are the official drivers for the R9 300 series and Fury X. The 15.15 currently do not support anything lower unless you download the hacked version which actually will work for the R9 200 series.

One more thing. The reviewed Fury X is a match to stock 980Ti (1000Mhz base/1075Mhz boost). Aftermarket 980Ti from ASUS, MSI and Gigabyte will have better power limit, cooler and quieter fans and much higher clock. MSI's 980Ti Gaming 6G have 200Mhz overclock at base and boost speed (1178Mhz base/1279Mhz boost). Overclocking GTX further to 1500Mhz at default voltage is also common and at those speed even a Fury X overclocked from 1000Mhz to 1125Mhz (Guru3D's overclock) will not stand a chance.

Why would you think the Fury would have less OC ability? I would think once Aftermarket GPUs are out it would be much more easily overclocked and get more benefit from it, with that fancy panse memory.

3 reasons:
1, GCN 1.2 don't overclock as well as Maxwell and Fury is a tweaked GCN 1.2.
2, Guru3D only managed a 12.5% overclock to 1125Mhz. They think they can get it up to 25% at 1250Mhz if the overclock app is more mature and they can give it more voltage.
3, If the reference card is water-cooled at $650, how much can board partner sell their air-cooled Fury X? $550? $600? You cannot replace a $100 AIO water-cooler with a $40 heatsink and sell it at the same price.

Then looking at 980Ti which overclock 30-40% at stock voltage on air and 50% on water, you'd see how overclocked Fury X is not a match for overclocked 980Ti.

The Fury X is stock clocked at 1050MHz so 1125MHz would only be only about 7% not 12.5%.

There is not going to be an air cooled Fury X but there will be an air cooled Fury which is probably going to be a cut down Fiji XT known as Fiji Pro that is slated for $550 in July.

Overall it is an OK card. It is nothing like the hype train was hoping for but it is not as bad as the negative people said it was. I personally think it is not priced right. If it was overall equal to and beat the 980Ti by say 5-10% then yes it is price right. But every review is showing it either losing or matching, with a few slight wins over the 980Ti, at most resolutions which to me says it should be a bit cheaper.

It is what it is though. I wasn't expecting much since it is still 28nm process after all.

You are correct about Fury X's stock clock being 1050Mhz and not 1000Mhz as I thought. That put Fury X in a worse light compares with 980Ti then as both have similar clock and performance and therefore instructions-per-clock. With 980Ti capable of running at much higher clock on air it has a distinct advantage over Fury X to overclockers.

 

[sp00]

Where is all the overclocking room at. Since that is what AMD fed us at E3?

No. 15.15

Would it be possible to get updated benchmarks with the latest drivers?

Looks like there is a performance difference with the new drivers.
Would it make a big difference?

 

[meribela]

This and the Guru3D review seems to be the only ones really praising the card. From the 290X yes it is. With the competition it is just a meh. It has a dark future ahead of itself. Lack of quick driver updates will not help. At least Tom didn't Gold the card for what it has "caught" up to.

 

[Gurg]

Where are overclocking results for all the 980ti, 980, Titan X as well as R9 390x and Fury X comparison?

Where the hell are the Overclocked benchmarks?

Who buys a GPU for more than 500$ and doesn't overclock it?

I am very curious. Is there any good reasoning behind you not including any overclocked GPU results in your benchmark?

We all know Maxwell 980ti OCs 1500Mhz quite easily on air, same with 980.

Can we please get some figures how far can newest iteration of GCN go? I am intrigued that you didn't even try to OC HBM memory either.

What's going on here? AMD paying loads of cash? Or is the software for overclocking GPUs not available, and furthermore, your old results of overclocking Maxwell and r9 390x are gone as well which is why you can't show them to us?

This is a terrible review compared to the one for the 390x last week. First the cards aren't identified. Then it obviously run at low settings due to the significantly higher frame rates than the cards in the 390x review. Then the cards are not the factory overclocks used for the 390x reviews. Then it doesn't include the 390x. WHY?

Why didn't Tom's just run the FuryX and drop its results into the 390x review charts? This isn't even close to apples to apples. Its more like playing a PGA tournament from the members tees allowing handicaps.

 

[meribela]

Glad to see some neck and neck competition. I breathed a sigh of relief as there's a 980 Ti in the mail with my name on it.

Where is all the overclocking room at. Since that is what AMD fed us at E3?

How did hardocp manage to have such drastically different results. What is going on here?

http://hardocp.com/article/2015/06/24/amd_radeon_r9_fury_x_video_card_review/11#.VYqy-0ay58E

Different settings, some different titles. Mostly GameWorks titles which are notoriously known to not work well on AMD cards

Uhh, appears to me Toms and [H]ard did the same titles, Tom' covered one or two more.

Try again?

As far as I can tell, Tom's may be the only reviews (gone through 10 reviews thus far) that is different from the rest. Fury is ahead of the 980Ti, which many shows it is not. Same for power consumption. I be weary of the reviews but more so Tom's.

 

[CaptainTom]

Why the hell is everybody disappointed ?

The price is good, the performance is good, the temperature is very good ( see 980ti / titan-x temps), the noise is good, power consumption is good. So what`s your problem ? I`ve seen some comments about Nvidia going HBM ... well they aren`t on HBM atm, why do you compare this to a future nvidia product that we don`t even know about ? Compare the present not the future. You can never please some people.

I agree people are acting odd, but honestly I was hoping for a 10%+ knock-out punch like the 290X delivered. However I think we all know that the Fury X will be at least 10% stronger within half a year if history is anything to go by.

 

[gamebrigada]

I see a lot of disappointment here. Are you disappointed this card didn't kick ass? I'm very happy with it. Look at the huge improvement over r9 290x. That's what really matters. This isn't their flagship for this series. Lets see what the dual gpu version of this will do.

 

[Eggz]

This is really good or just "meh" depending on how you look at it. Look at it purely from comparing to Nvidia, it's "meh," but it's really good given where AMD came from. The 290X (and now 390X) were two steps behind Nvidia, but Fury came out of nowhere to match Nvidia's top performer. I'm liking it from that angle.

 

[crisan_tiberiu]

For me it is very simple. Mathematics give you the answer. AMD cannot compete on 2 fronts, 1 against the giant Intel and 2nd with nVidia. It is all about the money. For intel to be beaten by a company witch has less than 10% of Intels money it is a "disaster".
For me, AMD is awesome and always was. Fury X, its an awesome card, unfortunatly i am not in the target (i cant even aford a GTX970, so i wont even come close to a GTX980Ti or Radeon Fury...)

 

[ccarroz]

same here, disappointment... guru3d has calculated tdp around 300w for this card. not good at all.. 4096-bit and this is it? once nvidia goes HBM i don't know how AMD will be able to compete with it.

I don't know if If I agree with this comment. HBM is only in its first gen and I am not sure we are saturating the memory bus as it is so outside or smaller form factor I am not convinced that HBM buys AMD a lot at this point. Either way, I think AMD will be close enough for their to be good competition and I am sure AMD will latch onto the value segment. As a disclaimer, I buy cards from both companies and am a current owner of a 980 TI. I am not trying to be a fan boy here.

 

[Vlad Rose]

Glad to see some neck and neck competition. I breathed a sigh of relief as there's a 980 Ti in the mail with my name on it.

Where is all the overclocking room at. Since that is what AMD fed us at E3?

How did hardocp manage to have such drastically different results. What is going on here?

http://hardocp.com/article/2015/06/24/amd_radeon_r9_fury_x_video_card_review/11#.VYqy-0ay58E

Different settings, some different titles. Mostly GameWorks titles which are notoriously known to not work well on AMD cards

Uhh, appears to me Toms and [H]ard did the same titles, Tom' covered one or two more.

Try again?

As far as I can tell, Tom's may be the only reviews (gone through 10 reviews thus far) that is different from the rest. Fury is ahead of the 980Ti, which many shows it is not. Same for power consumption. I be weary of the reviews but more so Tom's.

If you look at toms review, the Fury X did indeed lose to the 980ti @2560x1440, at approximately the same ratio that HardOp has on the titles that are the same. The numbers aren't exact between the 2 reviews there due to different settings.

At 3820x2160, which only toms did, shows a different result with the Fury X trading blows. HardOp, or most sites for that matter, did not test at that resolution; although they 'claim' it was a 4k test.

And to benchmark a title like Dying Light, which as they say "this game supports many advanced 3D effects including many NVIDIA created 3D features" is just atrocious on their part as it is known to cripple AMD cards in comparison. They may as well included Project Cars and Gotham Knight.

I'm not saying that the Fury X is more than what it is; honestly it's rather disappointing since AMD needed a home run to try to regain some market share back. Losing at anything less than 4k for resolution is really bad on their part. I will say though that Tom's benchmarks are pretty close to the other sites. People are comparing their 4k benchmarks to that of 2560x1440 from other sites and saying they're 'cheating' and/or 'being biased'.

 

[Johnpombrio]

As for overclocking, read the TechPpowerUp article. In that article, they tried overclocking the card and came up with 10% maximum o/c for the GPU (Roughly 5% improvement in gaming) and the memory clock was NOT ABLE to be overclocked at all without messing up the screen.

 

[Vlad Rose]

As for overclocking, read the TechPpowerUp article. In that article, they tried overclocking the card and came up with 10% maximum o/c for the GPU (Roughly 5% improvement in gaming) and the memory clock was NOT ABLE to be overclocked at all without messing up the screen.

Yeah, I'm thinking that the card is probably near its limits already for thermal headroom since they are using a liquid cooler on it; making OCing not much of an option.

 

[zodiacfml]

Lower prices soon please! I love the design!

 

[InvalidError]

45% more shaders, 60% more memory bandwidth, only 20-30% better benchmarks than the 290X. Either there is another 10-20% extra performance that can be squeezed out by future driver refinements or Fiji is having trouble making efficient use of all the extra shaders and bandwidth.

I am glad to be proven right about more than 4GB not being necessary when there is enough memory bandwidth to avoid duplicating resources across multiple memory channels.

 

[jimmysmitty]

45% more shaders, 60% more memory bandwidth, only 20-30% better benchmarks than the 290X. Either there is another 10-20% extra performance that can be squeezed out by future driver refinements or Fiji is having trouble making efficient use of all the extra shaders and bandwidth.

I am glad to be proven right about more than 4GB not being necessary when there is enough memory bandwidth to avoid duplicating resources across multiple memory channels.

That is a good question. I hope it is the former (more performance with driver improvements) and not the latter.

I also have another question.

Chris, do you know anything about the warranty that is being provided for the GPU? How long and if it covers accidental damage to other components like the Corsair Hydro series does in case of a water leak?

 

[Memnarchon]

45% more shaders, 60% more memory bandwidth, only 20-30% better benchmarks than the 290X. Either there is another 10-20% extra performance that can be squeezed out by future driver refinements or Fiji is having trouble making efficient use of all the extra shaders and bandwidth.

I am glad to be proven right about more than 4GB not being necessary when there is enough memory bandwidth to avoid duplicating resources across multiple memory channels.

Everyone who is realistic wouldn't expect more performance.
45% more shaders and 60% more memory bandwidth than 290X doesn't mean 40-50% more performance.
A great example is Titan X vs GTX 970. Near double (3072 vs 1664) shaders and more than 50%(due to 224bit 3,5GB) more memory but, "only" around 40% more perfomance...
Dispite this, its still a good card, already water cooled and a great performance and perf/watt jump from R9 290X.
AMD has already on the market, near 400GB/s GDDR5 (R9 390X). They could take the best samples to maybe 450GB/s. HBM has 512GB/s. Maybe using 8GB GDDR5 would be better (people would love 8GB) and cheaper (since HBM is more expensive)...

 

[childofthekorn]

Good review, how about that R9 Nano? Excited!

 

[kittle]

Interesting card and review .. but unless I missed it -- everyone except Chris is overlooking the elephant in the room -- WHERE am I going to PUT that extra radiator in my case???

Performance wont matter at all if it wont mount in the system. Maybe Chris can share pics of what he did?

 

[Guest]

Disappointing.