AMD Fury X Reviews

[Stuka87]

Did Anandtech receive a card to evaluate?

Yes but Ryan has been sick, so it will be a bit late.

 

[LTC8K6]

True, Fury does make significant progress over the 290X but look at how it does compared to the 390X since they have the same core clocks. At 1440p it's only 20% faster. With 43% more shaders and the added efficiency gains from GCN 1.1 to 1.3, I expected a lot better.

What if we can overclock a 390X better than a Fury?

 

[Fire&Blood]

Did Anandtech receive a card to evaluate?

Pretty sure they did. The review may not be up yet but it's in the benches:

 

[at80eighty]

even if they did, AT takes time to review anyway.

 

[Stuka87]

I am thinking that being able to adjust GPU core voltage should make a huge difference in overclocking results.

It looks like in an attempt to get power usage as low as possible, the voltage is run as low as they can get away with, which results in almost no overclocking headroom. Once the voltage can be adjusted, we should see better OC results.

 

[Silverforce11]

True, Fury does make significant progress over the 290X but look at how it does compared to the 390X since they have the same core clocks. At 1440p it's only 20% faster. With 43% more shaders and the added efficiency gains from GCN 1.1 to 1.3, I expected a lot better.

It's still GCN 1.2, basically big Tonga.

Also the same can be said for GM200, on the same charts, the 980Ti is only 28% above 980. At 4K, its still 28% above the 980. Fury X is 20% at 1440p and 25% above 390X at 4K.

That's my problem with this class of performance. It's too mild and definitely not worth the money for someone already on 970/980/R290X (non throttling ref models heh) class performance.

Would be nice to see some CF Fury X vs SLI 980Ti at 4K reviews to finish the round up.

 

[YBS1]

I am thinking that being able to adjust GPU core voltage should make a huge difference in overclocking results.

It looks like in an attempt to get power usage as low as possible, the voltage is run as low as they can get away with, which results in almost no overclocking headroom. Once the voltage can be adjusted, we should see better OC results.

I hope you're right, I'm worried the exact opposite is true. Meaning they already may have the card right up on the sweet spot in voltage and any further meaningful speed increases will only come with heaps of voltage.

 

[DarkKnightDude]

At least if AIBs go crazy on the Fury Pro we might get something great. Depending on pricing.

 

[tviceman]

I am thinking that being able to adjust GPU core voltage should make a huge difference in overclocking results.

It looks like in an attempt to get power usage as low as possible, the voltage is run as low as they can get away with, which results in almost no overclocking headroom. Once the voltage can be adjusted, we should see better OC results.

This is the catch 22. GM200 doesn't need a (worthwhile) voltage adjustment to get a 25% OC. Power consumption goes up linearly with performance. If Fury X is going to need a voltage adjustment to get >10% OC, then power consumption is going to be massive. Whats the point in getting similar / slightly less performance with 200+ watts more power consumption?

I hope you're right, I'm worried the exact opposite is true. Meaning they already may have the card right up on the sweet spot in voltage and any further meaningful speed increases will only come with heaps of voltage.

I think this will likely be the truth.

 

[tviceman]

Would be nice to see some CF Fury X vs SLI 980Ti at 4K reviews to finish the round up.

This is the only way buying Fury X at $650 makes sense. I'm guessing that crossfire will offer better performance at 4k than 980 TI, and will actually be able to drive good frame rates. But Fury X's overclocking is just terrible right now.

 

[monstercameron]

http://semiaccurate.com/2015/06/24/frame-rate-limiting-with-amds-fury-x/

For all the power nerds out there, frtc will save you a few watts!

 

[RussianSensation]

Maxwell has had 9 months of driver optimization under its belt. Fiji has 0. I imagine Fiji will come up a few percent on average over the coming months.

If Fury is ~10% slower than Fury X ala 290x vs 290 we could have another situation where the top end chip isn't worth getting but the 2nd tier chip is a great deal.

Good points but take a look at this and tell me how much more overclocking Fury X needs + drive improvement to catch up to a Gigabyte G1 980Ti? I think 1440P is more useful for a single card as neither of these is fast enough for 4K. The result is an absolutely destruction by the 980Ti. I was actually afraid this would happen if AMD's Fury couldn't overclock well.

Gigabyte G1 980Ti Gaming OC vs. Fury X @ 1135mhz

40% faster

41% faster

15% faster

18% faster

25% faster

22% faster

21% faster

22% faster

21% faster

19% faster

Sweclockers

Fury OC won 0/10 benchmarks against 980Ti OC at 1440P here. 980Ti OC leads Fury X 1135mhz at 1440P by a jaw dropping 24.4%

imho 980Ti easily wins in the most important resolution for these 2 cards unless buying them in pairs for 4K.

If we argue that Fury X can overclock way better with voltage, we can also say that 980Ti's full potential hasn't been revealed since we haven't seen MSI Lightning or EVGA Classified versions of the GM200 cards and NV didn't even have to release a fully unlocked GM200 after-market cards to win.

An after-market 980Ti has HDMI 2.0 (possibly higher resale value down the line/maybe easier to find a buyer when it comes time to resell), 6GB of VRAM for free as a bonus, overclocks with minimal increase in power usage (my guess is for Fury X to overclock well it needs voltage), and it's superior performance at lower resolutions mean that when it comes time to resell once again, someone with a 1080P-1200P monitor will want a 980Ti over the Fury X. Then there is the pump noise/coil whine noise coming from some Fury X cards which is not acceptable for a card at $650. Sweclockers has a video of the coil whine.

Honestly, AMD needs to find 10-15% improvement in the drivers or this card will need 1400mhz overclocks to compete with a 1500mhz 980Ti. But even if that happens, 980Ti has 'free' 50% more VRAM as a bonus.

I think AMD really needed to beat the Titan X by 5-7% to be on similar footing to after-market 980Ti cards. Right now at $650 it's practically impossible to recommend this card as it's having issues at lower rez gaming, gets killed at 1440P with overclocking and isn't fast enough for 4K on its own. It literally sits in no man's land against an after-market 980Ti.

I think AMD's AIBs and AMD are going to have to step up their driver and factory pre-overclocking game, otherwise price drops are required. AMD needs to release a fully unlocked Fury X version at $549 and allow AIBs to crank it to 1150mhz+ out of the box. Even then from the Sweclockers benchmarks an after-market Gigabyte G1 980Ti OC or EVGA SC+ ACX 2.0+ 980Ti OC crushed an 1135mhz Fury X. So that's not even a solution. A lot of gamers will just pay $100 extra for 20-24% more overclocked performance. I know I would.

What's more worrying is AMD's 596mm2 chip with HBM1 can't beat 601mm2 Maxwell with GDDR5. If NV improves perf/watt 2X and adds HBM2 on top of a brand new Pascal architecture, next gen AMD could be in serious trouble. The big thing here is that AMD did make a better reference card than GM200 as Titan X and 980Ti reference cards are hot and loud but NV was a step-ahead with after-market versions available shortly after launch.

 

[moonbogg]

The Fury is as loud as a 980ti reference and its slower. This water cooling solution didn't help with noise, only with heat and you can't even hardly OC the thing. Its not a bad card, but its kind of a dud really. No reason to get this over a aftermarket 980ti at all, or even a reference 980ti. I'd still buy reference 980ti's if I was doing it again. Throwing blocks on anyways.

 

[Silverforce11]

This is the catch 22. GM200 doesn't need a (worthwhile) voltage adjustment to get a 25% OC. Power consumption goes up linearly with performance. If Fury X is going to need a voltage adjustment to get >10% OC, then power consumption is going to be massive. Whats the point in getting similar / slightly less performance with 200+ watts more power consumption?

I think this will likely be the truth.

I agree with this fully. GCN is not a graceful overclocker when it requires vcore. It's good for doing some benchmarks, but certainly not worth 100W extra for 10% more performance. That just destroys efficiency right there.

Even with vcore, a 1.25ghz Fury X (~20% OC) will at best match a 1.5ghz 980Ti at 4K, but end up being 400W vs 300W.. with less vram.

AMD over-estimated the premium on cool & quiet and under-estimated the importance of overclocking potential.

 

[Stuka87]

The Fury is as loud as a 980ti reference and its slower. This water cooling solution didn't help with noise, only with heat and you can't even hardly OC the thing. Its not a bad card, but its kind of a dud really. No reason to get this over a aftermarket 980ti at all, or even a reference 980ti. I'd still buy reference 980ti's if I was doing it again. Throwing blocks on anyways.

No it isnt. Load sound levels for the fury are ~32dB, while the 980Ti reference is up ~48dB.

 

[blastingcap]

Fury X = what happens when you slash and burn R&D budgeting and lay off too many engineers.

Also, ouch. Usually HardOCP is an easy grader but... http://www.hardocp.com/article/2015/06/24/amd_radeon_r9_fury_x_video_card_review/11#.VYrEyUbIe4I

The Bottom Line

The new AMD Fiji GPU and Fury X video card looks awesome on paper, but has underwhelmed and disappointed us when it comes to real world gameplay. The AMD Radeon R9 Fury X feels like a proof of concept for HBM technology.

In terms of gaming performance, the AMD Radeon R9 Fury X seems like better competition for the GeForce GTX 980 4GB video card, rather than the GeForce GTX 980 Ti. GTX 980 cards are selling for as low at $490 today. This is not a good thing since the AMD Radeon R9 Fury X is priced at $649, the same price as the GeForce GTX 980 Ti.

(Editor's Note: AMD's Fury X is starting to show up at Amazon and Newegg this morning.)

Usually trying to decide between two video cards at the same price point is a wash, with very even and split performance. However, this is not the case this time with the AMD Radeon R9 Fury X and GeForce GTX 980 Ti. There is a definite pattern that leads to one video card being the best value for the money, and it is GeForce GTX 980 Ti, not the AMD Radeon R9 Fury X.

Limited VRAM for a flagship $649 video card, sub-par gaming performance for the price, and limited display support options with no HDMI 2.0 and no DVI port. To be honest, we aren't entirely sure who the AMD Radeon R9 Fury X is really built for? The AMD Radeon Fury X is a confusing product, like a technology demo not fully realized, a showcase for HBM only but with no real substance. The AMD Radeon Fury X looks to be a great marketing showcase, but its prowess starts waning when you consider its value to gamers and hardware enthusiasts.

 

[RussianSensation]

The Fury is as loud as a 980ti reference and its slower.

Ok this part is flat out incorrect. Fury X is quieter than almost all after-market 980Ti cards that have been out so far and it's world's quieter than a reference 980Ti.

http://www.computerbase.de/2015-06/amd-radeon-r9-fury-x-test/9/

980Ti OC reference is a > 50dBA jet engine and runs at 82-84C which means it's worthless for overclocking without headphones or blasting speakers high. When Fury X is overclocked, almost nothing changes since the rad keeps it cool and quiet. OTOH, 980Ti reference shouldn't even be in the conservation with after-market 980Ti and Fury X cards - the blower TX cooler on it is simply trash.

 

[Elfear]

It's still GCN 1.2, basically big Tonga.

You're right, I misread Tom's article and though Fiji had improvements to Tonga's architecture. Looks like it's the same underlying GCN 1.2 though.

Also the same can be said for GM200, on the same charts, the 980Ti is only 28% above 980. At 4K, its still 28% above the 980. Fury X is 20% at 1440p and 25% above 390X at 4K.

That's my problem with this class of performance. It's too mild and definitely not worth the money for someone already on 970/980/R290X (non throttling ref models heh) class performance.

Would be nice to see some CF Fury X vs SLI 980Ti at 4K reviews to finish the round up.

I'm right on the verge of thinking a 980Ti/Fury would be a good upgrade for me. With aftermarket 980Ti's getting really close to the 295X2, I'd really like to have the performance of my current setup without multi-GPU compromises. It's a hefty price to pay though for a 980Ti ($680) and a waterblock ($130).

 

[LTC8K6]

Ok this part is flat out incorrect. Fury X is quieter than almost all after-market 980Ti cards that have been out so far and it's world's quieter than a reference 980Ti.

http://www.computerbase.de/2015-06/amd-radeon-r9-fury-x-test/9/

980Ti OC reference is a > 50dBA jet engine and runs at 82-84C which means it's worthless for overclocking without headphones or blasting speakers high. 980Ti reference shouldn't even be in the conservation with after-market 980Ti and Fury X cards - the cooler on it is trash.

Well, we don't know what will happen regarding noise when Fury gets overclocked, do we?

 

[Grooveriding]

The only thing I saw that I expected is that 4GB is more than enough VRAM even for 4K. Otherwise the performance is meh. The card seems overdone with that watercooling and everything, but still slower than a 980ti at the same price. Disappointing.

 

[Head1985]

It's still GCN 1.2, basically big Tonga.

Also the same can be said for GM200, on the same charts, the 980Ti is only 28% above 980. At 4K, its still 28% above the 980. Fury X is 20% at 1440p and 25% above 390X at 4K.

That's my problem with this class of performance. It's too mild and definitely not worth the money for someone already on 970/980/R290X (non throttling ref models heh) class performance.

Would be nice to see some CF Fury X vs SLI 980Ti at 4K reviews to finish the round up.

But GTX980Ti have 37% more SP and its downclocked vs GTX980
GTX980TI avg boost 1140Mhz
http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_980_Ti/35.html
GTX980 avg boost 1240Mhz
http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_980/30.html
GTx980TI is faster than GTX980 by:
1080P-22%

2560x1440-25%

4k-28%

clock difference is 8.7%.If both have same clocks:
1080-22%+8.7%=30.7%
2560x1440-25%+8.7%=33.7%
4K-28%+8.7%=36.7%
So almost perfect scaling in 4K for maxwell.Now compare it to 390x vs FuryX
FURY X have vs 390x
+46% SP
+46%TMU
+30% memory bandwidth + HBM
SAME ROPS only 64
They have same 1050Mhz clock, but FuryX is only 23%(in 2560x1440 its only 19%) faster than 390x in 4k.I think thats because bad FIJI design with ONLY 64ROPS and not enough improved tessellation

 

[Silverforce11]

That [H] review is pure shill.

Witcher 3 with HairWorks, check.
FC4 with Enhanced Godrays, check.
Dying Light, check. (ok fine, but when their list of games is tiny, including a few NV biased titles determines the conclusion)

AMD would actually need Fury X to be 50% faster to compete when its crippled so badly.

Then his GTA V result is out of line against almost every other site. As is the BF4 results.

Compare to other sites that use a lot of games, Fury X matches 980Ti at 1440p/4K. It takes a custom 980Ti to beat it (& those custom 980Ti beats Titan X too).

 

[Stuka87]

Well, we don't know what will happen regarding noise when Fury gets overclocked, do we?

Most likely barely change. Its a 500W cooler, so it may go up 1-2dB, but nothing significant.

 

[tviceman]

I'm right on the verge of thinking a 980Ti/Fury would be a good upgrade for me. With aftermarket 980Ti's getting really close to the 295X2, I'd really like to have the performance of my current setup without multi-GPU compromises. It's a hefty price to pay though for a 980Ti ($680) and a waterblock ($130).

Why get a waterblock (unless you are a water jumkie). The EVGA SC+ is $680, 35db under load, and out-of-the-box is 12% faster than a reference 980 TI / 21% faster than a reference Fury X!

 

[RussianSensation]

Well, we don't know what will happen regarding noise when Fury gets overclocked, do we?

Yes, we do know from R9 295X2 overclocking results. This is 1x120mm rad cooling 1 chip. It's also common sense looking at the temperature results to see that even a 100-150W increase in power usage on Fury X will mean a way quieter card at load than a TX/980Ti reference. We also know that 980Ti/TX reference blower at overclocked states is a jet engine so basically unusable at 1.45-1.5Ghz overclocks for those of us who like quiet cards. Right now EVGA's SC+ 980Ti strikes a great balance of noise, performance, overclocking. Pretty much every brand agnostic user I've spoken to has either returned a reference 980Ti or put after-market cooling on it. The blower 980Ti is just garbage for OC, sorry.

Why get a waterblock (unless you are a water jumkie). The EVGA SC+ is $680, 35db under load, and out-of-the-box is 12% faster than a reference 980 TI / 21% faster than a reference Fury X!

Yup, that's the problem with Fury X right now. Using 2 of those to exhaust air out of the case is a nice bonus but for those of us who ran dual cards with similar power usage in a well-ventilated case, this is not a concern. Prices are going to come down on after-market 980Ti cards as competition between AIBs heats up. Zotac already has one for $660 or just $10 more than Fury X. MSI Gaming 980Ti is probably even cooler and quieter than the EVGA SC+.

Honestly I would have been OK if Fury X was 5-7% slower max overclocked as at that point it's close enough, but over 20% slower and having 50% less VRAM for almost the same price is just nuts. I don't know what AMD was thinking. During HD4000-5000 era they priced their cards too low and now they went the other way with 390/390X/Fury X pricing where the prices are simply too high. They haven't been able to find this nice middle-ground of price/performance and performance that HD7000 series had.

Maybe AMD can extract 10% more performance with drivers but making up against GM200's 25%+ overclocking headroom is probably not going to happen since 1135mhz Fury X would need to overclock close to 1.4Ghz to keep up without driver increases.

That [H] review is pure shill.

An easier way to put is foreign sites like Computerbase, Hardware.fr, Sweclockers, TechSpot and sites like AnandTech and TPU do such a good job, any review with 4-5 games tested of which most are NV-biased is just a waste of everyone's time. HardOCP practically has little positive to say about AMD cards since probably HD5000 era. When they covered up 960's 2GB VRAM issues against 280X and hardly discussed 680 / 770 2GB VRAM issues during their eras, and then they blamed AMD's performance/CF in GW titles squarely on AMD, with total disregard to what GW's is, that site wiped out 15+ years of reputation they had built.