AMD Fury X Reviews

[Jaydip]

I understand that, but going by spec, what nVidia lists on their site for the 980 Ti reviewed against what AMD lists for the Fury X.

I go to Newegg and all ref 980 Ti's have the same numbers, 1000 stock, 1075 boost.

I look at fancy factory OC cards and they have the OC specs; I know what I am getting.

So, I think it would be fair to benchmark apples to apples, publicly advertised specs per the mfg., AND disclose what your chosen boost settings are.

Also, why do your 980 Ti review numbers differ so much from the 980 Ti numbers in the Fury X review? Mostly 4K stuffs.

The by-line of "real world play testing" does not help when actually trying to make a fair comparison, when the testing for the same card (980 Ti) is different between two reviews.

[H] publishes this stuff, and should be as transparent as possible, less we end up bickering in forums.

Let me explain that to you, the boost NV advertises is the clock you guaranteed to get while running a game but that doesn't mean it can't boost higher. Provided the power/thermal targets are within bounds it will boost more that the listed spec but it is not guaranteed.

 

[Alatar]

I understand that, but going by spec, what nVidia lists on their site for the 980 Ti reviewed against what AMD lists for the Fury X.

I go to Newegg and all ref 980 Ti's have the same numbers, 1000 stock, 1075 boost.

The problem here is that you just don't understand how AMD's and NV's boost functionality differs.

AMD's spec sheets give you the maximum boost that the card can reach. You can throttle below this point but usually non ref cards or watercooled ones (fury) don't.

NV on the other hand gives you the base clock (lowest possible frequency during normal loads that don't cause safety throttling due to 99C temps or something) and base boost clock that's the lowest clock when the boost functionality is doing its thing. The higher number is definitely not the maximum frequency of any stock Nvidia cards. Unlike AMD, Nvidia just doesn't publicly state their maximum boost frequency.

If reviewers did what you want them to do they would have to manually downclock the nvidia cards so that they can't go over X MHz.

tl;dr NV numbers don't mean what you think they mean, stock cards boost higher than base boost freq.

 

[B-Riz]

The problem here is that you just don't understand how AMD's and NV's boost functionality differs.

AMD's spec sheets give you the maximum boost that the card can reach. You can throttle below this point but usually non ref cards or watercooled ones (fury) don't.

NV on the other hand gives you the base clock (lowest possible frequency during normal loads that don't cause safety throttling due to 99C temps or something) and base boost clock that's the lowest clock when the boost functionality is doing its thing. The higher number is definitely not the maximum frequency of any stock Nvidia cards. Unlike AMD, Nvidia just doesn't publicly state their maximum boost frequency.

If reviewers did what you want them to do they would have to manually downclock the nvidia cards so that they can't go over X MHz.

tl;dr NV numbers don't mean what you think they mean, stock cards boost higher than base boost freq.

I know how it works, I am pointing out the BS of it all; nVidia lists one thing on their site and then the real world review does not disclose ALL the testing specs for the hardware used.

I would not have guessed the 980 Ti boosted to 1201, I checked the nVidia site before posting anything here and it says 1075.

 

[Glo.]

http://www.tweakpc.de/hardware/test...ry_x_vs_geforce_gtx_980_ti_benchmarks/s03.php

They tested 20 games, and the benchmarks show the overall scores for the GPUs. Also they checked how AA affects the performance of the GPUs in different resolutions. Check the benches anyway. 290 faster than GTX970, 290X faster than 780 Ti. 390X faster than GTX980. Looks like situation of 7970 vs. 680 all over again. Nvidia cards are getting older quite faster than AMD ones. And it will be extremely apparent when DirectX 12 will come out.

Looks like, it is true what I have written before. For now GTX 980 Ti is the better option. But "future proof" option is the Fury X.

For me, its a tie. For now.

 

[YBS1]

I know how it works, I am pointing out the BS of it all; nVidia lists one thing on their site and then the real world review does not disclose ALL the testing specs for the hardware used.

I would not have guessed the 980 Ti boosted to 1201, I checked the nVidia site before posting anything here and it says 1075.

Why is it you are outraged at nVidia for outperforming it's stated specs, but not at AMD for performing below them? (Before anyone gets butt-hurt, I'm speaking of reference edition 290 series, not Fury, I suspect it never throttles down from advertised clocks).

For the record I don't suppose either way is right or wrong, the consumer just needs to be aware the pretty much work opposite of each other.

 

[boozzer]

http://www.tweakpc.de/hardware/test...ry_x_vs_geforce_gtx_980_ti_benchmarks/s03.php

They tested 20 games, and the benchmarks show the overall scores for the GPUs. Also they checked how AA affects the performance of the GPUs in different resolutions. Check the benches anyway. 290 faster than GTX970, 290X faster than 780 Ti. 390X faster than GTX980. Looks like situation of 7970 vs. 680 all over again. Nvidia cards are getting older quite faster than AMD ones. And it will be extremely apparent when DirectX 12 will come out.

Looks like, it is true what I have written before. For now GTX 980 Ti is the better option. But "future proof" option is the Fury X.

For me, its a tie. For now.

if it is a tie I would 100% go with fury x for the water cooler :biggrin: what I like is the absence of crap game project cars, a super outliner.

 

[B-Riz]

Why is it you are outraged at nVidia for outperforming it's stated specs, but not at AMD for performing below them? (Before anyone gets butt-hurt, I'm speaking of reference edition 290 series, not Fury, I suspect it never throttles down from advertised clocks).

For the record I don't suppose either way is right or wrong, the consumer just needs to be aware the pretty much work opposite of each other.

Not outraged.

Would a fair comparison be 980 Ti w/ launch drivers benched against Fury X w/ launch drivers, both limited to the max mfg. stated boost clock?

And all games getting a baseline treatment then eye candy turned up?

[H] is all over the place, even with the "better" 980 Ti, its launch review numbers are quite different than what shows up in the Fury X review.


If I was buying a 980 Ti or Fury X today, I would have specs for both memorized and mull over the pro's / con's of each.

 

[Jaydip]

Not outraged.

Would a fair comparison be 980 Ti w/ launch drivers benched against Fury X w/ launch drivers, both limited to the max mfg. stated boost clock?

And all games getting a baseline treatment then eye candy turned up?

[H] is all over the place, even with the "better" 980 Ti, its launch review numbers are quite different than what shows up in the Fury X review.


If I was buying a 980 Ti or Fury X today, I would have specs for both memorized and mull over the pro's / con's of each.

1. This is one of the most ridiculous statements I have heard here, so if a card launches 6 months later the other card should be tested with it's launch driver ?

2. I think you still don't understand boost, have a look at it.

3. [H] has apples to apples for a reason.

 

[B-Riz]

1. This is one of the most ridiculous statements I have heard here, so if a card launches 6 months later the other card should be tested with it's launch driver ?

2. I think you still don't understand boost, have a look at it.

3. [H] has apples to apples for a reason.

Since you have all the answers, figure out the 980 Ti launch review numbers vs. the 980 Ti numbers in the Fury X review.

Srsly.

The tests cannot even correlate due to changed settings; there is no standard, these "apples-to-apples" comparisons are not good measures of anything.

Their testing is flawed, and everyone defends it.

http://www.hardocp.com/article/2015...x_980_ti_video_card_gpu_review/9#.VY2b1PmjNcY

http://www.hardocp.com/article/2015...x_980_ti_video_card_gpu_review/9#.VY2b1PmjNcY

 

[Jaydip]

Since you have all the answers, figure out the 980 Ti launch review numbers vs. the 980 Ti numbers in the Fury X review.

Srsly.

The tests cannot even correlate due to changed settings; there is no standard, these "apples-to-apples" comparisons are not good measures of anything.

Their testing is flawed, and everyone defends it.

http://www.hardocp.com/article/2015...x_980_ti_video_card_gpu_review/9#.VY2b1PmjNcY

http://www.hardocp.com/article/2015...x_980_ti_video_card_gpu_review/9#.VY2b1PmjNcY

They have the best "testing" setup cuz unlike many sites they don't do canned benches. The variance of 1-2% is always expected, I am sorry that you can't see what you want to see in their benches(and no it was not a jibe at you at all). I understand that Fury-X disappointed many AMD fans but that fault lies with AMD not with [H].

 

[B-Riz]

They have the best "testing" setup cuz unlike many sites they don't do canned benches. The variance of 1-2% is always expected, I am sorry that you can't see what you want to see in their benches(and no it was not a jibe at you at all). I understand that Fury-X disappointed many AMD fans but that fault lies with AMD not with [H].

I am no fan of anything except cheap performance.

But their testing needs some normalization for comparisons sake and better transparency on all the card specs, irrespective if they should be givens or not.

And yes, I want to see numbers where the 980 Ti is limited to the 1075 that nVidia touts on their effing website; it does not say min 1075, max 1202.

http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-980-ti/specifications

nVidia should put this info on their site, I should not have to dig around reviews to get it.

http://www.anandtech.com/show/9306/the-nvidia-geforce-gtx-980-ti-review/17

 

[Jaydip]

I am no fan of anything except cheap performance.

But their testing needs some normalization for comparisons sake and better transparency on all the card specs, irrespective if they should be givens or not.

And yes, I want to see numbers where the 980 Ti is limited to the 1075 that nVidia touts on their effing website; it does not say min 1075, max 1202.

http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-980-ti/specifications

nVidia should put this info on their site, I should not have to dig around reviews to get it.

http://www.anandtech.com/show/9306/the-nvidia-geforce-gtx-980-ti-review/17

Actually they can't since it is impossible to predict the highest boost clock.

 

[AtenRa]

[FONT=&quot]https://youtu.be/iEwLtqbBw90?t=452

[/FONT]From the Video Review above,

“This card is amongst the loudest, most annoying cards I have ever heard”

Really ???? What the hell ???

He had to find something bad to say

 

[x3sphere]

[FONT=&quot]https://youtu.be/iEwLtqbBw90?t=452

[/FONT]From the Video Review above,

“This card is amongst the loudest, most annoying cards I have ever heard”

Really ???? What the hell ???

He had to find something bad to say

Some reviews complained about the pump noise too but I thought this was only affecting pre-production samples.

Apparently he didn't get his card from AMD, it came from a retail partner. If all retail units have this issue right now that's not good...

 

[boozzer]

Some reviews complained about the pump noise too but I thought this was only affecting pre-production samples.

Apparently he didn't get his card from AMD, it came from a retail partner. If all retail units have this issue right now that's not good...

that is some amazing leap at conclusions. :awe:

 

[x3sphere]

Fair enough, but I just looked on YouTube and it doesn't seem to be an isolated case.

https://www.youtube.com/watch?v=XfyQzroYnrI

Another Fury X video, he has noise issues too. I mean of course it is possible they both just got defective units, we will see as more people get cards in.

 

[sze5003]

I remember when the 7970ghz models came out. My first HIS model had bad whine. I returned it and got the one I have now.

 

[ocre]

I have never heard of pump noise issues in closed loop solutions. I never noticed my CPU cooler making that kind of noise. Could it be the fact that they dont have it in a case?

I just think it is wrong to test acoustics in the open like that. I especially think any test done right beside the motherboard are completely useless. The sound is not representative at all. A fan moving a lot of air will make noise but the rest of the design determines how that noise/sound is carried. At my work, we had the challenge of controlling sound and reducing dBa. The noise that had to be generated, it is unavoidable. But we had to dissipate the loudness and it is pretty amassing what can be achieved.

I think any sound or noise test in reviews should be done properly. That card should be in a case and noise monitored from a typical distance. Listening beside the motherboard or without a case is wrong. There is no chance of dissipating when designs are built with sound dissipation in mind.

I can tell you first hand, it is kinda of funny actually.

People will post some obscure review of the 980 reference and claim "look it is loud". But when i game i hear my CPU cooler fan ramp up louder than my 1450mhz 980. This is crazy because the LQ320 is a very quiet CPU cooler (reviews back this up), yet to me it is always louder than my 980. At 1400rpm, its loader than my 980 running at 50%.
See i always leave my fans on auto, but when i force the 980 fans up all i hear is air flow. My GPU at 62% is about as loud as the LQ320 at 1500rpms. Which isnt loud at all. It is just air, but the CPU cooler is a lower freq that seems to carry further

The point is
Open acoustic figures are worthless unless you have a motherboard is sitting right in front of you and out in the open. That moving air and coolers make sound, the engineering task is how that noise is handled. Testing the noise out in the open and/or directly next to the cooler is so wrong i would call the results worthless.

I probably never heard my closed loop cooler pump because I have it in a real PC. I am not running with my motherboard sitting on my desk beside my ear, inches in front of me, out in the open.

This is why there are always drastically different results when it comes to noise and GPUs. Some reviewers swapping out HW may have their motherboard sitting on their desk but those noise and acoustic results dont represent what they are passed off as.
In a case, at a desk, the results will be very different. See, those decibels are not the same. Two cards with exactly the same 45dBa recorded right next to a motherboard out in the open can end up drastically different when put into cases and sitting at a proper PC desk. They can be totally different, one sounding exceptionally quiet and the other sounding annoying and loud.
It is all depending on how the coolers were designed. Recording sound beside the motherboard, that is just the noise that needs to be dissipated.....
Those type test defeat the purpose and are useless to most people that will buy these cards

 

[Azix]

They have the best "testing" setup cuz unlike many sites they don't do canned benches. The variance of 1-2% is always expected, I am sorry that you can't see what you want to see in their benches(and no it was not a jibe at you at all). I understand that Fury-X disappointed many AMD fans but that fault lies with AMD not with [H].

Canned benches > not. We may not play benches but if it shows the graphics in the game well, its good. Some canned benches are very good showing worst case scenario and stressing cards well. A benchmark that involves variation is not as accurate.

As for boost clocks, he means what a real gamer might get would be different, I think. Eg. if nvidia sends out cards for reviews and those just happen to boost higher than an average card would.

 

[Jaydip]

Canned benches can't be trusted as they can be optimised by both vendors. I remember the bench in hitman and the actual gameplay differ quite a bit. Check various 980Ti reviews , different reviewers got different boost clocks.

 

[ocre]

Canned benches > not. We may not play benches but if it shows the graphics in the game well, its good. Some canned benches are very good showing worst case scenario and stressing cards well. A benchmark that involves variation is not as accurate.

As for boost clocks, he means what a real gamer might get would be different, I think. Eg. if nvidia sends out cards for reviews and those just happen to boost higher than an average card would.

If you don't understand the issue with canned benchmarks, you aren't knowledgable enough to be having a meaningful discussion about these methods in the first place.

As far as boost, there is nothing magical about it. If reviewers were getting cards that magically boost higher, then these cards would have to also use more power, be louder, and/or run hotter. See, that is how boost works. There is a temp target, power target, and fan profile. Boost clocks are a result of these measures. Turn up these targets n a maxwell card, in any system, you will have higher boost clocks.

You can't have one without the other. If reviewers are getting cards that boost higher, they have a higher targets. And since 1200mhz is not an issue on any of these cards, it makes no sense to make review cards more power hungry or hotter when they are being measured and recorded.

See, you don't understand boost either. Those boost clocks are not wild or out of the realm of expectations. No, not every single person who buys one will see those out of the box boost clocks but its not for the reasons you are thinking. The limiting factor is temps which is related to airflow. Cases with limited airflow may only boost to 1100 out of the box. Change either the fan, the temp/power target and boom....your boosting just as high. They could also improve the case airflow and the boost clocks will improve themselves.

All over the Internet, people buying maxwell cards, they all find clocks similar to what reviewers find. This is not some trickery. The advertised boost is not what you are thinking. See, not all work loads are created equal. And games are typically not the most demanding loads for a GPU. Boost is about a target power and instead of a fixed clock speed. See, one demanding app may use 180watts running at 1070mhz. That task may be very demanding but then another task which is much lighter and only uses 150watts at the same clock speed of 1070mhz. So, just like intel, nvidia decides to try to take advantage in loads that aren't as demanding. Boost will push up the MHz when the load allows it. This is just using untapped TDP. And since gaming loads are typically not as demanding, it works out really well.

You seem to think that the advertised boost clock is what the worst cards should reach when actually it is what you would boost in the most brutal and demanding gaming loads. Boost clocks are game sensitive. Some games allow higher boost than others. This can't be helped. Boost is designed around that fact. Some loads leave TDP on the table, why let it go to waste?

The boost clocks reviewers get in the games they review aren't out of the norm. Not everyone will have the same results because not every card is in the same environment. But they are usually pretty close and not too far off. There maybe some cards in cases that are really stuffy. But even then, get the temps down or change the temp/power target and you won't have any problem running at 1200mhz boost.

Nvidia was nowhere near aggressive when it comes to boost. Their target of 80degrees and auto fan settings in the driver, there is no reviewer out there running boost speeds far out of the reach of the cards on the market.

 

[DarkKnightDude]

Got my Fury X, has no pumping sound problems or anything. Its hilariously quiet compared with my 770.

 

[Hitman928]

Got my Fury X, has no pumping sound problems or anything. Its hilariously quiet compared with my 770.

DarkKnight, any chance you could upload a video? I'd like to hear a good unit in working order.

 

[DarkKnightDude]

DarkKnight, any chance you could upload a video? I'd like to hear a good unit in working order.

I don't have a camera on hand atm, though so far, there's nothing really to hear. I don't hear any sound from it at all. The only indication that its on is the LED lights on the side.

I'll give it a spin in some BF4 and maybe Crysis or Witcher 3.

 

[raghu78]

They have the best "testing" setup cuz unlike many sites they don't do canned benches. The variance of 1-2% is always expected, I am sorry that you can't see what you want to see in their benches(and no it was not a jibe at you at all). I understand that Fury-X disappointed many AMD fans but that fault lies with AMD not with [H].

hardocp has the most inconsistent testing methodology. Their definition of playable settings changes from review to review and depending on their mood. In certain reviews when they want to show a particular card in favourable light they chose certain settings and call it playable. In another review the same card is told to be not playable at the same settings. At times they have said Watchdogs required 3 GB at 1080p and 4GB at 1440p for Ultra settings and when reviewing GTX 960 2GB, the same Watchdogs is told as playable at Ultra settings at 1080p with 2GB VRAM. Thats inconsistency and hypocrisy at its best.

hardocp also has the worst game suite loaded with Gameworks titles and do not have a clue that it skews their results horribly. Other like hwc do a good job maintaining a broader testing suite with neutral titles and AMD Gaming Evolved titles. In fact hardocp justify it by saying they test the games which everybody is playing and do not bother which GPU vendor is associated with those titles. What rubbish. As if everybody plays only the 4-5 games which they test and the rest of the games are not played at all. Also at times they are overly sensationalist in their reviews like they have been with Fury X. hardocp also was one of the few sites which showed GTX 960 significantly better than R9 285 and on par with R9 280X while other sites showed R9 285 on par with GTX 960 and R9 280X much faster than GTX 960. All these were due to their test suite being completely favorable to Nvidia as it included mainly Gameworks titles. Their current suite consists of BF4 - AMD GE, GTA V - neutral and Witcher 3, Farcry 4 and Dying Light all of which are Gameworks titles. btw they specifically test Gameworks features which are badly implemented and a waste of GPU resources like Hairworks. Anyway trying to say hardocp is the best site for GPU reviews is the worst thing a person can say. They are so utterly clueless that Gameworks is specifically designed to cripple the competition even at the cost of hurting Nvidia's previous gen cards like kepler. As always you have people defending them because they have an extremely Nvidia favourable view with this entire Maxwell generation.