AMD Radeon R9 Fury X Review

H

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
TechSpot posted a review of the AMD Radeon R9 Fury X today. For comparison purposes, you can see our evaluation here.

On the heels of delivering its latest round of rebadges with the Radeon 300 series, AMD has launched what may be its most hyped product of 2015. The R9 Fury X employs the GCN 1.2 architecture as the R9 380 but doubles the SPU count, and touts a massive memory bandwidth of 512GB/s thanks to an exciting new memory technology known as HBM.
Click to expand...
 
I mean really, looking at the marks between the hardocp vs the techspot marks ,

it seems like techspot received a working card where as hardocp received something broken.

this will cause some confusion for sure as people debate which is better.
 
chaos4u said:
I mean really, looking at the marks between the hardocp vs the techspot marks ,

it seems like techspot received a working card where as hardocp received something broken.

this will cause some confusion for sure as people debate which is better.
Click to expand...

Something that should be kept in mind is that [H] uses real world gameplay measurements where as Techspot uses benchmark runs (either an in-game tool or a custom recorded one). Canned benchmarks can have significant differences compared to real-world gameplay results (I can't remember the percentage number from that article [H] did a few years ago). The settings between games might not always be the same as well. For example: No mention what Fur setting they use in FC4 and no mention of Advanced Options for GTAV. It looks fairly normal for a real world vs benchmark scenario.
 
yeah, i was taking that into account . still their numbers puts the fury closer to the ti than what hardocp was showing . but maybe it is whithin the margins
 
chaos4u said:
I mean really, looking at the marks between the hardocp vs the techspot marks ,

it seems like techspot received a working card where as hardocp received something broken.

this will cause some confusion for sure as people debate which is better.
Click to expand...

So... according to the techspot article, it looks like the FuryX is at least double the performance of a 7970. This is going from reviews that showed 7970s vs 290x cards.

If I had the money to burn, I would buy one right now.
 
Doesn't surprise me. The [H] review was the only one that trashed it. Everybody else gave it pretty high marks. I don't get it.
 
There was something about drivers with the Fury X and some reviewers got older drivers.
 
cyclone3d said:
So... according to the techspot article, it looks like the FuryX is at least double the performance of a 7970. This is going from reviews that showed 7970s vs 290x cards.

If I had the money to burn, I would buy one right now.
Click to expand...

For what it's worth, I did a run in Unigine Valley with my crossfired 7970GHz just before I pulled them to put the FuryX in there, there is around 100 points difference in favor of the 7970's. I'd say it's pretty much double the performance of a single 7970. But it's only one benchmark and 1080p.

Edit: I forgot about scaling so maybe not double...but pretty close
 
MacLeod said:
Doesn't surprise me. The [H] review was the only one that trashed it. Everybody else gave it pretty high marks. I don't get it.
Click to expand...

Benchmarks dont tell the story how you experience the games with the card.
Main issue with benchmarks and I stopped reading them for clues to what they be for me as they cant bench my set up anyhow.
 
Interesting, in the case of any reviews it can happen one site is out in left field. Many variables at play, so discarding the highest and lowest scores isn't such a bad idea in this case IMO.

Zuul said:
For what it's worth, I did a run in Unigine Valley with my crossfired 7970GHz just before I pulled them to put the FuryX in there, there is around 100 points difference in favor of the 7970's. I'd say it's pretty much double the performance of a single 7970. But it's only one benchmark and 1080p.

Edit: I forgot about scaling so maybe not double...but pretty close
Click to expand...

Good to know, I would guess as resolution obsessed the Fury X pulls ahead.
 
Brent_Justice said:
Cause I actually played games with it
Click to expand...

Well you played 5 games with it. A lot of the others do give a lot more of a cross section even if they are only canned benches. But like a lot of others say you gotta look at them all not just one to get the whole picture.
 
durquavian said:
Well you played 5 games with it. A lot of the others do give a lot more of a cross section even if they are only canned benches. But like a lot of others say you gotta look at them all not just one to get the whole picture.
Click to expand...

Not really. Games don't have a long shelf life in the general population. People play a game, they beat the game, they move onto the next game. Using old games for evaluating a new card is sort of foolish IMO. Maybe [H] should still be using Quake 1 as part of their evaluations to make sure they get the whole picture?

For a vast majority of the population, once a game crosses 60 FPS with minimal stuttering, there is no game play improvement. Therefore if the Fury X gets 90 FPS and the 980 Ti get 80 FPS both with max settings, it doesn't mean shit.

Secondarily, nearly all benchmarks do NOT evaluate any form of frame to frame stuttering AT all. This is probably one of THEE most important factors in game play besides a minimal FPS. The only way to evaluate this is either play the game or run frame time analysis on it. Since frame time analysis takes longer than playing a game and most review places already can't afford the time to actually play the games...do you think frame time analysis is actually done except in rare cases? Yep..what I thought.

Finally, throwing out a [H] review because it is an extreme is not correct. The [H] tests differently. Therefore, putting it in the same group as canned benchmarks is scientifically incorrect.
 
durquavian said:
Well you played 5 games with it. A lot of the others do give a lot more of a cross section even if they are only canned benches. But like a lot of others say you gotta look at them all not just one to get the whole picture.
Click to expand...

Have to make a decision, do you value canned benchmarks more, or do you value someone actually playing the games with the card more and telling you their real-world experiences hands on.
 
MacLeod said:
Doesn't surprise me. The [H] review was the only one that trashed it. Everybody else gave it pretty high marks. I don't get it.
Click to expand...

Every review I've seen is pretty much the same...People calling AMD out for over-promising...fake benchmarks... and under delivering...Where are these "pretty high marks" at?
 
There's quite a few reviews where the conclusion on the fury x acts like the 980 ti wasn't part of the benchmark suite. Confuses the shit out of me.
 
Very few games have canned benchmarks with them anymore. As I understand it, these other sites have a set run through of a part of the game that they record and measure with FRAPs or whatever. How is that wrong? I get the disdain for the canned benchmarks as theyre never right it seems. Crysis' benchmark utility would give you frame rates in the high 40's and youd think you were golden. Fire up the game and youre in the low 30's most of the time. Same with Metro's and Batman Arkham City's. But these sites are using actual in game play for their benchmark runs unless Im mistaken. That seems a fair way to do it.

Dont get me wrong, I like [H]'s way of testing highest playable settings but me personally, I look at the Apples to Apples benchmarks more. Id rather have 75 fps with a little less AA than 45 with all the AA trimmings.

GlazedDonut said:
Every review I've seen is pretty much the same...People calling AMD out for over-promising...fake benchmarks... and under delivering...Where are these "pretty high marks" at?
Click to expand...

Theyre not saying its God's own card or anything and their main bitches center around the price being about $50 too high but theyre hardly trashing it like [H] did saying it reminded them of AMD's shitty CPU past.

Guru3d gave it a "recommended" award.

PC World said, "No, the Fury X isn’t Titan-killer that Team Red fans hoped it would be—but it is a GTX 980 Ti equal. This is nothing short of a powerful, thoughtful graphics card that once again puts AMD Radeon on equal footingwith Nvidia’s gaming finest."

Overclock3D gave it the Gold award and said, "We're delighted to award the R9 Fury X our OC3D Gold and Innovation Award. It's good to have you back, AMD."

Tom's Hardware while mainly pointing out the problem with the thing costing the same as the 980 Ti still said, "In the end, AMD has plenty to be proud of. By combining a more resource-rich GPU and our first taste of HBM, it successfully leapfrogs the GeForce GTX 780 Ti, which first cast a shadow over Radeon R9 290X, and the GeForce GTX 980 that sat in a class of its own for several months, landing right next to GeForce GTX 980 Ti."

Look I know the Fury X isnt a 980 Ti killer and it priced too high but from the vast majority of the reviews out there, the only thing wrong with the Fury X is the price. If it was $550-600 we wouldnt be having this conversation. The fact that is priced at the 980 Ti level is kinda dumb in my opinion and hopefully prices will begin to drop to a more realistic point. But saying that the Fury X is so bad that it reminds you of Bulldozer is a little over the top IMO.
 
Last edited:
MacLeod said:
Very few games have canned benchmarks with them anymore. As I understand it, these other sites have a set run through of a part of the game that they record and measure with FRAPs or whatever. How is that wrong? I get the disdain for the canned benchmarks as theyre never right it seems. Crysis' benchmark utility would give you frame rates in the high 40's and youd think you were golden. Fire up the game and youre in the low 30's most of the time. Same with Metro's and Batman Arkham City's. But these sites are using actual in game play for their benchmark runs unless Im mistaken. That seems a fair way to do it.

Dont get me wrong, I like [H]'s way of testing highest playable settings but me personally, I look at the Apples to Apples benchmarks more. Id rather have 75 fps with a little less AA than 45 with all the AA trimmings.
Click to expand...

The problem is, it's not playing the game. You can't tell how a game plays and feels watching a video. I barely pay attention to pure FPS these days. I like reading about the user's experience with a product. To me that is far more valuable.
 
I am new here, can someone explain to me what "canned benchmark" is? I read quite a bit reviews: tomshardware, anandtech, techpowerup, guru3d, are those not reliable?
 
Derangel said:
The problem is, it's not playing the game. You can't tell how a game plays and feels watching a video. I barely pay attention to pure FPS these days. I like reading about the user's experience with a product. To me that is far more valuable.
Click to expand...

Yeah but thats going to lead a lot more to subjective opinions. Aside from microstuttering Ive never noticed a different "feel" between Nvidia cards and AMD card. My boy's rig has a GTX 670 in it and my 290x feel the same especially on games that are capped at 60 fps. I think subjective opinions in GPU reviews are good to include but I pay more attention to the objective numbers more. Even in the microstuttering, that is something that will show up on graphs just like it did when [H] discovered it a few years ago when AMD cards in Crossfire were getting higher average framerates but the Nvidia cards were running smoother because they were keeping a more continuous average where the AMD cards were dipping all over the place.

crislevinZTAMPj said:
I am new here, can someone explain to me what "canned benchmark" is? I read quite a bit reviews: tomshardware, anandtech, techpowerup, guru3d, are those not reliable?
Click to expand...

Some games come with a benchmark utility that will run some graphic scenes from the game and give you frame rates. Problem is that these utilities usually dont represent actual game play and your results will usually be way off from what youll experience in the game. And yes, the sites you listed are reliable. [H] just has a different methodology for GPU testing than the others as they test highest playable settings as opposed to just cranking all the settings to max and then running their benchmarks.
 
crislevinZTAMPj said:
I am new here, can someone explain to me what "canned benchmark" is? I read quite a bit reviews: tomshardware, anandtech, techpowerup, guru3d, are those not reliable?
Click to expand...

A canned benchmark is a term for a benchmark built into a game. It can also be used for custom made benchmark runs the sites use. Those kinds of results aren't unreliable per-say, they just aren't the full picture. It also just means you really can't compare the results H has to someone that just runs benches.
 
MacLeod said:
Yeah but thats going to lead a lot more to subjective opinions. Aside from microstuttering Ive never noticed a different "feel" between Nvidia cards and AMD card. My boy's rig has a GTX 670 in it and my 290x feel the same especially on games that are capped at 60 fps. I think subjective opinions in GPU reviews are good to include but I pay more attention to the objective numbers more. Even in the microstuttering, that is something that will show up on graphs just like it did when [H] discovered it a few years ago when AMD cards in Crossfire were getting higher average framerates but the Nvidia cards were running smoother because they were keeping a more continuous average where the AMD cards were dipping all over the place.
Click to expand...

I find more value in that style of objective review though. Whether it's hardware or software. FPS is fine and dandy, but I just get more out of knowing how something runs for a person or just what they feel about it over-all. I know what I like as far as smoothness and FPS goes and can easily do mental calculations based on what the highest settings get and quickly glancing at the chart and go from there. Highest playable settings tell me more about what my experience with a game will be then a pure benchmark run. Though that is just me and how I look at things. That isn't to say that canned benches have no value, just to me they're not what I'm really after.
 
Brent_Justice said:
Have to make a decision, do you value canned benchmarks more, or do you value someone actually playing the games with the card more and telling you their real-world experiences hands on.
Click to expand...

Sorry you missed my point, I wasn't that thorough. What I was saying as you have to look at more than one. Cant just look at one and think it tells you everything. There are positives to both methods. Your way gives an idea of in game but is very hard to make each run exactly the same so there is some error in the run. But since you run for a hefty duration likely those differences if meant to avoid are nullified somewhat. Canned benches give the exact same situation to any card that runs it, assuming one uses the same settings when different settings are allowed.

It isn't and all or nothing approach but a big picture way to look at it.
 
MacLeod said:
Very few games have canned benchmarks with them anymore. As I understand it, these other sites have a set run through of a part of the game that they record and measure with FRAPs or whatever. How is that wrong? I get the disdain for the canned benchmarks as theyre never right it seems. Crysis' benchmark utility would give you frame rates in the high 40's and youd think you were golden. Fire up the game and youre in the low 30's most of the time. Same with Metro's and Batman Arkham City's. But these sites are using actual in game play for their benchmark runs unless Im mistaken. That seems a fair way to do it.

Dont get me wrong, I like [H]'s way of testing highest playable settings but me personally, I look at the Apples to Apples benchmarks more. Id rather have 75 fps with a little less AA than 45 with all the AA trimmings.
Click to expand...

A recorded run through is still a canned benchmark, its just harder to cheat on.
 
Derangel said:
A canned benchmark is a term for a benchmark built into a game. It can also be used for custom made benchmark runs the sites use. Those kinds of results aren't unreliable per-say, they just aren't the full picture. It also just means you really can't compare the results H has to someone that just runs benches.
Click to expand...

MacLeod said:
Some games come with a benchmark utility that will run some graphic scenes from the game and give you frame rates. Problem is that these utilities usually dont represent actual game play and your results will usually be way off from what youll experience in the game.
Click to expand...

I see, thanks guys.

I would just say they still have values. They may not be able to tell buyers what kind of real life gaming experience they will have with certain games, but they can tell buyers, generally, where each card is on the hierarchy chart. When people makes purchases, that may be all they are looking for.
 
Canned benchmarks are valuable because they can set repeatable scores to judge your hardware upon. Reading reviews based on this can set expectations for how a given piece of hardware will perform.

Gameplay benchmarks, such as what is done here at [H], give a real-world experience in a particular game. This approach is less "repeatable" than canned benchmarks, but gives a real-world experience with hardware, which is valuable in making a purchasing decisiong.

Synthetic benchmarks are even valuable, as the can tell you the potential of the hardware given the software is written to take advantage of the hardware.

Bottom line is, if you use only one measure to make a purchasing decision, you are doing it wrong. I think there are very few here who only use one review site.
 
Hey steve did you notice any noise issues like TR and PCPer brought up?
 
MacLeod said:
Theyre not saying its God's own card or anything and their main bitches center around the price being about $50 too high but theyre hardly trashing it like [H] did saying it reminded them of AMD's shitty CPU past.

Look I know the Fury X isnt a 980 Ti killer and it priced too high but from the vast majority of the reviews out there, the only thing wrong with the Fury X is the price. If it was $550-600 we wouldnt be having this conversation. The fact that is priced at the 980 Ti level is kinda dumb in my opinion and hopefully prices will begin to drop to a more realistic point. But saying that the Fury X is so bad that it reminds you of Bulldozer is a little over the top IMO.
Click to expand...

[H] didn't trash the shit out of the Fury as bad as people are making this out to be. Was the review a bit harsh in its conclusion? Yes.

Was the conclusion unfair given AMD's track record on CPU development? No.

Was it fair to compare an over hyped card that was marketed as a "designed for 4K", that really isn't a 4K killer to a processor that failed to live up to the hype? Yes.

[H] is conclusion basically says HMB is cool, AMD screwed up the release of this product, lets hope they don't keep fucking up or they are going to end up in the same boat as the CPU team.

All the reviews show the same thing:

1. The 980ti is generally faster than the Fury X, and when you overclock the 980ti something you can't do on the Fury X (very well) the 980ti is a much better product.

2.The Fury X card runs cooler, than the 980ti.

3. The Fury X fits in a smaller case.

4. The Fury X is overpriced based on performance.

Does that sound like a good card?
 
My take away from the anand review is that there is still a lot of untapped potential in the Fury. CPU utilization is still way too high and without unlocking voltage it doesn't make sense to talk about OC potential yet. I can't wait to see what the first set of optimized drivers do followed by voltage unlocked OC potential. Feels like there is at least another 10% - 15% additional performance locked away in there.
 
Sharps97 said:
without unlocking voltage it doesn't make sense to talk about OC potential yet.
Click to expand...

.. Unless the reason they haven't unlocked voltage is because of lack of OC potential (maxed out from the factory)
 
DPI said:
.. Unless the reason they haven't unlocked voltage is because of lack of OC potential (maxed out from the factory)
Click to expand...

One can't help but wonder if they pushed the card as far as they could when internal testing showed them how weak they were against the 980 Ti. nVidia did something similar back in the GTX 480 days, releasing a card that teetered on the edge of unusably hot and unstable so they could get a win. Then they refined the architecture and came back with a much better GTX 580. Maybe AMD just needs a good re-spin on Fiji.
 
Sharps97 said:
My take away from the anand review is that there is still a lot of untapped potential in the Fury. CPU utilization is still way too high and without unlocking voltage it doesn't make sense to talk about OC potential yet. I can't wait to see what the first set of optimized drivers do followed by voltage unlocked OC potential. Feels like there is at least another 10% - 15% additional performance locked away in there.
Click to expand...

Don't worry, AMDs crack driver team is on the job!
 
durquavian said:
That crack driver team has 7970 performing at 780 levels far beyond the 680 its original competition.
Click to expand...

Uh, great. Maybe in a few years, Fury X will be the monster overclocking, 40% faster than Titan X, fastest video card on the planet it was said to be?
 
durquavian said:
That crack driver team has 7970 performing at 780 levels far beyond the 680 its original competition.
Click to expand...

Now all AMD needs is a time travel machine!

Seriously though, any number of things could change with time - games could start using heavier shaders for example which would hurt Kepler's IPC dependent scheduler and register limitations. There's no indication the same will happen with Maxwell - it's much easier to get a Maxwell shader core running at full speed.
 
thedocta45 said:
[H] didn't trash the shit out of the Fury as bad as people are making this out to be. Was the review a bit harsh in its conclusion? Yes.

Was the conclusion unfair given AMD's track record on CPU development? No.
Click to expand...

Absolutely it was. Bulldozer was slower than its predecessor. Thats ridiculous! Fury X's only problem is that it came out a couple weeks too late and is $50 overpriced.

Was it fair to compare an over hyped card that was marketed as a "designed for 4K", that really isn't a 4K killer to a processor that failed to live up to the hype? Yes.
Click to expand...

Depending on the review, the Fury is as fast as the 980 Ti at 4K so I say no. Especially considering the 980 Ti didnt exist when AMD was making those claims. Besides, everything is hyped. From cases to power supplies. I dont hold that against any company.

[H] is conclusion basically says HMB is cool, AMD screwed up the release of this product, lets hope they don't keep fucking up or they are going to end up in the same boat as the CPU team.
Click to expand...

No they pretty much said it sucked. Underwhelmed, disappointed, sub-par performance and so on.

All the reviews show the same thing:

1. The 980ti is generally faster than the Fury X, and when you overclock the 980ti something you can't do on the Fury X (very well) the 980ti is a much better product.

2.The Fury X card runs cooler, than the 980ti.

3. The Fury X fits in a smaller case.

4. The Fury X is overpriced based on performance.

Does that sound like a good card?
Click to expand...

It sounds like a card that should be $600 instead of $650 and it then it would be fine. Doesnt sound like the POS that a couple sites are making it out to be. You guys are making it sound like the Fury is 50% behind the Ti. Its very close depending on which review youre reading.

Fuck it, I officially dont care anymore. I guess Ill never understand the rabid AMD hate. First lackluster GPU release from them since they bought ATI and everybody's calling for the end of AMD. I dont get it and probably never will.
 
Last edited:
You must log in or register to reply here.
Back
Top