Unknown Unknown Author
Title: AMD Internal Fury X Benchmarks Revealed
Author: Unknown
Rating 5 of 5 Des:
Fury X Performance Revealed AMD’s launch of the 300 series and Fury X were supposed to have been a carefully orchestrated release of mater...

Fury X Performance Revealed

AMD’s launch of the 300 series and Fury X were supposed to have been a carefully orchestrated release of materials at specific times during the past week and going into next. The Fury X and various other cards are currently on display at E3, and AMD has held several conferences this week giving us pieces of information. Officially, we were supposed to only talk about what was shown at the press conferences, and today is the official launch of the 300 series. You can read our coverage of the 300 series, as well as our review of the Sapphire Tri-X R9 390X, but one thing we weren’t supposed to discuss until next week was Fury X performance. Except there was a “misunderstanding” or some such, the information got posted on the web, and now it has been declared fair game for everyone.

Here’s the thing: internal benchmarks, whether they come from AMD, Intel, Nvidia, or some other company, are never 100 percent trustworthy. They come from a biased source. Which is why most places will say, “take this with a grain of salt.” By tweaking settings and resolutions, performance can be made to favor a certain product more than it might at different settings. As long as we all understand that, posting the information is fine, but until everyone can publish our own independent benchmarks—which still won’t happen until June 24 for the Fury X—it’s too early to draw any final conclusions. We're not saying AMD has done that here and according to our preliminary numbers, we think AMD has some accurate numbers but we'll post our own benchmarks soon to verify.

Let’s use an example that we can talk about to clarify. Here’s a chart showing our testing results of the R9 390X against the GTX 980, and next to it is AMD’s own chart showing a similar comparison:

AMD 390X Internal vs External Benchmarks

Many of our results match up, like Batman: Arkham Origins and Shadow of Mordor. But we happened to run four games/tests that AMD doesn’t include for this particular comparison, and of those, three (GTAV, Tomb Raider, and Unigine Heaven) happened to favor Nvidia. 

We also had different results with The Witcher 3, because we enabled HairWorks and ran at the Ultra preset. Coincidence? We think not. And we’ve encountered similar behavior from Nvidia, so we’re not trying to make AMD out as the bad guys. This is simply Marketing 101, and it’s why we say, “Take internal benchmarks with a grain of salt.”

DETAILS and Statistics

With that introduction out of the way, then, here are the Fury X vs. GTX 980 Ti performance comparisons in all their glory. We’re including the complete pages, with all the extra information that sometimes gets left out. No, we didn’t run these benchmarks, and we’re not going to pretend we did. The test system, if you’re wondering, is an i7-5960X running stock clocks, which means there’s also a potential for the CPU to be a bottleneck in some cases—it’s why our GPU test bed is overclocked to 4.2GHz.

The following are taken from the reviewer's guide that accompanies the Fury X:

Amd Fury X Benchmarks

Amd Fury X Benchmarks 2

Amd Fury X Benchmarks 3

Amd Fury X Benchmarks 4k

Amd Fury X Benchmarks 4k 2

And what we see is that the Fury X wins in every single benchmark that AMD ran—at least at the selected settings and in the selected games. Notice how the 390X was compared to the GTX 980 in Alien Isolation, Battlefield Hardline, Civilization Beyond Earth, Project Cars, and Skyrim, but in the Fury X vs. 980 Ti they’re missing. In their place we have a few other titles, however: Assassin’s Creed Unity, Battlefield 4, Crysis 3, and Hitman Absolution. That doesn’t mean Nvidia would have performed in every missing game—e.g., we know Hitman Absolution at least clearly favors the 390X over the 980—but the decisions on what to include and exclude aren’t simply made at random.

But there’s still only so much you can do with customized settings to “improve” performance, so taken at a higher level one thing is certain. Even at 4K, where the 4GB HBM compared to 6GB GDDR5 might present some issues, AMD’s Fury X is still quite competitive with the 980 Ti. There will undoubtedly be times where it’s slower, and there will also be times where it’s faster. The two cards cost the same, so that’s to be expected—if the Fury X were significantly faster than the 980 Ti, no doubt it would carry a price premium. The naming also suggest that perhaps AMD was hoping to go up against the Titan X rather than the 980 Ti, and perhaps Nvidia spoiled their fun by shipping 980 Ti a couple weeks ago. C’est la vie! However, you can compare numbers with our original Titan X review.

Tying it up with a bow

There are still other questions to consider, of course. Even if performance is basically tied, or perhaps even favors the Fury X, do you want a graphics card with a closed loop cooling solution? It can work wonders for cooling with one GPU, but if you wanted to run 2-way or 3-way CrossFireX, you need a case that can actually handle all of the CLCs. The Fury X also has a slightly higher TDP, though 250W vs. 275W isn’t nearly as problematic as the 390X 275W vs. the 980’s 165W. And even though we aren’t seeing any major issues in the AMD provided benchmarks with 4GB HBM vs. 6GB GDDR5, the omission of GTAV (which can use nearly 6GB at 4K with all setting maxed) suggests lack of memory can still be a problem at times. Not that you need to max out every setting in every game, of course.

The takeaway right now is that, generally speaking, we expect the Fury X to match up well against the 980 Ti. AMD even said that Fury X will be the fastest single GPU card when it comes out, which seems to indicate the company may be aiming to even beat Titan X.

We’ll have fully independent results next week, at which time we can finally declare a winner—or a tie, which seems far more likely. Until then, happy gaming, on whatever hardware you currently happen to run.



From maximumpc

from http://bit.ly/1G8JMr0

Advertisement

 
Top