Review Index:
Feedback

AMD Radeon HD 7990 6GB Review - Malta Gets Frame Rated

Author: Ryan Shrout
Manufacturer: AMD

Battlefield 3

Battlefield 3 (DirectX 11)


 

Battlefield 3™ leaps ahead of its time with the power of Frostbite 2, DICE's new cutting-edge game engine. This state-of-the-art technology is the foundation on which Battlefield 3 is built, delivering enhanced visual quality, a grand sense of scale, massive destruction, dynamic audio and character animation utilizing ANT technology as seen in the latest EA SPORTS™ games.

Frostbite 2 now enables deferred shading, dynamic global illumination and new streaming architecture. Sounds like tech talk? Play the game and experience the difference!

Our Settings for Battlefield 3

Here is our testing run through the game, for your reference.

View Full Size

Our first test result shows that the best case scenario, the new Radeon HD 7990 is just a bit slower than the HD 7970s in CrossFire and faster than both the GTX 690 and the GTX TITAN across the board.

View Full Size

Our observed frame rates though, after removing any runt frames that take up an insignificant amount of space on the screen, are much lower and you can see that the HD 7990 suffers in the same way as the HD 7970 CrossFire configuration does.  Both NVIDIA solutions though maintain consistent frame rates compared to the previous result.

View Full Size

The black and orange lines that represent the AMD HD 7990 / HD 7970 CF results are both indicative of the of the alternating high and low frame times that result in runt frames and artificially inflated frame rate numbers.  The green line that is the GTX TITAN is the best overall result in this regard with tight frame times and the GTX 690 is a bit more variant but is still doing better than the AMD options.

View Full Size

Taking away the runt frames, the average frame rates both the HD 7990 and the HD 7970s is just over 100 FPS while the GTX Titan hits 120 and the GTX 690 jumps to 140 FPS. 

View Full Size

By far the least variant configuration here is with the GTX Titan card as even the GTX 690 sees some degradation here.  Both the HD 7990 and the HD 7970s in CrossFire are much worse relationally.

 

View Full Size

View Full Size

The observed and FRAPS reported frame rates again differ with the CrossFire solutions (which the HD 7990 is still built around) while the NVIDIA options maintain the same result, even the dual-GPU GTX 690.

View Full Size

The comparison between the NVIDIA and AMD results is even more dramatic at 2560x1440 with the GTX Titan producing by far the best result of the four.  Even those the frame times are higher resulting in lower frame rates, there is basically no frame time deviation.  The black and orange lines though are quite the opposite.

View Full Size

The average frame rate differences expand at this resolution as well - the new HD 7990 is only able to get 58 FPS on average for the full benchmark run while the Titan hits 75 FPS and the GTX 690 95 FPS. 

View Full Size

Our graph that shows the potential for stutter gives the GTX cards the benefit once again with even the GTX 690 and its pair of GK104 GPUs a solid result, leaving the HD 7990 and the HD 7970s well behind.

 

View Full Size

View Full Size

Though our HD 7970 CrossFire results didn't pan out correctly, the HD 7990 was able to give us some solid data.  Observed frame rates are MUCH lower...

View Full Size

Our problems with Eyefinity continue with the HD 7990, as we see more dropped frames than runts.  The GTX 690 is definitely a more variant experience than the GTX Titan but both are very playable experiences. 

View Full Size

Just so you can see what the HD 7990 result looks like on its own, with the drop data included, I have shown you above. 

View Full Size

Even though the Titan has a lower average frame rate (the 50th percentile) it actually keeps its frame rate a bit longer than the GTX 690 that has more variance to its frame times.  The HD 7990 is barely able to breach the 30 FPS mark.

View Full Size

Because our frame variance data doesn't have runts to judge differences by, the HD 7990 Eyefinity 5760x1080 result actually appears to be good, though with that one major caveat. 

 

April 23, 2013 | 10:05 PM - Posted by Yang Ji (not verified)

Very nice review Ryan. FPS does mean everything. Great methodology.

April 24, 2013 | 06:05 PM - Posted by Anonymous (not verified)

FPS won't mean everything anymore as soon as amd has a single feature like PhysX or TXAA that nVidia does not - then we will hear how IT MATTERS and is of utmost import, and definitely makes the purchasing decision for everyone...

I can hardly wait... since AMD is about 5 or 6 massive features behind nVidia, and all those of course do not matter at all, only fps, which as we all now must admit instead of just me saying it for years, amd FAILS AT cf FPS.

Man I tell you, I am so, so sick of it, I have been so sick of it for so long, for so many years, and now finally, I bask IN THE HOLY GLORY OF THE TRUTH ! AMD SUCKS VERY BADLY !

I told you all so for years, while I got kicked and stomped with lies and fanboy emotes gone wild, man is it ever good to be totally freaking vindicated.

I will add, no thanks and a pox on all you amd wackadoos that screamed for years about your amd loser cards. I have recieved zero apologies from all of you, and in fact, have seen EXACTLY ZERO APOLOGIES ON THE WEB FROM ANY OF THEM ! SITE OWNERS, AND ALL THE POSTERS !

Thus we can be certain it will all happen again.
Sigh.

April 25, 2013 | 06:03 PM - Posted by Anonymous (not verified)

no.

April 28, 2013 | 05:10 PM - Posted by Anonymous (not verified)

You deserve nVdia.If it makes it easier to have special names for things that all modern graphics cards are capable of.

When nVidia gets Full directX11.1 support we'll talk.

April 23, 2013 | 10:07 PM - Posted by Boxes (not verified)

Well I was waiting on getting my GPU(s) until this the 7990 was reviewed to make sure I wasn't missing out on a new CFX patch. Now the 780 is coming out in May....

I CANT TAKE INTEL HD4000 ANY LONGER MAN.

What do with $1200? Dual 680 4GB or Single Titan OR Wait for 780 and pick up a couple of those.

April 24, 2013 | 12:25 AM - Posted by Ryan Shrout

You could buy a single GTX 680 and then contribute to our Indiegogo project?  :)

http://www.indiegogo.com/projects/pc-perspective-video-set-and-productio...

April 24, 2013 | 12:44 AM - Posted by Prodeous (not verified)

Ohh Ryan....

Spamming your own site :(

Still If something, instead of a 680 it should be a Titan and then contribute a bit to PcPer :P

April 24, 2013 | 12:37 AM - Posted by Prodeous (not verified)

If you can't wait any longer to spend the money I would say Titan, as both dual GPU solutions still are not as smooth as a single card. And far more power hungry.

April 23, 2013 | 10:27 PM - Posted by derz

You can probably kill a man if you fling some of those cards at some unfortunate sap.

April 24, 2013 | 12:41 AM - Posted by Anonymous (not verified)

This testing methodology is deliberately created to make Nvidia look good.
I watch other reviews and the frame times don't look that bad.
Here Nvidia looks great and AMD look very bad. Plus a saw reviews where average and minimal fps are quite tight.
Others also used the
AMD Catalyst Frame_Pacing_Prototype v2 For Radeon HD 7990

and it showed nice reduction of frame latency. In time it can only get better

April 24, 2013 | 12:49 AM - Posted by Prodeous (not verified)

Testing methodology isn't deliberately created to make Nvidia look good, it was to show the issues associated with dual GPU.

The only reason NVidia cooperated was that they knew they had the issue fixed where as AMD ignored it for some time.

If Nvidia still had the issue, they would never participate in developing this testing methodology.

Luckily this method of not looking at FPS (started up by TechReport site) opened our eyes to something we have seen for some time but simply accepted it. And it Forced AMD to actually do something about it. And I have to say, they did an amazing job on it.

April 24, 2013 | 02:46 AM - Posted by rrr (not verified)

Read the last page, PCPer mentioned that driver too, just as a reminder it exists and shows promising results. But it isn't user ready yet, so doing entire review on it would be stupid.

April 24, 2013 | 07:31 AM - Posted by Josh Walrath

Read Ryan's other article here:  http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-AMD-Improves-CrossFire-Prototype-Driver

April 24, 2013 | 09:13 AM - Posted by bystander (not verified)

If you want another view, look at THG's version: http://www.tomshardware.com/reviews/radeon-hd-7990-review-benchmark,3486...

The result is the same, though their definition of a runt frame is far more forgiving (less than 2% of the screen, or 20 lines worth). However, they added objective testing at the end, with several different people taking blind tests. All of them found the 690 was smoother.

April 27, 2013 | 04:07 PM - Posted by Anonymous (not verified)

Also note Ryan is using REALLY old 314.07 drivers.

NV has 314.14, 314.22 and even 320's out now (heck they had a few others in there also). Not sure why he keeps using an ancient driver other than saving a lot of retesting.

NV has much higher scores in almost every game now.

Start testing with a new driver please. Every other review I've seen uses 320. You are at least a few months behind.

So AMD would look even worse vs the latest drivers.

April 24, 2013 | 01:30 AM - Posted by arbiter

7970 in CF is better deal if you are going to overclock the card. Seeing how power envelope and constraints is limited on this card, there is very little overclocking headroom.

April 24, 2013 | 01:55 AM - Posted by Mountainlifter

Great Review Ryan and team.

Now what about those 780 rumors that just hit?

April 24, 2013 | 02:46 AM - Posted by Howie Doohan (not verified)

Message to AMD, fix your damn drivers! I agree they are headed in the right direction but they're still miles behind Nvidia on this issue.

April 24, 2013 | 06:49 AM - Posted by FenceMan (not verified)

Ryan,

I love your site, podcasts, and reviews. I cannot for the life of me figure out why you have once again railroaded AMD without trying any of the publicly available solutions to fix your issues? Can you at least acknowledge that Radeon Pro exists? I understand it is a third party program, but we are PC gamers, we tweak things, is it really that far of a stretch to imagine that in the real world we take advantage of Radeon Pro? I sometimes spend hours tweaking game settings before I actually start playing the game, I am sure most of your readers do the same, why do you continue to ignore the obvious?

April 24, 2013 | 07:30 AM - Posted by Josh Walrath

Those things really didn't fix much and introduced other issues (Ryan covered that in an article that had the videos you could download).

 

Also, this!  http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-AMD-Improves-CrossFire-Prototype-Driver

April 24, 2013 | 08:26 AM - Posted by Ryan Shrout

For this review I didn't put a high focus on the RadeonPro solution because I knew and was playing with the prototype drivers from AMD - that is the REAL solution.  I will still tryout RadeonPro eventually but that is just a hack or patch job in reality.

April 24, 2013 | 08:44 AM - Posted by Fenceman (not verified)

All due respect call it what you like it works??

Really love to see in depth report based on your experiences which I recognize are greater than most of us.

April 24, 2013 | 09:16 AM - Posted by bystander (not verified)

RadeonPro fixes the problem by limiting FPS. However, it has at least a couple draw backs you will see noted by different people.

1) It does lower your average FPS to work.
2) If a game has different average fps in different areas, you have to lower it to the worst case, lower FPS even more (skyrim users have to limit FPS for outdoors, even though indoors has much higher FPS)

April 24, 2013 | 09:00 AM - Posted by Fenceman (not verified)

http://www.overclock.net/t/1363712/pcper-frame-rating-part-3-first-resul...

Its fair to say that AMD currently requires more tweaking, its not fair to say issues cant be corrected, just my opinion. Look forward to the podcast later.

April 24, 2013 | 08:14 AM - Posted by FenceMan (not verified)

Maybe I just dont get or see it.... I had 7950 crossfire and picked up a GTX 690 somewhat cheap on Ebay and got frustrated very quickly when I couldn't get it to work as well as what I had. Tomb Raider was brutal if I turned on the hair and Bioshock took a lot of tweaking to get me where I was with the 7950's. Just sold the 690 (little profit bonus) and ordered a couple of Vapor X 7970's....

April 24, 2013 | 09:20 AM - Posted by bystander (not verified)

Tomb Raider had problems with the 314 drivers on Nvidia, the drivers that were current on release. Going back to 310 make it run great. I believe the most recent 314.22 drivers fixed the issue as well. Bioshock infinite is having troubles with both cards.

Anyways, the point isn't about FPS, but performance. Tomb Raider is particularly bad on crossfire by all accounts: http://www.tomshardware.com/reviews/radeon-hd-7990-review-benchmark,3486...

April 24, 2013 | 10:02 AM - Posted by Victah Von Doom

Hey Ryan i just wanted to say congratulation on the 10k mark you and your team truly deserve it. And don't let these trolls influence you man you're doing an amazing job. You write it we read it. Im going to contribute $10 this weekend.

No homo
i love you man

April 24, 2013 | 10:54 AM - Posted by Fenceman (not verified)

It is possible to disagree without being a "troll".....

April 24, 2013 | 02:31 PM - Posted by Tim Verry

Thank you for your support!

April 24, 2013 | 11:10 AM - Posted by Victah Von Doom

Yes It is possible to disagree without being a "troll" but the real question is are you in a position to disagree.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.