AMD Radeon HD 7990 Review: Dual GPU Comeback

By on April 23, 2013, 11:00 PM

The current generation AMD GPU series collectivelly known as "Southern Islands" were released over a year ago, with the beginning of its rollout in January 2012. Sixteen months later, the Radeon HD 7000 series is still very much relevant, as AMD continues to release new models under the same GPU family.

Still AMD feels they need to offer an ultra-expensive graphics card as well and therefore today they are releasing the Radeon HD 7990. Although it's been over a year since they launched the Radeon HD 7970, we are just getting an official dual-GPU version, in truth they have been around for some time. Products such as the PowerColor Devil 13 HD 7990 6GB, HIS Radeon HD 7970 6GB IceQ X2 and Asus ROG Ares II have been doing the rounds for quite a while now.

The Radeon HD 7990 it takes a pair of 7970 GPUs with overclocked cores (from 925MHz to 1000MHz), while boosting the GDDR5 memory from 1375MHz to 1500MHz. These clock speeds happen to match the 7970 GHz Edition, though the 7990 doesn’t feature a Boost clock, so we feel the GPUs are better compared to the standard 7970.

Read the complete review.




User Comments: 222

Got something to say? Post a comment
donutpolice said:

Time for all Nvidia fanbois to go jelly....

Darth Shiv Darth Shiv said:

Frame latency is an interesting measurement. Do you also bench latencies of different inputs in LCD panels?

madboyv1, TechSpot Paladin, said:

What a monster of a card, but I'm totally not jelly, don't have the money for at $1000 GPU setup. =p

Also, I think the Temperature chart on page 8 is mislabeled as a second power efficiency chart. =o

1 person liked this | St1ckM4n St1ckM4n said:

Someone please explain how the card is consistently clocking better FPS than xfire'd GHz cards? Internal connector better than external...?

The test bench mobo supports x16 lanes on both PCI-E slots, so it's not this...

Staff
Steve Steve said:

Someone please explain how the card is consistently clocking better FPS than xfire'd GHz cards? Internal connector better than external...?

The test bench mobo supports x16 lanes on both PCI-E slots, so it's not this...

Not sure if consistently is the best word to describe it. The 7990 was slower in Tomb Raider and Resident Evil 6 by 6%, 5% slower in Far Cry 3 and 3% slower in Medal of Honor while the performance in Max Payne 3, DiRT 3 and Crysis 3 was the same. It was only 1% faster in Battlefield 3 and Sleeping Dogs (so you can call that the same) while the only game it was actually faster was Hitman Absolution. So by consistently you mean one game

dividebyzero dividebyzero, trainee n00b, said:

Frame latency is an interesting measurement. Do you also bench latencies of different inputs in LCD panels?

Apples and oranges comparison. For a start, display latency is constant and shows itself by input lag, while GPU latency shows up by stuttering and inconsistent frame rates due to widely inconsistent frame rendering.

An extreme example here:

While the average for tri-crossfire looks pretty good, the actual user experience paints a different story:

With AMD Radeon HD 7970 GE 3-way CrossFire we had a very bad experience in this game. It wasn't so much to do with raw performance as it was extremely bad stuttering. It was terrible, taking into account the game's natural stuttery behavior and combine that with the inconsistent framerates we experience with CrossFire equaled a horrible experience in this game

St1ckM4n St1ckM4n said:

@Steve 1% is 1%, I know it's very close (and probably in the realms of testing anomalies) but it's still 1%. The GHz card can even boost, so this is why I'm surprised the 7990 clocked better at ALL.

Staff
Steve Steve said:

@Steve 1% is 1%, I know it's very close (and probably in the realms of testing anomalies) but it's still 1%. The GHz card can even boost, so this is why I'm surprised the 7990 clocked better at ALL.

1% is certainly well within the error testing margin so it's not really any different from the other tests where the 7990 and 7970 GHz CF cards delivered the same performance. Again there was only one game where the 7990 was genuinely faster, why that was isn't exactly clear. Most likely do to the 7990 using a slightly newer driver.

alex1234 said:

AMD should have solved the crossfire latency issue first and then released this card. May be AMD is getting ready for their 8000 series. The game bundle coming with this card is extraordinary from AMD. I like the back plate, Nvidia should include those things in their high end cards(atleast). Nvidia is making a lot of profit compared to AMD.

ghasmanjr ghasmanjr said:

Another excellent review but I have one question to ask:

Why did you test sli 660s? I was expecting to see sli 680s because that is the "single card configuration" for the 690 which is the direct competition of the 7990, the dual card configuration of the 7970.

ghasmanjr ghasmanjr said:

Btw, HOLY SHIT to the free games bundled with this thing. That is one impressive sales incentive.

Staff
Steve Steve said:

Another excellent review but I have one question to ask:

Why did you test sli 660s? I was expecting to see sli 680s because that is the "single card configuration" for the 690 which is the direct competition of the 7990, the dual card configuration of the 7970.

Unfortunately we don't have an endless treasure trove of graphics cards, if we had a second GTX 680 handy we would have included SLI results. The GTX 660 Ti SLI cards were included because ... you guessed it we had two

It is very hard to get as many high-end graphics cards as we do and hold onto them, getting two is all the more harder. Still you have the GTX 690 there so it shouldn't be all that difficult to work out where a pair of GTX 680's would be sitting on those graphs.

Alpha Gamer Alpha Gamer said:

Steve, all the games used in the test have great scaling under dual gpu solutions. Did you already know that beforehand and, therefore, conducted the tests or was it it just a happy coincidence?

My point is: are there really that many games so well optimized out there so that we can consistently expect that kind of scaling ?

Staff
Steve Steve said:

Steve, all the games used in the test have great scaling under dual gpu solutions. Did you already know that beforehand and, therefore, conducted the tests or was it it just a happy coincidence?

My point is: are there really that many games so well optimized out there so that we can consistently expect that kind of scaling ?

Certainly with AAA titles yes. These days Crossfire and SLI support is common and most games are supported, most quality games that is.

Ranger12 Ranger12 said:

I'm still impressed with how 660ti's in SLI perform.They're consistently right up there competing with the top cards despite costing half as much.

ghasmanjr ghasmanjr said:

Unfortunately we don't have an endless treasure trove of graphics cards, if we had a second GTX 680 handy we would have included SLI results. The GTX 660 Ti SLI cards were included because ... you guessed it we had two

It is very hard to get as many high-end graphics cards as we do and hold onto them, getting two is all the more harder. Still you have the GTX 690 there so it shouldn't be all that difficult to work out where a pair of GTX 680's would be sitting on those graphs.

I understand and I appreciate the clarification. I'm just a little interested/biased towards seeing those results because I have 680s in sli

cldmstrsn cldmstrsn said:

I totally would be jealous if crossfire was better than SLI. But until AMD can get equal or better performance from their Xfire I will be sticking with Nvidia. So no jelly's here haha.

Guest said:

$1000 worth of microstutter

Guest said:

Coming back and forth from AMD and Nvidia I totally understand frame latency now. It's pretty obvious on AMD in most games. The new drivers that Nvidia released yesterday are supposed to push fps in popular games even more.

amstech amstech, TechSpot Enthusiast, said:

It's no secret that my personal loyalties are with Nvidia. Over the past 5 years GTX's have been smoother cards with more features/less problems. It was nice to see the 7970 have the success it did but I still never considered it the single GPU king during that period because it came with an asterisk; it never definitively beat the 680 outright and still it loses stock for stock in games.

7970 owners are only fooling themselves with thier performance claims...these results are devious and don't represent the stuttery gameplay or frametime results. Simply put, they are not as good as GTX's and never were.

They are bang for your buck GPU's and that has not changed.

A GTX is the best GPU money can buy and offers the smoothest, most feature rich gaming experience money can buy. Period.

JC713 JC713 said:

What happened to the Prototype driver and the 13.5 driver? Toms Hardware used those.

cmbjive said:

Well, that about settles it.

I'm getting two 660 TI's for SLI. As someone mentioned above it is surprising that these two cards perform almost as well as the $1,000 cards and cost several times less.

Fbarnett Fbarnett said:

Well, that about settles it.

I'm getting two 660 TI's for SLI. As someone mentioned above it is surprising that these two cards perform almost as well as the $1,000 cards and cost several times less.

Why would you do that 1 680 would do better and I got news for you a lot of games don't work right with crossfire or sli

Don't think that 2 cards are better than one look at benchmarks and you will see 1 680 will beat out 2 660ti.

JC713 JC713 said:

Well, that about settles it.

I'm getting two 660 TI's for SLI. As someone mentioned above it is surprising that these two cards perform almost as well as the $1,000 cards and cost several times less.

Because, the 660Ti is a 680 with disabled cores, TMUs, and ROPs. It performs on par because it happens that some games are good at scaling. That is not true for a lot of games. With crossfire and SLI, there is a higher chance of graphical corruption. Fbarnet is right in that aspect, but a 680 is a better option because it can play most games on highest settings @ 1080P. If you arent playing on >1080P, dont SLI.

cmbjive said:

Well, that about settles it.

I'm getting two 660 TI's for SLI. As someone mentioned above it is surprising that these two cards perform almost as well as the $1,000 cards and cost several times less.

Because, the 660Ti is a 680 with disabled cores, TMUs, and ROPs. It performs on par because it happens that some games are good at scaling. That is not true for a lot of games. With crossfire and SLI, there is a higher chance of graphical corruption. Fbarnet is right in that aspect, but a 680 is a better option because it can play most games on highest settings @ 1080P. If you arent playing on >1080P, dont SLI.

I will be gaming at 1080P, but you and the poster before raise some good points. I'll do some more research on both options because I'm not going to be getting a new GPU until sometime in late summer/early fall (hopefully before either Xbox 720 and PlayStation 4 launches).

1 person liked this |
Staff
Steve Steve said:

What happened to the Prototype driver and the 13.5 driver? Toms Hardware used those.

I got them last minute but so far we have only seen improvements in just a few games. At this stage we will likely wait for an official driver before providing a complete review.

JC713 JC713 said:

I got them last minute but so far we have only seen improvements in just a few games. At this stage we will likely wait for an official driver before providing a complete review.

Meh hopefully they fix that. I have faith in AMD. They really need to restructure their driver team. That is what is holding them back :/.

JC713 JC713 said:

I will be gaming at 1080P, but you and the poster before raise some good points. I'll do some more research on both options because I'm not going to be getting a new GPU until sometime in late summer/early fall (hopefully before either Xbox 720 and PlayStation 4 launches).

Wait for the 700 series. They are supposed to be released in late May/early June. The 760Ti is rumored to use the same amount of cores as the 670 did, etc. I am gonna go into much detail because it isnt official and there is a lot info that is irrelevant.

Staff
Steve Steve said:

Wait for the 700 series. They are supposed to be released in late May/early June. The 760Ti is rumored to use the same amount of cores as the 670 did, etc. I am gonna go into much detail because it isnt official and there is a lot info that is irrelevant.

I am confident that they will. Shouldn't be much longer now and they will have a finished driver.

Wait for the 700 series. They are supposed to be released in late May/early June. The 760Ti is rumored to use the same amount of cores as the 670 did, etc. I am gonna go into much detail because it isnt official and there is a lot info that is irrelevant.

Yes samples are about a month off.

JC713 JC713 said:

I hope they change their Catalyst Install Manager soon. It has literally been the same for years. They really need to introduce something like nVidia's simple installer. A lot of AMDs interfaces look like they are from the XP era.

JC713 JC713 said:

Are we gonna get a retest with the 13.4 WHQL Steve? [link]

Staff
Steve Steve said:

Are we gonna get a retest with the 13.4 WHQL Steve? [link]

No because that is just a certified 13.3 beta driver. The retest wont happen until the final Latency driver is released, we wont see in real improvements from AMD until then.

JC713 JC713 said:

No because that is just a certified 13.3 beta driver. The retest wont happen until the final Latency driver is released, we wont see in real improvements from AMD until then.

True, it just has the WHQL tag on it. But something notable that AnandTech brought up is that the 690 works under a 300W power envelope and the 7990 works under a 375W power envelope. That means the 690 should theoretically produce 20% less heat. But, the 7990 also brings 3GBx2 instead of the 690s 2GBx2. So they are basically equal at this point with the latency issues. I wanna see if AMD can solve these issues to combat the 700 series if the rumored release date is true.

VitalyT VitalyT said:

Increasing the board size and pumping up power consumption isn't really a progress.

Someone needs to educate AMD the notion of ergonomics.

For the first time in my life I downgraded graphical card inside my desktop PC from the 3-year old energy-sucking monster HD 5870 to a fanless HD 6670 that's half the size.

LNCPapa LNCPapa said:

It's also less than half the potency.

VitalyT VitalyT said:

It's also less than half the potency.

It is not. I ran my own tests on it, and it performs quite well. See reviews about the card on the internet. Windows 7 dropped performance index from 7.8 to 7.1, but for what I need it is still plenty. The only game I ever play is SC2, which still gives me about 30fps in 2560x1600 with almost everything on maximum, which is enough for such game.

The main reason I did such change though was to get my new monitor (DELL U3014) to use its DisplayPort 1.2 in its full capacity, which the old card couldn't support.

Also HD 6670 is the best money-performance choice today if you have PCI Express 2.0 motherboard. It costs about $90, tiny fanless design - great for a quiet and ergonomic system:

1. Amazon: [link]

2. NewEgg: [link]

There is a good reason it is one of the most highly rated graphical cards

1 person liked this |
Staff
Steve Steve said:

Increasing the board size and pumping up power consumption isn't really a progress.

Someone needs to educate AMD the notion of ergonomics.

For the first time in my life I downgraded graphical card inside my desktop PC from the 3-year old energy-sucking monster HD 5870 to a fanless HD 6670 that's half the size.

You must be taking your crazy pills today. As LNCPapa said its less than half as fast and that's fact.

[link]

Note where the 6670 is sitting (at the very bottom) while the 5870 is in the middle. Its also 3x faster in COD Black Ops II.

As for value for money the 6670 is a terrible choice.

For the same money this consumes less power and is much faster...

[link]

VitalyT VitalyT said:

As LNCPapa said its less than half as fast and that's fact.

And that fact comes from which source exactly? Made it up all by yourself - didn't you? I have both cards here, and I put both through multiple tests. Perhaps you need to get your "facts" together.

As for value for money the 6670 is a terrible choice.

99% of people who bought it think it was the best choice they made. See all the excellent reviews that card received. You must be that unhappy 1%, which is fine by me, as long as you're trying to sell your 1% as a fact to everybody else here.

You must be taking your crazy pills today.

I may have mood swings at times, but you must be this crazy without any pills to write such nonsense...

Staff
Steve Steve said:

You mean to tell us that the Radeon HD 6670 which as a 128-bit wide memory bus, produces 64GB/s bandwidth in its strongest form and has a texture fill rate of 19.2GT/s is faster than the 5870. The 5870 which has a 256-bit wide memory bus and a memory bandwidth of 153.6GB/s along with a texture fill rate of 68.0GT/s? So far everything seems to be at least double, if not more

If you are going to hit us with newegg.com user reviews then we might as well give up now. Yes some people bought the 6670 knowing no better and they were happy enough to make mention of it on the site they bought it from, that does not make it a good graphics card and that certainly doesn't make it good value.

The 6670 is a slug and shouldn't be used for gaming, there were far better solution available upon its release and there are certainly much better options available now.

There isn't a single game where the 6670 is faster than the 5870 but if you can provide us with one link to a professional review that says otherwise I will gladly take a look.

You might have missed the COD Back Ops II link so here is another for you...

[link]

When testing Far Cry 3 the 6670 was so slow we dumped it, the 6750 is included instead and again notice where that is sitting on the graph opposed to the 5870...

[link]

Care for some Diable III maybe? 38fps vs. 96fps!

[link]

I'm impressed you can play StarCraft 2 maxed out at 2560x1600 since the 5670 which isn't much slower than the 6670 averaged just 18fps. Are you sure its the 6670 you have there?

[link]

VitalyT VitalyT said:

Did you even read what I wrote initially? I said I downgraded my video card, I said Windows 7 Performance Index dropped from 7.8 to 7.1, etc,... where on earth did you see me saying anything what you wrote here??? That is crazy...

Staff
Steve Steve said:

Did you even read what I wrote initially? I said I downgraded my video card, I said Windows 7 Performance Index dropped from 7.8 to 7.1, etc,... where on earth did you see me saying anything what you wrote here??? That is crazy...

Sorry my mistake then I though a few posts back LNCPapa said the 6670 is half as fast as the 5870 and you fired back fight away with "It is not. I ran my own tests on it..."

Though you also said...

And that fact comes from which source exactly? Made it up all by yourself - didn't you? I have both cards here, and I put both through multiple tests. Perhaps you need to get your "facts" together.

I just provided those facts at your request for some facts. Now all the facts are here.

VitalyT VitalyT said:

Lost in translation then.

To digest, I bought it because I needed DisplayPort 1.2 to work in full capacity with the new monitor, and for a fanless tiny graphical card that fits into PCI Express 2.1 and under $90 that card is a good choice, and all who bought it agree. And it is easily the most ergonomic card also. There is not better choice that supports Express 2.0, they better ones require Express 3.0, just too much of a change, if you already have a good working system

VitalyT VitalyT said:

To your update, I only objected to LNCPapa saying that the card is half the power, because it is absolutely not. It is slower, of course, but not by half, not even close. From my tests, about 15% or so, but that wasn't comprehensive. And Windows 7 dropped it to 7,1 from 7.8, which seems about right.

Staff
Steve Steve said:

To your update, I only objected to LNCPapa saying that the card is half the power, because it is absolutely not. It is slower, of course, but not by half, not even close. From my tests, about 15% at most, but that wasn't comprehensive. And Windows 7 dropped it to 7,1 from 7.8, which seems about right.

And this is where we disagree. In fact if you look at my links provided you will see that the 6670 is in fact 3.4x slower in COD, 3.0x slower in Warfigher, presumable more than 2.2x slower in Far Cry 3 and 2.5x slower in Diablo III.

[link]

[link]

[link]

[link]

You are obviously please with your 6670 and that's fantastic. However there are far better options for $100 and that is my opinion and I have also done a bit of testing.

Also as a hot tip the Windows 7 performance index under no circumstances should be used to measure GPU performance, especially for gaming.

VitalyT VitalyT said:

Well, I do not play FPS games, and as I wrote earlier its performance in SC2 gave good results, which was all I cared about.

Just to return to my first message, I only mentioned this downgrade because the article was about the most anti-ergonomic product AMD ever made, perhaps, and for that reason I resend it being anything good or associated with progress in PC graphics. It's like trying to promote a V12 engine today that though pumps out 1000bhp, gives 10mpg and weights 500kg, just as ridiculous.

Time to go green people!

Guest said:

Steve plz include amd prototype driver perf improvement in the test...waiting for it

Guest said:

Not really that great of a comeback if it's just several percent better than already all too old GTX 690. I even expected more. Is AMD slipping here as well?

It will hold for some time, but come June GTX 780 will probably prove to be a better investment.

1 person liked this | Burty117 Burty117, TechSpot Chancellor, said:

It's also less than half the potency.

It is not. I ran my own tests on it, and it performs quite well. See reviews about the card on the internet. Windows 7 dropped performance index from 7.8 to 7.1

Just going to stop there, are you seriously using the windows index as a performance measure? my core i5 HD3000 in my laptop gets 6.3, enough said.

Also HD 6670 is the best money-performance choice today if you have PCI Express 2.0 motherboard. It costs about $90, tiny fanless design - great for a quiet and ergonomic system:

No, its not, not even close, as per the link Steve so kindly gave you, for that price range the best money-performance choice is an Nvidia 650 and what the hell has PCI Express 2.0 got to do with any of this? if you have a PCI Express 3.0 motherboard its backwards compatible and if you check comparisons PCI Express 3.0 doesn't make much if any difference in games anyway. And if the graphics card supports PCI Express 3.0 it is backwards compatible to PCI Express 2.0 motherboards? why are we even talking about this?

There is a good reason it is one of the most highly rated graphical cards

No its not, in fact across the internet the world over you'll find its very low in the ratings, only on places like new egg will it have a half decent score because it was purchased by someone who... well by a particular set of people

the card is half the power, because it is absolutely not. It is slower, of course, but not by half, not even close. From my tests, about 15% or so, but that wasn't comprehensive.

So the small amount of testing you did shows only a 15% decrease? have you got a dual core AMD Athlon in your rig or something?! every single test across the internet will clearly show you've put in a graphics card with almost a third of the performance.

1 person liked this | GhostRyder GhostRyder said:

It's no secret that my personal loyalties are with Nvidia. Over the past 5 years GTX's have been smoother cards with more features/less problems. It was nice to see the 7970 have the success it did but I still never considered it the single GPU king during that period because it came with an asterisk; it never definitively beat the 680 outright and still it loses stock for stock in games.

7970 owners are only fooling themselves with thier performance claims...these results are devious and don't represent the stuttery gameplay or frametime results. Simply put, they are not as good as GTX's and never were.

They are bang for your buck GPU's and that has not changed.

A GTX is the best GPU money can buy and offers the smoothest, most feature rich gaming experience money can buy. Period.

Thats a very biased response, your basic logic for the GTX cards being better than AMD cards is as such

"We beat them in one benchmark, therefor we are better than them"

The HD 7970 GHz edition is a better bang for buck card because the power differences is noticeable and as drivers get better, the card has only gotten better. The same can be said for the GTX series of cards, but AMD has been making huge strides (Including now beating them on BF3) to make their cards perform better.

I personally have a Desktop with dual HD 6990's and they perform very nicely. I would have bought the GTX 590 but I could not get one at the time and I got these for a good price. My laptop has a gtx 675m card and I love it as well. However I would not go off on a tangent and flat say I prefer either in one direction totally.

On the note of the GTX cards having the smoothest gaming experience, I dissagree, the minor difference in the frame deliverance test does not show a huge difference in most cases which im sure many will be fixed with updates. The GTX 690 has been out for 6 months roughly and your comparing it for better performance in games that its made updates to run and improvements on for that 6 month period. Lets wait and see what happens in a few months after its release and see where the performance stands (Though even on a pre-release its performing in most thigns better FPS already).

Im not trying to start a fan-boy war or anything, but comments like that annoy me because its logic is just "1 thing is better so its better". Overall, I see the HD 7990 as better and will only get better with time as usual after reading the benches shown here. Now with time, we will see what happens or whose cards go where. If your going to make a judgement like that, you better have more fact to back it up instead of 1 thing that a card excels at over the like 5 other things cards do.

Thats my 2 Cents

VitalyT VitalyT said:

I personally stopped buying NVidia graphical cards since 2009 when my then $600 NVidia 280 GTX after 1 year of working perfectly gave up a stinky smoke and left me without desktop for 3 weeks. I never tried to overclock it or change any parameters in it, it just died during normal desktop work, emitting toxic smell all over the room. When I put it apart I saw one of the chips on the card (not the GPU) caught fire for reasons unknown.

Load all comments...

Add New Comment

TechSpot Members
Login or sign up for free,
it takes about 30 seconds.
You may also...
Get complete access to the TechSpot community. Join thousands of technology enthusiasts that contribute and share knowledge in our forum. Get a private inbox, upload your own photo gallery and more.