AMD Radeon HD 7990: Eight Games And A Beastly Card For $1,000

[JJ1217]

[citation][nom]bartholomew[/nom]A Price tag of $849.99 would had been quite aggressive & increased its value significantly.[/citation]

Yep, with that horrible coil whine, and not very nice design (Like come on, that thing is uglier than Powercolor's Devil13), I don't know why AMD are pricing it at the same point as Nvidia. With the single GPU's, they seem to know that holding a larger market share (Selling more graphics cards are more important than making more money), works, however this dual GPU card selling for $1000 seems to signify to me that they've forgotten that.

 

[CaptainTom]

Yeah I gotta agree with others here, USE 1080p!!! Most people can not afford 1440p (And we all know 4k is gonna be here soon anyways).

 

[de5_Roy]

i call the testing methodology suspect. one of the 'gamers' is seen wearing his shades backwards in a photo. how can they point out the 7990's worse performance if they can't even wear their glasses where their eyes are supposed to be. they should be wearing their glassses right first. amd is clearly better here and toms is out to make them look bad.
in the second half of 2013, Months from this launch (could easily be december) and nearly 2 years after 7970 launch, amd will release better drivers for crossfire and 7990. that makes amd competitive. and the game bundles add so much value to the card.
edit: just like 'real gamers and enthusiasts' here pointed out, 1080p testing should have been performed. and 'real gamers and enthusiasts' don't care about 75w more power at stock. it doesn't even matter if boutique builders don't pick up this card and put them in their builds. oems that buy gfx cards in large numbers don't drive sales, 'real gamers and enthusiasts' do.

 

[alidan]

[citation][nom]donquad2001[/nom]this test was 99% useless to the average gamer,Test the card at 1900x1080 like most of us use to get a real ideal of what its like,only your unigine benchmarks helped the average gamer,who cares what any card can do at a resolution we cant use anyway?[/citation]
congrats, dumbest thing i read here in months outside of spam... you would hear me clapping if you were here.

people who spend 1000$ on a gpu, do not have a single monitor setup or a single 1920x1080
in fact, most people have a 2560x1600 monitor, a step up from what was used in the test to run games if they are dropping a k on a gpu

 

[sarinaide]

1) Thank the heavens you have reverted to the old school format, It was just an irritation before.

2) On what basis are we being told the GTX690 is better, clearly not performance and power consumption at $1000, no enthusiast will give the slightest iota about, most will be running watercooling aka fish tanks.

3) Accoustics is objective, not like any gamer will really notice that when they hook up BF3 to Home Theatre sound arrays or have $200 headsets, nor do I honestly believe people buy a graphics card to sit and listen to it, then there is PSU and CPU noise along with water cooling solutions so its really much of a muchness.

4) Prototype drivers tested at PCPer seemed to reach the conclusion that AMD's Catalyst Beta drivers achieved far more than expected, most were so close to Nvidia's own that it was said to be a success, and being prototype they will get better.

 

[JJ1217]

[citation][nom]sarinaide[/nom]1) Thank the heavens you have reverted to the old school format, It was just an irritation before.2) On what basis are we being told the GTX690 is better, clearly not performance and power consumption at $1000, no enthusiast will give the slightest iota about, most will be running watercooling aka fish tanks. 3) Accoustics is objective, not like any gamer will really notice that when they hook up BF3 to Home Theatre sound arrays or have $200 headsets, nor do I honestly believe people buy a graphics card to sit and listen to it, then there is PSU and CPU noise along with water cooling solutions so its really much of a muchness.4) Prototype drivers tested at PCPer seemed to reach the conclusion that AMD's Catalyst Beta drivers achieved far more than expected, most were so close to Nvidia's own that it was said to be a success, and being prototype they will get better.[/citation]

I actually disagree with that.

2) The 690 I believe is more balanced, and is more for the person who is inclined towards Nvidia rather than AMD. Not all people care too much about performance -- otherwise people would only be getting 7970's because its the current performance king (Not counting Titan of course)

3) Here you say Acoustics is objective, then you start going on about how people shouldn't care about it and how it doesn't matter. That is sort of going against what you're saying, don't you think?

 

[abbadon_34]

[citation][nom]cangelini[/nom]Power is the one thing I didn't have time for. We already know the 7990 is a 375 W card, while GTX 690 is a 300 W card, though. We also know AMD has Zero Core, which is going to shave off power at idle with one GPU shut off. I'm not expecting any surprises on power that those specs and technologies don't already insinuate.[/citation]

Amen. So tired of the focus on power usge. Who spends a grand on a video card to count pennies on their power bill ?

 

[hunshiki]

[citation][nom]cangelini[/nom]Power is the one thing I didn't have time for. We already know the 7990 is a 375 W card, while GTX 690 is a 300 W card, though. We also know AMD has Zero Core, which is going to shave off power at idle with one GPU shut off. I'm not expecting any surprises on power that those specs and technologies don't already insinuate.[/citation]

ZeroCore only works if there is no output. E.g.: The display is turned off.
Which I bet rarely ever would happen to a 7990 user.

 

[blubbey]

[citation][nom]donquad2001[/nom]this test was 99% useless to the average gamer,Test the card at 1900x1080 like most of us use to get a real ideal of what its like,only your unigine benchmarks helped the average gamer,who cares what any card can do at a resolution we cant use anyway?[/citation]

Not sure if moron......

 

[cangelini]

[citation][nom]hunshiki[/nom]ZeroCore only works if there is no output. E.g.: The display is turned off.Which I bet rarely ever would happen to a 7990 user.[/citation]
It wouldn't be any more frequent or infrequent than a single-GPU config. If you step away from your desk for 30 minutes and the display turns off...well, you get the idea. Of course, perhaps you were being sarcastic and suggesting that anyone with a 7990 would want to be in front of their monitor all day? ;-)

 

[sarinaide]

[citation][nom]JJ1217[/nom]I actually disagree with that.2) The 690 I believe is more balanced, and is more for the person who is inclined towards Nvidia rather than AMD. Not all people care too much about performance -- otherwise people would only be getting 7970's because its the current performance king (Not counting Titan of course)3) Here you say Acoustics is objective, then you start going on about how people shouldn't care about it and how it doesn't matter. That is sort of going against what you're saying, don't you think?[/citation]


Define "balanced", since nobody actually owns a HD7990 "Malta" yet its hard to base it on balance having no experience, and since its current generation past experiences don't count due to changed circumstances.

At a $1000 its priced in the enthusiast bracket and enthusiasts care about performance so yes performance is the fundamental basis for having this card, since we can rule out the average joe soap owning this, or anyone that is building a system for under 2K we can basically say this is not a part that will ever be used by the mainstream market. Again enthusiast level is all about performance and the 7990 beats the 690 in plenty metrics nothwithstanding its negitives of course.

I used accoustics for again this is not a mainstream or budget builders part and enthusiasts 9.5/10 run custom cooling solutions which are a) noisier and b) will likely have this card under cooling to. Since this is a performance card the heat and noice is acceptable for its performance offerings. For a gamer they will likely be to busy enjoying its performance rather than worry about how much noice it makes.

Accoustics, power and heat are more relevent to the budget market or SFF such as myself. Toms recently benched the 650ti boost and the 7790 and proclaimed a staggering victory to Nvidia, which I object to that, the 650ti boost used more power than the incumbant 7850 and thus more heat while the 7790 used the same power as the 7770, produced less heat and accoustic while coming within 10% of the 650ti and 7850 performance on again less power. For ITX gaming builds or budget builders with smaller PSU's these factors come into play.

And lastly there is a general reluctance to regard that this generation now pretty much belongs to AMD in the way the previous generation belonged to Nvidia. Back onto this part despite the power and accoustics is delivering the top performance expected, throw in 8 games to the mix and you are getting tremendous value on top, good on AMD for actually catering for the end user. Whatever our allegiences to brand, this is the alpha card on pure performance and no "feelings" go beyond metrics and the metrics back it up.

 

[weatherdude]

Wow, great article Angelini. You did a great job explaining frame time variance so that I can understand better now. Before, despite my science and math oriented brain, I was kinda running into mental roadblocks. I have a question about the FCAT analysis though. Why is the runt frame threshold 20 vertical lines?

The recordings of the fan coolers were done in an open test bed, yes? I'm not sure but would a solid case mitigate the problem of the coil whine? Although the test gamers noticed the noise and the systems were behind the desk so maybe not.

 

[Memnarchon]

Hmmm, it seems that the new prototype driver is on the right way at the practical FPS, but they still need a lot of work at the frame time variance.

 

[silverblue]

The 7990 is a very interesting prospect. On one hand, it has unmatched compute ability, more RAM than competing solutions and uses noticably less juice than unofficial 7990s or 7970 GEs in CrossFire. On the other, NVIDIA clearly offers a more stable gaming experience for less power usage, and the 7990 cannot go to an idle state as low as that of the 7970 GEs in CrossFire. If the bundle games didn't matter or wouldn't fetch a decent price, the 690 would be the better choice right now - the extra RAM and wider memory bus that the 7990 sports must be responsible for a good amount of that extra power, and there's always that feeling that the 690 is just made better. If you want something a little more future proof and far more capable at compute than the 690/Titan without quite the noise and power consumption of the 7970 GEs in CrossFire, the 7990 makes more sense.

Personally, I'd like a Titan.

 

[christop]

I would like to see how it stacks up when it is not running a prototype driver.

 

[spentshells]

Are the runt frames avoided with vsync?

 

[virtualban]

Nvidia has put a delay in frame display for quite a while, to deal with the issue frame stuttering of multi card configurations. I wish some future article will test and measure the delays in the graphics pipeline.

 

[flexxar]

[citation][nom]timw03878[/nom]Here's an idea. Take away the 8 games at 40 bucks a piece and deduct that from the insane 1000 price tag.[/citation]
AMD doesn't pay $40 per game. That's the retail value. They pay significantly less. Probably around $10. AMD wins because they then market it to you as a $40 value. The game companies win 2 fold. One, because they sell copies to people that wouldn't have otherwise bought their game. Two, they get to brag about the number of copies sold and therefore end up selling more copies. It's actually a really great model to do it this way and not possible to do as you suggested.

 

[ericjohn004]

Wow, another great article by Tom's Hardware. It sure seems like Nvidia has the higher quality parts here, or at least the drivers.

And for anyone who thinks this article is useless or has any complaints.. what's your problem? Are you such a fan of AMD graphics that you hate on an article just because you don't like facts? Either way, AMD seemed to fix some of these problems with the driver update, although still not as good as Nvidia.

 

[bystander]

I couldn't help but think that your definition of a runt frame is too small, which the time variance chart would back me up. Given 20 rows is only 2% of the screen, you are only ignoring the slivers. Pcper was counting a runt frame as a frame that is 20% the size of the average 10 frames prior. That seems a bit more practical, though I can understand wanting to refine it, but it does address two issues: As your FPS get higher, a normal frame is smaller, and I'd consider even 10% of the screen to be pretty pointless. It also kind of combines your frame variance idea (which I like a lot) with the practical FPS.

Kudos on the frame variance chart. That makes complete sense to me, and a very good idea.

 

[Adam Lovelace]

I'm very surprised that they took the time and energy to test eight compute benchmarks, but didn't test bitcoin mining, which is an important gpu application and has been an AMD selling point.

 

[jamie_1318]

Very good article, super thorough testing. Dropped frames appear to be a real game changer. Very impressed by your survey.

I have a few comments about the display of the information though. Your graphs are too busy, both bar graphs in the openCL section, and the line graphs for frame-times. For both graphs you need to shave off extra cards/data sets, and highlight the most relevant cards. In the line graphs, you need to choose some colors that will accentuate the most important cards. I'd assume for most of them you would want to keep GTX690, 7990 and 7990 experimental being the most obvious and the others should be desaturated. Also, most of the time the dotted lines being hardware FPS was nearly identical to practical FPS and should be removed from the graph. The cards that the hardware and practical FPS were significantly different would be a lot more obvious.

Once again very good article, a lot of new and very practical results here. Really like the direction you are going with GPU benchmarking.

 

[jamie_1318]

Darn it new forum format, can't find edit/"view in forum" option anywhere, should say GPU, not CPU.

Edit: found post in forum, still don't see "view comments in forum" on the news article.

 

[bystander]

[citation][nom]sarinaide[/nom]1) Thank the heavens you have reverted to the old school format, It was just an irritation before.2) On what basis are we being told the GTX690 is better, clearly not performance and power consumption at $1000, no enthusiast will give the slightest iota about, most will be running watercooling aka fish tanks. 3) Accoustics is objective, not like any gamer will really notice that when they hook up BF3 to Home Theatre sound arrays or have $200 headsets, nor do I honestly believe people buy a graphics card to sit and listen to it, then there is PSU and CPU noise along with water cooling solutions so its really much of a muchness.4) Prototype drivers tested at PCPer seemed to reach the conclusion that AMD's Catalyst Beta drivers achieved far more than expected, most were so close to Nvidia's own that it was said to be a success, and being prototype they will get better.[/citation]

2) The 690 had similar FPS, but far less frame variance, resulting in smoother frames.
3) How many gamers use headsets? It matters for those who care about it.
4) Did you read the Pcper article? It came to the same conclusion. The 7990 has improved drivers, but is still behind Nvidia quite a lot.

 

[bystander]

Meh... they should have spent their time making better drivers than releasing this card.

Not sure what they're thinking pricing this at $1k.

They are different teams, so I'm not sure it would have made a difference on drivers if they pulled the plug on the 7990. The good news is they are working on drivers. The prototype drivers are a definite improvement.