The Best Graphics Cards: Full AMD and Nvidia GPU Comparison with Latest Drivers

I'm sorry but that is just not true. They make GHZ edition cards for 7970 that are overclocked... just like they make overclocked cards for GeForce.

I'm not talking about boost clocks. They should grade the non reference cards from GTX such as FTW or MSI, ASUS, ETC Variant.

Your statement is just not fact. I am sorry.
 
7950 boost is a reference card....not individually released SOC card
for eg---gtx 680 has 1006Base Clock (MHz) and 1058 boost clock
Similar case for 7950 boost 800 base clock to 925 mhz boost clock
it's called 7950 boost as amd later fixed it with boost clock by a bios update...
 
I've become so hopelessly tired of seeing people do this. Please read this entire post and analyze yourself... see if you fall into this category and try to figure out why - and change.

Many people do research on something for one day to a few weeks prior to making a purchase of a new product (in this case GPUs.) They make their purchase (hypothetically) at the beginning of the year and hope that it's the best bang for the buck or overall performance well into the year and possibly even in to next year. Once that investment has been made they will defend that item (or defend the purchase?) anytime a differing opinion comes up.

When the 680 came out it was faster than anything else out... but things have evolved a bit. ATI has done a wonderful job with drivers (which is really out of character for them) and they even released a new reference set of cards... the Boost cards. The 7990 does not have a reference model so it's hard to include it in benches without making everything else under the sun an option.

Now, keeping all of that in mind you need to pay attention to what games you play, how each performs on each chipset, your resolution and IQ settings and number of displays, if you're using SLI or Crossfire, your refresh rate, limitations of your motherboard or CPU bottlenecking, driver versions and updates, and several other things to figure out what kind of results you might see. Just because the card you and I bought last month is no longer the fastest thing on the block is no reason to resist progress. ATI turned things around and I'll be the first to admit that they're king of the hill right now. I'd still buy 680s today simply because I really like the difference PhysX makes, but I'll tip my hat and admit that a reference 7970 Ghz is faster than the 680 in most titles that matter... today. Drivers could change this - or even a 680 Boost Edition updated reference, but don't be a hater because something else is faster now. This is an ever evolving area of computing and you guys have got to embrace the change and give credit where credit is due. I've made some bad purchases in the past and several with little foresight - we just need to learn from those or accept that we'd do it again and not argue the contrary.

Some of you act as if that purchase you made is a direct investment into the company's welfare. You're one person who bought 1 to 4 cards from them - they don't know you nor do they care if you defend them from everyone else. As long as people don't put out garbage there's a good chance they'll do well and be back for the next series of cards... with or without you.

Oh yeah, one more thing. Steve is one person that none of you pay to do these reviews. You should really be happy with what he's giving you for free and glean what you can from these reviews. If you're serious about this stuff you can make some of the simple deductions that are asked, yourself... such as the occasional "I need 690 numbers but you only have SLI 680 results" or the like. Many of you guys also assume that every piece of hardware that has ever crossed reviewer hands stays there forever. Many devices have to be sent back after review and then if it is to be included in future numbers a new one has to be PURCHASED. Just out of curiosity, how much money have YOU given to TS for these purposes?
</rant>
 
Great article!! So many games tested with a wide range of videocards and on the latest drivers. This is a good summary of where the current generation stands against the previous generation as both companies have had time to improve their drivers. The only thing I would have done is replaced some of those older titles like Aliens vs. Predator and Dirt 3 with newer games such as Hitman Absolution, Assassin's Creed 3 or Dirt Showdown. Also, the article mysteriously omitted the existence of some game bundles that sweeten the deal depending on what games people are looking to buy.

AMD HD7950/7970/7970Ghz cards come with free copies of Far Cry 3, Hitman, Sleeping Dogs and a 20% discount on Medal of Honor: Warfighter. NV GTX670/680 cards often come with Borderlands 2 and Assassin's Creed 3 coupons. Lower end GPUs such as 7850/7870 get Far Cry 3, while other NV cards get Assassin's Creed 3. Something to consider for prospective buyers.

Regarding the complaints that non-reference clocked GTX600 cards were not included, this article didn't use non-reference clocked AMD cards either. Both the 7950 Boost (V2) and 7970 Ghz are official SKUs. If Nvidia released a GTX680 with 1267mhz GPU boost, then could be included.

This doesn't change the conclusion for enthusiasts though as the fastest single-GPU card this generation, the Asus Matrix HD7970 Platinum, is still faster than any after-market GTX680. Also, for gaming above 1920x1200, GTX680 is much slower.

http://www.youtube.com/watch?v=sJaoY0-kfk8

http://www.hardwarecanucks.com/foru...70-3gb-matrix-platinum-edition-review-20.html

GTX690 is a more elegant solution than the power hungry HD7990. Although I feel not many people who follow GPUs would be likely to spend $1,000 on these GPUs given that it's winding down of this generation and HD8000/GTX700 are around the corner.

Glad to see both companies have been working hard to provide prospective gamers more value via price cuts, driver updates and game bundles.
 
I can appreciate this article to a degree.. My problem is I own a 670 FTW and I have benchmarcked better results then you are showing in some of the games you tested (BF3 being one of them) The fact that you show two versions of the 7970 , 9750 (7970 / 7970GHZ | 7950 7950 Boost) and only show a reference gtx680 / 670. makes me think you lean tword AMD anyway and this article is Bias.

No offence Kyle but if you are getting different results on your GTX 670 that means absolutely nothing. First of all where and how are you even testing Battlefield 3? Assuming you are testing a completely different part of the game which you likely are, are you then comparing your GTX 670 data with say a Radeon HD 7950? Because if you aren’t what are you using as a measuring stick?

You then go on to complain about the inclusion of the very real Boost cards from AMD and to top it off suggest that we are bias for only showing the reference GTX 680 and GTX 670 cards, like there is an alternative. As others have clearly pointed out the article only includes official spec graphics cards operating at their official specifications.

I'm sorry but that is just not true. They make GHZ edition cards for 7970 that are overclocked... just like they make overclocked cards for GeForce.

I'm not talking about boost clocks. They should grade the non reference cards from GTX such as FTW or MSI, ASUS, ETC Variant.

Your statement is just not fact. I am sorry.

I’m not sorry, you are just wrong. Call the 7970 GHz Edition an overclocked card all you will (in fact it is) but it’s also an official model in the 7900 series. You can buy GHz Edition cards and they cost $50 more than the standard 7970. There is no spec for an overclocked GTX 680 or GTX 670, there are just factory overclocked models and comparing factory overclocked graphics cards was not the goal of this article.

Furthermore if you bothered to even read the article you would have found that we didn't recommend the 7970 GHz Edition as it is simply an overclocked graphics card that AMD are asking a premium for. It was the standard 7970 that we felt was better value than the GeForce GTX 680.

Great article!! So many games tested with a wide range of videocards and on the latest drivers. This is a good summary of where the current generation stands against the previous generation as both companies have had time to improve their drivers. The only thing I would have done is replaced some of those older titles like Aliens vs. Predator and Dirt 3 with newer games such as Hitman Absolution, Assassin's Creed 3 or Dirt Showdown. Also, the article mysteriously omitted the existence of some game bundles that sweeten the deal depending on what games people are looking to buy.

Mate I hear ya but Assassin's Creed 3 and Hitman Absolution were released after much of our testing was done. That said both games are being added to our data and an in-depth performance article based on Hitman Absolution will be released in the coming days. We did include Dirt Showdown for a little while but it was so heavily in favor of AMD that we felt it skewed the results too much.
 
It is funny how a GPU that was re-released and then called "The GHZ Edition" is somehow just a standard 7970, like the original 7970 doesn't exist. Regardless of why I still get some humor there.
As far as the results, review is top notch and looks nice.
Setting PhysX on low in Borderlands 2 is a big opportunity wasted to say the least, critical mistake there.

The big 4 seem to be in the same range of performance, with no bandwitch advantage at 1600p or lower for the most part. Money is no object on my builds now (not always this way :D) so as nice as it is to brag about a couple frames for the price and maybe a GPU a tad faster, I'll take Nvidia's drivers, folding ability, TXAA, PhysX and less issues (from my experience) easily.
 
It is funny how a GPU that was re-released and then called "The GHZ Edition" is somehow just a standard 7970, like the original 7970 doesn't exist. Regardless of why I still get some humor there.

The reason is simple, marketing.

Setting PhysX on low in Borderlands 2 is a big opportunity wasted to say the least, critical mistake there.

No it would be a critical mistake to test with PhysX set to anything other than low. The AMD cards do not render the PhysX effects they offload them to the CPU so how would that make for a fair or accurate comparison?
 
"The AMD cards do not render the PhysX effects they offload them to the CPU so how would that make for a fair or accurate comparison?"

I don't give a donkey's balls how well the Radeons run PhsyX, or what needs to be done for them, or what their results are. I paid $400 for my GTX and I want the GTX PhsyX results. If the Radeons can't play, remove them from the chart entirely.
 
"The AMD cards do not render the PhysX effects they offload them to the CPU so how would that make for a fair or accurate comparison?"

I don't give a donkey's balls how well the Radeons run PhsyX, or what needs to be done for them, or what their results are. I paid $400 for my GTX and I want the GTX PhsyX results. If the Radeons can't play, remove them from the chart entirely.

That's intelligent and would be highly useful for the reader. Thanks
 
In Oz the 660 Ti and overclocked editions are priced in many online stores higher than HD7950/Boost/OC eds which makes the HD7950 OC and Boost cards great value considering the HD7950 Boost seems to be pretty comparable to a 670 in a lot of the benches you show.
 
What is with the hate, and moaning from some of you in this thread? Some of you whiners need to take your meds :-& ...It kinda reminds me of a COD vs BF3 forum... in other words, childish!

I think this article was a well done, thank you Steve! It's nice to see some performance results a few months after gpu release because, as users have stated, drivers mature. I'm really surprised, and happy that AMD is at the top of the charts. Not because I'm a fanboy, (my main rig has a 670 in it) but because competition is VITAL for us, the average consumer! I hope to see AMD kick it up in the cpu arena next architecture change, and give intel a reason to lower their prices, but, I wont be holding my breath :)
 
I know most people only use their GPU for gaming but this doesn't take into account CUDA cores and using them for rendering. This really changes the picture if you consider any sort of rendering (video editing, 3d Modeling, etc). Honestly it makes it hard to consider AMD without a major discount.

I really think you have no idea what you're talking about, do you? CUDA cores are the basic computing units NV chips are made of and yes they are used in all rendering operations if you're using HW acceleration. And yes all games use HW acceleration. 3D games and are among the best rendering benchmarks you can ever find. As far as video editing and 3D modeling is concerned, these types of apps, with a few exceptions, are hardly representative of what modern GPUs can do.
 
Great article. My wife got me a Asus Radeon HD 7970 for Christmas. Can't wait to install that bad boy and play BF3 on Ultra. But I gotta say, my current 5780 is an amazing card.
 
May want to take a second look at the Borderlands 2 settings. It's clear you've hit a framerate cap that the game set (vsync, running windowed/windowed fullscreen).
 
May want to take a second look at the Borderlands 2 settings. It's clear you've hit a framerate cap that the game set (vsync, running windowed/windowed fullscreen).

You would think so but if I look at the ground for example the frame rate hits a little over 80fps on most cards.
 
What I see,borderlands 2 not such a demanding game...at 1200p,even gtx 460 or 6850 is doing pretty well with physx off...with physx on you may need a gtx 560 to get 40-50+ fps .....or a good cpu(core 2 quad or i3 2100) is enogh with 6870 to offload the physx rendering to the cpu and get 50+ fps avg .....my friend running bl2 with i3 2100+6870 on physx high and 45-50+ fps on most occasions...very good performance:D
 
amstech,

"It is funny how a GPU that was re-released and then called "The GHZ Edition" is somehow just a standard 7970, like the original 7970 doesn't exist. Regardless of why I still get some humor there."

By the same logic, all these are factory pre-overclocked versions with a premium and no one had a problem with them:

Radeon X850 XT PE vs. X850 XT (exact same) or X800XT PE vs. X800XT
GeForce 4 Ti 4600 vs. 4400 vs. 4200 (exact same GPU, different clocks)
GeForce 6800 Ultra Extreme (UE) vs. 6800 Ultra vs. 6800 GT (exact same GPU different clocks)

Again, if you don't like AMD's name of HD7970Ghz, ignore it and look at the price levels.

Newegg sells 1Ghz HD7970 with 3 free games for $360; that gives you GTX680 level of performance for way less $.
http://www.newegg.com/Product/Product.aspx?Item=N82E16814202008

Beyond that point, if you are spending $450, there are plenty of 1100mhz HD7970 cards and really cream of the crop Asus Matrix Platinum. No matter how you slice it, GTX680 lost the performance crown this generation. Don't be bitter that NV delivered a mid-range Kepler GK104 and priced it for $500 because that means next generation they might give us GK110 (or something way faster than GK104). You lose some, you win some. Like you said if you want PhysX natively in BL2, get an Nvidia card.

I think GTX680 is a great card but when people start defending a slower product that costs more and comes with less free games, you really have start to wonder about objectivity. That's not even touching performance with mods (Skyrim + ENB), 8xMSAA, or strictly looking at 2560x1440/1600P. And really, at high resolutions, GTX680 is not a contender but people spend extra for a GTX680 4GB. Marketing FTW.

http://www.legionhardware.com/articles_pages/gigabyte_geforce_gtx_680_4gb,4.html
 
I'm sorry but that is just not true. They make GHZ edition cards for 7970 that are overclocked... just like they make overclocked cards for GeForce.

I'm not talking about boost clocks. They should grade the non reference cards from GTX such as FTW or MSI, ASUS, ETC Variant.

Your statement is just not fact. I am sorry.

HD 7970 Ghz and HD 7950 boost are cards competing with GTX 680 and GTX 670. there are factory overclocked HD 7970 Ghz cards like ASUS HD 7970 Matrix Platinum(1100 Mhz), MSI HD 7970 Lightning Boost edition (1150 Mhz) and HIS HD 7970 X Turbo (1180 Mhz). so if you want a GTX 680 Lightning to be compared then bring a HIS HD 7970 X Turbo or HD 7970 Lightning boost edition.

http://www.techpowerup.com/reviews/HIS/HD_7970_X_Turbo/
http://www.hardwarecanucks.com/foru...-7970-3gb-matrix-platinum-edition-review.html

it won't change the equation. the HD 7970 Ghz is faster. Once you bring in voltage overclocking it will be even more in favour of HD 7970 Ghz and HD 7950 boost.



http://www.overclockers.co.uk/showproduct.php?prodid=GX-181-MS&groupid=701&catid=56&subcat=938
 
"The AMD cards do not render the PhysX effects they offload them to the CPU so how would that make for a fair or accurate comparison?"

I don't give a donkey's balls how well the Radeons run PhsyX, or what needs to be done for them, or what their results are. I paid $400 for my GTX and I want the GTX PhsyX results. If the Radeons can't play, remove them from the chart entirely.

its very clear to see you want to defend your purchase. no need for that. enjoy your GTX 670. its still a very capable card. and since you really like Nvidia's exclusive features its all the more better. why so serious . ha ha ha
 
Hd 7950 boost is the best price/perf card in my opinion .....much lower price than gtx 670 but going par or beating gtx 670 even in higher resolutions and demanding graphics games as hitman absolution,metro 2033,alan wake,alien vs. predator,moh war-fighter,skyrim,sleeping dogs,sniper elite v2 etc etc

.........7950 boost beats gtx 680 on demanding games as crisis warhead,dirt showdown,sniper elite v2,hitman absolution,sniper elite,warfighter,alien vs predator,sleeping dogs,etc etc

(considering the proven fact that metro 2033 and hitman absolution are the two most graphically demanding games out there until now...... )

also see this link-

http://www.legionhardware.com/artic...hz_edition_7950_iceq_xsup2_boost_clock,1.html

......amd 12.11 beta never settle driver is legendary for amd gcn cards and expect more as gcn is a ground up architecture unlike Kepler (which's refined Fermi) and code optimization is still in works for gcn cards.....

......and even the difference between 7970 and 7950 boost is very less(5-6 fps avg) and it even gives hard time to gtx 680 on specific occasions until now....


......even I am not taking into account the much greater overclocking potential of amd gcn than kepler cards


.....so,in this criterion 7950 boost is the best valued card in my opinion...(y)(y)(y)
 
Hd 7950 boost is the best price/perf card in my opinion (y)(y)(y)

absolutely without a doubt. the HD 7950 boost with power option maxed out will run at 925 mhz consistently without throttling and gain performance which would actually put it slightly ahead of stock GTX 670. HD 7950 is 3 - 6% slower than HD 7970 at the same clocks.

http://hexus.net/tech/reviews/graphics/34761-amd-hd-7950-vs-hd-7970-clocks/?page=3

its easy to see that a overclocked HD 7950 will give an overclocked GTX 680 a run for its money. In fact its clear that an average clocking HD 7950(1.15 Ghz) will match or exceed a GTX 680(1.3 Ghz). A golden HD 7950 (1.25+ Ghz) will beat even a voltage unlocked GTX 680 lightning at 1.35+ Ghz.
 
amd 12.11 beta never settle driver is legendary for amd gcn cards
Pity that owners had to wait a eleven months to reap the benefits. Who remembers that it took AMD some time to actually get a whql driver out for Tahiti, or that VCE -a key bullet point in Tahiti's press kit - was basically fubar'ed for five months. It would have been legendary if 12.1 was the driver in question.
and expect more as gcn is a ground up architecture unlike Kepler (which's refined Fermi) and code optimization is still in works for gcn cards.....
Not really. Closer to a gutted (of FP64 shader cores, ECC,sacrificing shader hot clock for higher shader count ) Fermi . Each manufacturer taking a leaf out of the others book. Nvidia going small die, perf/watt, perf/mm^2, primarily gaming orientated (traditional ATI/AMD strongholds), while AMD add compute at the expense of die size, perf/watt and perf/mm^2 (trademarks of Nvidia since G80)
 
Ihave some questions:

1. Was that min fps, average, max, what?
2. Was Bf3 benchmark made in MP? (cause its kinda useless if it wasnt)
3. Why dont you do a benchmark about which card is for which cpu? I mean, can someone put 680gtx on a Phenom II x4 and have a same gpu usage as someone with i5 3570k?
This cpu dependency with gpu manufactures is a living nightmare for any gamer so clearing it up would mean so much. I if you takke this tas, please do Mp games cause they are much more demanding than Sp.

PS
I read your review Bf3 cpu vs gpu performance, thats not enough, frames in Sp and MP cannot be compared
 
I game at 2560x1600. I have Crossfire 6870's but even if I disable Crossfire a single 6870 will run 99.9% of games maxed out at this res. With AA + AF as well.

Almost all PC games and game engines are made with consoles in mind, so they're made for ancient 7 year old hardware and will easily run on any decent graphics card. Which isn't a good thing as far as I'm concerned. Consoles are seriously holding back PC graphics, and you now get people that actually think games like Portal 2 look good because they're so used to crappy graphics that this low graphical standard has now become acceptable, when PC hardware could do so much better and isn't even beginning to be stressed with stuff like this.

I'm not just talking about graphics either, the same goes for gameplay physics which can massively benefit from more processing power in PC's.

Stopped reading the rest of your post when you said a single 6870 is enough for 2560x1600. Not even a 6970 can do that res without turning settings down and its the fastest single card in the 6 series generation.
 
Back