All ATI's current graphics cards tested
FOR MOST OF year 2007, Nvidia had a firm lead on the PC graphics card market in both performance and brand power. During the 8800GTX days, the Green Goblin's lines ruled both the high end-gaming and workstation performance charts, and the 55nm "bumped" dies fiasco was still far in the future. That hit in 2008, as we wqell know.
Meantime, DAAMIT has stolen the thunder. Its 4870 series of cards, especially the HD4870X2, took the performance lead from the Nvidia GeForce GTX280, while the lower end and midrange 4800 and 4600 series cards offer all the 3D and HD video support of the top cards on a lower budget - important in these crisis times. Nvidia needed to create a limited-run kludge or a masterpiece, depending on who you ask - the GTX295 - just to claim the performance crown here.
With a price difference of a factor of four between the highest and lowest end cards, how much does the extra investment buy you in terms of synthetic card 3D performance? Is there the optimal price/performance choice? And yes, how does the top entry from Nvidia, the chief competitor, perform?
Here we had a look at five AMD Radeon reference cards: HD4870X2, HD4870 and HD4850 at the high end, as well as HD4830 and HD4670 in the midrange, not to forget an Nvidia GeForce GTX280 card in the shape of Asus' top-end overclockable edition.
We ran the test on our favourite new platform, the Intel Core i7 965 Extreme with three channels of memory on an Asus P6T Deluxe board - using this cream-of-the-crop processor ensures that the graphics card performance is never slowed down by the CPU.
Cards on the table
With the Windows Vista 64-bit OS, right now the only choice besides 64-bit WinXP to make full use of 64-bit desktop under the Microsoft platform, we ran the 3Dmark Vantage test - it has dedicated CPU and GPU scores there. The CPU scores on GTX280 are affected - read: skewed - by the Nvidia PhysX support, so we focus on more objective (if you accept them as such) graphics scores in this comparison:
As you can see, for a cost spread from US$100 at the bottom to US$500 at the top, you do get roughly that performance proportion too - at least in the 3Dmark.
On the high side, the Radeon HD4870X2 wins the run hands down against the Nvidia GTX280. For a little more than half its price, the HD4870 offers exactly same proportional amount of performance. If you're on a budget, the HD4830 offers all the bells and whistles, including the half-a-gig memory size of the HD4870 at nearly half the price but without the excessive heat of the 4850 and 4870 reference designs.
And, if REALLY on a tight budget, the HD4670 will give you sufficient speed to play entry level 3-D games on up to HDTV resolution, plus with all the high definition video support of its bigger brethren - but definitely no Crysis and such there, at any resolutions. Don't even bother trying it.
In summary, the middle of this roundup may be the best price/performance choice, but for the discerning users with casings big enough to dispel the heat, the HD4870X2 is the graphics card champion for the year 2008. No, we don't count the GTX295 until there are more of them than just a handful for a few Newegg shelves. µ
Share this:
Business Intelligence Software
Customer Relationship Management (CRM)
Enterprise Accounting Software
Enterprise Resource Planning (ERP)
Enterprise Systems Management
Inquirer is a biased media. The editors are more like ATI's salesmen than 3rd party professionals.
They made up all kinds of stories to scare people away from nVidia, and persuade them to buy a piece of ATI card.
Good job! AMD/ATI must pay you well.
how can the inquirer claim skewed figures for nvidia, but then compare only a non reference GTX280 against reference ATI cards - while persistantly claiming complete ATI dominance? sounds like one of the most retarded tests ever on this site.
physx is an advantage for nvidia, and not an unfair one, merely consumer incentive! GTX260 216 of course doesnt get a mentiion or a test, because clearly its now a match for 4870 1GB at the same price, AS WELL as having the physx advantage.
finally the inquirer attempts neutrality on its GTX295 stance, then at the end of the article blatantly sides with the view its a limited run 'kludge'! what is it with some people? if the inquirer werent so anti nvidia, they would acknowledge the GTX295 as a serious performance contender, regardless of its probably limited run. after all, its a top end card, you dont sell millions of them
First off those scores are not a true gaming representation. Lets see frames per second in real games like Toms Hardware shows. The 8800GT beat the HD4870X2 at 1920 x 1200 in COD4.
http://www.tomshardware.com/charts/gaming-graphics-charts-q3-2008/Call-of-Duty-4-v1-6,743.html
The GTX 280 single card was top dog over 3 way SLI and crushed the the HD 4870X2 card in Crysis.
http://www.tomshardware.com/charts/gaming-graphics-charts-q3-2008/Crysis-v1-21,749.html
Second, ATI had to use two GPUs (HD4870X2) to try and beat the single GPU GTX 280 juggernaught and not to forget twice the power consumption.
Happy New Year
The 8800 GT is in a SLI configeration that eat out the ATI card mentioned.
This is some good propaganda. What kind of kickbacks is the Inq getting from big red these days?
People get a grip... Virtually every online review of graphics cards shows the ATI cards dominating the Nvidia cards. That's reality.
If you like Nvidia buy their products. If you like ATI buy their products. Whining doesn't change the fact that ATI has delivered the dominant graphics solution. Maybe Q1 Nvidia will deliver a better product, who knows or even cares? You people need to get a life.
Tomshardware? LOL, those people keep the same old results from old benchmarks made months ago!! A real webpage like Anandtech showed that the HD 4870 1GB is overall as fast or slighly faster than the GTX 260 216 which sometimes is dangerously close to the GTX 280 in performance which is the fastest single GPU card on the market, period. What I see is that the GTX 280 is a bit overpriced and for the small performance difference, most people will buy the GTX 260 216 or the HD 4870. PhysX is not a selling point, most games using it shows no change in gameplay and doesn't use the nVidia GPU acceleration, thus it runs in the CPU, at least DX10.1 gives you a bit more eye candy and performance, the main selling point is the usable things today (Like Video acceleration and quality) and a very competitive price and features, thanks to ATi, we don't have to pay $650.00 for a decent card like it happened before with the 8800Ultra and the GTX 280. PS; Vantage is the worst way to show the power of the videocard, specially when vendors optimize the driver like crazy to increase the score and use it as a selling point...
Look up invidious - then you Nvidea fanboys will be xposed for what u r - a bunch of chip rot* loving paid pumpois.
*See numerous Charlie expose' articles on Nvidious' coverup of massive chip rot prollems.
regardless of whether people think or feel the inquirer is biased. This is an utterly shocking article. the testing is hardly indepth. the synth app used is useless and the settings and comparison products are so in favour of ATI that its meaningless.
Stick to news guys, everytime I see a hardware "compare" or some such article on the inq I cringe before I open it.
I like to see GPU test comparing with cards I've currently got, not what the latest is.
I've no idea what these scores really mean as I've no idea what my own card would have got. So how can I truly compare?
Sure the new cards are no doubt faster than my own. But by how much? Is it worth me spending cash on getting one of these new cards?
I use:
http://www.techarp.com/showarticle.aspx?artno=88&pgno=0
It allows me to see my own card and then I can compare it with the latest and greatest.
wow you must be a dumb retard. PHYSX wasnt meant to change your gameplay. its more on enhancing the visual quality of the game experience. its an added extra. extra eyecandy at the cost of performance. its like DX10, completely unnecessary, you'll realize that you're playing with dx10 is when it gives you the feeling that the game you're playing is slower than dx9 mode. eats performance but adds little to the gameplay experience.
i own a 4850 btw.
The scores look ok, nothing surprising there. I had a ASUS 280GTX and it was bad out of the box, Tried 3 more same thing. You cannot sell broken hardware. The shop suggested what I now run, two ASUS 4850 1Gig in xfire(790core/1200Mem and that is auto tune), with a ASUS 9600GT 512MB($45 used) and it seems to be working just fine for PhysX.
If you really want to piss off the Nvidiots you really should do a folding@home comparison. Then you will see some " WOW HOLY CHIT" scores on at least one brand.
Although I own ATI for graphics I still believe for "PURE GAMING" Nvidia is the fastest if you can get one that works.
PMPL @ the guy who suggested DX10.1 is currently more useful than GPU physx acceleration. name three high profile games that use DX10.1? ATI failed miserably the get developers to incorporate 10.1 in their games, there are NO major game titles that use the API. it weas even removed from assassins creed.......whereas like everyone else, i have seen the benefit of physx on a whole host of top name titles now, including some up and coming major games such as cryostasis and mirrors edge. cryostasis in particular looks incredibly demanding - physx acceleration in the vista demo shows how useful it is - ALREADY. trust ATI to take ages to sort anything out, 10.1 has been on their cards for over a year and theres STILL no major titles and thus incentives for it, and several MAJOR pc games have ran crap on ATI hardware. how come it took ATI over a YEAR to get crysis running properly on its hardware? how come GTA4 is still pathetically slow on ATI hardware, a 512 8800GT can easily beat out a 1GB 4870!!!
W3resmycar? Are you fucking retarded asshole? I didn't disrespect you brat, so STFU and eat your cheeseburger. Cryostasis looks fine, but look how pathettic the water moves in the table, looks like Hectoplasm from the Ghostbuster cartoons, dense, whitish and ridiculous. Tell me which games utilizes GPU PhysiX besides UT 3? None, you have to code for that, is not like the old games that used AGEIA PhysX automatically run on nVidia GPU's. So do us a favor and jump over that bridge with a fork in your mouth and rid us of your fanatism. For Plasticman; DX10 ran slower because when it was used, developers used more eye candy, hence slower, common sense eh? The main advantage of DX10 is that all those effects currently used cannot be done in DX9 without a huge performance hit, so if DX10 is slower, running the exactly effects will be a pretty slide show in DX9 because of the overhead. Far Cry 2 benefit greatly of DX10.1 anti aliasing performance, that also benefits the nVidia hardware through the use of the NVAPI for feature query, which once nVidia removes during time like they did with the GeForce 7 hardware, will runs slower. ATi cards ages better, simply as that, try to compare the X800XT vs GeForce 6, GeForce 7 vs X1900 and you'll see that newer games runs much faster in ATi hardware than on nVidia hardware, an advantage that wasn't there before. DX10.1 is a MICROSOFT feature, not ATi feature, Crysis always ran fine on ATi hardware, the issues were in Crossfire, and please, show us a review were the crappy console port GTA4 runs faster on your pathettic 8800GT than on the much more powerful HD 4870, stupid. Number the Game of the Year and triple A games which uses PhysX? UT3? LOL, Ghost Recon? LOL, you are funny. Talk about real and good games like Bioshock, Gears of War, Call of Duty 4 which runs great on ATi hardware. And about performance, a game from WIMTBP games like Far Cry 2 which isn't the best game of the year, yet run faster on ATi hardware than nVidia hardware? LOL, go back and play Wii kid.
multisampled depth buffer read. PMPL i was waiting for you to say far cry 2, but as we all know, nvidia's cards implement the useful DX10.1 features and they run on nvidia DX10 hardware nin far cry 2 - even though it isnt actually 10.1 compliant!! bwahaha. proves my point- ATI have no actual advantage with DX10.1 hardware on any major game. whereas hardware physx is already important in several games. crysis ran crap on any ATI hardware for the past year. it was NOT just in crossfire, every test showed how poorly 38/48 was in crysis, right up until a month ago and the latest drivers. its taken over a YEAR to get crysis working properly! thats a fact, dont spin your crap here. i can post tests so dont argue with me. GTA4? my pleasure. hows three tests?
http://www.pcgameshardware.com/aid,670889/Reviews/GTA_4_PC-_Benchmark_with_7_Ati_and_Nvidia_graphics_cards/?page=2
http://enthusiast.hardocp.com/article.html?art=MTU5Niw1LCxoZW50aHVzaWFzdA==
http://enthusiast.hardocp.com/article.html?art=MTU5Niw1LCxoZW50aHVzaWFzdA==
in all three tests, nvidia clearly have the edge, with even last generation cards like the 9800GT (its just a 8800GT in disguise) beating out the 1GB 4870. even though GTA4 is supposed to like more memory
now please be quiet, you clearly dont have a clue
Just tried GTX295 on Google Shopping and it says the product does not exist, so reasonable to leave it off.
Lining up all of the ATI cards with the fastest NV card available is 'exactly what it says on the tin' - an ATI graphics round up.
Google link for checking... http://www.google.co.uk/products?hl=en&q=gtx295&um=1&ie=UTF-8&sa=N&tab=wf
plasticman, stop being a jerk, HardOCP is an nVidia sponsored webpage which is so nVidia biased. GTA4 sucks, is a very badly ported console game and the performance difference isn't that great, I expected more reviews from great websites like anandtech, but GTA4 doesn't worth the time, and neither you. ati hardware usually shines in more demanding games which can put a strain on their superscalar shaders, so stop spreading your crap here, you rotten chip lover hahaha, put some 8x FSAA levels and see who's gonna sink first, drashek
dont be a moron borderline. unix asked for tests, because i produced three tests that prove my point while unix denies it? hardOCP follows the trend of the other tests, its therefore going to be pretty accurate. who is talking crap there then? im very happy to have proven unix wrong.
http://en.hardspell.com/doc/showcont.asp?news_id=4632&pageid=3550
generalisation like ATI hardware shines in more demanding games.....what a load of crap. generalisations stink, i hate it when people think that somehow comments like that are valid. utter nonsense. get a grip on reality boyo
for plasticman, DX10.1 is far more useful than you think in the long term. Games like Far Cry 2 uses the NVAPI which is a driver feature query, so it can use the multi sampling read buffer in the game, in the long term, with driver improvements toward other hardware, it can be removed to reduce the bloatware and unnecesary codes, that would mean that the feature wouldn't work, and were not talking about years here. Look at the 9800GX2, with newer games, is running slow, and is not getting updates at all, sure current drivers will work for it, but what about fixes and performance improvements? none!! nVidia loves to improve drivers in newer hardware, and loves to abandon the previous king of the hill, I wouldn't personally invest on something that wouldn't be supported in a driver level in less than 2 years, 79x0GX2 cough cough, ati in some releases in 2008 fixed an issue with the X1950 series of card!! It may take long, but at least is done, better later than never
problem is carlos, where are all the DX10.1 games? seriously? ATI have had this in their cards for over a year, and still.......nothing. by the time DX10.1 becomes remotely useful, it will be OBSOLETE along with the hardware. you know it, i know it. windows 7 and DX11 is going to be available this year, tesselation, SM5.0, and GPGPU support - all the arguments about 10.1 will go up in smoke because the hardware will be retired in favour of the next generation DX11 cards, which we know nvidia are ready to support!
you point to the 9800GX2, but what about 3870X2???? dont point the finger just at nvidia. its the same, it runs like a turd. whats to say 4870 X2 and say GTX295 wont be the same? nothing. these cards have finite lifespans. nvidia know that pc gamers dont really keep hold of their hardware for more than a couple years, which is why ATI scrambling about with 10.1 is as good as pointless. it'll be history inside a year. whereas physx implementation will simply grow and grow.
I have two 4830's in xfire, they are excellent. Great performance at 1920 x 1200 on any game. Will do for me for a while :)
"whereas physx implementation will simply grow and grow. "
PhysX will be a flash in the pan like Glide was.
The future will be an open standard like "Open CL" (ATI as well as Nvidia will support this), and I suspect the ATI cards will have an advantage in this particular area.
The Inq has really got to stop this kind of bias if you want to be taken seriously as reviewers.
I hate nVidia as much as the most informed tech people, as well as Apple, and whilst I agree with the fact that ATI's cards are generally more worthwhile this generation than nVidia's, your review does not show that.
Have you forgotten how you previously bemoaned the uselessness of apps like 3D Mark Vantage given their ineffectiveness of showing real world situations? In spite of this you managed to use a crappy benchmark to show that ATI does better in this single situation on a cost basis. Bravo. You've brought nothing new to the table, so why even bother?
I highly recommend your New Year's resolution is to learn to make proper graphics and hardware reviews like HardOCP do. You need multiple reviewers and multiple subjective tests on cards: there is no magic "one benchmark fits all" app.
Feel free to expose all the dodgy goings-on you wish with companies, but don't focus on just nVidia and Apple. I want to know if AMD/ATI et al mess up too so I -- as a consumer -- can benefit from your hard work.
The INQ dismisses 3DMark as a tool not meant to reflect real-world gaming performance of a card... There have been several articles that the INQ has posted dismissing all releases of 3DMark...
Then why the hell the same guys of INQ now use 3DMark for this test? Is it because it's the only test that DAAMIT has an advantage?
Oh well... Too much for unbiased reporting.
Open CL is trademarked by apple. its first uses are in macs. nvidia and ATI will indeed support the standard, but then you honestly think microsoft's continual evolution of direct X and nvidia's ongoing implementation of CUDA coupled with physx will allow Open CL to completely corner the market? as an apple owned trademark? LOL. i think not. it will be business as usual for a good few years, not least because pc gaming is forced into a rut by the consoles until their generation ends. CPU accelerated physics will probably die, but nvidia's GPU acceleration is set to propogate. we are on the verge of a GPGPU revolution FFS, and nvidia have positioned themselves accordingly
My apologies plasticman if I disagree with your rosy view of CUDA.
With only 1 (Nvidia) supporting CUDA, but all 3 (Nvidia, ATI, Intel) supporting Open CL, it becomes very hard for me to believe that CUDA has a rosy future and that Open CL is a lost cause.
What future game programmer will support CUDA in their game and at the same time support Open CL, when Open CL by itself will support all the new generation video cards from all manufacturers. It does not make economic sense.
I agree that CUDA is a nice feature for a very select number of Nvidia cards, but its future looks bleak from what I can see.
i agree with plasticman, open Cl being owned in name by apple and using it on pc? if anyone thinks microsoft in particular would alow Open Cl to corner the market without mirosoft extending their direct X API like they said they will are a bit shortsighted. these open sets are a wonderful idea, but are sadly dying out. Open GL as a graphics API is as good as dead, even poor old ID software recognising it because it dated so badly and ended up such a mess. Open CL would only be the same, someone like microsoft with somethin to lose would step in and offer their solution, Open Cl hasnt a hope in suceeding
I dont get what you guys are up in arms about, but the did you read the title? The author is testing how ATI's graphic card stack up against each other. That means he does not need to put in a bunch of benchmark, heck he can even put only one, its enough to demonstrate how much performance you get for one card relative to another. He threw in the nvidia card, mind you the top one in the line as a point of reference not to prove how bad Nvidia is. If you want a long drawn out review go to anandtech or xbitlabs. Use some common sense, you all buy specialize computer parts for a specific job (gpu, cpu...ect) then why not go to site that devoted especially to reviews and quench your anemic frenzy. Read the site's heading it does say news, reviews, fact, and fiction. So if you dont like it, dont come, you are not adding anything positive to the community by opening your mouth. You are in fact adding to the author's credibility and drawing attention to his article; in effect you are drawing loyalist to defend the author just by flaming him. What Charlie or the inquirer did was provided a valuable service to the community whether this is debateable is up to anyone. However I do admit that sometime Charlie goes overboard when he bags on Nvidia but that doesnt mean you ignore the facts that he presents. In the end the inquirer is only trying to help the consumer, so take his arguments, develope your own and make a decision. You are accusing the author for trying to degrade a brand yet you are doing the same thing, its like the kettle calling the pot black, and two wrong doesnt make a right so why hate?
How did he show the Nvidia card is sooooooooooooooooo bad? I don't see full listed prices, and I'm under the impression that it is less than the 4870x2 and more than the 4870. Which is exactly how it scored here.
I see no problems technically with this article. 4870x2 is fastest "card" and the middle of the pack cards provide best value (even if he didn't show his math..... which he should if he wants to be a real reviewer)
----correction.. The title should be "some ATI cards tested". I was totally ready to laugh at the 4350's scores etc. But he clearly did not test ALL current ATI cards.
-Bounty