ATI Radeon HD 4890: Playing To Win Or Played Again?

[americanbrian]

So in the GTX 275 review you mention the need to keep in mind that it is using beta drivers.

Does the same apply here? or does the 4890 run on an official release of Catalyst Drivers?

 

[cangelini]

This applies here as well (the betas are listed on the system config page). Fortunately, ATI's beta drivers behaved better than Nvidia's and didn't cause any erratic behavior that'd justify a warning.

 

[arthemos]

What kind of a "review" is this?
do you realy think that all agmers are that stupid to belieb these numbers Mr. Chris?

Coming from anandtech and xbitlabs, which are sites with well know funder/s, I can tell you that this review has nothing genuine in it, and forgive me if im being rude Mr. Chris.

Compare the following numbers to yours and see for yourself what im talking about.


http://images.anandtech.com/graphs/radeonhd4890_040209033751/18715.png

http://images.anandtech.com/graphs/radeonhd4890_040209033751/18718.png

http://www.xbitlabs.com/images/video/radeon-hd4890/stalkercs.png

http://images.anandtech.com/graphs/radeonhd4890_040209033751/18721.png

http://images.anandtech.com/graphs/radeonhd4890_040209033751/18724.png

http://www.xbitlabs.com/images/video/radeon-hd4890/codww.png

http://www.xbitlabs.com/images/video/radeon-hd4890/crysiswh.png

 

[cangelini]

Accidentally hit the submit button before you had the chance to make a valid point?

 

[ManusDeus]

"Accidentally hit the submit button before you had the chance to make a valid point?"

Appears to be the case. My perspective is that chasing the edge of the performance curve is quick way to spend a lot of money while putting up with sub-par drivers. My personal plan is to buy mid- to mid-upper range on the first price drop and then SLI or CrossFIre another board or two when the price drops. BUt in the end it is "what ever floats your boat". It is well-known in vendor circles that you charge early adopters a premium over you plan for the mid-market. IF you have the cash and no other priorities for it, why not. Me? I wait--I have other competing demands for my coin.

 

[rambo117]

great!! im glad THG added stalker CS to the benchmarks! just seeing crysis as the most taxing was getting a little boring.

wow, my 3870s keep lookin older and older....

 

[cangelini]

Absolutely Manus; I agree completely. That's why, in both cases, I decided that I would *personally* rather exploit the value of a 4870 or GTX 260 for significantly less. If I were instead looking at the GTX 285, the GTX 275 might make a great play on near-similar performance at a substantial price break.

 

[randomizer]

[citation][nom]arthemos[/nom]What kind of a "review" is this?do you realy think that all agmers are that stupid to belieb these numbers Mr. Chris?Coming from anandtech and xbitlabs, which are sites with well know funder/s, I can tell you that this review has nothing genuine in it, and forgive me if im being rude Mr. Chris.Compare the following numbers to yours and see for yourself what im talking about.[/citation]
You're assuming that both Anandtech and Xbitlabs are correct. You're also assuming they used the same settings and test bed as TH (whether you think you are or not). If they used a different software/hardware setup, your point is nullified because you're comparing apricots to bananas.

EDIT: FYI Chris, he had links in his comment which he formatted using URL tags incorrectly. Here they are again:


http://images.anandtech.com/graphs/radeonhd4890_040209033751/18715.png

http://images.anandtech.com/graphs/radeonhd4890_040209033751/18718.png

http://www.xbitlabs.com/images/video/radeon-hd4890/stalkercs.png

http://images.anandtech.com/graphs/radeonhd4890_040209033751/18721.png

http://images.anandtech.com/graphs/radeonhd4890_040209033751/18724.png

http://www.xbitlabs.com/images/video/radeon-hd4890/codww.png

http://www.xbitlabs.com/images/video/radeon-hd4890/crysiswh.png

 

[arthemos]

[citation][nom]cangelini[/nom]Accidentally hit the submit button before you had the chance to make a valid point?[/citation]

I linked to some tests from anandtech and xbitlabs, but for some reason the charts didn't appear, and i couldn't edit my post to readjust it.

 

[arthemos]

[citation][nom]randomizer[/nom]You're assuming that both Anandtech and Xbitlabs are correct. You're also assuming they used the same settings and test bed as TH (whether you think you are or not). If they used a different software/hardware setup, your point is nullified because you're comparing apricots to bananas.EDIT: FYI Chris, he had links in his comment which he formatted using URL tags incorrectly. Here they are again:http://images.anandtech.com/graphs [...] /18715.pnghttp://images.anandtech.com/graphs [...] /18718.pnghttp://www.xbitlabs.com/images/vid [...] lkercs.pnghttp://images.anandtech.com/graphs [...] /18721.pnghttp://images.anandtech.com/graphs [...] /18724.pnghttp://www.xbitlabs.com/images/vid [...] /codww.pnghttp://www.xbitlabs.com/images/vid [...] ysiswh.png[/citation]



Great job randomizer.appreciated.

These tests are all done on max settings at 1920*1200 res. with 4x/16x. aa/af.

 

[randomizer]

[citation][nom]arthemos[/nom]I linked to some tests from anandtech and xbitlabs, but for some reason the charts didn't appear, and i couldn't edit my post to readjust it.[/citation]
If you're going to use URL tags, you need to format them like this (without spaces in tags):

[ URL=www.blahblahblahblah.com]Some text to actually show up[ /url]

You missed the bolded part

 

[arthemos]

[citation][nom]randomizer[/nom]You're assuming that both Anandtech and Xbitlabs are correct. You're also assuming they used the same settings and test bed as TH (whether you think you are or not). If they used a different software/hardware setup, your point is nullified because you're comparing apricots to bananas.


Well, at first you must agree with me that there's a huge difference between this review and the two reviews i linked to.

The reviews were done on x58 platforms with overclcoekd Core I7s, and the latest drivers from both camps were used.Where did this variation is results come from?

Fortunaely for me, I have friends who have the same pc hardware interest as me.So i have had the chance to test the latest graphics cards from both ATI and Nvidia on my gigabyte x58 UD4 board equipped with a core I7 920 clocked at 3.6 GHz (200 * 18) and 6gigs of OCZ 1600MHz ram, by swapping them with my Asus 4870 1gig.

From my experience, i can tell you that a stock clocked radeon 4870 1gig is faster than a stock clocked gtx260+ and it's in the same performance level of a stock clocked gtx280.

In stalker clear sky, the 4870 does significantly better than the geforces even without the DX10.1 patch that Mr. Chris thinks that it should not be used to keep performance comparison consistent.


Most harwdare sites these days are just marketing tools for who pays.I just miss the old tomshardware and the great old guys whom i trusted and enjoyed viewing their articles.

 

[ManusDeus]

arthemos finally got his formatting correct for the HTML parser (no biggie dude--we've all been there at one time or another). My takeaway is that he demonstrates that gfx performance testing is a very, very complex pursuit: main boards have BIOS differences especially "tweakable" ones with voltage, clock, CAS, etc parameters, etc., etc. ad infinitum, games are tuned for boards with specific drivers, drivers evolve, GPU vendors focus on certain games at the expense of others, and so on and so on and so on. Not that my statement de-legitimizes his point Not at all--facts are facts--for those specific cases detailed cases. But I've seen variances that big between Gigabyte, ASUS, and MSI main boards alone. In the end, I think you may convince yourself to buy any of the top boards at any time and be able to prove it's the best with carefully selected benchmarks. You pays your money and you convince yourself you made the right choice and you are happy. But never forget that all vendors need to be treated well enough to keep them sending evaluation boards, their distributors and e-store partners to keep advertising, and them to keep providing data for articles. So a lot of this is manufactured "tempests in a teacup". You are the audience for the advertising and not the funding behind these sites. Remember that well lest you be taken in for a fool.

 

[randomizer]

[citation][nom]arthemos[/nom]Most harwdare sites these days are just marketing tools for who pays.[/citation]
Do you have any credible sources or are you just going to use results variation as your only evidence? Lots of people say what you just said, but nobody seems to have proof. Some benchmarks are definitely "helped along" by the manufacturers, like NVIDIA did with the GTX295 mandating certain games in the previews. But there was no pay, just rules.

 

[Ken168]

Not sure if this has been mentioned (didn't bother to read all the comments).

Even though the HD4890 is only slightly faster than HD4870 or Geforce 260, however, it appears that the review failed to point out that at 1920 resolution, the 4890 run as fast as both cards running at 1680 resolution (if not faster).

 

[cangelini]

[citation][nom]arthemos[/nom]Well, at first you must agree with me that there's a huge difference between this review and the two reviews i linked to.The reviews were done on x58 platforms with overclcoekd Core I7s, and the latest drivers from both camps were used.Where did this variation is results come from?Fortunaely for me, I have friends who have the same pc hardware interest as me.So i have had the chance to test the latest graphics cards from both ATI and Nvidia on my gigabyte x58 UD4 board equipped with a core I7 920 clocked at 3.6 GHz (200 * 18) and 6gigs of OCZ 1600MHz ram, by swapping them with my Asus 4870 1gig.From my experience, i can tell you that a stock clocked radeon 4870 1gig is faster than a stock clocked gtx260+ and it's in the same performance level of a stock clocked gtx280.In stalker clear sky, the 4870 does significantly better than the geforces even without the DX10.1 patch that Mr. Chris thinks that it should not be used to keep performance comparison consistent.Most harwdare sites these days are just marketing tools for who pays.I just miss the old tomshardware and the great old guys whom i trusted and enjoyed viewing their articles.[/citation]

arthemos,

While I appreciate the fact that you *have* feedback, I think it bears mentioning that while I provide configuration, settings, and driver versions for all of my data, you're trying to sell the crowd using "From my experience..."

I'm not sure if you actually read the piece and saw the full page dedicated to DirectX 10.1 results in Stalker, but I can assure you that all of the decisions made within the body of this piece were made to best establish equal footing between platforms. Those were broken out separately with the specific goal of demonstrating quality and performance benefits on ATI's architecture.

If you truly believe that reputable sites are mere marketing tools, then you'd seem to be wasting your time reading them. But I can assure you that myself and the team of writers I work with are passionate about this stuff. I was sitting shoulder to shoulder with Tom back when ATI's first Radeon launched as SharkyExtreme's hardware editor and that enthusiast spirit is still very much alive here.

 

[ManusDeus]

"Do you have any credible sources or are you just going to use results variation as your only evidence? "

You are stretching my words. I never said the vendors directly pay for benchmarks. It is far, far more subtle than that. Here is one example of how it works: If you diss a vendors latest products more than they like or more often than they like, your eval boards will come later--meaning other sites get their articles out before you--and you lose page view ad revenue. Another way: you diss vendors X's latest board and one of their main e-tailers wants to run a special ad campaign on it--teh vendor doesn't provide matching funds to the e-tailer so they don't place as big an ad on your site and give more ad revenue to your competitor. You'll learn quick only to slightly nip, not bite, the hand that feeds you.

No strong-arm tactics, no conspiracy--just a subtle reminder as to whose hand feeds you. And you'll wise up or your editor will bench you. Similar to how White House press conferences work--you ask uncomfortable questions, you don't get picked to ask a question next press conference--it's self -correcting if you wish to stay employed.

It's not a grand nefarious conspiracy--there is no illegal payola. It is just how the media and vendors work together--in every industry. You just need to know whose interests the articles really serve and learn to view all article with a skeptics eye. The Devil is indeed in the details.

 

[randomizer]

[citation][nom]ManusDeus[/nom]You are stretching my words. I never said the vendors directly pay for benchmarks.[/citation]
I was quoting arthemos

 

[ManusDeus]

[citation][nom]randomizer[/nom]I was quoting arthemos[/citation]

No worries. these are mostly little "while lies" anyway. The vendors largely produce good products and the differences are nuanced and depend on a lot of complex parameters: CPUs, main boards, BIOS & settings, memory timings, game tuning, and much, much more. You cannot blame vendors for ensuring they get the best press for their new products (they are in this to make profit) nor can you blame the media for keeping themselves employed and in the black (they are also in this to profit). And by in large the information is presented accurately--it is just so complex that its on the level of world peace--there are many sides to every story. You just need to watch things carefully and not get sucked into the hype.

 

[Guest]

Seems like I'd want to buy a 4890 single card only when playing Crysis, World of conflict, or Stalker at highest detail settings; since there the extra framerate could dictate if a game would run smoothly or not.
True raising the fps from 30 to 33 does not do much in playability; but any boost of fps that could topple 25fps could mean playable framerates!

Stalker seems very memory based. Looking at the performance difference as memory decreases is amazing! Clearly this game is made for 1GB of VRAM. We can see the same thing with GTA.
869MB of ram on the GTX260 card seems to also show a drop on highest settings.
The 4890 also seems to have a much better power consumption than the 4870 X2!

 

[Lans]

The idle power was disappointing.

Physically, the GPU is fractions of a millimeter larger due to additional capacitors that clean up power to the chip.

That would suggest there were attempts to optimize power but is not supported by the numbers in this review... Hopefully a retail board will show something better? Plus I am waiting for OC'ed numbers as well.

$220 does seem a lot more appealing to me although I generally just try to keep up with tech news and buy later.

Power envelop seems to allow for a 4890X2 (not sure about a 4890X2 OC version). > I'll just wait to see the next batch of 4890s.

 

[arthemos]

[citation][nom]cangelini[/nom]arthemos,While I appreciate the fact that you *have* feedback, I think it bears mentioning that while I provide configuration, settings, and driver versions for all of my data, you're trying to sell the crowd using "From my experience..." I'm not sure if you actually read the piece and saw the full page dedicated to DirectX 10.1 results in Stalker, but I can assure you that all of the decisions made within the body of this piece were made to best establish equal footing between platforms. Those were broken out separately with the specific goal of demonstrating quality and performance benefits on ATI's architecture.If you truly believe that reputable sites are mere marketing tools, then you'd seem to be wasting your time reading them. But I can assure you that myself and the team of writers I work with are passionate about this stuff. I was sitting shoulder to shoulder with Tom back when ATI's first Radeon launched as SharkyExtreme's hardware editor and that enthusiast spirit is still very much alive here.[/citation]


Thanks for your respond Mr. Chris.Tomshardware is a reputable, well know site.As a long time visitor of this site, im concerned about the reliability and the credibility of this site.Because what separated
and distinguished tomshardware along these years, is not only its professional articles and reviews, but also its unbiasedness and neutrality.

I would like to suggest a new graphics cards testing method that will make a lot of difference and transform gpu reviews into an exciting and far more informing experience.

Why don't we get video reviews instead of those dead charts?

You make a separate video that shows the benchmarking process of each gpu in a game.The video shall show the setup details and clocks on the monitor and in wild if possible, and it shall then show the benchmark running with the fps shown clearly during the test run.After that, you combine the clips of the tests of each particular game in one video, either by putting the clips running simultaneously beside each other or one following the other.......

By using this way, showing the performance level of a graphics card will be more detailed and alive., and it will kill any doubts related to the reliability of the review.

by using this new method, tomshardware will finally step the level up and separate itself vastly from the other so called hardware sites.

I hope you take my suggestion, Mr. Chris, into a deep consideration.And i hope to see this new method applied to tomshardware reviews as soon as possible.

Arthemos.

 

[randomizer]

You do realise that video would be about a week long? :lol:

 

[Onus]

Arthemos, it appears that what you're asking to see in a video is what they are describing and graphing in the article. Are you saying that THG might be liars, so without the video they cannot be trusted?
I understand your comments about not biting the hand that feeds you, but even subtle pressure on reviewers to distort facts will not prevent the ultimate truth from coming out. When it does, if it differs remarkably from reviews, then the reviewing site will take a credibility hit, and the vendor may lose business for dishonesty. An honest review, however, is one way of letting the market know how a new product should be priced, i.e. where on the totem pole it sits. As such, accurate reviewing may bruise a few egos, but doesn't do any real damage.

 

[ManusDeus]

" if it differs remarkably from reviews"

That's the crux of the whole issue, the areas of difference are rarely (actually never) marked because you are right, adding a little sizzle to the steak is expected--outright fabrication is a lose-lose proposition and manipulation isn't tolerated by the market.

As I said before, most vendors have good engineering teams, they produce quality products, and the devil is in the details--but he's pretty small and tuned for specific games. It's a competitive market and we are blessed to have so many choices with all the segmentation around price points with two great architectures.
The fact is that Tom's hardware and the other enthusiast sites won't stay in business if they seriously diss a major vendor in a market of two big players--the ad revenue fall off would be a killer. Fortunately, they don't have to make that choice because both are by in large straight-shooting firms.

Arthemos is just wanting more detail (a picture is worth a thousand words) but he won't get it. The level of detail provided currently is competitive and that's what this is all about. TH competes with other enthusiast sites--they'd got broke focusing on such detail because it would slow down review releases and their competitors would get the newest releases reviewed first.--meaning more page views boosting ad revenues quickly, allowing them to charge more for ad space/impression/etc. And so on. They cannot please us all. TH does a great job trying to give the majority of us what we need. I visit often and will continue to do so.

(You can tell I had time on my hands to pontificate this morning while some compiles finished ;-) take it with a grain of salt or whatever. But the Tom's hardware team does a good job staying objective while providing cutting edge testing and analysis. They will honestly state there is no good buy at a price point if they think there isn't. What more should we expect? Nothing in my book. Just be aware that they have to find the good in each vendor's product--fortunately it isn't hard to find most of the time. Caveat emptor.