Print 58 comment(s) - last by ToeCutter.. on Mar 26 at 4:30 PM

ATI Radeon 4890 shows its face

DailyTech has been able to "acquire" an ATI Radeon HD 4890 video card from one of our sources. Although we don't disclose names in order to protect our sources, we are able to say that the card arrived from Taiwan.

The Radeon HD 4890 features the RV790 core, which we are able to confirm runs at a core clock of 850MHz. We have heard that it is capable of more, but that will be up to individual board partners. The chip is not just an overclocked RV770, as there are a few tweaks and modifications that we are still investigating.

The card also features 1GB of memory via eight GDDR5 chips from Qimonda, which declared insolvency in January. The reference card we have is set to 3900 MHz effective, and we expect most of ATI's board partners to do the same. This provides 124.8 GB/s of bandwidth.

Samsung could emerge as a source for GDDR5 chips if Qimonda is forced to halt production. It recently announced 50nm production of GDDR5 chips, which it claimed was capable of 7 Gb/s.

Pricing and launch dates are still being adjusted at this time. AMD recently tried to lower prices on its Radeon 4870 and 4850 cards, but its board partners believe that the performance of the Radeon 4870 is too good  to lower prices further.

The Radeon 4890 will probably be priced moderately above the 4870, but still far below NVIDIA's GTX 280. The GTX 280 should not be confused with the mobile GTX 280M part, which is actually a rebadged GTS 250/9800 GTX+.

There are new drivers coming out next week that will have increased performance for the Radeon 4890 cards, which will invalidate all current benchmarks that have leaked out

More details will be unveiled next week at the Game Developers Conference. DailyTech will be testing the card this weekend.

Comments     Threshold

This article is over a month old, voting and posting comments is disabled

RE: Please, please, please
By Jedi2155 on 3/19/2009 1:42:48 PM , Rating: 1
I think 3DMark still has its place and use in for comparing graphics cards. Game engines have too many quirks to indicate relative performance in games, and the 3dmark suite will allow a card to take reach is full potential whether that potential is used in future games or not but it does shed some light that would be applied to possibility a future game that might use it that you can't use the current set of available games to test on.

I think a well rounded benchmarking suite has all the bases covered, including 3dmark numbers but having only 3dmark numbers do not help very much.

RE: Please, please, please
By Rob94hawk on 3/19/2009 1:47:05 PM , Rating: 5
People don't buy vid cards to run 3dmark. It's more for "bragging rights."

RE: Please, please, please
By quiksilvr on 3/20/2009 4:12:54 AM , Rating: 5
Dude: "He baby, Josh is running an HD 4650...heh..I got me a 4890!"
Chick: "I have no personality! Show me your penis!"

Sigh...if only life was that simple...

RE: Please, please, please
By spread on 3/21/2009 12:52:38 PM , Rating: 4
It really is. Replace "4890" with "money", and "I have no personality" with "you sure there's no alcohol in this?"


RE: Please, please, please
By PhoenixKnight on 3/21/2009 1:13:48 PM , Rating: 3
Last year a friend of mine went to Fry's to pick something up and one of the salepeople was hitting on her and started bragging about his computer to her. Needless to say, she wasn't very impressed.

RE: Please, please, please
By DopeFishhh on 3/20/2009 11:34:38 AM , Rating: 2
Ideally 3dmark is meant to act like a common measuring stick for performance. If I bought a new card I'd like to see how it fared against the reviews I looked at, 3dmark seems like one of the better options especially if I didn't have any of the games in the review.

Where 3dmark (and others) fall down is that they don't give you measurements that are universal enough for people to be able to translate them into scores for other games.

Theoretically if they were, I could run a benchmark test and be able to calculate system performance in any game past, future or present with reasonable accuracy. Not only would that make benchmarks useful but they could easily be the only performance metric you need consult prior to purchase.

RE: Please, please, please
By Proteusza on 3/19/2009 1:48:41 PM , Rating: 4
Generally speaking synthetic benchmarks have the lowest validity of all benchmarks, since performance in synthetic benchmarks is only indicative of performance in the same synthetic benchmark.

Besides, you say game engines have their quirks, havent you realized that synthetic benchmarks such as 3DMark themselves have rendering engines which would themselves have quirks?

Future games using technology in synthetic benchies? As I said, synthetic benchies themselves have rendering engines, so it really doesnt matter, cos you could quite easily find a rendering engine more advanced than them (Crysis is a case in point).

Whenever I see synthetic benchmarks in a review I always sigh and skip them, since I dont know about you guys but I never play this "3D Mark" game.

RE: Please, please, please
By MadMan007 on 3/19/2009 2:16:18 PM , Rating: 5
Pretty much all of the arguments you listed can have either 3DMark or blank game inserted and still be valid, different games don't always have corresponding results either. 3DMark is just another benchmark like any game. It does provide some low-level information that games don't though like the specific shader tests and whatnot. It's sort of like HDTach for hard drives in that you an idea things are working right but doesn't always correspond to real-world program use.

RE: Please, please, please
By The0ne on 3/19/2009 3:18:41 PM , Rating: 2
Test apps just needs to be done better that's all. I wouldn't want to be testing hardware on games solely simply because of the unreliability of the game itself. I rather have a proper tool that allows you to perform your test consistently, effectively and accurately rather than relying on a game, mostly, aren't consistent in the content being tested (weather, effects, position, etc.) That's what tools are for.

And yes, you don't get to play on the tool *roll eyes* you do play the game.

RE: Please, please, please
By just4U on 3/20/2009 1:21:04 AM , Rating: 1
I always reference 3dmark scores as a gauge. It's designed to test your systems performance so it's relevant for what it's used for.

RE: Please, please, please
By VooDooAddict on 3/19/2009 2:35:22 PM , Rating: 5
I think the next step for synthetic benchmarks would be licensing game engine tech.

I'd be much happier with 3dmark if it's different "game" tests were snippets of:

Epic's UnrealEngine
ID'd Latest Engine
Valve's Latest Engine
One of the MMORPG Engines

RE: Please, please, please
By plonk420 on 3/19/2009 3:17:01 PM , Rating: 2
why not test those SEPARATELY, as well as the GPU/CPU-crushing FutureMark? Epic/ID, and especially Valve are far from bleeding edge. and everyone "feels" that Crysis/et al are far from efficiently programmed...

RE: Please, please, please
By christojojo on 3/19/2009 5:37:46 PM , Rating: 2
plonk420 said, "why not test those SEPARATELY, as well as the GPU/CPU-crushing FutureMark? Epic/ID, and especially Valve are far from bleeding edge. and everyone "feels" that Crysis/et al are far from efficiently programmed... "


inperfectdarkness said, "i'd rather see a benchmark of the serious engine & the lithtech engine; rather than id's POS."

So what you two are saying is only test game engines that are Coded perfectly and only sold by certain companies?

That makes perfect sense when you talk about "wonderfully coded for console" games like Call of Duty World of barely going ten minutes before crashing" and the like. Yep these days of console ports and and Beta gold it definitely is wise to only target test on games that are perfectly coded.

Seriously, the games tested should be on games that we as hobbyists and professionals can show our non tech friends and family performance on games they play too.

If we are going for perfection of my own personal flavor they might as well stop at testing Dark Forces and Outlaws.

I am sorry for the rude reply but reading the ideal tests with dream games makes as much sense as fake benchmarks.

RE: Please, please, please
By The0ne on 3/19/2009 7:41:41 PM , Rating: 4
While I agree with your statements there still has to be a point where one draws the line when one encounters a game like Crysis. It is a good example of how poorly this game was coded although the end result are good. So you have all other games run 50fps and this one game alone runs at a mere 10fps due to? And it's look prettier how? And it's a good/great/so-so game even though...??

So really, when put into various perspectives Crysis shouldn't be use as a benchmark at all. All it tells you is that the game is so crappy it'll take several generations of video hardware advancement to be playing at higher fps. Are you really benchmarking the video hardware or are you benchmarking how poorly this game is to be able to bring down video cards that run any other games out there just fine.

RE: Please, please, please
By sleepeeg3 on 3/19/2009 10:21:49 PM , Rating: 2

RE: Please, please, please
By inperfectdarkness on 3/19/2009 3:21:43 PM , Rating: 1
i'd rather see a benchmark of the serious engine & the lithtech engine; rather than id's POS.

if this is a 45nm chip, & they make a mobile version within 2 months--i'm going to jizz in my pants.

...because if it is, it will almost certainly destroy any semblence of the gtx-m series being top-dog; especially in x2 configuration.

all i really want is a 15.4" laptop with quad core, wuxga & a 45nm 1gb mobility 4890. the force is strong with this one. i can FEEL the "win" eminating from it.

RE: Please, please, please
By Natfly on 3/19/2009 4:03:39 PM , Rating: 2
This is supposedly a 55nm chip. The next process node that AMD will use for its graphics chips will be 40nm.

By inperfectdarkness on 3/21/2009 8:06:10 AM , Rating: 2
so the 4860 is 40nm...but this is 55nm? where's the logic in that?

RE: Please, please, please
By TA152H on 3/19/09, Rating: -1
RE: Please, please, please
By Alpha4 on 3/19/2009 6:07:41 PM , Rating: 1
Alpha4_RS_ @


RE: Please, please, please
By Chocobollz on 3/20/2009 3:45:33 AM , Rating: 2
performance in synthetic benchmarks is only indicative of performance in the same synthetic benchmark

I found that your statement is contradicting with your own thoughts. I mean, you're saying that synthetic benchmark is only indicative of performance in the same benchmark, so what do you think about a real game benchmark? IMO, it is just the same concept, it just using a different engine, that's all. You can't just compare performance between different game because every game made different, just as the case with 3DMark. You can't compare its score with another game because it's using a different engine.

Besides, you say game engines have their quirks, havent you realized that synthetic benchmarks such as 3DMark themselves have rendering engines which would themselves have quirks?

Ok, so what's the point then? If both have the same quirks, then both can't be used as benchmarks, is that what you mean?

I myself prefer 3DMarks to quickly measure performance between different graphics cards than using some games I don't even play. If you play Crysis for example, do you expect everyone else to play it too? Not everyone play Crysis but anyone could just download 3DMarks to measure their card's performance against other cards.

I found this debates of whether using synthetics or real world benchmarks as a useless debate. IMO, 3DMark itself is a game, it just it can't be controlled/played by the user, so it just as the same as running a game. It may not reflect real world performance because it is not a real world benchmark, it is synthetic , for God's sake.

"If they're going to pirate somebody, we want it to be us rather than somebody else." -- Microsoft Business Group President Jeff Raikes

Most Popular ArticlesAre you ready for this ? HyperDrive Aircraft
September 24, 2016, 9:29 AM
Leaked – Samsung S8 is a Dream and a Dream 2
September 25, 2016, 8:00 AM
Yahoo Hacked - Change Your Passwords and Security Info ASAP!
September 23, 2016, 5:45 AM
A is for Apples
September 23, 2016, 5:32 AM
Walmart may get "Robot Shopping Carts?"
September 17, 2016, 6:01 AM

Copyright 2016 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki