AMD Radeon HD 4670: Ruling from Top to Bottom
by Derek Wilson on September 10, 2008 12:00 AM EST- Posted in
- GPUs
Test Descriptions and Settings
Age of Conan
For Age of Conan, we opened up the graphics settings and clicked the high button. For added fun we went into the advanced page and upgraded to SM 3.0, though nothing else on this page was changed. We left AA off, as the graphics on this one were a bit tough already, and getting a good experience might require dropping down to medium settings. It's a shame that AoC doesn't allow us to test at 800x600. On an MMO, screen space can be pretty important, so we do understand the decision. Our tests consist of a swim towards Tortage Island and a quick run to the jungle. The test is a straight line run and very repeatable. We used FRAPS to record the average frame rate.
Enemy Territory: Quake Wars
We recorded a custom net timedemo for this one. All of our tests are done using the highest possible settings with 4xAA. The Doom 3 engine is getting a little outdated and just about everything can run it well. One of the more interesting issues is that we need to quit out of the game every time we change resolution or our scores get really crazy.
Race Driver: GRID
This game plays surprisingly well at lower resolutions on these cards. We used the ultra high settings, but disabled AA here. Some of the shaders this game uses really benefit from AA though, and the 4670 is actually able to handle AA pretty well on this one at 1280x1024. The test for GRID is the first straight (right into a wall) of a track. We start at the back of the pack and start FRAPS running as soon as we take off. FRAPS is stopped when we hit the wall.
The Elder Scrolls IV: Oblivion
This test has been with us in some way or another for a very long time. The game is DX9, and we use Ultra High default settings for this one. For our AA/AF testing, we force those options through the driver control panel. The test is a straight line run with FRAPS. Even though there's a little age on this game, it is still a great title to get people hooked on gaming. It's open and fun and it doesn't look bad at lower resolutions with the settings cranked up. And these little cards can deliver as well.
Crysis
Yes, the obligatory Crysis test. This one is done using Medium settings across the board and no AA. We were running in DX10 mode and 64-bit as well. Our test is the built-in GPU benchmark run 3 times, and our score is the average of the second two runs. The game does lose some of its luster under Medium settings, but it is still very playable. Depending on the card you have, you would want to spend some time seeing what settings you could get away with pushing up to High.
The Witcher
For The Witcher we crank everything way up. We would have done more AA tests with this one, but there is this annoying new trend of limiting the maximum resolution AA can be enabled on based on framebuffer size. Aside from being annoying in our testing, architecture has a lot to do with AA performance and framebuffer is sort of a naive metric to use. In any case, this is another FRAPS test, but we benchmark one of the early in-engine cut scenes
| Test Setup | |
| CPU | Intel Core 2 Extreme QX9770 @ 3.20GHz |
| Motherboard | EVGA nForce 790i SLI Intel DX48BT2 |
| Video Cards |
ATI Radeon HD 4870 |
| Video Drivers | Catalyst Press Driver 8.9 Beta (Radeon HD 4670) Catalyst 8.7 (Radeon HD 4850, 3850) Catalyst 8.8 (Radeon HD 4870, 3870) ForceWare 175.19 (9600 GSO) ForceWare 175.16 (9500 GT) |
| Hard Drive | Seagate 7200.9 120GB 8MB 7200RPM |
| RAM | 4 x 1GB Corsair DDR3-1333 7-7-7-20 |
| Operating System | Windows Vista Ultimate 64-bit SP1 |
| PSU | PC Power & Cooling Turbo Cool 1200W |
Facebook
Twitter
RSS
90 Comments
View All Comments
sskk - Friday, September 12, 2008 - link
OMG, my first thought was this is a look alike site and I've been duped. I mean no way this review belong to anandtech, this review is good for the "epic failed" site, PLEASE PULL IT NOW and write a new one, from scratch I might add.xeutonmojukai - Thursday, September 11, 2008 - link
I'm having a nice time imagining these in a triple crossfire config with a DDR2, X48 mobo, overclocked, liquid cooled (finally, a somewhat-gaming card that can use cookie-cutter water blocks), and probably a Q9300 thrown in. Something tells me that could actually make a great midrange system. Space for dual HDTV tuners and a PCI-E sound card too? Sign me up!djfourmoney - Tuesday, September 16, 2008 - link
In HEXUS' review its equal to a HD4850 in Crossfire. Since triple isn't common I would say that you would pick up some additional performance and distance yourself even further from HD4850.Honestly that is what I am planning. Not triple but dual HD 4670. I ordered a Sapphire 512mb HD4670 from NE and UPS bungled and lied about the delivery. It should be here tomorrow.
But two HD 4670 1GB on a 790GX with a 5000+ BE for now and a Phenom later, plus my two TV tuners (ATI TV Wonder 650 and Cat's Eye 150 PCI) on my Gateway 24" monitor I envision fairly powerful multi-media PC that will play every game on the market.
GRID with AAx4 on Ultra @1920x1200 - 33fps according to Legit Review
ATI's goal was 30fps with current games, I'd say Mission Accomplished
HEXUS saw increase of 80% over one card in many games including GRID, Crysis and COD4.
So I'm with you man, two are better than one when they are affordable as this and take less power to run.
I just wish mine was delivered on time, no thanks to UPS...
djfourmoney - Thursday, September 11, 2008 - link
Those resolutions are on TV's and most of the public still separates their PC viewing from TV viewing. Only gadget freaks, geeks and tech nuts are even doing the HTPC thing and have their PC's connected to their DLP, Projector, LCD or Plasma.So there's some use for that info, but c'mon Tom's hardware has covered the whole resolution issue before and breaks it down between say 1280x768 (720p) and 1920x1200 (1080p) is something like 17% more pressure put on a GPU to render images.
But anyway....
Yes a 9600GT is faster, but by how much and does it really make a difference in gameplay?
I say NO it doesn't not even a little bit. From 30 to 60 that's a fairly big step in performance, but if your talking like in another review 50.3 vs 47.0 in GRID @1280x1024 AAx4 that's NOTHING and only for Gamers to argue about, which I won't stoop that level, its immature for one thing. Its a my dick is bigger than yours nonsense, when it serves the same purpose, its strictly a men's argument and doesn't pass the giggle test with most people.
At the END of the day, guess what?
Even at $95 the 9600GSO is roughly the same performance if not slightly less in every bench I have seen thus far and its $16 more.
Even at $99 the 9600GT is slightly faster in some benches, even in others, some games are optimized for Nvidia GPU's, others ATI, GRID for ATI, Crysis for Nvidia. Nevermind the HD4670 is faster than 2900XT, HD3850 256MB, 2600XT, HD3650, the last two cheaper than the HD4670.
Its target was the 9500GT no matter how the media and Nvidia want to SPIN THIS. It crushes it, totally and completely and its not a argument of a difference of 2-3fps, its more like 10-15, even 20fps in some cases, you can't even play some games with the 9500GT, it won't reach 20fps in many games, even at mild resolutions.
In fact saving $19 over the 9600GT buys -
Splinter Cell, Company of Heroes and a few other games on Direct2Drive
Flatout 2, Midnight Outlaw - 6 hours till sunup, NFS: Pro Street, Jurced 2 HIN
For true gamers you more than likely have many of these games, already. The last game I bought for my PC was Pro Race Driver....
So FOR ME and people shouldn't let other people tell them what they should buy if they don't totally agree.
I can pick up this card, plus GTR2/GT Legends for $19 at Fry's and I also wanted to pick up GRID and NFS: Pro Street. Wow $140 spent and I even have enough for another HDD.
Value is value and this product is about VALUE.
strikeback03 - Thursday, September 11, 2008 - link
[quote]Its target was the 9500GT no matter how the media and Nvidia want to SPIN THIS. [/quote]Umm, they never target a specific card, they target a price point. At $80, the 4670 is clearly the winner, partly because there is no competition as the 9500GT has rightly dropped below $70.
[quote] So FOR ME and people shouldn't let other people tell them what they should buy if they don't totally agree. [/quote]
Obviously. Less conspiracy-minded people might see though that they are pointing out that for the cost of a few lunches at whatever your fast-food restaurant of choice is, you can pick up substantially better performance
djfourmoney - Thursday, September 11, 2008 - link
Other than the fact that the CIA is behind JFK and the FBI is behind many murders of people including MLK and 9/11 is a total and complete LIE, I'm not a conspiracy theorist.I rather save money on the product than to save money on food. If I can buy two things with the same amount of money I rather do that.
Happy Meals vs $19 Games, hell I found NFS: Pro Street for $9 at Best Buy and GTR2/GT-L for $9 as well. I'll be busy for the next few months...
Hrel - Thursday, September 11, 2008 - link
Anandtech, can you please start running your benchmarks in widescreen resolutions; that's where the industry is. 1280x800,1366x768, 1440x900, 1680x1050, 1920x1080. That would be great.Also, it seems like this card, run at 1280x800 with AA on would almost always outperform the 3870; and with the low power requirements it would be great as a laptop add on, using the expressbox. I'd love it if you could do a current article on external notebook graphics, even if it was only using the expressbox since I don't think any other exteranl graphics are currently for sale; I could be wrong though.
Anyway, this looks like a great card, and a positive step in incresing the lowest common demoninator for graphics performance; in 5 years they can design games to run on this card on min settings and low resoluiton; while tuning the setting to go all the way up to fully utilize whatever hardware is out then. ATI, thank you for the 4000 series.
frozentundra123456 - Wednesday, September 10, 2008 - link
I think this is quite an appealing card for the price and power usage. It is much better than the 9500GT or the HD3650 which are the closest competitors in terms of power usage. It will also blow away the HD2600 pro that I bought a few months ago at a similar price point, and has the same power supply recommendations. To me this is progress: better performance per watt and per dollar.Granted, spending about 50 dollars more could buy a much more powerful card, but it would also use more power, be bigger, and require a better power supply with an external power connector.
I am wondering what the HD4650 and possible 4400 series cards will be like. Considering the low price and power comsumption of the HD4670, I really question the need for any lower line cards.
TheJian - Wednesday, September 10, 2008 - link
You should have just ran it against a 9600GT. When you can get one for $95 at newegg (2 of them at that price) this article is pointless. They even have a huge selection of them for $99. Just tell us the GSO sucks, is pointless, and that the GT beats them all for $95. Could have saved a lot of moaning and groaning. :) The GT should have at least been in the article because they have a ton of them below $100 at newegg alone (like 10 or so).For $15 difference, it seems foolish to buy a $80 4670. Even more dumb to even suggest I'd buy a $100+ GSO model when I can absolutely dominate it for $5-15 less. Buy a 9600GT...geez.
JarredWalton - Wednesday, September 10, 2008 - link
Just something to consider in relation to antialiasing:"Small rabbit hole here: the real long-term solution to image quality is not AA, it is increasing DPI (dots per inch). Decreasing the size of a pixel will do a lot more to make an image look smooth than any amount of antialiasing could. What's the analog in the real world? Compare those duplo and lego castles to a sand castle. Many more grains of sand that are much smaller mean a very very smooth appearance with no AA needed. Display technology has severely fallen short over the past few years and we still don't have desktop LCD panels that really compete with top of the line CRTs from 7 or 8 years ago."
There's truth in that, but you also have to consider a few other things. One of the major problems with high DPI displays is that there are MANY programs that don't recognize DPI well, so you get itty bitty text that's practically illegible. I run a 30" LCD, and at 2560x1600 you get a .25mm pixel pitch. In comparison to other normal desktop LCDs, that's the smallest pixel pitch you can currently get (DPI is 101). It's about the same as a 17" 1440x900 laptop display, only instead of sitting 18 inches away on my lap, mine sits about three feet away. Let me tell you, it *does* make a difference.
There are plenty of websites for example that for a specific font size, and by default it's way too tiny for my eyes. Even MS Word has problems, so that I usually run at a 150% magnification. Also, those beloved flash movies - and movies in general - that come at 512x384 or 640x480 are hard to watch.
Very high DPI displays sound like a great idea - and they'll become more practical at some point - but right now I think .27 to .30 mm dot pitch is the sweet spot for most eyes. It's one of the reasons I think 27" LCDs are often a better choice for people than 30" (of course you can run the 30" display at both 2560x1600 and 1920x1200, although you get some blurriness at non-native resolutions). Those with better eyes may not find it a problem, but personally I don't think ultra-high DPI provides a great user experience right now.