It's been since the holidays that we've done a GPU buyers guide. It never seems like the right time to do a new GPU buyers guide, as NVIDIA and AMD have been pushing aggressively back and forth for leadership in the market place. When new parts or tweaked cards haven't been coming out, prices have been adjusted quickly to maintain tight competition.

Now is no exception. There are a couple spots in our line up where we will have to make recommendations based on what we know about what's happening in the market place. In competitive reviews, we try very hard to look only at that exact time slice to make our recommendations. In our buyers guides we like to be a little more flexible and take a more retail and market place view rather than the heavily technology and performance based focus of our GPU reviews.

Starting out, we're looking at the roughly $75 market where we split our recommendation between the 4670 and the 9600 GT. Prices have compressed more over the past few months, and the 4670 comes in low enough to cover many needs at very little cost. You can always spend less on graphics and get less, but if you want more than 2D, the 4670 and 9600 GT are where you should start looking.

$75 Recommendation: ATI Radeon HD 4670

 

  ATI Radeon HD 4670
Apollo $64.99
Gigabyte $79.99
Sapphire $69.99

 

And we've got the GeForce 9600 GT. Just a little more performance in some games, maybe a little less in others, with roughly the same cost. But if you want any more than that, you'll want to wait about a month.

$75 Recommendation: NVIDIA GeForce 9600 GT

  NVIDIA GeForce 9600 GT
Apollo $74.99
Gigabyte $67.99
Sparkle $89.99
PNY $97.99

 

For our ~$100 price point (plus or minus a bit) we are going to strongly recommend that people wait for about a month. This price point will be shaken up a bit in about that time and we really aren't comfortable recommending anyone purchase something in this market until sometime in early May. This may or may not further compress the sub $100 market, but there really isn't much more room down there, so we don't expect much change except at right around $100.

$100 - $200 Recommendations
Comments Locked

91 Comments

View All Comments

  • triclops41 - Sunday, April 5, 2009 - link

    I think it should be mentioned that the 4670 doesn't require external power, therefore it can be used on pretty much any computer because you only need a very baseline power supply and a PCI-E slot.

    the 4670 is comparable to the 9600gt in power, but surpasses it in quietness, efficiency and flexibility of implementation.

    My secondary PC is a basic dual core dell with a 4670, and it runs any game at 720p full settings without a hitch. Didn't even have to swap out the crappy Dell stock PSU.
  • DBissett - Wednesday, April 1, 2009 - link

    So you eyes find what appear to be the BOLD section headings, and then it takes a minute to figure out that the relevant text is ABOVE what appears to be the BOLD heading and not below it. So now you're scanning backwards to read what you want in the review. WHY? Get a better editor on these articles. Eyes don't need this kind of exercise.
  • KhadgarTWN - Wednesday, April 1, 2009 - link

    4870 could perform better then GTX 285 in what circumstance?
    let alone in the REAL GAMING review GTX 285 sit on par with 4870X2 with about 2/3 of power consumption. (OCP)
    And how about GTX 260?
    And how about the situation when SLI/CF failed?
    And how about the PSU?

    The only one silly site "cite" about 4870 may catch up GTX 285 is this one....

    Anandtech's recommandation is more and more like a reversed one.
  • 7Enigma - Wednesday, April 1, 2009 - link

    Let's see....

    Age of Conan: 4870X2 kills 285 @ 1680X1050, 1920X1200, 2560X1600

    Call of Duty World at War: 4870X2 destroys 285 @ 1680X1050, 1920X1200, >10fps @ 2560X1600

    Crysis Warhead: Close on the low end resolutions but 4870X2 >15% faster at 2560X1600

    Fallout 3: Close on low end resolutions but 4870X2 14% faster at 2560X1600

    I got bored after 4 straight benchmarks in the 285 review showing the 4870X2 was better. Me thinks you need to go back and check your facts....
  • strikeback03 - Wednesday, April 1, 2009 - link

    Didn't say 4870x2, but 4870. I'd assume that was a response to the second paragraph on the last page: "Yes, the GTX 285 does offer the highest performance of a single GPU solution in many cases. But the fact that the 4870 costs just a little more than half as much, and in some cases performs higher, leaves a sour taste in our mouths."
  • 7Enigma - Wednesday, April 1, 2009 - link

    I was refuting his comment that "in real-world" it is similar to the X2, which it in no way is. reviews now have so much ambiguity in what they decide is [hard] they fail to have much actual scientific comparisons. I used to use them as another source for reviews but after the past couple fiasco's (C2D launch, last 3 or so major gpu launches), I believe they are (and very smartly so) pandering to the flamewars so they can generate ad revenue instead of actual good reviews (I was banned after that comment over there).

    As for Anand's comment it holds water:

    "Yes, the GTX 285 does offer the highest performance of a single GPU solution in many cases. But the fact that the 4870 costs just a little more than half as much, and in some cases performs higher, leaves a sour taste in our mouths."

    Just with another quick look at the 285 article here's what I can see:

    -Age of Conan: 4870 1gig beats GTX 285 at all resolutions

    -Fallout 3: At 1680X1050 and 1920X1200 4870 1gig beats GTX 285

    -G.R.I.D.: At 1680X1050 and 1920X1200 4870 1gig beats GTX 285

    If you just browsed the original review and looked at the bar graphs which only showed the highest resolution (which IMO is not the proper one to show since we are talking single GPU setups where at most people are running 24" displays, I actually think I posted a comment in regards to that choice since many people don't look at the detailed data), you would come away with the 4870 1gig only beating the GTX 285 in Age of Conan. Once you see that unless you are running a 30" display they pretty much split the games selected in 1/2 at 22-24" display resolutions, I think Anand's comment is actually going easy on the GTX 285.

    If I was to write that sentence it would be much more painfully truthful, "Yes, the GTX 285 does offer the highest performance of a single GPU solution in 4 out of 7 cases at 22-24" display resolutions. But the fact that the 4870 costs just a little more than half as much, and in 3 out of 7 cases performs better, leaves a sour taste in our mouths."

    Disclaimer: I own a 4870 512meg, but could care less "defending" it against an Nvidia product. I only care about performance/$, and more importantly factual data.
  • san1s - Wednesday, April 1, 2009 - link

    rofl
    grid? age of conan?
    ati biased games, all of them
    if a gtx 260 beats a 4870x2 in the nvidia biased dead space, would "it leave a sour taste in our mouths?"
  • 7Enigma - Thursday, April 2, 2009 - link

    Cherry pick all you want, and claim bias as well, the simple fact stands that when you look at the data the 4870 at practical resolutions for its price range/target audience does quite well against the GTX 285 (that would be 24" and smaller displays). BTW, you forgot about Fallout3 or is that also an ATI-biased game?

    If you feel the games are biased take it up with Anandtech, I'm simply going over the data to refute the baseless claims.
  • san1s - Friday, April 3, 2009 - link

    "feel the games are biased"
    I don't feel the games are biased, the ARE biased toward ati cards
    in a real life scenario, the games you play aren't half nvidia favoring half ati like the selection on anandtech reviews
    looking at the top games, the overwhelming majority are "the way its meant to be played""
    if you want to go for "practical resolutions" and "target audience" then when don't you support testing with the target games that people that pc gamers play?
    "does quite well against the GTX 285"
    the cards are not int he same league, different price points
    performance does not increase at the same rate as price
    for example, you can buy a $100k porsche 911 turbo, or you can step it up to a $200k gallardo
    do you get a 200% increase in performance? no
    same thing with graphics cards, and just about anything else you buy

  • 7Enigma - Wednesday, April 1, 2009 - link

    To Anand and the crew:

    This just goes to show the damage those bar graphs can cause when people go back to compare data from previous articles. A casual glance would have you saying the GTX285 pretty much destroys the 4870 1gig in 6 out of the 7 tests (by destroy I mean overall with the bar graphs the 4870 loses the vast majority). This is a false assumption as I described in the previous comment.

    I would like to ask again that either the bar graphs get removed from future articles (I'd personally like this but I think the multi-resolution broken line graphs are not easy to read for people unaccustomed to looking at data), or the team comes up with a practical and more importantly transparent decision on what resolution gets selected for that bar graph. Here's my personal opinion:

    -In single GPU tests (ie the GTX 285 article) the bar graph is either 1680X1050 or 1920X1200. 99% of the people purchasing this card are going to be using a 22-24" monitor (possibly still a 19" even!) and so the majority are going to benefit from seeing the big chart showing one of those 2 resolutions. You also need to sum up succinctly the complex data a bit better in the written portion after the figures. You can break it down anyway you like, but please also have an overall last sentence to the big picture effect. Sometimes you do this, but other times I'm left puzzled what was mentioned when a bigger point should be mentioned.

    -In mainstream card reviews (4850, 8800 (or whatever they are now rebadging them :), and lesser cards) the bar graph should be 1280X1024 or 1680X1050. I think you'll probably find the majority of people are probably split about even between 19" and 22" displays that are looking at these products. Again we want to see higher resolutions for things such as SLI/X-fire, but that is not what these products are predominantly geared towards.

    -In high end/multi-gpu on a card review go hog wild. Sure the large majority of people buying these probably don't even have 30" displays, but that's really who should be purchasing these in the first place. Having 200fps with a 4870X2 running on a 19"-22" display is just stupid, but at least you will be able to show this with data.

    Just my 2 cents. Keep up the great work guys, you are the definitive leader of tech sites.