back to article 16-card GPU bangs-per-buck mega shoot out

We've looked at a host of graphics chips, and while each new GPU generation has its fans who'll dash out and buy it no matter what, plenty of punters prefer to wait and see whether they're worth the money. In short, they ask, what level of bang will we get per buck? Or would be be better off spending a lot less on an older GPU …


This topic is closed for new posts.
  1. Anonymous Coward

    Hang On...

    Surely the 'ideal' line is close to the horizontal - shit loads of performance for sod all dosh?

    Mine's the one with Sapphire HD 4850's in each pocket...

  2. Jonathan

    But who plays 3D Mark?

    I dont know about you guys, but I play games like Crysis, Race Driver: GRID and Supreme Commander, not 3D Mark. How well will these cards run those games?

  3. Keith Goddard
    Thumb Down

    3d Mark Sucks

    You need far more than just 1 test in 3dmark to base any kind of comparison, this is a very lazy article.


  4. Brian Miller

    + 1 for 3dmark issue

    While the results are pretty much how it stands, only using 3dmark06 is a bit of a fob off.

    Also, seeing as the new king is in town, where is the 4870x2 R700 GPU. eh??

    I think an update is in order.

    At least you only have to run a single test.

  5. Mark Brew

    Single cards for me - and what about filtering?

    Dual card systems are unstable, whether it be crossfire, SLi, or X2 flabours of cards. i've treid them all and i'm now flogging my 2 8800GTX OCs in favour of a single GTX 280.

    I use dual monitors also which is flakey on two cards, if it works at all.

    You also disregard the filtering ability of these cards which is important as the 4850s with their paltry 512Mb cannot filter at high resolutions, wheras the GTX 260 and 280 eat it for breakfast due to the higher RAM onboard which again is where the extra money is going.

    I suggest people disregard this article and look at the card that suits their needs.

  6. Bad Pritt

    Re: 3d Mark Sucks

    Although it is indeed a rather lazy way of comparing the cards, the conclusion is still valid given the huge margin by which the 4850s came out on top. I checked some other websites for reviews (dan's data was especially enlightening) and the figures seem to be fairly representative for average gaming performance. The consensus of online reviewers is that if you want a cheap, single card you should be looking at an 8800GT, 9600GT or 4850, while for top performance you really should go for the dual 4850s (or maybe 4870s later).

  7. Anonymous Coward
    Thumb Down

    Article full of obvious flaws

    This article obviously was not subjected to any editorial control.

    1. GeForce 6800GT = £140? Perhaps in 2004.

    2. Where's the ATI 4870 X2? Tom's Hardware Guide just declared it the world's highest performing retail graphics card.

    3. "Games such as Crysis or Elder Scrolls: Oblivion load the whole system quite heavily and make it tricky to judge the effect of the graphics card in isolation." BUT the whole point is that you need to take the entire system into account!

  8. Adrian Jackson

    I could give you an antique graphics card for free...

    That gives *infinite* bangs per buck, and thus will be the greatest graphics card of all time in all future comparisons.

  9. Timbo


    Quite agree.....the ideal best bang per buck, should be closest to the "cost" baseline, whilst being as far away as possible from the "performance" baseline.

    Hence the line should NOT be at a 45 degree from the origin - coz that's just a random "assumption" of the "ideal".....the actual ideal angle would depend on the scale used for each baseline.....a "shortened" X-axis base line would cause the line to be at a completely different angle compared to, say, a "shortened" Y-axis.....

    Isn't it

    isn't it



  10. b

    oops, the 4870X2 was just released..

    missed that little puppy eh?!

    reports of average 25% gains over the GTX280 and peaks of 50%.

    it should be noted that you can now get the rather excellent 8800GT (G92 core) for around £80 now, which is incredible value.

    you can also get a 4850 for around £115 and a 4870 for around £180 (as spotted on amazon, but that's not a recommendation, just a point of reference, tho personally i do like buying from them). tho looking at this chart, i'd say the 8800GT for £80, which tucks in quite close to the 9800gtx and 4870, was the real steal when it came to budget gaming and what i'd recommend in this time frame.

    personally i've got a 9800GTX, which is a nice stable card which i got for about £180.

    2x4850's in SLI seems to be the generally accepted current fave setup for power/value for gaming...for now..tho single card i'd go for 8800GT/4870 depending on budget..



    relevant stuff and nonsense (all welcome!):

  11. adnim

    Great with news poor with reviews

    Don't give up the day job Reg ;-) I use dedicated sites for hardware reviews, I use Reg for my daily dose of IT news. The right tool for the job as my missus always says.

    Reg I will happily do professional, real world hardware reviews for your site for a fee which is open to negotiation. You have my email address.

  12. Anonymous Coward
    Thumb Down


    Why bother having an Intel Skulltrail rig, and not posting its specs?

    Also, totally agree with Keith. I actually feel sorry for anyone who stumbles across this review and makes any decision based upon it, as it isnt really worth the hosting space it takes up.

    Very disappointing El reg, please dont do such articles unless you're going to post full spec list of test rig, along with at least 4 modern game comparisons......

    Also - 8800GT's can be had for £88 not £125, and if you're quick this week, pick one up for £80!

    If your prices were correct, that would skew the graph VERY differently!

  13. Peter Gathercole Silver badge

    45 degrees?

    No, not necessarily 45 degrees. Do a least square regression (or similar), and then all of the cards below the line show better than the average, and those above are worse. The further a card is from the line, the more extreme the perforance vs. price is from the 'average'. But the line could be any angle, depending on the scales you choose.

  14. Anonymous Coward

    Ignoring upgradability

    Surely whether the rig is SLi\Crossfire is also important? For the same (or slightly worse) price and or performance you should always choose the single card! Since then in the future as prices come down you can buy the second card and boost the performance...

  15. xjy
    Paris Hilton

    E&M and junk shop junkies

    @ everybody: The "ideal" line referred to is one where the relation of price to performance equals out, ie 45 degrees. The distance from this line can then be inspected by eye, and the relevant conclusions as to performance and price can be drawn.

    Also: not everyone reading the Reg eats the whole of Exchange & Mart for breakfast or knows all the best junk shops/fences in town. If you've got lots of time on your hands, or incredible enthusiasm, or decades of experience, you're bound to find better deals. My brother frinstance has never paid full price for anything in his whole life, and he's over 60. Hasn't made him any happier ;-) Dad was the same E&M and junk shop specialist.

    What this review (and others like it) does is point you in the right direction. You are not being infantilized and ordered what to buy. It shows you what's hot and what's not, and what prices are nicest.

    Beat the review and you get bragging rights, but if you're so fucking clever, what are you doing wasting your precious time reading this crap? Time is money, or kills, or a few more levels... ;-)

    Paris, cos her time is precious too...

  16. Anonymous Coward

    Breath of fresh FLAME

    It's great to see someone use a shorter format for a hardware review. I'm not buying the flames roasting this as lazy work.

    The 3dmark tool used is a good middle of the road benchmark. The author acknowledged it's shortcomings and made a valid case for it's use, to give an apples to apples comparison on the different boards. Same for the lack of exact specs test board.

    I don't think the choice of a singe benchmark is a bad decision either. There are a dozen sites that will give you 15 page reviews of a single card. By the time you finally click past the 13th page of boilerplate the card you were reading about is already obsolete junk.

    Not to mention the kilowatt epochs of power we waste each year running all of these frame rate tests on everything from Crysis to Pong.

    Waste of the authors time running the tests, waste of the readers time sifting through it. Keep these things short and sweet. 2 pages long, with a cost/performance graph with something like the frames/cost chart that can be compared with later articles with newer models. Then just knock them out quick and keep them coming.

    If we need more information, we can find it pretty easy someplace else. It's nice to see Hardware coverage on the Reg, but i'd rather see a mix of real articles on hardware as well as the reviews. If the short form lets you guys keep up with the latest models, and still write more in-depth stories that's a win-win.

    Oh, and Leo, say hello to Sisyphus for me. Your mother must be heartbroken. We all make bad choices in life, but hardware reviewer? What did you do in your last life, club seals?

  17. Anonymous Coward
    Anonymous Coward

    What about

    The Age of Cohen performance benchmark - FPS - the only thing worth knowing now.

  18. Jamie Kephalas

    What I really want to know is

    Will it blend?

    Mine's the one with an AGP card... PCIe? What's that!?

  19. Scott Butterworth

    Dual GTX280's for me thanks

    Dual GTX280s for me. Best $600 I spent. Blows everything and anything in your little graphs away. Contemplating a 3x280 config once the prices start to drop.

  20. A

    Not bad, but

    What I'd like to see are:

    * Bang/Watt figures

    * Bang/dB figures

    The best graphics card in the world is next to useless if it sounds like you're stuck inside a Vulcan's engine whenever you use it, similarly if you need a small nuclear reactor just to feed it there might be better options for the average user ;)

    Overall a nice review though; not sure about the 45° part, but the rest of it is useful for a quick summary.

  21. alistair millington
    Thumb Up

    I like 3d mark

    The dragon and the air ship particularly. :)

    The review however is okay, it doesn't compare for example twin of every card and singles of them all... Or those that a re from different manufacturers but claim to be the same card. So you have to take what is in it as read and work with it.

  22. Lightnix

    3Dmark, isolated graphics - not quite

    "Games such as Crysis or Elder Scrolls: Oblivion load the whole system quite heavily and make it tricky to judge the effect of the graphics card in isolation."

    I find it annoying that you missed out the fact that 3DMark 06 is a lot more CPU bound than many games.

This topic is closed for new posts.

Biting the hand that feeds IT © 1998–2021