Oooh, new GPUs from AMD - however these ones aren't going to be breaking any performance records, they are both priced below $60.

The Radeon HD 4350 and 4550 are slotted in between integrated graphics and the set of hardware we took a look at recently in our Radeon 4670 review. The Radeon HD 4550 will run you around $45 - $55, while the Radeon HD 4350 will be priced at $39. Generally speaking, if you're a gamer you're not spending any less than $150 for a graphics card - so these GPUs are mostly for enabling hardware Blu-ray acceleration or providing a boost in performance over games running on integrated graphics.

  ATI Radeon HD 4870 ATI Radeon HD 4850 ATI Radeon HD 4670 ATI Radeon HD 4650 ATI Radeon HD 4550 ATI Radeon HD 4350 ATI Radeon HD 3870
Stream Processors 800 800 320 320 80 80 320
Texture Units 40 40 32 32 8 8 16
ROPs 16 16 8 8 4 4 16
Core Clock 750MHz 625MHz 750MHz 600MHz 600MHz 600MHz 775MHz+
Memory Clock 900MHz (3600MHz data rate) GDDR5 993MHz (1986MHz data rate) GDDR3

1000MHz (2000MHz data rate) GDDR3

or

900MHz (1800MHz data rate) DDR3

500MHz (1000MHz data rate) DDR2 800MHz (1600MHz data rate) DDR3 500MHz (1000MHz data rate) DDR2 1125MHz (2250MHz data rate) GDDR3
Memory Bus Width 256-bit 256-bit 128-bit 128-bit 64-bit 64-bit 256-bit
Frame Buffer 512MB 512MB 512MB GDDR3 or 1GB DDR3 512MB 512MB or 256MB 256MB 512MB
Transistor Count 956M 956M 514M 514M 242M 242M 666M
Die Size 260 mm2 260 mm2 146 mm2 146 mm2 ? ? 118 mm2
Manufacturing Process TSMC 55nm TSMC 55nm TSMC 55nm TSMC 55nm TSMC 55nm TSMC 55nm TSMC 55nm
Price Point $299 $199 $79 $69 $45 - $55 $39 $199

 

 

While the Radeon HD 4670 was quite impressive with 320 stream processors, the same number that was in last year's Radeon HD 3870, the 4350 and 4550 only have 80 SPs. That's twice the number of SPs in AMD's 780G, the current highest performing IGP solution on the market. In our Radeon HD 4670 review we found that the GPU was fast enough for pretty much all current generation games at resolutions up to 1280 x 1024, but with only 1/4 the shader power of its $75 brother we don't have high gaming expectations from these cards.

Both the 4550 and 4350 are mated with a 64-bit memory interface and either a DDR2 or DDR3 frame buffer. With very little memory bandwidth, and very little processing power you need to have good expectations for these cards.

The competition from NVIDIA is a little blurrier; while the GeForce 9400 GT is priced more in line with where we expect these cards to end up, NVIDIA does have one trick up its sleeve. The GeForce 9500 GT, paired with 256MB of DDR2 memory (the same type you'd get on your desktop) is priced at around $65 - $70 but is currently available with a $15 - $20 mail in rebate, bringing it down to the about same pricepoint as the Radeon HD 4550.

The Perfect HTPC Cards
POST A COMMENT

56 Comments

View All Comments

  • ThePooBurner - Tuesday, September 30, 2008 - link

    If yo are going to test something as a gaming alternative for those who want a bit of a boost in their gaming, try actually setting the games to settings people might use. Also, testing with just 2 games, both of which either were or are system breakers seems a bit... dumb.

    I would prefer the cards be tested with actual settings to see what they can ACTUALLY do and give a better indication of performance. It's been my experience that turning on some of the eyecandy and turning up the resolution for a slightly reduced framerate is preferred over looking at something ugly for hours. Also, why test the 780G instead of the 790G? or why not both?

    Here is are test setups i would suggest that would be more reasonable and representative of what people would actually do with these cards.

    1) 1024x768 Med Quality (as in the second down from the highest setting), no AA, Triliner/8x filtering.
    2) 1024x768 High Quality, no AA, Triliner/8x filtering
    3&4) Same as above but with 1280x1024(or 960).

    These tests are much more inline with the testing that happens on their big brothers and allows for a much more quantitive comparison of what you are getting. Automatically testing everything at junk levels is worthless to people. It gives us no basis for comparison. We need apples to apple and orange to orange testing if we want to see what exactly the fruit we are looking at is. Also as a point of comparison it should include the 4850 or the 3850. Those are both very well performing cards that will give a good idea of how much more you would get for more money. But as this review currently stands it is of little to no worth to anyone.
    Reply
  • superflex - Tuesday, September 30, 2008 - link

    Using a $1450 processor, $240 mobo, $300 RAM and $400 PSU to test a $40 GPU is assanine. That does no service to the HTPC end user. Reply
  • strikeback03 - Tuesday, September 30, 2008 - link

    It is their standard graphics test bed, you can't very well make comparisons if you are changing non-tested hardware around.

    They probably shouldn't have bothered posting anything on these though until they were done with the HTPC portions, judging by all the complaints in the comments.
    Reply
  • deruberhanyok - Tuesday, September 30, 2008 - link

    Yeah, it would be completely irrational to have a second standard test system for mid-range parts and a third for these lower end ones. I mean, what do you think this is, a tech website? Reply
  • PAPutzback - Tuesday, September 30, 2008 - link

    Something along the lines of it being tested in hybrid mode if it is capable with an HTPC type setup. IE, low watt fanless PSU, and low watt processor.

    The new game should not be benchmarks scores but more about what is the leat wattage I can make a PC and still have all the function of a Media HTPC. No gaming.

    I was really disappointed seeing as Anand has a theater blog on the site and this would of been a perfect setup to test along with his theater.
    Reply
  • whosthere - Tuesday, September 30, 2008 - link

    Derek,
    You guys published an article on G45 a few days back, then why didn't you show any G45 numbers in this article and still showing the antiquated G35 numbers?


    Reply
  • fic2 - Tuesday, September 30, 2008 - link

    Looking forward to the article on how well a 4870 runs on an Atom based computer.

    Got to go google for articles on how well a Landrover does on an F1 track.

    What a waste of an article. Best thing that can be said is that trees died for it.
    Reply
  • archer75 - Tuesday, September 30, 2008 - link

    The ATI cards have problems with some receivers and TV's. It has to do with the EDID and is documented over at avsforums.com
    This affects some Onkyo, Denon and Yamaha receivers.
    Reply
  • madspartus - Tuesday, September 30, 2008 - link

    Honestly why did you pair a $40 graphics card with a qx processor at 3.2 ghz...it just doesnt make any sense.

    Then showing power consumption differences on that system using a qx processor and 1200W power supply where the power used by the video card is little more than a unmentionable blip.

    next time would you try testing it in a system we might actually use this hardware in, like HTPC which you said yourself.

    maybe compare the power consumption to an IGP solution...and use a ~300W power supply etc.

    All those graphs were of no value to someone who wants to evaluate this thing for HTPC.
    Reply
  • duploxxx - Tuesday, September 30, 2008 - link

    I agree, keep the review in reality and put this in a 780-790 mobo with x2 6000 or 6500 and an intel p45 with e5xxx - e7xxx series.

    that would give a realistic platform
    Reply

Log in

Don't have an account? Sign up now