Sign in with
Sign up | Sign in

Benchmark Results

Radeon 4870: Does It Matter Who Made Yours?
By

3DMark Vantage

Rather than execute the entire benchmark, which also measures CPU performance, we ran only the GPU and Feature tests. All test options except resolution (which we changed to 1920x1200) were left at their default values.

Far Cry 2

We measured Far Cry 2 performance using the built-in benchmarking tool. Screen resolution was set to 1920x1200, anti-aliasing was set to 4x, and all options were set to their highest values (either High or Very High).

Supreme Commander: Forged Alliance

We measured frame rates using a saved game and FRAPS. Screen resolution was set to 1920x1200 with 4x AA and Render Sky enabled, and all detail levels were set to High.

Unreal Tournament 3

We measured frame rates using the UT3 Benchmark tool in DirectX 10 mode. We used the Defiance level with 12 bots and texture and level details set to five. We set the game’s screen resolution to 1920x1200. Brightness and field of view were left at their default values; reduced gore, subtitles, and hardware physics were all unchecked; and texture detail and world detail were set to Max. V-sync was disabled.

World in Conflict

We used the game’s built-in benchmarking tool to measure frame rates with the game set to a resolution of 1920x1200. We used 4x anti-aliasing and anisotropic filtering, turned on all features, and set all quality values to High or Very High.

Ask a Category Expert

Create a new thread in the Reviews comments forum about this subject

Example: Notebook, Android, SSD hard drive

Display all 48 comments.
This thread is closed for comments
  • -8 Hide
    MrMick , December 23, 2008 6:30 AM
    Once you’ve found the GPU that best fits your needs, does it really matter which vendor sells you the card? We take a look at AMD's reference 4870, Sapphire's, and Palit's to get a better sense for what differentiates graphics cards.

    Radeon 4870: Does It Matter Who Made Yours? : Read more
  • 3 Hide
    wavetrex , December 23, 2008 6:58 AM
    You can raise the fan speed of the reference card very easy using the 8.11 and 8.12 drivers.

    My totally stock card runs idle at 45 degrees with 35% fan speed ( inaudible ) and at 55 degrees full load ( in Crysis ) with 45% fan speed (audible). During the summer I'll probably raise the speed by 5-10% to compensate for the increased ambient temperature.

    non-reference is overrated...
    Reference has a problem with bios settings and NOT with the cooler itself.
  • 5 Hide
    Proximon , December 23, 2008 7:42 AM
    The Palit may be the best in a well cooled case, but regardless I'll take the Toxic. Rear venting, 48-60C temperatures at a whisper quiet 34% fan speed.

    Good article though, especially since many here in the US have been curious about Palit lately. You might have added more about Vapor-X. My understanding was that ATI intends to make it reference design.
  • 1 Hide
    Anonymous , December 23, 2008 10:03 AM
    I think what really matters about the vendor is at what quality the card was built. For example, If the vendor used a poor quality BGA process, one vendor may have more problems with artifacting in 3D applications later on down the road vs a different brand may have a higher percentage of product that makes it several years if need be without a single glitch.
  • 0 Hide
    montyuk , December 23, 2008 10:10 AM
    talking about third party hsf - i bought the force3d dht which whilst it has a fantastic heatsink, decided to couple it with a fan that does not allow the fan speed to be controlled by software (yes i have tried the usual), so ive ended up replacing their fan with a new one which can be controlled.

    surely their engineers must have realised that the fan was stuck at 100%?
  • 0 Hide
    Tekkamanraiden , December 23, 2008 10:25 AM
    Hmmm I wondered why my 3870 pc/mac hybrid card was blue in color. Now I know why.
  • 0 Hide
    V3ctor , December 23, 2008 10:48 AM
    I put mine on water... 29ºc idle - 41º full... I just hope ATi uses the same pcb layout in the HD5800... That way I can re-use the waterblock
  • 6 Hide
    Malovane , December 23, 2008 11:16 AM
    Exactly... how many times can you beat that horse to death? The cooling is fine on the reference 4870's. Just manually set the fan speed with the new catalyst control center and it will run cooler than the previous generation of ATI's products.
  • 0 Hide
    Shadow703793 , December 23, 2008 11:38 AM
    Good write up. But I don't care much for the software, but rather the cooling & Ocing. Tbh, you can find smiler software for totally free (ie VCL, Zoom,etc)
  • 2 Hide
    customisbetter , December 23, 2008 12:46 PM
    I have the Palit in question and its awesome for two reasons.
    First, the card is physically shorter than the reference card so it actually fits in my case. The second is the amount of cooling comes from ONE fan. I unplugged the smaller fan due to noise and set the large fan to 45%. Its totally quiet and never goes over 70 degrees C. I highly recommend that card and whenit becomes obsolete, the HDMI and Display port connectivity will be great for a HD Media box.
  • 0 Hide
    malveaux , December 23, 2008 2:23 PM
    Interesting.

    Essentially, there's no difference amongst them.

    Cheers,
  • 0 Hide
    Niva , December 23, 2008 2:23 PM
    80 degrees won't burn you, it is hot to the touch though. I enjoyed this article, I still run a Tyan made ATI 9700 Pro card in my old PC which stays on all the time, that card is a champ!
  • 2 Hide
    Anonymous , December 23, 2008 3:14 PM
    OMG, Catalyst 8.10, I didn't even bother going through the benchmarks.
  • 5 Hide
    cl_spdhax1 , December 23, 2008 3:14 PM
    add fanspeed control to catalyst for most cards.
    written by spyral,
    "START -> REGEDIT (RUN AS ADMIN) -> HKEY_LOCAL_MACHINE -> SYSTEM -> CurrentControlSet-> CONTROL -> CLASS -> {4D36E968-E325-11CE-BFC1-08002BE10318} -> 0000 (be sure to tweak your CURRENT drivers key MINE IS 0000)

    THEN LOCATE AND EDIT THE VALUE DATA FOR EACH BELOW TO (2)

    FanSpeedPercentActual_NA
    FanSpeedPercentTarget_NA
    FanSpeedRPMActual_NA
    FanSpeedRPMTarget_NA

    SO AFTER THEY WILL ALL HAVE DATA VALUE 2


    EXIT -> REBOOT -> GO TO CCC -> OVERDRIVE"

    under the overdrive menu, if it shows up, it works. if it doesn't, then your card doesn't support it. simple.
  • 0 Hide
    Transsive , December 23, 2008 3:35 PM
    I have an Asus 4870. I think it's pretty much the same as the reference board.
    I set the fan speed at 30% and it usually sits at 55-70 degrees.
    It has a dual slot cooler with rear exhaust.
    750 core, 900 memory.
  • 4 Hide
    yyrkoon , December 23, 2008 3:38 PM
    malveauxInteresting.Essentially, there's no difference amongst them.Cheers,


    Actually, there is a big difference in name brands. It's known as "customer relations". All companies can have lemons in any part, and it is their willingness to stand by their respective parts that counts. Sapphire is one of these companies that *will* stand by their products, and give customers very little hassles during such unfortunate circumstances *if* a RMA is imminent. However, I am not saying there are no other brands out there that may practice the same objective, but now days, with shoddy workmanship around every corner, I tend to not stray from something that works( and works well ).

    Driver support is another gotcha, but mostly now days ATI, and nVidia products are covered by their respective 'parent' companies. Overall system stability with said products installed would be another. But in the later situation here, there can be more than a single variable at fault.
  • 0 Hide
    Anonymous , December 23, 2008 4:03 PM
    I can't say the voltage regulation of the Palit model is a step up, it's just cheaper and allows them to sell the card for less. Personnally, I'd stick with the Volterra system.
  • 1 Hide
    JonnyDough , December 23, 2008 6:07 PM
    These benchmarks show one thing. Nothing.

    The slight fluctuations determined by the benchmarks show that a change of FPS by .7 or even two FPS mean absolutely nothing. Half a frame isn't going to make or break the playability of a game. For all we know a tiny background process decided to run at the exact moment the FPS snapshot was taken.

    Stick to differences of steady 5 FPS differences or more, or don't even bother writing these types of "articles."

    The difference is minimal at best, and even an OC'd card will barely make any difference over a stock card. 100mhz does little for gaming today. Maybe it made a difference in the Commodore era, but it makes very little difference now.

    Anyone with a brain knows that the difference between vendors is the over-clockability, but the following are really the deciding factors:

    1. Price.
    2. Silence.
    3. Included adapters.
    4. Included software.
    5. Consumer relations.

    Any 100mhz increment or frame rates over any other similar card is next to pointless.

    Tell us something we didn't know.
  • -2 Hide
    ideapete , December 23, 2008 7:17 PM
    Be aware that most if not all new graphics cards Nvidia or ATI ( Or a subset like powercolor ) all have poor OGL support and if you use the cards for modeling with sketchup or 3dMax or similar CAD programs they are useless. If you are a modeler get recommendations from your software supplier or your gray hairs will increase accordingly

    We recently tried ATI Pwrclr AGP 3650 and the performance and support was atrocious
  • 0 Hide
    avatar_raq , December 23, 2008 7:17 PM
    Michael, I just wanted to say don't pay attention to cranky people that impose destructive criticism, you should have got used to them by now.
    Your article is so good and it answers questions I've had in mind for a long time about the diffrences between the available 8800 GTs for example, I guess the same results apply there, right?
Display more comments