ATI Performance Comparision. 3850/4850/5850 benchmarks!!

Discussion in 'Video Cards & Monitors' started by AMD2400, Jul 21, 2010.

  1. AMD2400

    AMD2400 Member

    Joined:
    May 28, 2004
    Messages:
    659
    Location:
    Sydney
    Aye guys,

    Recently got a 5850 to play with and realised i had a 4850 and 3850 lying around, so i thought i'd do a comparision on a current mid range computer. Don't have anything earlier but even with the 3850, things start to get slow to a point of wasting my holidays :lol:

    [​IMG]
    The Cards. In a Sandwich.

    Introduction

    Since the introduction of the 3850, the mid - high end graphics card market started to change. ATI Graphics cards in this segment moved to become more affordable, and new graphics card buyers had more of a issue when choosing nvidia and ATI. I recall at one point the 3850 was around ~$120 brand new, while the 8800GT's where going for twice that (Granted the 3850 was EOL at that point).

    Two iterations on and we now have the 5850.

    [​IMG]

    Since the 3850 and 4850, the 5850 has grown longer, and more power hungry, but in turn performance in some situations has more then doubled from the 3850.

    [​IMG]
    The Family Shot

    Performance History:

    Performance wise the initial 3850 was lacking in performance to the 8800GT/S that it competed with, especially when Anti-Aliasing was enabled. Today the 3850 can still play a lot of games smoothly with medium detail, and with a secondhand price of around $40-50, its quite a decent card for the money.

    With the introduction of the 4850, AA and outright performance was bumped a notch which competed quite well with the 9800GT/9800GTX+. The 260GTX/260GTX 216sp however was one step ahead, and competed more with the 4870/4890 series of ATI cards. With the 280GTX/285GTX being the card of choice for outright single card performance.

    The current generation 5850 cards offer a nice performance bump over the 4850, and finally Crysis is near playable on very high with a single card. The 5850 slots in between the 460GTX and 470GTX performance wise, and also price wise, offering great alternative.

    That said 5850

    [​IMG]

    The 5850 1GB GDDR5 Direct CU i have here (Thanks ASUS :Pirate:) is the newer non reference model with ASUS'es own HSF design. My desaturated phase aside, this is the one of the blackest 5850's you can get. Even the rear IO is some dark metal. It also comes with red stripes on black action which seem to be all the rage these days.

    [​IMG]

    You get a large box filled with adapters and a DVD with drivers and manual. Standard stuff, you really don't need anything else.

    The cooler on this thing is pretty quiet, (probally a bit louder then my notebook HDD RAID 0 setup) and cools a bit better then the reference design. It reached a max 67 degrees (ambient of 18 degrees) with a low hum in the background running FurMark. It's no suprise given the two huge heatpipes this thing has.

    [​IMG]
    Heatpipes!

    Enough talk, now to some benchmarks!

    Test setup:
    AMD 955BE @ 3.2ghz which is really STOCK.
    4GB Kingston DDR3
    ASUS M4A89GTD 890GX
    2x250GB WD2500BEVT in RAID 0 (I like...silence)
    Corsair HX520watt PSU

    [​IMG]

    Graphics cards:
    Sapphire 3850 256mb
    ASUS 4850 512mb ROG EDITION
    ASUS 5850 1gb GDDR5 Direct CU

    FurMark
    Settings:
    Benchmark 6000ms with 4xAA / No AA
    Resolution: 1024x768

    [​IMG]

    We see the 5850 taking a huge lead over the 3850 and 4850, OpenGL performance greatly improved over past generations.

    3Dmark 06
    Settings: Pre registered standard tests.
    Resolution: 1280x1024

    [​IMG]

    Here we see a large point increase of nearly 4k with the 5850 from the 4850.

    Crysis
    Settings:
    Game: All Very high
    Resolution: 1920x1200
    Mode: GPU Timedemo

    [​IMG] [​IMG]

    With and WithoutAA enabled, we see a near 2x performance gain from the 4850 in AVG FPS, and almost 6x performance gain from 3850 against the 5850. The 3850 was a very slow slideshow with AA enabled. Would probally perform similar to the 890GX'es 4290.

    DiRT2
    Settings:
    Game: Ultra where possible
    Resolution: 1920x1200
    Mode: In Game Benchmark.

    [​IMG] [​IMG]

    We see a incremental performance gains from 3850 to the 5850. We can particularly see the hit the 3850 takes when AA is enabled.

    NFS SHIFT
    Settings:
    Game: Max Settings
    Resolution: 1920x1200
    Mode: FRAPS measured on Tokyo Circuit 3 laps while keeping 7-8th place (middle of the pack) With alot of bumping and sliding :thumbup:

    [​IMG] [​IMG]

    Again we see incremental performance gains from generation to generation. With AA enabled the 3850 almost becomes unplayable.

    COD:MW2
    Game: Max Settings with 4x AA
    Resolution: 1920x1200
    Mode: FRAPS Measured on the stage: "The Enemy of my Enemy" played through once without dying. :tired:

    [​IMG]

    Here the 3850 with 4x AA was playable, showing that you dont need the biggest and baddest card to play MW2. :thumbup:

    Conclusion:

    As we can see, 2 iterations on from the 3850, there have been large performance gains. Although the X850 series are not the fastest graphics cards you can buy, they are decently priced against their competitors. We also see the incremental performance gains between each iteration quite healthy. Hopefully the Radeon 6000 series will give us a much larger performance gain then what the 5850 did against the 4850.

    [​IMG]
    3850's and its incrementally more powerful spawns.

    I would like to thank ASUS again for lending me this 5850, its been a rock solid and quiet card!

    Let me know what you think guys! doubt this comparison is up to your standards :Paranoid:

    AMD2400.
     
  2. mrpiccolod

    mrpiccolod Member

    Joined:
    Feb 8, 2009
    Messages:
    1,014
    Location:
    Brisbane
    thanks for the review mate, some great pictures there too :)
     
  3. Frontl1ne

    Frontl1ne Member

    Joined:
    Mar 10, 2008
    Messages:
    3,886
    Pretty good review, it's nice to see how far graphics cards have come :p

    I'm surprised, they only lent you the 5850? I thought they would've given you one :shock:
     
  4. tek_01

    tek_01 Member

    Joined:
    Dec 4, 2005
    Messages:
    2,212
    Location:
    SW Sydney
    Good review :thumbup:
     
  5. Archy

    Archy New Member

    Joined:
    Jan 23, 2010
    Messages:
    1,312
    Location:
    Glenhaven, Sydney
    I doubt they'd give a free card to anyone who wants to start posting reviews on a forum. It's a nice review, but I'm sure they don't just give them away.
     
  6. impact_player

    impact_player Member

    Joined:
    Jul 12, 2006
    Messages:
    1,168
    Nice work AMD2400, good read.

    It's a shame there aren't a few more reviews that are like this, because I'm generally more interested in how a prospective purchase will improve on what I've already got.

    I remember that the 3850 was at two points the card to have when you wanted bang for buck; first, when the 8800GT skyrocketed in price in the shortages just after launch, and also when it was cut down to about $110 @ ITESTATE and sold with a decent game when that was cheaper than 8600GT cards were going for.

    When I bought mine in late 2007, I remember some poor sucker in the line in front of me buying an 8600GTS for about $50 more.

    And I still remember being wowed by the 4850 when I got one a week before they were supposed to launch. I almost think that pips the R9700Pro as my favourite card I've ever owned.

    Here's hoping that the GF104 reinvigorates competition between nVidia and ATI as I have my doubts that the 5000 series ATI cards cost any more for ATI to make than the 4000 series did due to improvements in process & good ol' Moore's Law.
     
  7. mevereyn

    mevereyn Member

    Joined:
    Dec 29, 2009
    Messages:
    1,009
    Encore! :lol:

    Actually, that could be quite cool - how about the '70s?

    NO, not those 70s... :sick:
     
  8. mAJORD

    mAJORD Member

    Joined:
    Jun 4, 2002
    Messages:
    9,648
    Location:
    Griffin , Brisbane
    nice comparo

    The 3850 was a great card.. low power consumption, so could be shoved in nearly anything, but 'enough' punch for top games at the time.

    Really limited in 256MB form, but the 512MB model was great value. Still using one in a gaming box in living room, and takes everything thrown at it still
     
  9. hvalac

    hvalac Member

    Joined:
    Jul 6, 2005
    Messages:
    2,179
    Location:
    Parmelia, W.A, 6167
    imo the 3850 isn't playable with 4xaa in MW2, having a minimum of around 4 fps in any game would completely render it unplayable. Other than that nice benches I was actually just wondering how much the ati series cards have improved over the past few generations.
     
  10. impact_player

    impact_player Member

    Joined:
    Jul 12, 2006
    Messages:
    1,168
    Yes, but as I recall the AA hardware was broken on the 3000 series cards - turning it off made a phenomenal difference in performance.
     
  11. mAJORD

    mAJORD Member

    Joined:
    Jun 4, 2002
    Messages:
    9,648
    Location:
    Griffin , Brisbane
    Correcto, Leave AA off they do OK on 1650 Res
     
  12. CirCit

    CirCit Member

    Joined:
    Apr 4, 2002
    Messages:
    143
    Friends dont let friends bench on trashbins

    unless they offer or form part of a cooling solution
     
  13. Archy

    Archy New Member

    Joined:
    Jan 23, 2010
    Messages:
    1,312
    Location:
    Glenhaven, Sydney
    Eh, it's out of the way and there's no harm done.
     
  14. swong

    swong Member

    Joined:
    May 11, 2005
    Messages:
    460
    Location:
    Sydney
    Great comparison review.

    Very useful for those who have an older card (such as myself) and are wondering if it was worth upgrading to a new gen card.

    Not many reviews on the net compare cards between generations, and those who have older tech have to get the figures from seperate older reviews in order to make a comparison.

    Good work :)
     
  15. Aranarth

    Aranarth Member

    Joined:
    Jul 20, 2004
    Messages:
    4
    Location:
    Michigan, USa
    Hey guys, I'm looking at this from the USA and I'm missing the graphs.

    Is there a server issue or something wrong on my end?
     
  16. budzilla

    budzilla Member

    Joined:
    May 6, 2006
    Messages:
    2,669
    Location:
    Right There->
    Probably your end, it's photobucket hosted if that causes problems for you.

    AMD2400, I just finished talking to Asus and they told me to tell you to send out that 5850 to me when you're done with it.
     
  17. IPW

    IPW Member

    Joined:
    Dec 3, 2002
    Messages:
    1,194
    Location:
    Gold Coast
    So I wonder how this compares to my 4670?

    It would shit all over it I'd imagine, even the 4850 should be better yes?

    I still can't justify spending money just so I can have a prettier sc2 experience though :(

    Maybe I could get some crossfire action going or something.
     
  18. LINUX

    LINUX Member

    Joined:
    Sep 18, 2001
    Messages:
    3,062
    Location:
    Dudley, Newcastle
    Interesting that FurMark shows a massive bias towards the 5850 that isn't really repeated in any other test. I was trying to work out if the performance improvements between generations were linear or exponential, as a Moore's Law* adaptation would have you believe. Unfortunately, it looks to be much closer to linear in all the non-AA gaming benchmarks show, except for MW2.

    The raw processing power may well be exponential as rendering one frame requires (overhead)+(processing time), where overhead is the time is takes the CPU to upload a new frame's data to the GPU. Assuming the overhead is roughly consistent between generations (ie: CPU and PCI-E bound) and that it's significant compared to the processing time (say, 30-50% of the time taken per frame) then an exponential increase in raw GPU calculation performance would be required for a linear increase in FPS, so maybe that's still happening.

    I guess it's too hard to make a sound judgement without some seriously biased synthetic tests. Something like fractal calculations or Folding@Home would be a better test for raw performance. Basically a test where the frame processing time is at least an order of magnitude longer than the CPU overhead.

    *Yes, I know it's only for transistor count, hence the "adaptation" qualifyer.
     
  19. mAJORD

    mAJORD Member

    Joined:
    Jun 4, 2002
    Messages:
    9,648
    Location:
    Griffin , Brisbane
    A 4670 is approximatly a HD-3850-512MB ( so not this one tested) and a HD3870 performance, but with better AA perf.
     
  20. miicah

    miicah Member

    Joined:
    Jun 3, 2010
    Messages:
    6,128
    Location:
    Brisbane, QLD
    Those are great looking photos :thumbup:

    Pretty interesting that a 3850 still chugs along with modern games and realisticly who is going to run 1920x1080 with that card? I imagine at 1440 or 1650 the FPS would be very much playable.
     

Share This Page

Advertisement: