1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Hardware Nvidia (Zotac) GeForce GTX 285 1GB

Discussion in 'Article Discussion' started by Tim S, 16 Jan 2009.

  1. RotoSequence

    RotoSequence Lazy Lurker

    Joined:
    6 Jan 2004
    Posts:
    4,588
    Likes Received:
    7
    Maybe I missed it in the article, but why does the actual core have a heat spreader on it? I thought they avoided those on GPUs (when most people don't and aren't supposed to swap out the HS since its a factory job) since chips can get the heat off better without them (silicon -> paste -> metal -> paste -> metal is less efficient than silicon->paste->metal)...
     
  2. Tim S

    Tim S OG

    Joined:
    8 Nov 2001
    Posts:
    18,881
    Likes Received:
    78
    I'm still working on getting readings from other cards as I spent all of Friday recreating the issue in a completely different environment to my own (it's not just limited to my test systems), but I've run a 280 and a 260-216 and the problems are still there, but not nearly as profound. There's a lot of DC component noise (the delta is about 2V at 4.3KHz, which is pretty high) and there's more than what I'd call typical resonation coming from both of the 6-pin PCIe power connectors.

    The resonation is heading straight back up to the power supply and causing it to sing in a rather erratic fashion in some scenarios. I think it's a problem with how the PWMs are switching between modes because the pic I posted on Twitter is actually when the card is switching states (idle/load) - the sound you get out of the PSU is akin to something you'd associate with 'EXTERMINATE!!!1', or that's the easiest way to describe it at the moment.

    The problem is that this high frequency noise being pushed back to the PSU could actually cause it to shut down (a poorer quality unit may pop) - at least, that's how the issue on the 280 was discovered originally by some system builders in Eastern Europe. I believe it's an issue with the firmware on the card, but I could be wrong - I'm still waiting for Nvidia to acknowledge that there's a lot of component noise coming from the PCIe power connectors, but they've gone a bit quiet on me since reporting the issue.
     
  3. Tim S

    Tim S OG

    Joined:
    8 Nov 2001
    Posts:
    18,881
    Likes Received:
    78
    The core is huge in simple terms. I reckon it's about 412mm² based on the 576mm² 65nm GT200 and directly shrinking to 55nm, but Nvidia hasn't gotten specific - the heatspreader essentially prevents it from chipping. Bigger pieces of silicon are generally easier to damage, so it's a layer of protection from a heavy-handed enthusiast (like Harry :p).
     
  4. konstantine

    konstantine New Member

    Joined:
    18 Jan 2009
    Posts:
    51
    Likes Received:
    1
    Tim. I have some preservations about the review.

    Firstly, In Grid, the 4870 gets lower fps in this review than in the original 4870 1Gig review.I mean from 78 fps at 1920 res. on max setts to 69.4 fps in this review at the exact same setts and res.While the 280gtx ups from 66 fps to over 70.4 fps in this review.
    So what's the deal with this decrease in the performance of the 4870 in Grid?

    Secondly,From my personal experience, i can tell you that crossfire doesn't scale in crysis until you hit 1920 res. on max setts with 4x aa.so how did you get X-fire to scale that well at 1680 res.?
    And lastly, I wonna thank you for the review.
     
  5. Tim S

    Tim S OG

    Joined:
    8 Nov 2001
    Posts:
    18,881
    Likes Received:
    78
    Hey, thanks for signing up to comment! :)

    We have started testing GRID with a different track where we found that relative GPU performance scaled much better and is also a lot more consistent. That is probably the reason for the perceived drop in performance. Nvidia has also done quite a bit of work in its drivers between this review and the original 4870 1GB review - release 180 increases performance quite a bit.

    With CrossFire, the simple answer is I don't know. The more complex answer is going through the process of getting those numbers. We did a fresh Vista SP1 installation on the test system and cloned the hard drive once we had installed all of our games, save games, test applications and OS updates - we have not installed any graphics card drivers before cloning.

    This means that the system is in exactly the same state before testing any graphics card - we literally install a new graphics card (or pair for CrossFire/SLI) and then install the driver. We then reboot, set the resolution, turn Aeroglass on, check that everything is running at the correct settings and then begin running our benchmarks. Each benchmark is run at least three times and the values are checked, double checked and then triple checked to make sure they fall in line with our expectations. If they don't, we'll re-run the tests and hunt out the bugged scores, passing information back and forth with the hardware vendor until we work out the bug (we don't pass our demos onto third parties to stop 'cheating' even though the industry is hopefully past that nowadays).

    It's as simple as that. :)

    In the past, we had a lot of scaling problems with CrossFire and SLI because we were using a different method to what I've described above (it involved using system restore and removing registry entries for old driver installations), but I'd definitely recommend reinstalling if you go from one to two GPUs. The problems associated with uninstalling/reinstalling drivers disappear as soon as you don't do just that.

    Hope this helps,
    Tim
     
  6. konstantine

    konstantine New Member

    Joined:
    18 Jan 2009
    Posts:
    51
    Likes Received:
    1
    Thanks Tim.
     
  7. biebiep

    biebiep New Member

    Joined:
    12 Dec 2007
    Posts:
    101
    Likes Received:
    3
    Man, looking at these numbers. The 8800GTX was my BEST buy EVER :').
     
Tags: Add Tags

Share This Page