Holering wrote:
@ obobskavich:
There's no way a 5800 card can run cool enough without unbearable noise, unless you add a third party cooler (zalman comes to mind). Guru3d.com has a review on the original 5800. Pretty rare card but it does suck.
When I say the card runs passively, I mean the cards runs passively (as in, silent). The following #s are based off of my card: With room ambient at around 17-18* C the card reports the GPU at around 30-40* C in passive 2D mode (the board sensor, which I believe to be fairly close to the GPU, usually reports a few degrees above the motherboard's reported case ambient, until the card goes into 3D at which point it reports pretty close to the GPU's reported temperature - I've touched the board while its running 3D and I'm skeptical the "board" is at >50* C). If it switches into 3D (e.g. start up a game) the fan will turn on, and I've never seen it go much over 60* C even looping 3DMark or Aquamark - normally it runs in the mid-50s (most games don't apply a constant 100% load to a GPU, the load goes up and down, and the temperatures go up and down in response). It runs cooler than my 5900XT (which uses a single-slot cooling solution similar to what some GF4 Ti cards used), cools down faster than my 5900XT, and is quieter at idle than my 5900XT (it's also faster than the 5900XT, at least by a little bit).
Under load it's a toss-up as far as which is potentially more intrusive - the 5800 Ultra has a consistent fan noise, the 5900XT's fan adjusts itself up and down in response to heat. When they "start out" the 5900XT is quieter, but by the time the 5900 gets into the 60-70* C range its fan is running at 100% and is whinier (it has a smaller fan). The noise is on-par or quieter-than most modern cards (Radeon 4870X2 is much louder and runs much much hotter, for example), but unlike most modern cards the fan isn't cycling RPMs - the fan is either all-on or all-off so it's easily enough tuned out. When it drops 3D clocks it goes back to passive cooling after a 2-3 minute cool-down run (and it is a timed delay). I stated in another thread that the WildcatVP 880Pro (all the time; its fan never turns off) is louder than the 5800 Ultra at load, and I did mean that. The "unbearable noise" complaint really isn't anything I've ever understood - I will say that a lot of the acoustic measurements you will find in old reviews should just be thrown out with the bath water though. Usually there is no weighting, calibration, etc associated with them - for example I found a TechReport measurement that asserted both FX 5800 Ultra and Radeon 9800 Pro ran at around 48 dB (at unknown distance) when idle. That's hokey imo - the 5800 Ultra is silent, and the 9800 Pro is pretty-near silent (you have to work to hear that fan (at least on my 9700P you did)). So that means they're probably picking up background noise in their office (or the computer's case fans, PSU fans, etc), not the cards themselves. Modern acoustic measurements tend to be at least somewhat better, but still usually amount to little more than just holding an uncalibrated SPL meter "near-ish" the computer with other associated fans/etc running (yes, taking accurate SPL readings is hard work, which is why I'm providing relative examples and not trying to pin it down to dB values).
As far as aftermarket cooling, the 5800 Ultra (I can't speak to the 5800nu; from photos it looks very similar to the 5900 Ultra though) has a relatively unique cooling mount, and replacing it with something else would likely require at least a modicum of customization. IXBT also cautioned against removing Flow FX from a working 5800 Ultra, as parts of the contact plate are reportedly glued or adhered to the card and there apparently is a risk of unintentionally de-lidding the NV30 itself, or damaging the RAM chips. Personally I think Flow FX does a fine job in terms of thermal performance and it also provides some rigidity to the board (so there's no "warping" of the PCB), and I have no cause to disassemble it as a result; 3D noise could always be lower (lower noise is always better, isn't it?), but it isn't "unbearable" by my standards, even in a relatively quiet residential setting. Yes, you are likely going to be aware that the 5800 Ultra is running in 3D (here's another thing that a lot of reviews don't mention: you usually aren't playing games with the sound muted; having the radio on in the background is usually enough to drown out the 5800U's fan IME, if you were blasting away in UT or Quake with headphones you probably wouldn't even notice the card), and yes it would obviously be "better" if it (or any card) ran passively through its entire duty cycle, but that isn't real-world. I think it took a very hard rap back in 2002/3 mostly because it was the first card to use a 2-slot blower, and compared to mostly passive competitors, it stands out quite a bit. By modern standards, however, it would likely sit more towards the middle of the pack, and probably be praised for being able to cycle the fan off during 2D operation.
Performance is well documented in various reviews, and for DX7/8 it does just fine (in some cases it will actually outrun the FX 59x0 and Radeon, which TechReport speculated was due to the higher clocks influence on TnL emulation). I personally wouldn't lean on a GeForce FX *or* a Radeon 9 for serious DX9 gaming. For that I'd much rather have the 8800GTX you mentioned earlier. 😀
To vetz: I'd add that some Radeon 9 cards are universally keyed, but some are not. On the 3.3V key conversion for GeForce 6, do you know if it's dictated by the HS-bridge? (that is, can the 6600GT be converted to universal, or is it only the AGP-native 6800s?)