Page 13 - Heat, Power and Noise levels

Le pouvoir "The Power"

We'll now show you some tests we have done on overall power consumption of the PC. Looking at it from a performance versus wattage point of view, the power consumption is really not bad. Our test system is a Core 2 Duo X6800 Extreme Processor, the nForce 680i SLI mainboard, a passive water-cooling solution on the CPU, 2GB memory, DVD-ROM and WD Raptor drive. Have a look:

The methodology is simple: we look at the peak wattage during a 3DMark05 session with hefty IQ settings to verify power consumption. It's a good load test as both GPU and CPU are utilized really hard here. Please do understand that you are not looking at the power consumption of the graphics card, but the consumption of the entire PC.

So let's analyze. We had a total system wattage peak at roughly 400 Watts with the HD 2900 XT card, which is really a lot. We simply place a wattage meter in-between the PSU and power socket. It's not the most objective way to test as you have to consider PSU efficiency as well, but it's the best thing we can do, though.

In my view the Radeon HD 2900 XT requires you to have a 500 Watt power supply unit at minimum if you use it in a high-end system, and I think that's barely on the safe side. Also recommended is 35 AMP's on the 12 volts rails for stable power distribution.

I say this because the card itself will use a lot of AMP's on that 12-volts line. If we take the reference peak 215 Watt: 12 Volts that would be 18 AMPs for just one HD 2900 XT. And then you have the rest of your system to feed.

Notice that the card uses one 6-pin power connector and one 8-pin, I'd like to strongly advise a PSU with dual 12-volts rails here. You can also connect 2x 6-pin connectors yet overdrive (overclocking) will be disabled. If you wan to do it really right; check out this PSU review please.

Why does the Radeon HD 2900 XT have two 6/8-pin power connectors? Well, the PCI Express spec allows for 75W from the edge connector plus 75W from each external 6-pin power connector. Then 75 Watts from the PCI-Express bus. The trick here is that the 8-pin connector can draw 150 Watts.

So mainboard 75 Watts, 6-pin connector 75 Watts, 8-pin connector 150 Watts. That's a total of 300 Watts at your disposal for a card that (non-overclocked) peaks at 215 Watts.

If you have dough to spend and opt the guru path of righteousness by doubling up towards two cards in your system -> Crossfire, then you should end up with a 800 Watt or better PSU with a 45 AMPs 12 Volts rail. Again, definitely check out this PSU review, please.

There are many good PSU's out there, please have a look at our many PSU reviews as we have loads of recommended PSU's for you to check out in there. What would happen if your PSU can't cope with the load?:

• bad 3D performance
• crashing games
• spontaneous reset or imminent shutdown of the PC
• freezes during gameplay
• PSU overload can cause it to break down

Les températures de la carte graphique

As with any graphics card we test, we also take a look at heat build-up in the GPU. This is a fairly complicated process to track properly and the one tool to do it with is Rivatuner. Since Rivatuner does not support R600 (yet) we had to make sure that the programmer (Unwinder) got his R600 board quickly. 48 hours prior to this article release he received the board and was able to do his fancy stuff and got HW monitoring available to me. So thanks to our much appreciated programmer Unwinder for his help here!

 Let's have a look. They're all overclocked and the temps are varying very little. We measured at a room temperature of 22 Degrees C.

I inserted a good number of NV products here as well. We only recently started tracking GPU temps in this manner.

Much like NV's flagship product, the 8800 Ultra, the R600 (HD 2900 XT) also shows a new level of hotness. The cooler alone already revealed that it needs to dissipate a lot of heat. In fact the cooler alone uses 25 Watts to operate at 100%

Anyway, the Radeon HD 2900 XT is running really hot. The card was doing temperatures off the chart as it reaches 90 degrees C. Now, much like the Ultra, the weird thing is that the product can deal really well with it. Despite that load temperature it even overclocked much higher, as we'll show you later on. It's however a lot of heat dissipation in your system, though.

The beauty of the cooler design is that this heat is pushed outwards of your PC as exhaust at the connector side.

Noise Levels coming from the graphics card

When graphics cards produce a lot of heat,  that heat usually needs to be transported away from the hot core as fast as possible. Often you'll see massive active fan solutions that can indeed get rid of the heat, yet all the fans these days make the PC a noisy son of a gun. I'm doing a little try-out today with noise monitoring, so basically the test we do is extremely subjective. We bought a certified dBA meter and will start measuring how many dBA originate from the PC. Why is this subjective, you ask? Well, there is always noise in the background, from the streets, from the HD, PSU fan etc etc, so this is by a mile or two not a precise measurement. You could only achieve objective measurement in a sound test chamber.

The human hearing system has different sensitivities at different frequencies. This means that the perception of noise is not at all equal at every frequency. Noise with significant measured levels (in dB) at high or low frequencies will not be as annoying as it would be when its energy is concentrated in the middle frequencies. In other words, the measured noise levels in dB will not reflect the actual human perception of the loudness of the noise. That's why we measure the dBa level. A specific circuit is added to the sound level meter to correct its reading in regard to this concept. This reading is the noise level in dBA. The letter A is added to indicate the correction that was made in the measurement. Frequencies below 1kHz and above 6kHz are attenuated, where as frequencies between 1kHz and 6kHz are amplified by the A weighting. 

We start up a benchmark and leave it running for a while. The fan rotational speed remains constant. We take the dBA meter, move away 75 CM and then aim the device at the active fan on the graphics card.

There's no other way of saying this ... the card is very noisy. With the fan in idle it's really okay. But every now and then the RPM of the fan will quickly spin faster, and that makes a lot of noise. Even in desktop mode this happens. With Vista; Aero will definitely stress the graphics card some more, so I predict you could get irritated by this. AMD has acknowledged this and stated that they will fix it soon.

In Idle you can expect a dBa level of roughly 48. This already is moderate. But as mentioned, the card's fan will spin up even when you are doing nothing, and then we hit a 53 dBa volume level coming from the PC. This is very audible.

We did some stress testing though, and the RPM of the fan will remain at roughly 30-40%. With the help of Rivatuner programmer Unwinder we got a little deeper into this matter. If the GPU core hits 100 Degrees C a safety feature will kick in and the fan will rotate 100%, and man I tell you; its like a helicopter landing in your room. We did a small test. We forced the fan to 30% (desktop mode) at 48 DBa, then ran 3DMark 06. Believe it or not but after a minute the core had reached 100 Degrees C and the 100% fan RPM kicked in. So in theory in warm circumstances (a really hot summer), this might happen to you.

I'll leave it up-to board partners like HiS to solve this issue. They have an excellent reputation with it, but I'm not really thrilled about this reference fan.