Amazing power characteristics for ATI RV530

[John Howard]

Hi guys, long-time lurker that finally joined SPCR.

Have a look on the upcoming ATI RV530
http://www.anandtech.com/video/showdoc.aspx?i=2532&p=2

Perhaps one of the more interesting aspects of RV530 is the low power consumption. Reference 64-bit boards of RV530 using GDDR3 at 500MHz core clock speeds have a power consumption of just 25W. 128-bit boards were estimated at 40W and X1600XT boards with all the goodies were estimated by AIBs to be 55W.

25W!! WOW

Finally we can have a Silent video card that capable of outputting Dual DVI or HDMI. Perfect media center and gaming card.

[darthan]

Ummm, I don't see what is so special. That 25W card is not going to be much of a gaming card. Beat the Radeon 9250? yes. Beat the Nvidia 7200 (if they release one)? No. The X1600XT and any 256-bit cards they release should be great gamers in terms of price/performance especially but the 25W card has an utterly crippled memory interface. It will do dual monitors just fine (hasn't every card for the last few generations?) but getting a current low end dual-dvi card will do you just as well for media center capabilities. If you're looking for good gaming capability you really have to look at the mid range cards. Of course, a 55W gaming card ought to be silenceable by the experts (and the not so experts) around here.

[John Howard]

Well, I got my eyes on the X1600 XT 60W



As you can see the 7800GTX consume about 113W and the x850xt PE 129W comparing it to the 60W of the x1600xt.

That's amazing in my opinion, considering that this card might outperform 7800gt and x850xt PE.

Let's not forget that this card has AVIVO and H264 acceleration.

[Elixer]

Your method for calculating power is incorrect. More than just the video card is loaded when a graphics benchmark is run. Also, video cards use power at idle.

[John Howard]

Your method for calculating power is incorrect. More than just the video card is loaded when a graphics benchmark is run. Also, video cards use power at idle.

You're right they use power at idle. I'll try to find another chart.

[rpsgc]

Try this one



60W for the X1600XT is a lot!

[John Howard]

There must be something wrong with the xbitlabs chart.

While generally our power consumption measurements are very precise, this time we would recommend to consider them cautiously, as NVIDIA GeForce 7800 GT graphics card behaved very strangely during the measurements. Our multimeter, indicated that the graphics card consumes nearly nothing from 3.3V lines (about 2.1W), which is something that we see for the first time.

Also Nvidia state that the power consumption of their 7800GTX is 110W

http://www.gamergod.com/article_display ... cle_id=231
http://www.digital-daily.com/video/gefo ... ndex02.htm
http://www.cdrinfo.com/Sections/News/EM ... wsId=14260

Besides, you can tell from the previous chart that the power consumption of the 7800GTX is more than 113W not less. The difference between idle and load is 113W and we do not know how much watts the 7800GTX consume at idle. So it should be 113W + idleWatts = total consumption.

Can anyone shed some light on this?

[bredbored]

Also Nvidia state that the power consumption of their 7800GTX is 110W

I think they actually state that the consumption *could* be as much as 110W in some unspecified circumstance. The xbitlabs chart is a measure of *actual* power consumtion running a *specific* benchmark; it is not necessarily equivalent.

Besides, you can tell from the previous chart that the power consumption of the 7800GTX is more than 113W not less. The difference between idle and load is 113W and we do not know how much watts the 7800GTX consume at idle. So it should be 113W + idleWatts = total consumption.

John, can you re-read this and check your logic, because I can't see how you're reaching your conclusions. You have multiple variables in the equations and don't know what value to put in any of them - for example, to simplify - the xbitlabs chart tells us three things;

1. idleCPU + idleMemory + idleGPU= 127W
2. loadCPU + loadMemory + loadGPU=240W (where "load" represents a specific benchmark)
3. maxGPU=110W

So, all we know is that

a. loadGPU<=maxGPU=110W
and
b. idleCPU + idleMemory + idleGPU + 113W = loadCPU + loadMemory + loadGPU

i.e. nothing very conclusive about loadGPU.

Kingsley