According to an article at Dailytech, a system with 8800 GTX draws only 13W more at load than one with X1950XTX.
http://www.dailytech.com/article.aspx?newsid=4812
They didn't mention when was that power consumption measured, in 3DMark or some game. And haven't tested it with Oblivion.

weird...doesn't it need 2 6-pin power connectors?

is daily tech meant to be a reliable site?


the half life 2 lost coast numbers look waaaay off

i have an 1950xtx and x6800 combo and i get 97fps on half life 2 lost coast at 1920*1200 (full detail 16*AF 4*AA - 6*AAhas some rendering bug.. )

there score is 60fps at 1600*1200

sounds really suspect.....

The prototype/preproduction model had the two connectors. It's quite possible they improved efficiency for a newer revision. Wonderful news if it's true.

I can still see the TWO power connectors in the pictures. Here is the link:
http://images.dailytech.com/nimage/2873 ... _G80_3.jpg

Hmm... here's a crazy thought...
What if the second power connector is actually just a power pass-through for a second card; like an SLi Master/Slave Edition or something?
300w power draws don't make any sense at all, but making it easier to connect up two cards certainly does.
Hmm again...

probably just needs more current i guess, and might have put a strain on lower end psus without quad 12v lines

Probably just a lame attempt at marketing. "Ooo, my nvidia card is sooooo powerful it needs 2 6pin power connectors!" My psu is bigger than yours!

Or maybe they anticipate that most power supplies like thermaltakes don't have enough power for two of these monsters so you have to buy a power supply that has 4 6pin connectors to make sure users don't use it on thermaltake psus.

the difference in your system and the test system is the CPU. They used the new Core 2 Extreme QX6700. you have a X6800. Games are not yet multi-threaded enough so it becomes Ghz dependant. The x6800 has a higher clock speed than a QX6700.

also the fact that different demos might have been used.

How does drawing 13W more than the most power hungry card on the market (X1950XTX) come to be qualified as "not very power hungry"? The 50W extra it draws at idle doesn't make it look too good either.

You must be joking, right? It was expected to draw close to 200W, making it very difficult to cool quietly.
The X1950 XTX is listed by Xbitlabs as 125W, so the 8800 GTX should be ~138W. Considering the beefy cooler it should run reasonably quiet.
Being a card that's 60-90% faster than the X1950 XTX it has good performance / watt. You won't be able to cool it passively, but did you expect that?
IMO it's a good sign for future mid-range cards and even high-end ones when manufactured on the 65 nm process.

half life 2: lost coast is a demo there's only one way of benchmarking it (which is what i used)

admittedly there is a slightly lower clock on the cpu but as the results show its not cpu limited (as evidenced by the huge difference in the nvidia numbers) i would say this shouldn't be making the difference in the results...

my personal feelign is maybe there's some problem with the ati board and nvidia mb (i remember there was a problem with intel 975x boards and the nvidia 7950 cards) thats makign it run slower.

i emailed the author...haven't heard back.

i'm worried about the sheer size of this thing, it's pretty tight with my x1950xtx and antec solo...

at least the power connectors seem to be in a better place...

i wonder if they'll every get sli wortking with 975x

maybe they did it on vista (for dx10 support?)

Maybe some psu's while having enough power, don't have enough power on a single rail to power the card.. thus allowing you to use two different rails to power the card..

I know very little about psu's but i can see logic in the setup if that were the case, what do you think ?

Someone explained the 2 PCI-e connectors in the comments after the article. Truthfully they only really needed one PCI-e connector, but as a safety issue they put in 2. By regulation there can only be 20A max going through one 12V line. So I guess at some point it might hit close to 40A? So it might not be so Watt hungry but it's probably still very Amp hungry.

Or maybe the second PCI-e connector is for the built in physics card. Personally I'm stick with ATi all the way! It's still CANADIAN to me! lol

It's a 12 volt line. Power = Amperage * Voltage, so if it's amp hungry, it has to be watt hungry too.

Taking 20 A is a thermal disaster already (240 watts is significantly more than the hottest Prescott), there's no way in heck this thing gets anywhere close to 40 A.

ehm not W hunger but A hungry?? ehmm i dont know but last time i did physics W and A and V are all related:D

Eh, if it's amp hungry it's watt hungry.. P= U * I. 40A from 12V would be 480W.

18 or 19A is enough to warrant a second connector (because not all PSUs can actually provide the rated 20A on a single rail - safer to design with cheap parts in mind). If it were approaching 40A they would have installed a 3rd power connector.

This is what I've been saying...

High Five! (You all have to see the Borat film, it's great)

if the numbers pan out and the noise is no worse than my x1950xtx (which is pretty good at idle, terrible under load), i'd get 1 or 2 of these in my system...

A test was made in my country: E6600 @ 3.6 GHz and AtiTool + dual Prime95 on an 8800GTX stock, peak 331W.
PSU Sirtec 500W.
Silent cooling and very hot.
If we consider ~ 70% PSU efficiency =>232W delivered by PSU. Aprox. 100W for CPU, MB, MEM, HDD and we remain with 132W, just a little bit more than an X1900XTX.
I say a thumbs up for nVidia for power consumption for this kind of GPU.
BTW the card is severe bottlenecked even by a E6600 @ like hell.

Below a picture with an Kentsfield @ 3.3 GHz, 8800GTX @xxx/xxxx (NDA in action here ) on Phantom 500.

Nice info, thank you Burebista! I'm not planning to buy it, since it wouldn't fit in my P150 (why, oh why, didn't Antec make the P150 just 3 cm longer?).
Question: how does the 8800 GTS behave? It should be 40% slower and at least 40% 'cooler' than the GTX. Would it feel bottlenecked on my new C2D E6600?

Basing to that preview it consumes ~38W (45W * 0.85) more than x1950XTX which means ~70W as idle consumption (X-Bit Labs got 33W for X1950XTX) which is just ridiculous!
Even more when you compare that to TDP values of Conroe CPUs!

Also remember that new features of DX10 apparently differ quite much from DX9 so untill we have good DX10 stress software that measured power consumption means very little for max consumption.

That's a good point, because AFAIK the geometry shaders are not used by any current software. But DX10 will only be available on Vista.
Once the OpenGL extensions for geometry shaders are going to be exposed in the drivers those will be available under XP too.
My guess is that those won't contribute a lot to the power consumption. Most of it should be because of the highly clocked vertex / pixel shaders.

I found a first article on performance and other stuff:
http://www.theinquirer.net/default.aspx?article=35604
Apparently the geometry shaders are executed on the stream processors too.

AFAIK there's still an embargo on it today...

165W as peak consumption.
http://www.beyond3d.com/reviews/nvidia/g80-arch/index.php?p=03
Apparently still with DX9 stress tools so still higher is possible...

TDP of card is 185W.
http://enthusiast.hardocp.com/article.html?art=MTIxOCwyLCxoZW50aHVzaWFzdA==


I would expect embargo continue at least to next month when it comes to availability.

Some sites are saying that the 8800GTX is surprisingly quiet.. be interesting to find out just how quiet.