No molex connector on X800 PCI-Express?

[tahrikmili]

All photos I've seen so far seem to support that.. am I the only one noticing? Isn't nVidia working on a PCI-Express part with a non-standard molex connector?

*scratching my head at nVidia*

Yalaz Ozkanli

 

[AlphaWolf]

It seems that there are some of both from ATi. One with the 2x3 connector and some without any connector at all. I expect that the pro will not require, but the XT might.

There have also been pics of a 2x3 to molex adapter so it doesn't seem it will be an issue where you need a new psu.

 

[Lezmaka]

The 2x3 is on some newer power supplies, it's just like the 12V 4 pin that was added for Pentium 4 motherboards, but with 6 pins so people don't get confused. I believe Anand has a picture of the adapter in their one of their NV45 articles, and you plug 2 standard drive molex connectors into it.

 

[Tim Murray]

hmmm... so you could say that the PCI Express X800 XTPE takes TWO molex connectors... just like the 6800 Ultra, eh?

(okay, that was stupid. I apologize.)

 

[Dave Baumann]

http://www.beyond3d.com/previews/ati/pcie/

 

[MasterBaiter]

Doesn't the PCI-X (x16) interface supply around 75 watts of juice? I could see maybe the Pro not needing it, but I think the XT is borderline. :?

 

[Elminster]

Doesn't the PCI-X (x16) interface supply around 75 watts of juice? I could see maybe the Pro not needing it, but I think the XT is borderline. :?

that's right. xt is need ,pro is none

 

[gokickrocks]

http://www.vr-zone.com/?i=861&s=1

seems it doesnt require the connector to be plugged in...

 

[WaltC]

I would prefer to have an external power bus connector, myself. We've come a long way since having to depend on motherboard power regulation circuitry, which was a pain for 3dfx and nVidia both at one time, and I'd hate to go back to those days--especially now that the 3d cards require so much more juice than they did 4-5 years ago. I consider an external connector a plus (just not two of them, of course, but nVidia shouldn't need two for the 6800U PCIe version...)

 

[Entropy]

What's interesting is that PCI-Express extended the allowable power draw over regular AGP. Given how early the gfx IHVs were reported to be collaborating and having input into the specification, it would appear that they simply didn't anticipate how much power their products would come to require. Or it may be that the PCI-SIG didn't wan't to make designs capable of supplying 150-odd Watts of power over the connector mandatory. I'm betting on the lack of anticipation though, as extending the power spec of the PCI-Express connector would be pointless if the IHVs figured they would connect straight to the PSU anyway.

 

[JCLW]

Does this mean that every PCI-E slot on the board will be capable of supplying 75W?

The part of the PCI-E connector nearest the edge of the case does all the power/SMbus/etc stuff, and the other part carries all the data.
On side A pins 2 and 3 carry 12V power, and pins 9 and 10 carry 3.3V power. On side B pins 1 and 2 carry 12V power, and pin 8 carries 3.3V power. So a total of four 12V power connectors and three 3.3V power connectors. I wonder how much power is available on the 12V pins.

Anybody know how much current the average motherboard trace can handle? I wouldn't think it would be much over 1A.

edit: found a document that suggests a 1mm trace will carry 3A 50cm.

 

[Dave Baumann]

So a total of four 12V power connectors and three 3.3V power connectors. I wonder how much power is available on the 12V pins.

*sigh* We'll try again shall we - http://www.beyond3d.com/previews/ati/pcie/

Presently a PEG x16 connector can supply up to 75W to a graphics card (9.9W on 3.3V and 66W on 12V), however high end graphics cards already exceed these power specifications. The connector is under development by the SIG in order to provide a defined standard behind providing extra power to PCI Express graphics card

 

[vb]

What's interesting is that PCI-Express extended the allowable power draw over regular AGP. Given how early the gfx IHVs were reported to be collaborating and having input into the specification, it would appear that they simply didn't anticipate how much power their products would come to require. Or it may be that the PCI-SIG didn't wan't to make designs capable of supplying 150-odd Watts of power over the connector mandatory. I'm betting on the lack of anticipation though, as extending the power spec of the PCI-Express connector would be pointless if the IHVs figured they would connect straight to the PSU anyway.

They didn't draw these specs for 5% of the market or what 9800/5900U holds. they did it for mid range cards.

it should be a shame if 5700 PCI-E still requires it.

 

[WhiningKhan]

Anybody know how much current the average motherboard trace can handle? I wouldn't think it would be much over 1A.

edit: found a document that suggests a 1mm trace will carry 3A 50cm.

I would expect using power plane rather than fixed-width traces.

 

[Entropy]

What's interesting is that PCI-Express extended the allowable power draw over regular AGP. Given how early the gfx IHVs were reported to be collaborating and having input into the specification, it would appear that they simply didn't anticipate how much power their products would come to require. Or it may be that the PCI-SIG didn't wan't to make designs capable of supplying 150-odd Watts of power over the connector mandatory. I'm betting on the lack of anticipation though, as extending the power spec of the PCI-Express connector would be pointless if the IHVs figured they would connect straight to the PSU anyway.

They didn't draw these specs for 5% of the market or what 9800/5900U holds. they did it for mid range cards.

it should be a shame if 5700 PCI-E still requires it.

Not too sure about that. Cards exceeding AGPs 25W limit has been sold for less than two years (in consumer space). At the time the PCI-Express spec was laid down, I'm not at all sure that ATI and nVidia foresaw that competitive pressure coupled with process woes would push power consumption to the levels we see today.

Nobody likes having to design to fit a very wide power envelope - it's a pain in the ass. The existance of high power consumer gfx cards (and CPUs for that matter) adds cost and design restrictions to all upgradeable computers, regardless of whether the machines actually ships with those power hogs. (That's why I put in the other option, that the PCI-SIG might have refused a higher power specification because they wanted to exert a pressure on the gfx IHVs not to go too far in an undesireable direction.)