ATI Brings back external power supply?

Heh, the latest from the "all knowing" ;) Inquirer:

http://www.theinquirer.net/23040206.htm

We're hearing that the R300 is so power hungry that currently it needs an external power supply unit (PSU) to keep it juiced up, and consumes a rather more than adequate 28 watts.

This could mean any of the following:

*R300, or some variant of it, has always been designed to require external power supply, ala 3dfx Voodoo5 6000.
*R300 on 0.13 micron would not require a power supply, but ATI might be bringing out a 0.15 version that does, due to lack of available 0.13 capacity.
*The R300 prototypes that exist (perhaps on 0.15 micron) require a power supply, while the production boards, 0.13, are not expected to.

Of course, I skipped the usual "It could mean that The Inquirer is just full of sh*t!"
 
May be this thing is coming with a TruPower Unit delivering SmoothCurrent.

At least an internal power connector à la Voodoo 5 wouldn't be the worst idea for high end graphics cards anyway. R200 and NV25 are so power hungry that they need very complicated (and therefore expensive) board designs using the 12V, 5V and 3,3V power supply from the AGP to balance the current over the lines.
 
This is somewhat suprising, ATI is consistantly less power hungry than it's nVidia counter-parts. I suppose since it's being designed by a different team -ATI west- they're not as big on conserving power.
 
The normal AGP slot by itself is specified to be able to deliver 25 watts, which most recent motherboards have little trouble delivering - beyond that, there is AGP PRO (up to 110 watts), molex connectors (~50 watts?), and external power supplies (no real upper limit, but expensive)

An external power supply sounds like total overkill for just 28 watts - IIRC, Voodoo5 5500 drew something like 37 watts, and did just fine with only standard AGP and a molex connector.
 
Assuming the Inquirer story is correct on power consumption:

Molex power for the R300 is a bad idea because a second set of voltage regulators would be needed. The first regulator would be used for 3.3V off the AGP slot and a second would be for 5V from the molex. Not only would the 5V regulators take up extra PCB space, it would also make the board run much hotter.

If the R300 does need extra power, an external adapter wouldn't be the worst way to go. I'd bet if this was the case right now for targetted GPU frequencies, ATi would lower the speed in order to make it run on standard AGP power.
 
Modern graphics cards (Geforce3 and higher) already today require 2 full sets of voltage regulators for voltages that the AGP slot cannot supply directly - one for the GPU core (~1.6V) and another one for the DRAM I/O (2.5V for standard DDR1). So I don't see how that is a problem. And the power loss through a switching regulator is typically something on the order of ~10%, which would amount to ~2-3 watts in this particular scenario.
 
I wouldn't read too much into that news blurb. If it's true they probably just have something wrong with the samples. Some transistors are probably switching when they shouldn't be.
 
Back
Top