12V amperage and voltage vs. PCIe link width

[CoinOperatedBoy]

tl;dr version: Do I need 26A on the 12V rail to run a recert 9800 GT? My OCZ GameXStream 700W has 18A per rail. I realize that total 12V amperage is probably 56A-ish. Regardless, this card is stuck at 1x PCIe link width in my current setup and I don't know why.

Long version:
Unsure whether I should throw this in the video card or PSU forum, but I figured I could get a more detailed answer here. I recently had an 8800 GTS 320MB die on me and EVGA replaced it with a recertified 9800 GT. I've had sporadic performance with the new card, and I recently started getting "Attempt to reset the display driver and recover from timeout failed" blue screens related to nvlddmkm.sys. There are many suggested fixes for this floating around and I've tried most of them. When I am actually able to boot into Windows, the card is almost always running with PCIe link width 1x instead of 16x.

I had this happen with another card before when I didn't have the PCIe frequency locked to 100MHz, but that's not the issue this time. I've tried adjusting the frequency both above and below 100MHz, leaving it on auto, and giving the PCIe voltage a bump in BIOS. Nothing gets me back to 16x link width.

EVGA suspects the power supply. I have an OCZ GameXStream 700W and the rest of my system specs are in my sig. The wattage should be more than sufficient, but the specs for the 9800 GT demand 26A on the 12V rail... This OCZ shows 18A for each of four 12V rails. I've monitored my voltages via software (HWMonitor, OCCT) because BIOS only shows "OK" instead of actual values and I don't have a DMM. My 12V voltage fluctuates pretty widely between idle and load (from 12.42V idle to about 11.75V under heavy load), but the values fall within ATX specs (11.4V to 12.6V).

So my question is: Do I honestly need a PSU with 26A on the 12V rail to power this thing? The 8800 GTS that this card replaced has the exact same power requirements and I ran it fine with this PSU for over a year and a half, unless the PSU was what finally killed it. Also, would my voltage fluctuations be severe enough to cause instability?

 

[HOOfan 1]

I'll sing the same old song.

The 26A suggested rating, is not for the card itself. It is for the system as a whole. EVGA/nVidia try to think about how much 12V power a fairly high end system running the 9800GT would use.

The 9800GT has a single 6 pin PCI-E connector. 6 pin PCI-E power connectors should pull no more than 75W or ~6A from the power supply. Then the 9800GT is allowed to pull up to 75W, or another ~6A from the PCI-E slot on the motherboard. That is no more than 13A, and the PCI-E slot on the motherboard and the 6 pin PCI-E power connector are pulling from different 12V rails to begin with.

Lastly, if you were trying to draw more power from a single rail, than that rail is capable of pulling, the Power Supply would shut down, NOT show instability problems. The current limits on those rails are set by circuits, which shut the power supply down, if too much power is drawn. If those circuits were not there, you would be able to draw a full 50A or 600W from any single connector. Since it is circuits that are designed to limit the current draw and not some sort of limited capability of the components, you cannot blame instability problems on the rail divisions of your Power supply.

Edit: Unless you measure your voltage with a proper multimeter, then you may as well ignore them.

If your voltage fluctuation was indeed that high, then it is horrible. However, software readings are usually wrong, and the farther you get from nominal voltage, the more wrong the readings become.

 

[CoinOperatedBoy]

Thanks, I appreciate the response. I understand that power requirements are usually estimates for whole system usage and that 26A would not actually be devoted to the 9800 GT. I read the sticky, I get that this PSU probably does not actually have multiple sources and the combined 12V amperage exceeds the recommended total system-provided amperage. I guess my confusion stemmed from there being multiple 12V rails rated at 18A each and I wasn't sure if that would affect the card's behavior in any way because I didn't know what the card's actual current draw would be. So thank you for confirming that the rail split would not have caused my issues.

I'll try to track down a multimeter to take accurate readings. However, if the software measurements were even remotely correct and my fluctuations were that pronounced, would you say it's possible that that could cause some of my symptoms?

 

[HOOfan 1]

Originally posted by: CoinOperatedBoy

I'll try to track down a multimeter to take accurate readings. However, if the software measurements were even remotely correct and my fluctuations were that pronounced, would you say it's possible that that could cause some of my symptoms?

I suppose, if EVGA really thinks it is the PSU. It could be other things too though, like bad memory, bad motherboar, a faulty card.

 

[Modelworks]

Buy a meter. Nobody working with pc hardware should be without one. It is as necessary as a screwdriver
You can get one at walmart or any auto parts store for under $20.

 

[CoinOperatedBoy]

Thank you both for urging me to get a meter. I picked one up and checked my voltages as directed in the forum sticky. My 12V voltage actually fluctuates between 12.27 and 12.34 depending on load, so you were dead on about the software monitors being drastically inaccurate. It looks like amperage and voltage fluctuations are not causing my video card issues. I'll see what EVGA recommends from here. I appreciate the input.