Titan X review - worth looking at for new power draw testing

[wvjohn]

http://www.pcper.com/reviews/Graphics-C ... umption-Te

of course the card is stupid fast, but they handbuilt a PCI "riser slot" to measure the voltage.

I never realized that you could stress a mobo by oc on a vid card.

[FlyingPenguin]

Yeah, they and another site built rigs to test PCI-E current draw a back in June when someone else did some tests that seemed to suggest a the AMD RX480 was drawing too much current from the mobo. Turns out it was, and that forced AMD to write a driver revision that better balanced the load by drawing more from the direct power supply cables.

Seems like the power controller in these cards can draw power as needed from any connector, but the RX480 was exceeding the max rating for the PCI-E bus. No one fried a mobo that I'm aware of because modern mobos are actually over engineered, but it's conceivable that over a long period of time it could have melted the connections or damaged the onboard power electronics.

So since no one ever looked for these things before, and they already have these rigs, they look for it now.

Here's the original article on the RX480 on PcPer.com: http://www.pcper.com/reviews/Graphics-C ... eon-RX-480

The guys at PcPer are great at this kind of stuff. They were also one of the first to test framerates by using external hardware (framerating) instead of relying on an internal software benchmark.

Pcper also has a really great hard drive guy, Allyn Malventano, who really knows his stuff. I go out of my way to read any of his reviews.

I like the guys at PcPer so much that I setup a Patreon account to support them with a donation every month.

[Err]

I didn't know that the voltage through the PCI-e slot increased when overclocking a card either. You'd think manufacturers would force the extra power from the power supply.

[FlyingPenguin]

It's all done through the firmware. The onboard voltage controller can pull power from any of the sources, and usually balances things. Trouble is that there's this trend lately to sell a card as "lower power" and to convince you of that they reduce the number of power connectors or go to smaller ones, but then they ride a lot closer to the current limits on everything. Also, it seems, sometimes they fudge the actual current specs.

What gave away the problem on the RX480 was that the card seemed to be drawing more power than it was rated for (it was). Since normally all you can do is measure the total system wattage at the PSU input on the mains side (and fudging a deduction for what the rest of the PC is drawing) there was no accurate way to tell exactly how much current the vid card alone was drawing, until they came up with these rigs that measure the current at all the connectors including the PCI-E slot. Then they discovered that not only was the RX480 drawing more current than speced, it was also drawing too much from the PCI-E slot.