Single-Card Performance Testing (ATX Case)
We begin our benchmarks we start with a single card in our big, well-cooled ATX case. This is the best-case scenario for any video card, but it's best for open-air cards, as their waste heat is quickly diluted and exhausted out of the case. Below you can see the interior of this ATX setup, with a single GTX 1080 Founders Edition card installed.
We aren't going to be presenting idle numbers, as they just aren't that interesting. The open-air card is always silent (and we mean that quite literally, as the fans shut off below 50 °C). The Founders Edition card does not have a zero-RPM mode, likely because its enclosed shroud could potentially cause unnecessarily-high thermal conditions at idle. It's designed to push waste heat outside of the case, so it wouldn't do a great job simply dissipating it without air movement.
And this all leads us to our first benchmark, running a 4K Stress Test with our GTX 1080 cards running at reference speeds. The first thing we want to point out, and this is really important, is that our cards are set to the exact same clock rates. Despite this fact, they do not run at the same clock rates, nor do they perform the same. This is due to the very real effect of thermal throttling, which is built into the Nvidia Pascal design. Nvidia became extremely aggressive with its Boost mode with the Pascal generation of GPUs, in part because the architecture of Pascal just wasn't all that different from Maxwell that came before it, so to provide the performance gains that enthusiasts crave (and that justify high prices), Nvidia rusn Pascal at the ragged edge of stability. That's how we jumped from an average factory clock rate of 1200MHz in 2015 to 1750MHz just two years later. The smaller 16nm manufacturing node in use with Pascal is in part what allows this tremendous jump (and the nearly-linear 50% boost to performance over previous generation cards with the same basic layout), but Nvidia's aggressive Boost is likely just as important, because it allows cards to hit very high clock rates when thermal conditions allow, along with serious down-clocking when they do not. We know people can get all up in arms about throttling, but it's just something you're going to have to live with in modern GPUs if you want maximum performance in short bursts.
So, here we see that our open-air ACX card is superior nearly across the board. The only area where it falls slightly behind is in CPU temperature, where our massively-overclocked, water-cooled eight-core CPU runs just 1 °C, which in the end is completely meaningless. Under no circumstances would the Founders Edition card be the ideal pick in this situation. The ACX performs faster despite being set at the same clocks due to the Founders Edition hitting the 80 °C throttle point. And because its fans can run slower, it's much quieter. In fact, we should note that it's running at the noise floor of our test system - the CPU's liquid cooler and the case's four fans are actually what account for the 40 dB sound reading.
When we push our cards to a +190Mhz overclock (which amounts to around 12-13%), the situation barely changes. The cards are still running their fans at the same levels, and only the ACX card increases in temperature. Even so, it's still well below the 80 °C throttle point that limits the performance of the Founders Edition. We should note that you can get a bit more performance out of this card by ramping its fans way up to keep the GPU under 70 °C, but things get loud in a hurry. One of the serious misconceptions about open-air cards is that they are universally quieter than the Founders Edition. This simply isn't true. Two large fans running at the same rotational speeds as one small fan will in fact be much louder. Their saving grace is that they don't need to run as fast, and users of such cards would be wise to take advantage of this fact unless they're just chasing benchmarks. For long-term gaming, losing 10-20MHz of speed in exchange for a much quieter gaming experience seems like an obvious choice to us.
OK, the next step is to see if the ACX open-air model can keep its advantage when bottled up inside a compact ITX chassis!