Single-Card Performance Test (ITX Case)
Here you can see a photo that illustrates the layout of our test system's RVZ02 chassis. The video card resides in its own compartment, while the rest of the system lives below, separated by an internal wall. Note, however, that while these two compartments are nominally separate, plenty of exhaust heat from the video card area can directly reach the rear of the motherboard, which in turn can heat up the CPU pretty effectively. One reason we decided to use this case as our test platform is that it has no active cooling. That means the noise floor is essentially 0 dB (the CPU fan is nearly silent, and on the opposite side of the case from where our sound meter is placed), and the power supply has a zero RPM mode, so it's silent as well (note, however, that the current shipping version of the SX500 PSU we used has actually dropped that functionality). In the end, all the noise we'll be measuing is going to be from the video cards themselves.
All of this is great for noise levels, but it also means there's no way to get around a video card that isn't up to the task of cooling itself. Furthermore, while it is vented, a lot of heat ends up getting recirculated if exhausted into the case, particularly because venting is minimal on top, where open-air cards direct most of their exhaust. We'll see in a moment how that plays out in the benchmarks.
But first, we'll provide a closeup shot of the two video card installations, side-by-side. This isn't the smallest ITX case out there, and it can theoretically fit very large video cards. That being said, you'll quickly run into issues with most current after-market cards, because they aren't just longer, they're also a lot taller. While your eye might trick you into thinking they can fit, you need to be able to drop them in from above to insert them into the PCIe riser slot, so you'll lose the use of the mid-mounted VGA brace that makes sure the card doesn't place too much strain on the slot. You'd also need to remove an internal brace in the corner of the case, which could cause problems with load-bearing if something is stacked on top of the case.
So how do the two video cards behave in this unusual environment? Let's take a look:
What do we have here? Again, the open-air ACX card is able to boost a bit higher at the same reference clock rates, translating to slightly better performance. But in other respects, things are quite different than they were in the ATX case. First, the open-air card is coming dangerously close to the 80 °C throttle point, which would render it no faster than the Founders Edition. On the flip side, the noise readings are massively different: 33dB vs. 50dB. And no, this isn't a mistake. If you look back to the previous page, you'll see the ACX measured at 40 dB, while the Founders Edition measured at 44dB. How could one now be much quieter and the other much louder in this test despite the same fan levels? It all comes down to the test setup. The ATX case we used had lots of noise-generating components in it. Our ITX case has none. Therefore, the 33 dB measurement, taken 6 inches from the video card fans, represents the noise generated by the ACX card, pure and simple. It's the only noise-generating component in our entire test bed (nominal noise from the CPU fan isn't picked up by the microphone due to being on the other side of the case). Our Founders Edition, on the other hand, has nothing to muffle its noise. Our sound meter is just six inches from its intake, whereas in the ATX case the card was buried deep inside a case, with the sound meter over a foot away. That makes all the difference in the world. In other words, put your ear up to Founders Edition cooler when it's running at the 80% level, and you're going to hear it loud and clear!
We think just about everyone would conclude that the open-air card was again superior based on the graph above, but what happens when we push our cards to their maximum overclocks? Does the ACX cooler still dominate?
Uh, oh, what do we have here? The open-air EVGA ACX model can no longer stay below the 80 °C thermal throttle point. It's still running a bit faster, but its clock rate advantage is now cut in half. The continued existence of a performance delta between the two cards may come down to the degree of temperature fluctuation and how that affects Nvidia's boost algorithm. In this test, the ACX card jump around between 79 °C and 80 °C, which means it was flirting with higher speeds the whole time. The Founders Edition, on the other hand, was pretty well pegged at 80 °C. Remember how we said two big fans aren't always quieter than a single small fan? You can see that in action here, as the ACX produces much more noise at an 80% fan level than it did at 50%, nearly catching up to the noise level of the Founders Edition. Perhaps just as importantly, the open-air ACX card is heating up the interior of our case substantially, meaning our CPU is now running 5 °C hotter, despite an ample cooler. That's why builders of Small Form Factor systems really need to be careful about their choice of video cards, even if the cards themselves perform well enough in a cramped environment. We were running a 4K load, which means the bottleneck was most definitely the GPU. That made a lot of sense for testing the effectiveness of a GPU cooler, but if, for example, you had an application that loaded the CPU enough to have the bottleneck switch back and forth between the CPU and GPU, you could find the CPU running very hot in an ITX system.
Overall, we still think the open-air video card is superior, but just barely. For users looking for a low-risk, set-and-forget solution (which includes nearly all ITX System Integrators, by the way), the GTX 1080 Founders Edition is going to be the better pick.