Back in early 2015, The Tech Buyer's Guru published a unique look at video card cooler performance, comparing two very different versions of the top-of-the-line GeForce GTX 780 Ti 3GB video card to see how they compared in terms of thermal and acoustic performance. In that review, we determined that the reference "blower-style" Nvidia cooler, which had been around in one form or another for over a decade, provided consistent, proven performance, even if it wasn't always the best overall solution. Well, a lot has changed in the intervening years, so we're back at it to provide an even more comprehensive analysis.
This time around, we're going to be using the GTX 1080 GPU as our test chip. Released in May 2016, it's still among the most powerful chips on the market. In fact, it's almost twice as fast as the GTX 780 Ti 3GB released in late 2013, while using a lot less power, which is relevant when it comes to determining the effectiveness of a given cooling solution. To bring more attention to its efforts, Nvidia's actually gave a name to its blower-style design, which it now calls the "Founders Edition." While adhering to the same basic design as blower designs of the past, the shroud has been tweaked to give it an edgier look, and the internals have been upgraded to provide even better performance. As a counterpoint, we'll be using a custom-designed "ACX" open-air cooler from one of Nvidia's largest board partners, EVGA. Both will be strapped to our GPU of choice, the GTX 1080, which
Now, we want to make something very clear here: there are lots of custom designs out there, and enthusiasts could argue until the end of time about which is best, but we chose the EVGA model for two simple reasons: it carries no price premium, and it will fit anywhere the Founders Edition can fit. And that's a critical point, because in our previous comparison article, we chose a big model from MSI that couldn't be installed in one of our system, a compact ITX unit. This time around, you're going to be seeing exactly how these two cards perform in a variety of scenarios, including on their own in both a well-cooled ATX case and a passively-cooled ITX case, as well as in dual-card SLI setups. Hold onto your hats, because we think all you cooler fans you might just be blown away! OK, OK, puns aside, we've designed this review to be as thorough as possible, which is why this time around, we're calling it our "Definitive Analysis." Frankly, we think we're going to be able to put this issue to rest, so no, we're not going to be running a follow-up to this article any time in the near future!
All right, with that all clear, we can return to our regularly-scheduled program!
We used two differen test beds to assess the capabilities of the two video card designs. Our ATX-based test system, which allowed testing in both single-card and dual-card (SLI) configurations, was as follows:
- CPU: Intel Core i7-6900K
- Motherboard: Asus X99-Pro/USB3.1
- RAM: Corsair 4x8GB Vengeance LPX DDR4-3200
- SSD #1: Samsung 950 Pro M.2 512GB
- SSD #2: Samsung 850 Evo 1TB
- Case: SilverStone Primera PM01
- Power Supply: EVGA Supernova 1000 PS
- CPU Cooler: Corsair Hydro H100i v2
- Operating System: Windows 10
This big tower system offers up tremendous cooling capability, courtesy of the SilverStone PM01's three front-mounted 140mm intake fans and rear-mounted 140mm exhaust fan, along with the Corsair 240mm liquid CPU radiator with dual matching 120mm fans. It's basically an optimal environment for a hot-running GPU, as it has tremendous airflow, and the CPU's thermal envelope is almost entirely separate from the rest of the system. Cool air from the front fans hits the radiator, which is mounted in the upper-front corner of the case, meaning there's very little consequence to filling your case with waste heat from video cards that aren't "team players", i.e., open-air models.
For our other test bed, we chose another SilverStone-based system, but it couldn't be more different from the hulking PM01. The compact RVZ02 has been a mainstay in our DIY PC Buyer's Guides, in part because it offers such a unique take on PC design. It's actually one of the few compact ITX cases that can fit video cards larger than Nvidia's reference design, although as we stated, our EVGA ACX model conforms to the same basic dimensions as the Founders Edition. Here are the system's full specs:
- CPU: Intel Core i7-6700K
- Motherboard: Asus Z170I Pro Gamer
- RAM: Corsair 2x8GB Vengeance LPX DDR4-3000
- Solid State Drive: Samsung 850 Evo 500GB
- Case: Silverstone Raven RVZ02B-W
- Power Supply: Silverstone SX500-LG
- Optical Drive: Samsung Internal Slim DVD Burner
- CPU Cooler: Scythe Big Shuriken 2 Rev. B
- Operating System: Windows 10
This system is notable in that the SilverStone RVZ02 case is so slim (at under 3.5") that it offers no active cooling whatsoever. There isn't a single fan mount in the chassis, so all cooling comes from the CPU fan, the power supply fan (which is essentially isolated from the rest of the system), and the GPU cooler. To help maintain thermal control, SilverStone wisely separated the two main heat generators into separate compartments of the case, divided by an internal wall. As we'll see, however, heat can and does spill over when the going gets rough!
In terms of fan settings, we created a custom profile to provide the best opportunity for our cards to shine. We set them to run at 50% fan levels until they hit 80 °C, at which point they ramped up to 80%. When the GPUs hit 83 °C, the fan ramped up again to 100%. To be clear, we chose these points because they are "pain points" in the GPU's thermal programming. While in the past GeForce-based cards would throttle once at a certain temperature (typically 70 °C), and then again when they hit a temperature limit (typically 83 °C), GPUs based on the Pascal design actually throttle a number of times on their way from idle to peak load. We've seen Nvidia's GPU Boost drop clock speeds at 60 °C, 70 °C, 80 °C, and again at 83 °C, and while aftermarket software suites like MSI Afterburner appear to allow you to tweak the final throttle point slightly, we think most users would be surprised to learn that in the end, Pascal GPUs actually don't listen very well to such commands. No matter what, these GPUs throttle at 83 °C, which means you definitely want to keep them below that level.
We used three GTX 1080 samples, as follows:
All three of our test cards ran at exactly the same clock rates (1607 MHz Base Core Clock, 10,000 MHz Memory Clock). Well, that's the theory, at least. In reality, despite running equal clock settings on all our cards, they didn't necessarily boost to the same levels. That will be clearly explained in detail in this article. By the way, as always, we purchase our video cards at retail - these are not cherry-picked press samples. We strive to avoid any sort of bias when covering competitive markets like video cards. If you support this type of review, please help support this site by using our product links the next time you're ready for an upgrade.
To make things even tougher for the two tested coolers, we ran our GTX 1080 GPUs not just at reference settings, but also in a heavily-overclocked state. We settled on a stable overclock of +190MHz on the core, and +800MHz on the memory, which allowed our cards to run at around 1950MHz/10800MHz during our benchmark sessions. Note that you might see all sorts of wild claims about what the GTX 1080 GPU can hit (starting with Nvidia's absurd boast that the Founders Edition could sustain 2114MHz), but in our findings, testing four retail-purchased GTX 1080 samples, not a single one could do that. In fact, our earliest GTX 1080 sample, purchased in July 2016, was a terrible overclocker, unable to maintain stability with anything more than a +163MHz/500MHz overclock, translating to around 1925MHz/10500MHz in game. We obviously left that sample out of this analysis entirely, but it serves as a stark reminder that there's a big difference between the anonymous, unsubstantiated claims you'll see on forums and what hardware can actually achieve when you start to look at aggregate, substantiated numbers.
To put these coolers to the test, we used a very intensive, but still realistic, GPU benchmark: the 3DMark Fire Strike 4K Ultra Stress Test. Unlike power viruses like FurMark, this test renders an actual game-like scene, and uses multiple assets within the GPU to maximize load in a balanced fashion. It's not just hammering one element of a GPU to produce massive thermal spikes, it's forcing the GPU to render a complex 4K scene using all the GPU has to offer. This stress test runs for approximately 15 minutes, and that's critical to our tests, because it takes about that long for GPU temperatures to level out, meaning it gives you a very good sense of what a video card will sound like when you're actually playing a game for an extended period, not after 30 seconds of a benchmark loop. In all, we ran the 3DMark Stress Test over a dozen times, which took around four hours in total, and that doesn't count all the hardware changes, data tabulation, graphing, photography, and analysis that went into creating this article. So as you can probably imagine, this was most definitely a labor of love, but it's the kind of thing we're willing to put time into to come up with answers than no one else can give you. We hope you agree that our findings are indeed definitive.
All right, with that introduction out of the way, it's time to get into the nitty-gritty of our results, starting with a review of installation!