ProsEffective cooler; significant factory overclock; excellent EVGA service and warranty
ConsMinimum fan speed is too high at idle; fan resonates at certain RPMs
The EVGA GeForce GTX 780 Superclocked ACX 3GB is a high-end, semi-custom, factory-overclocked video card from veteran video card manufacturer and exclusive Nvidia-partner EVGA. It offers one of the highest out-of-the-box overclocks of any GTX 780 model on the market (967MHz versus 863MHz base clock, and more importantly, an impressive 1020MHz boost clock, versus 902MHz on reference models). It measures 10.6" long (listed specifications suggesting that it's 10.5" long are incorrect - the PCB itself is 10.5" long, but the cooler adds another 1/10 of an inch). It uses a dual-fan cooling arrangement, similar to cards from many of EVGA's competitors. Interestingly, this is one of the very first such coolers from EVGA, which has typically stuck with reference designs or high-end water coolers. It's not particularly stylish, but to its credit, it doesn't use flashy colors or huge protruding heatpipes to convince you that it's fast. Let's just say it speaks softly but carries a big stick.
The unique features of this EVGA GeForce GTX 780 are its cooler and its factory-overclock, both mentioned above, but it's how they are interconnected that really makes this card shine. A GTX 780 will begin to throttle its clockspeeds at 79C. Giving a card like this a factory-overclock, especially one as significant as EVGA provides, might not provide the on-paper benefits without the help of the ACX cooler. It's the secret sauce that allows EVGA to push the clocks and deliver improved performance simultaneously. In our testing, the in-game boost clock on this card was almost always pegged at 1110MHz, well above the 967MHz base clock and even the 1020MHz rated boost clock. Only on a few occasions did it slip down to 1071MHz - this was when temperatures exceeded 77C. A simple adjustment of the conservative fan profile took care of that, however, and given that the cooler is extremely quiet at load, boosting the fan speed to keep the GPU under 79C was not an issue.
What was an issue, however, was the unnecessarily-high minimum fan speed on this card - 39 percent, which translates to a somewhat buzzy experience. Given that the card idled at around 33C, we would have loved to drop that fan speed down quite a bit, but alas, for some reason EVGA has programmed the BIOS with a minimum higher than we've seen on any other card we tested. That's unfortunate, and the issue was only compounded by the fact that the fans resonated at certain lower RPMs, including from time to time at the 39 percent level. This is a problem we've read about in user reviews, and frankly, it's disappointing. We've already contacted EVGA to RMA the card, and as always EVGA is extraordinarily committed to keeping their customers happy, but we hung on to the card long enough to write this review.
Overclocking, Power Use, and Temperature
The test bench used for this article consisted of an Intel Core i7-4770K@4.4GHz, an Asus Gryphon motherboard, 8GB of DDR3-1866 RAM, Windows 8.1, and GeForce Drivers 331.82
The stock in-game boost of the 780 Superclocked ACX was 1110MHz core, with a 6000Mhz memory clock. We simulated a reference 780 by downclocking it to 1006MHz core (in-game), and we were also able to overclock it all the way to 1215MHz core memory (in-game). Keep in mind that the in-game boost is higher than the advertised boost, but can vary with both temperature and power use. At out-of-the-box settings, in-game boost did drop as low as 1071MHz when the card exceeded 78C. We never saw it exceed the 100% power limit at default speeds. To allow the card to really run, we set the power limiter in EVGA PrecisionX at the maximum (106%), which also set the maximum temperature to 95C. This avoided the temperature throttle that may occur depending on the case setup and ambient temperature.
We found that clockspeeds also varied with voltage, but not in the way that you might imagine. We were able to use PrecisionX to manually set the voltage to 1.187v, up from the default 1.162v. Interestingly, the card auto-overclocked itself when presented with more voltage. At as-shipped settings, simply upping the voltage boosted the in-game clock to 1137MHz, rather than 1110MHz. With a +109 core setting, which took the card to 1215MHz in-game at default voltage, the extra voltage pushed the card to 1241MHz. Keep in mind that we were not changing the overclock manually, just the voltage, so again, this is an auto-overclock. Unfortunately, that 1241MHz core clock was not stable in Crysis 3, and we couldn't dial in anything between 1215MHz and 1241MHz - the card just took itself straight to 1241Mhz at numerous core settings between +109MHz and +129MHz. Therefore, we consider the highest stable overclock to be 1215MHz in-game, and this was achieved only without extra voltage. We were also able to overclock the memory to 6800MHz, up from the as-shipped 6000MHz. While it would run in some games at 7000MHz, we witnessed reduced performance, so we consider that an unstable memory overclock.
At as-shipped settings, power use on our test bench peaked at 335W, using the 3DMark Fire Strike Combined test. Downclocked to reference 780 settings, it peaked at 325W, and at the maximum stable overclock of 1215/6800MHz, it peaked at 365W. Idle power draw was a very low 44W. Extra voltage did cause the power use to spike further, but as already noted, the over-volted overclock simply wasn't stable in all games, so we don't consider it a relevant measurement.
With regard to temperatures, we found the card to idle around 33C in an ambient of 70F, and it maxed out at 67C in our open test bench. When we closed the case up, however, the card did indeed push up against the 79C temperature limit imposed by the Nvidia, which is why we adjusted this upward using PrecisionX, just to make sure we were always at maximum boost.
Now, a word on noise. We don't use a decibel measure, because much of the experience depends on the ambient noise, case fans, the case setup, and even the game being played. So we'll just use a subjective measure. The card is exceptionally quiet at load, in fact one of the quietest cards we've tested. The problem for the ACX cooler is that it's not particularly quiet at idle, and this can be traced directly to its minimum fan speed: 39%. In our opinion, that's far too high, especially given how cool the card runs at idle. EVGA really should have allowed the user to reduce fan speed further to lessen noise. Another issue we had at lower fan RPMs is that the cooler resonated in a very unpleasant way. This occurred most often at low fan speeds and when the fans were increasing or decreasing in RPM. We actually had a more pleasant experience just locking the fans at a higher RPM to avoid the resonance issue. Overall, we'd consider the fan noise to only score an "average" versus competing cards. It's too bad that its exceptional full-load noise is marred by sub-par idle and partial-load noise.