Table of Contents
We finally got our hands on a GTX 980 thanks to the fine folks at EVGA. They were kind enough to send along their Superclocked version featuring the new ACX 2.0 cooler. Based on the NVIDIA Maxwell GM204 GPU, the GTX 980 promises less power consumption than the previous flagship Kepler based GPUs and performance that meets or exceeds them. This is also our first chance to look at EVGA’s new ACX 2.0 cooler that is said to run cooler, quieter, and use less power than previous versions. Let’s get started and find out what the latest NVIDIA GPU technology and EVGA have in store for us.
Specifications and Features
Here are the specifications as provided by the EVGA product page. When we compare the GPU’s base and boost clocks, we have a 140 MHz base/151 MHz boost clock factory applied overclock. The memory also has a mild 10 MHz overclock over reference design cards. The card also supports up to four monitors; and with the 4 GB of onboard GDDR5 memory, it should have no problem supporting those multi-monitor setups.
|Resolution & Refresh
A quick look at GPU-Z confirms what we see above and provides additional information as well. As you can see, the GTX 980 Superclocked ACX 2.0 comes outfitted with 4 GB of GDDR5 memory that sits on a 256-bit interface. ROPs and TMUs come in at 32 and 128 respectively, while the CUDA Cores sit at 2048. The GPU base and boost clock speeds of 1266 MHz/1367 MHz are confirmed as well, but we know from experience the actual boost clock will be higher than the advertised value. For this particular graphics card, the actual boost clock reaches 1404 MHz when the card is under load.
Ok, wait a minute here. Early returns show the GTX 980 performing as well or even better than the GTX 780 Ti. How can that be with a reduction in CUDA cores, memory bus width, memory bandwidth, and an 85 watt reduction in TDP when compared to the GTX 780 Ti? There are several reasons the NVIDIA GM204 GPU manages to pull this off. The first being the huge increase in base/boost clock speeds of almost 250 MHz. Another major improvement is each of the four schedulers now have their own group of CUDA cores and execution resources, which means they don’t have to share them among each scheduler as Kepler based GPUs do. There has also been an increase in cache up to 2 MB, which NVIDIA claims will reduce requests to the onboard GDDR5 memory.
One of the major selling points of EVGA’s version of the GTX 980 is their new ACX 2.0 cooler. Here is what EVGA states the benefits of the ACX 2.0 cooler are when compared to the reference design and competitor offerings.
- 0% Fanspeed when card is idle (under 60C)
- 26% Lower GPU Temperature (vs reference)
- 36% Quieter (vs reference)
- 250% Lower Fan Power Consumption (vs reference)
- 400% Longer Lifespan (vs competitor dual fan cooling solution)
When looking at the above comparisons, we see a pretty substantial drop in the amount of power the ACX 2.0 cooler draws. When dealing with NVIDIA’s Power Target limitations, lower power consumption by the fans means more available power for the GPU. The fan motor itself has gone through an upgrade and now consists of a 6-slot, 3-phase motor that promises greatly improved efficiency. The fan blade number has been increased from 9 on the ACX 1.0 to 11 on the ACX 2.0. Not only has the number been increased, but a “Swept” design has been introduced. What this means is the curve of the blade angles away from the direction of rotation. This feature is said to reduce wind resistance and turbulence, which increases fan speed while still reducing noise. Also worth noting is by default, the fans will not begin spinning until the GPU core temperature reaches approximately 60 °C.
EVGA sent along a set of “see-through” fan motors for an ACX 1.0 versus ACX 2.0 comparison. It’s pretty easy to see the design changes we mentioned by simply having a look for yourself.
The retail box does a good job of letting the potential customer know what the GTX 980 Superclocked ACX 2.0 is all about. An overlay flap on one end of the box front highlights the advantages of the ACX 2.0 cooler. Around back, we have a detailed list of all the features, specifications, I/O connections, and a small window to confirm the serial number on the card matches the one on the box. The box sides contain additional marketing blurbs and a multilingual list of a few high level features.
Inside the the outer carton, you’ll find a large plastic clam-shell that houses the card. Sitting on top of the clam-shell are the accessories, which include the following.
Before we take a more in-depth look at the EVGA GTX 980 Superclocked ACX 2.0 graphics card, here are several pictures taken from various angles.
One thing I would have liked to see is the inclusion of a factory installed backplate. Even if it came at the expense of a few accessories the majority of people will never use. A graphics card of this size would definitely benefit from having a backplate installed for the rigidity and PCB protection it offers. Luckily, EVGA should be offering a backplates as a separate purchase soon.
The EVGA GTX 980 Superclocked ACX 2.0 Up Close
Once the ACX 2.0 cooler is removed, we get a more detailed look at how it’s designed. The thermal paste was well applied and making excellent contact with the GPU core and cooler. All five heatpipes travel across the top of the cooler’s base plate and terminate at each end of the fin stack they are assigned to. The plastic shroud’s lettering across the top edge will illuminate once the system is powered on. Here are several pictures of the cooler showing the design of the fin array, heatpipes, and base plate area.
A thermal plate is attached to the card’s top side and includes thermal pads over critical components.
Once the card is completely naked, we see this EVGA offering is based off a reference design PCB. Given the factory overclock EVGA applied, it’s probably been binned to ensure it operates at advertised speeds. As you peruse the images below of the PCB, you’ll see a 4+1 power phase design and the use of Samsung K4G41325FC-HC28 GDDR5 memory modules. You might also notice the two open pads where a couple more power phases could be added, which means there is additional headroom available on the reference PCB should card manufacturers want to take advantage of it. You also see the two SLI bridge connectors meaning quad-SLI is a possibility. NVIDIA has yet to adopt an all PCI-E multi-GPU solution like AMD’s XDMA engine provides, so physical bridges are still needed for now.
For display connectivity, we have one Dual Link DVI, one HDMI, and three DisplayPort connections available. Times are changing as far as how display connections are being incorporated. It makes perfect sense to add more DisplayPort connections as they offer more functionality than other connection types, but can easily be converted to HDMI or DVI with cables or adapters if needed. With three available DisplayPort connections, for the first time, a three G-Sync monitor Surround setup can be accomplished from a single card.
Performance and Overclocking
|Test System Components
|ASUS Maximus VII Formula
|Intel i7 4790K Devil’s Canyon
|G.SKill TridentX DD3-2400 MHz 2x8GB @ 1866 MHz 9-9-9-24
|Samsung EVO 500 GB SSD
|Corsair HX1050 Professional Series
|EVGA GTX 980 Superclocked ACX 2.0
|EKWB Supremacy EVO Water Block – 360 mm Radiator – MCP35X Pump
We’ve got a good selection of comparison samples to include in our charts today, which includes a GTX 780 Ti and R9 290X. We’ll adhere to the Overclockers.com GPU test procedure that’s been in place since the Haswell platform was released. If you’re not yet familiar with our methodology, then click on the link provided for additional information. For quick reference, below is the down and dirty version of what we do.
Minimum System Requirements
- i7 4770K or i7 4790K @ 4 GHz
- Dual Channel DDR3 @ 1866MHz 9-9-9-24
- GPU @ stock and overclocked
- Monitor capable of 1920×1080
- 3DMark Vantage – DirectX 10 benchmark running at 1280X1024 – Performance preset.
- 3DMark 11 – DirectX 11 benchmark running at 1280X720 – Performance preset.
- 3DMark Fire Strike – DirectX 11 benchmark running 1920X1080 – Standard test (not extreme).
- Unigine Heaven (HWBot version) – DX11 Benchmark – Extreme setting.
- Batman: Arkham Origins – 1920X1080, 8x MSAA, PhysX off, V-Sync off, The rest set to on or DX11 enhanced.
- Battlefield 4 – 1920X1080, Ultra Preset, V-Sync off.
- Bioshock Infinite – 1920X1080, Ultra DX11 preset, DOF on.
- Crysis 3 – 1920X1080, Very high settings, 16x AF, 8x MSAA, V-Sync off.
- Final Fantasy XIV: A Realm Reborn – 1920X1080, Maximum preset.
- Grid 2 – 1920X1080, 8x MSAA, Intel specific options off, Everything else set to highest available option.
- Metro Last Light – 1920X1080, DX11 preset, SSAA on, Tessellation very high, PhysX off.
EVGA Software – PrecisionX 16
In order to get any overclocking done, a good software utility is required. PrecisionX 16 is a very capable utility to help you get the most from your graphics card. It provides complete monitoring capabilities, voltage adjustment, clock speed adjustments, fan control, and a variety of other settings as well. PrecisionX 16 is easy to navigate and understand right from the start. Good stuff here.
For those wondering about the maximum Power Target and available voltage control, the answer is 125% Power Target and +87 mV GPU voltage. There is also a KBoost feature that will force the GPU to constantly run at the boost clock speed.
Using the PrecisionX 16 software, I was able to reach a 24/7 stable overclock of 1340 MHz base/1441 MHz boost on the GPU and 1900 MHz (7600 MHz effective) on the memory. With this overclock in place, the actual boost clock held steady at 1491 MHz during the benchmark tests. A 24/7 stable overclock means the settings are able to complete all our benchmarks with no stuttering, artifacts, or any other abnormalities.
Beginning with the synthetic benchmarks, we see the GTX 980 Superclocked ACX 2.0 winning all three of the Futuremark tests. The HWBot Heaven benchmark favored the GTX 780 Ti Classified. Once the GTX 980 Superclocked ACX 2.0 was overclocked, it scaled nicely and really put the hammer down on the competitor cards.
As we head into the game benchmarks, it’s interesting to note how the GTX 980 Superclocked ACX 2.0 and the GTX 780 Ti evenly split victories with three wins each and one tie. Doesn’t get much closer than that. So, all the hype about GTX 980 graphics cards performing equal to or better than the GTX 780 Ti seems to be valid.
During the NVIDIA Surround testing, I expected the GTX 980 Superclocked ACX 2.0 to nudge out the GTX 780 ti Classified because of the additional onboard memory. While the advantage was small in most cases, the GTX 980 Superclocked ACX swept the field in all five game titles tested. As we usually see in our Surround/Eyefinity testing, three of the five game titles run above the minimum 30 FPS we like to see. Metro: Last Light and Crysis 3 fall below that magic number, but so has every other card we’ve tested to date. Keep in mind, we keep all the game settings maxed out for our Surround/Eyefinity testing. All the test results below were recorded with each of the samples at their stock settings.
Power Consumption and Temperature Testing
The power consumption numbers are what make the above benchmark results even more impressive. The maximum total system draw recorded was a mere 354 watts, and that was with the GTX 980 Superclocked ACX 2.0 overclocked and under full load. Going back and looking at the power consumption of the GTX 780 Ti Classified, we saw roughly 550 watts total system draw when it was overclocked and under load. That’s a pretty amazing reduction in power usage considering the performance the GTX 980 Superclocked ACX 2.0 brings to the table. A good quality 550 watt PSU with adequate +12 V amperage would be plenty to power a system with this video card.
Equally impressive as the power consumption numbers is the performance of the ACX 2.0 cooler. Keep in mind the fans do not begin spinning until the GPU temperature reaches approximately 60 °C. When the system is being used for things that don’t demand a high GPU load, it’s likely the fans will never even begin spinning. With the fans set to auto control, the highest recorded temperature recorded was 77 °C. On the other end of the spectrum, the ACX 2.0 cooler threw out some great temperature readings with the fans set to 100% speed.
Pushing the Limits
The most I was able to get and complete a run of 3DMark Fire Strike was 1370 MHz Base/1471 MHz boost, and the memory set to 1925 MHz (7700 MHz effective). At that setting, the actual boost clock speed climbed up to 1518 MHz. I also overclocked the CPU to 4.8 GHz and set the system memory to 2400 MHz. The end result was a score of 13293… Not too shabby!
The NVIDIA GTX 980 based graphics cards bring a new level of efficiency and raw graphics power never before seen. EVGA did a great job giving the reference design PCB a facelift by giving it a stout factory overclock and adding the ACX 2.0 cooler. The EVGA GTX 980 Superclocked ACX 2.0 not only stood tall against the GTX 780 Ti, but did so against what could be the best one of them all… EVGA’s own GTX 780 Ti Classified.
The low power consumption is pretty amazing for a graphics card that throws out these kind of performance numbers, and the performance of the ACX 2.0 cooler is more than able to keep the card cool under the most demanding scenarios. With the fans not spinning until the GPU reaches 60 °C, you’ll have the benefit of silent operation when performing tasks that require less GPU power.
As far as pricing goes, the EVGA GTX 980 Superclocked ACX 2.0 sells for $579 at Newegg and $569 at EVGA’s online store. There are several other GTX 980’s price the same, but none of them are factory overclocked as high as the EVGA offering. In fact, there are GTX 980’s selling for a lot more that aren’t factory overclocked as high. Any way you slice it, the price is right where it should be.
If you’ve been looking for a GTX 980 with top notch performance, a terrific cooling solution, and a stout factory applied overclock, then the EVGA GTX 980 Superclocked ACX 2.0 may very well be what you’ve been waiting on. Without question… Overclockers approved!