A couple of weeks ago, GIGABYTE very kindly sent in one of its latest and greatest graphics cards: the GIGABYTE G1 Gaming NVIDIA GeForce GTX 1080. It's a veritable behemoth of a GPU, which you'd expect given that the GTX 1080 is the most powerful single GPU card available to gamers today.
Amongst all the drama of E3 and everything else, I've been putting the card through its paces. So if you're in the market for a GTX 1080 and you're wondering how GIGABYTE's card fares, you've come to the right place.
The machine I use for testing in the office has some fairly standard parts -- not the absolute top of the line equipment, but some good medium-to-high end gear that the majority of PC gamers are likely to own or be able to afford.
CPU: Intel Core i7-4790K @ 4.0GHz RAM: 16GB Corsair Dominator DDR3-RAM @ 1600Mhz GPU: MSI R9 390X Gaming 8GB PSU: Corsair HX850i 850W HDD: Samsung 850 EVO 1TB SSD Motherboard: ASUS Z97I-PLUS Mini-ITX (Intel Z97) Keyboard: Cherry MX 3.0 Mechanical Keyboard Mouse: ZOWIE FK2 Headphones: Audio Technica ATH-M50x
And for the GTX 1080 itself, here's every bit of detail direct from GPU-Z:
It's worth adding that all of the below figures were achieved using the card's stock performance, and as a result you may get slightly better returns with overclocking. I'll have a report back on the card's overclocking capabilities in a later article. You can also click on any of the images below to enlarge.
3D Mark Fire Strike
Futuremark's synthetic Fire Strike benchmark is an industry standard for graphics cards, and if you're hunting around for leaked benchmarks on new GPUs then Fire Strike figures are probably the first ones you'll see.
The figures below are the averaged result across three runs. Note that as more and more graphics cards are added, this table will be updated: we've currently got reference GTX 1070 and 1080 cards in the office, and by the end of the week we'll have testing figures for the reference version of AMD's Radeon RX 480. Stay tuned for that.
The gap here is pretty much expected. While my testing rig doesn't have the latest generation CPU, fastest motherboard or the fastest RAM, the distance between our R9 390X and the 1080 is around 60 to 65% in the standard and extreme Fire Strike tests.
At Fire Strike Ultra, where the rendering resolution is yanked up to 4K, the improvement is less stark at 43.72% (rounded down). It just works out price-wise: the MSI R9 390X will set you back around $460 to $540 at retail, while the G1 Gaming GTX 1080 is selling for around $1140.
The Talos Principle
Croteam's first-person puzzler was one of the first titles to implement Vulkan support, although we're not including it in our tests here as it's currently largely unoptimised.
The game, however, still has an excellent in-game benchmark that's representative of real-world gameplay. It also scales remarkably well.
The settings that was used for all resolutions: Ultra settings for CPU, GPU, GPU Memory, and Level Caching; Fullscreen, 2.4x 3D Rendering MPIX, and 4x MSAA. All figures have been averaged out over three runs.
Unsurprisingly, The Talos Principle runs exceptionally well on even medium-range GPUs. The R9 390X has no trouble running Croteam's puzzler at 4K, and should official VR support ever be patched in it should be incredibly playable on a wide range of systems.
There doesn't appear to be a huge distance between GIGABYTE's GTX 1080 and the 390X at 2K and 4K, but percentage-wise (41.57% at 2K and 38.35% at 4K) it's about the same gap as was seen between the two cards in the Fire Strike Ultra test.
Ah, DOOM. Id has already done a damn good job of optimising the game for current hardware, and that's only set to get better whenever Vulkan support is officially patched in. Which was supposed to have happened already, given that DOOM came out over a month ago.
But no matter. DOOM doesn't actually have an in-built benchmark, so the figures were generated by running through the game's third mission, Foundry. FRAPS was used to measure the minimum, maximum and average framerate over the course of a couple of minutes, and I ran a lap throughout the course opting not to kill any enemies along the way. I'll record a video some time and upload it here so you can replicate the precise course yourselves.
DOOM returned the most variance between tests, but you'd expect that given that it was a real-world benchmark. All scores were averaged out over three runs, as always. The Ultra preset, with chromatic abberation enabled and FXAA (1x) anti-aliasing was used for all resolutions.
It's here where we start to see the biggest jumps, with a 83.13% improvement at 4K between the 390X and GIGABYTE's GTX 1080. Frame rates were 63.1% better at 2K between the two cards, while 1080p produced the least appreciable difference between the two.
Ashes of the Singularity
As it stands, Ashes of the Singularity is the only DX12 title currently available in our testing stable. We've reached out to manufacturers and vendors to have Hitman and Rise of the Tomb Raider added to our roster, and when those become available and I've had the time to update everything those figures will be included.
Ashes is particularly interesting as its one of the fewer games where AMD cards perform substantially better than average. It's here where the price difference between GIGABYTE's GTX 1080 and the 390X is the most difficult to swallow.
The extreme preset and 4X MSAA anti-aliasing was used for all resolutions, while the benchmark was set to GPU focused each time. Scores shown below are averaged out across three runs.
A 21.03% improvement in performance is hardly worth it for a card that can cost more than $680, if you're buying the cheapest available models of both boards. The 38.24% jump in FPS at 4K is a little more representative of the gap between the two GPUs, and mirrors results in other games at the same resolution.
Ashes is one of the few titles to make extensive use of asynchronous compute. NVIDIA has tried to update their architecture with Pascal to better take advantage of the technology, while AMD has been factoring in the power of asynchronous compute into their GCN architecture for years. When it comes down to the raw performance stats, it's clear that AMD still has a bit of an advantage here -- although whether that matters to gamers in the end depends on how heavily developers utilise the technique.
If you're looking to purchase a GIGABYTE GeForce GTX 1080 G1 Gaming card today, it'll set you back $1139 from major retailers across Australia according to a StaticICE search.
What do you think? Does the GTX 1080 still appeal to you and your budget?