NVIDIA might be crowing about how their GeForce GTX 1080 is the new king of graphics cards, but for many gamers their interest lies more in the smaller Pascal sibling — the GTX 1070.
We don't have an Australian price or release date yet. But after spending just over an hour with the card, I have something that's equally useful: benchmarks.
NVIDIA set up a room with two PCs for media to play with at their annual Computex pre-briefing event. One was a 12 CPU Intel i7-5930k running at 3.5GHz with 32GB RAM and a Samsung SSD on an ACER monitor running at 3840 x 2160 (4K), while the other was an Intel i7-6700K running at 4.0GHz with a 32GB RAM, and a ACER G-Sync monitor at 2560 x 1440. Both were also running the latest NVIDIA WHQL drivers (368.22).
Both machines, obviously, were sporting a GTX 1070 Founders Edition. It's worth remembering that when it comes to purchase time, although whether non-Founders Edition models will be available at launch is another matter entirely.
But I digress: the GTX 1070. The games available for testing included Fallout 4, DOOM, The Witcher 3, Rise of the Tomb Raider, the full 3DMark suite of tests, Unigine's Heaven benchmark, the EVGA Precision X utility, and, weirdly, Obduction — the crowdfunded spiritual successor to the Myst series from Cyan worlds.
The first PC wasn't properly authenticated online, however, so I was only able to test the 3D Fire Strike demo and The Witcher 3. Every other game wouldn't launch without connecting first, which prompted me to switch.
That switch meant that I was only able to benchmark the majority of games at 1080p and 1440p, although for transparency's sake I'll provide the data I was able to collect and note which system it was from.
But before we get onto that, here's a reminder of what's under the GTX 1070 and how it compares to its bigger brother.
|Specifications||GeForce GTX 1080||GeForce GTX 1070|
|Memory Clock/Type||10Gbps, GDDR5X||8Gbps, GDDR5|
|Memory Bus Width||256-bit||256-bit|
|Teraflops (rated)||8.9 TFLOPS||6.5 TFLOPS|
|Pricing||$USD599/$1179-$1299 (Founders Edition)||$USD379/$USD449 (Founders Edition)|
With that out of the way, let's get onto the benchmarks.
3D Fire Strike Extreme
Fire Strike Extreme is the 3D Mark fighting demo (my name for it, anyway) running at 2560 x 1440 on DirectX 11. Most gaming machines will score anywhere from the low-to-medium thousands, while the GTX 1080 broke past the 10,000 mark on stock clock speeds.
Machine 1: i7-5930K PC
Score: 7888 Graphics score: 8399 Physics score: 14634
Machine 2: i7-6700K PC
Score: 7826 Graphics score: 8409 Physics score: 13370
The figures are averaged out from two runs, both of which were slightly faster on the i7-5930K machine than the i7-6700K rig — although the difference is fairly miniscule anyway.
But what's important — even though this is a synthetic benchmark — are the results in relation to the GTX 1080 and, more specifically, the GTX 980 Ti and the TITAN X. The GTX 1080 still has approximately 25% or so advantage over the GTX 1070, something we largely expected after the first figures leaked.
As for the TITAN X and the 980 Ti? Well, they just got supplanted. Given that most retailers are still selling the TITAN X for $1499 and the 980 Ti is going for around $900, it's hard to imagine who would purchase either of the Maxwell cards at those prices.
I've seen some Fire Strike Extreme benchmarks where the 980 Ti and the GTX 1070 are about even, although given the systems aren't identical it's hard to put any stock in that. What it does mean is that there's little reason to upgrade to a 1070 from a 980 Ti — and if retailers cut prices heavily enough, it might still be worthwhile picking up a 980 Ti provided you don't have any aspirations beyond 1440p gaming.
Rise of the Tomb Raider (DirectX 12)
There aren't a great deal of games supporting DX12 right now, and even fewer that have an inbuilt benchmark. Fortunately, Rise of the Tomb Raider has both. It's a canned tool that shouldn't be taken as indicative of what things are like once everything around Lara starts blowing up, but it does run through some of the game's different environments and its useful given the testing conditions.
For this and the remainder of the games, only the i7-6700K PC was used. And if you want a point of reference for where these results sit in the wider scheme of things, here's how Rise of the Tomb Raider performed (using DX11) at launch.
2560 x 1440, Very High preset
Mountain peak: 83.58 fps (min 53.44, max 111.23) Syria: 67.38 fps (min 46.51, max 79.90) Geothermal Valley: 67.11fps (min 49.14, max 80.16) Overall: 73fps
1080p, Very High preset
Mountain peak: 123.87 fps (min 82.06, max 191.61) Syria: 96.70 fps (min 49.08, max 118.86) Geothermal Valley: 92.70 fps (min 72.24, max 111.37) Overall: 104fps
Two rounds of applause here: one to Crystal Dynamics and DirectX 12 for the quality optimisation, and a second to NVIDIA for comfortably positioning the GTX 1070 as the 1440p card of choice.
With more and more gamers considering a 27" monitor as their sole or primary screen, having a card with enough headroom to guarantee 60fps at 2K is a useful selling point. It remains to be seen how much of a value proposition it is once retailers starting driving down 980 Ti and TITAN X stock, but the numbers look good for now.
There's no automated benchmark for DOOM, but the game does provide a handy live feed of performance statistics that can be viewed in the upper right corner.
Given that, the best I can provide here given the testing environment is a rough outline. There's no automated benchmark or replayable demo for testing, so I simply played for a few minutes at each setting running through the game's third mission. I played through until I cleared out the first gore nest, trying to follow the same order of explosions and glory kills along the way.
2560x1440, TSSAA (8TX), Ultra preset
Max: 125fps Average: 75fps During combat: 61-62fps
1080p, TSSAA, Ultra preset
Max: 140fps Avg: 110-120fps Low: 100fps
Again, the testing conditions are far from ideal. But it does give you an indication of how well the 1070 holds up at the most common resolution in gaming and its successor. And don't forget: DOOM doesn't have Vulkan support yet, and when it does those figures will only get better.
Here's our benchmarks for DOOM when it launched. Keep in mind the SMAA anti-aliasing technique used isn't as rigorous as TSSAA. That's not as good as having a hardline figure, of course, although the fact that DOOM doesn't have a set benchmark or demo to replay means you have to allow for a greater margin of error.
In time, I'll be able to get my hands on a GTX 1070 and put it through a more rigorous testing procedure. But my main takeaway from my time testing the GTX 1070 was that I wanted more time to test.
It's an incredibly good card, even though it doesn't meet the lofty heights of its big brother, the GTX 1080. Question is: how much will NVIDIA charge? And how good will that performance be in comparison to the other silicon on the block?
The author travelled to Computex 2016 as a guest of Intel.