Between finding out the GTX 1070 Ti was a thing and actually getting my hands on one, I spent a lot of time trying, and failing, to determine where this 4K-bothering card sat in Nvidia’s overall strategy. Its position in the Nvidia hierarchy is obvious – between the GTX 1070 and the GTX 1080 – but other than sharing 8GB of memory, it seems to be more of a toned-down 1080 than a souped-up 1070. After all, it has far more cores than the GTX 1070 – 2,432 of them compared to the 1070’s 1,920 – but just 128 fewer cores than the (ostensibly) beefier GTX 1080. Why, then, would you not just take the tiny step further to a GTX 1080?
It gets weirder, too: Nvidia seems to have a strange rule against its partners setting their GTX 1070 Tis up with factory overclocks, meaning that you can only buy at stock speeds (1607 MHz base, 1683 MHz boost). Workarounds have already been found for this (overclocking isn’t outlawed per se), but it seems an awful lot like Nvidia’s scared of the 1070 Ti sapping GTX 1080 sales. That, or they just wanted to give a central digit to AMD’s Radeon Vega RX 56 by outperforming it with a less ‘important’ card. I dunno, basically. But to help cure my ignorance, I’ve got Zotac’s take, the GeForce GTX 1070Ti AMP Extreme.
This particular model skirts around Nvidia’s sort-of overclock ban by offering a “factory tested” 150MHz boost as a downloadable profile for the firm’s Firestrom software, but that’s still small enough for it to be broadly representative of the GTX 1070Ti as a whole.
First up for testing is Doom. Unsurprisingly, it runs superbly at both 1,920×1,080 (1,080p) and 2,560x,1,440 (1,440p), holding together even when the screen is rammed with baddies. Upping all the way to 4K (3,840×2160) has a clearly noticeable impact, but the GTX 1070 Ti still keeps smoothness well up to pacey shooter standards. Most importantly, it’s stable: there’s no real change between quiet scenes and massive shootouts. That’s partly a testament to Doom’s optimizations, but also a promising showing for the graphics card.
Switching to Hitman, which is routinely filled with hardware-punishing NPCs, Nvidia’s card held firm. At 1080p, there were slight drops from perfection during the busier scenes, but these were tiny slips rather than sheer drops. 1440p had a single lurch during a rooftop scene, but it passed immediately, and wasn’t repeated during a 4K run. Again, high marks at all resolutions.
Middle-earth: Shadow of War(dor) proved to be its first big test. Ignoring the menu’s warning that we lacked the VRAM for Ultra quality textures, as well as everything else on full, we launched into 1080p to find that everything was…fine, actually. The same was true at 1440p – you could easily wring more frames out by turning down some settings, but when you’ve already got a good balance between high fidelity and smoothness, it makes sense to stick with Ultra.
It wasn’t until we tried 4K that the GTX 1070 Ti started to struggle. It still ran at what we’d consider an acceptable pace, but big moments like a fort siege clearly strain the GPU. Expect to wave goodbye to some luxury settings after all, if you’re going to be playing at this res.
Total War: Warhammer II ticked along well, though understandably not as smoothly as our more focused benchmarks. Zooming in to within a spear’s length of a unit can be the fastest route to stuttery misery in RTS games; not so here, with both 1080p and 1440p staying suitably smooth on the battle map. On the campaign map, there was a touch of lurching on 1440p, albeit only in the sense that performance went from ‘great’ to ‘pretty good’. 4K was the only res where maxed-out settings is anywhere near too much; there was noticeable jittering on both battle and campaign maps, with more distinct gulfs between highs and lows on the latter.
Wolfenstein II: The New Colossus was basically another Doom, in that it was lovely at 1080p, still lovely at 1440p, and slightly less lovely (but still lovely) at 4K. Rise of the Tomb Raider, with all its varying weather, particle, and Lara Croft hair effects was a slightly tougher test: not at 1080p and 1440p, yet again, but at 4K, smoothness took a relatively big hit. That said, the GTX 1070 Ti still kept things perfectly playable, so it’s not a huge concern.
The Witcher III was second only to Doom in evoking the “Hey, I switched from 1080p to 1440p and could barely tell the difference” feeling, which is to say that the GTX 1070 Ti excels at both. 4K was really good as well; it’s not that you can’t see a performance hit, but you won’t be compelled to sack graphics quality either.
Lastly, we have Assassin’s Creed Origins, which didn’t have issues conveying its particularly stabby recollection of Egyptian history. At least, not at 1080p: 1440p wobbled a teeny bit during transitions into more open areas, with the same happening on a clearer scale at 4K. It still runs fine, but switching off some of the fancier visuals would be wise here.
No surprises, then: the GTX 1070 Ti is slightly better than the GTX 1070 and RX 56, and slightly worse than the GTX 1080. That doesn’t preclude it from being a good card, mind; it would have to be to come so close to Nvidia’s original Pascal flagship, and while it can’t run every single game flawlessly at 4K, not even the GTX 1080Ti can do that.
I should say that I’m not terribly keen on this AMP Edition’s general design – it’s thicker than a Gears of War protagonist’s neck, and despite the massive radiator and tripled-up fans offering good cooling, the latter still go like the clappers.
Here’s the real issue, though: this particular GTX 1070Ti card costs £500, and a decent partner-customised GTX 1080 costs £495. There are, of course, cheaper 1070Ti options out there, such as the £420 version you can buy straight from Nvidia, but this doesn’t come with any of the handy stuff like easy overclocking software, and its coolers are nearly as sophisticated as other partner cards. The best deal I’ve seen is EVGA’s £439 version, but to repeat an earlier question, when you’re spending that much in the first place, what’s an extra £60 to get the superior GTX 1080?