The new RTX 3070 is apparently better than a 2080 Ti, at least according to Nvidia. I’m going to put this claim to the test by comparing both in games at 4K, 1440p, and 1080p as well as content creator workloads to see what the differences are.
Let’s start with the spec differences, the 3070 has more CUDA cores, higher base and boost clock speeds, but 3 gigabytes less memory, so it will be interesting to see if that affects games at 4K. The 3070 costs significantly less than the 2080 Ti, half as much if we go by Nvidia’s $1000 USD MSRP, and even more if we work with the $1200 or so that it was actually available for.
Test PC Setup
This is the system that I’m using to test out both graphics cards, it’s got an overclocked i9-10900K at 5.2GHz with 32gb of DDR4-3200 CL14 memory in dual channel.
We’ll first dig into the gaming results, then check out cost per frame, power draw, and content creator workloads afterwards.
Starting out with Microsoft Flight Simulator, the 2080 Ti was ahead of the 3070 at all resolutions, though it’s not by much, just 3-4 frames on average. This equates to the 2080 Ti being about 8% faster at 1440p, the second biggest change at this resolution out of the 11 titles tested, but realistically it’s not a huge change at just 4 FPS difference. Red Dead Redemption 2 was tested using the games benchmark tool. This time the 3070 was ahead, with a 4% lead at 1080p and 1440p, then 6% higher at 4K, again not big differences, but that said, this was the game that performed the best on the 3070 out of the ones I’ve tested here. Battlefield 5 was tested in campaign mode by running through the same mission on both graphics cards. The 2080 Ti was back in front here, but like before not by a huge margin, regardless, winnings winning, and the 3070 was 6% slower at 4K.Shadow of the Tomb Raider was tested with the games built in benchmark. Both graphics cards were able to surpass 60 FPS with the highest setting preset which is a nice result, but the 2080 Ti was 6% faster than the 3070 at 1440p and 8% faster at 4K. For Control I’ll start with RTX off results. There’s basically no difference between the two at 4K, margin of error stuff really, the differences were a little larger at the lower resolutions, but like most other games, not really the sort of stuff you’re going to notice a whole lot while actually playing. With RTX on and DLSS enabled the frame rates at 4K and 1440p increase a little with both cards, and both GPUs also get closer together now with close to no difference regardless of resolution, the 2080 Ti was just one to two frames ahead for the averages. I’ve tested Metro Exodus with the game's benchmark tool, and the differences here were above average when compared to the other games tested, but yet again the gap between the two doesn’t appear that large, the 3070 was 8% slower at 4K and 7% slower at 1440p. Things are even closer together in death stranding, especially at the higher resolutions where the 3070 was just 4% slower than the 2080 Ti. The Witcher 3 had the biggest difference out of all 11 titles tested, but even so it doesn’t actually look that much with it layed out here on the graph, the 3070 doesn’t appear far behind and was 10% slower at 1440p, or 12% slower at 4K. Assassin’s Creed Odyssey was tested with the built-in benchmark, the average frame rates are very close together here. Interestingly the 1% low from the 3070 is higher than the 2080 Ti. This was also the case in The Witcher 3, though the difference was much less pronounced than it was here. Call of Duty Modern Warfare was tested in campaign mode. There were above average differences here with the 3070 around 6% slower at 1440p and 4K. Rainbow Six Siege was tested using the games benchmark tool with Vulkan. This was another game where there were basically no differences of significance between them.
At 1080p on average over all games tested the 3070 was 3% slower than the 2080 Ti in average FPS. Most games covered were a little ahead on the 2080 Ti, though in a lot of cases the difference was small.
Stepping up to 1440p and the 2080 Ti now has a slightly larger lead, with only Red Dead Redemption 2 doing better on the 3070, again though, for the most part the differences weren’t too large.
At 4K the 2080 Ti is a little further ahead now, I wouldn’t be surprised if its extra VRAM was helping out in some cases here, even if it is slower than the 3070’s memory, it’s still got more capacity which should have an advantage at higher resolutions with higher settings, but of course this is game dependent.
I guess to be fair to Nvidia, I am using an Aorus 2080 Ti so it is clocked higher than stock, meanwhile I’m using a founders edition 3070 as that’s just what I’ve got available. A 3070 from a board partner like ASUS or MSI or whoever is probably going to have a slight overclock, so the differences between the two might be a bit closer together. With the 3070 slightly overclocked, I could see the difference being much closer.
Cost Per Frame
It doesn’t matter much though as the price difference is so large. The 2080 Ti had an MSRP of $1000 USD, but was generally only available for $1200. I’ve decided to add both price points into the cost per frame graph for argument’s sake, though there’s not too much point as the 3070 is demolishing the 2080 Ti in terms of value. This makes perfect sense, as we just saw the 3070 was only a little slower on average, but it’s less than half the price, so quite simply the value is just far better.
2nd Hand 2080 Ti?
You could of course look at buying a 2nd hand 2080 Ti, I’ve seen them going for as low as $500 USD, personally I’d probably be more inclined to get the 3070 assuming it’s available, as you get the nice extras like warranty and knowing it hasn’t been used in a mining farm or something. The only exception would be if I really needed the extra VRAM the 2080 Ti offers.
The 2080 Ti is less efficient too, drawing around 16% more power in Control at 4K when compared to the 3070, and the frame rates were within 1 FPS of each other at this resolution, so basically no performance difference either.
Performance Per Watt
This is further illustrated by comparing performance per watt, the 3070 is just able to get more done with the same amount of power, which you would expect with the new and improved architecture.
Content Creation Benchmarks
It’s not all just about gaming though, I’ve also compared both in content creator and compute focussed workloads. I’ve tested DaVinci Resolve with the Puget Systems benchmark, and the 3070 was scoring around 10% higher than the 2080 Ti, so a much bigger difference when compared to the games covered previously. The performance gap is even larger between them in V-Ray, the 3070 was scoring more than 30% higher than the 2080 Ti in this one, which makes sense as these sorts of tests can better leverage the compute power available compared to the games. The improvements keep on coming, Blender was tested with the Open Data benchmark, and the Classroom test was completing 60% faster on the RTX 3070, though the shorter BMW test was 21% faster, less of a difference but still quite significant.
So basically the 3070 is the clear winner, it uses less power and is more efficient, it’s cheaper than the 2080 Ti and the gaming performance isn’t that far behind, it’s quite similar, and as noted if you do get a 3070 from a board partner that’s overclocked out of the box the performance gap might be much closer together.
I’d only really suggest the 2080 Ti if you can get an excellent price, probably second hand if you’re willing, and really only if you need the extra VRAM, perhaps for 4K gaming. Otherwise, in most cases the 3070 makes more sense.
So if the 3070 is the winner here, then the question really becomes is it worth paying $200 more to go up from the 3070 to get the 3080? I’ve compared the 3070 and 3080 over in this review.