Skip to main content

Long read: The beauty and drama of video games and their clouds

"It's a little bit hard to work out without knowing the altitude of that dragon..."

If you click on a link and make a purchase we may receive a small commission. Read our editorial policy.

Nvidia GeForce RTX 3070 review: welcome to the new mid-range

Neck-and-neck with the RTX 2080 Ti for half the price.

Until September 2020, the RTX 2080 Ti was the fastest consumer graphics card on the market and retailed for $1000 - often more. Now, Nvidia's new RTX 3070 is here, with Team Green promising the same level of graphics performance at $500, half the price. That's a bold yet simple proposition that will no doubt spur a lot of upgrades, so does it hold water? We've been testing the 3070's gaming performance and power efficiency for the past week to find out.

For what it's worth, Nvidia certainly seems to be expecting a lot of demand. The RTX 3070 was originally set to go live in mid-October, but the launch was delayed by two weeks to build up inventory of the new cards and bulk up measures against automated purchases. The RTX 3080 and 3090 both sold out immediately and remain near-impossible to find six weeks post-launch, so let's hope that availability is improved this time around.

So assuming for a moment that cards are available for purchase in any kind of reasonable time frame, how does the RTX 3070 deliver the kind of performance gains Nvidia is claiming?

Watch on YouTube

The answer lies in architectural improvements, mostly, including a die shrink from TSMC's 14nm process to Samsung's 8nm. That allows for many more transistors to be packed into a smaller space, unlocking both performance and efficiency gains. Combined with second-generation ray tracing and third-generation tensor cores, some seriously impressive gen-on-gen improvements can be achieved - as you can see from our generally complimentary RTX 3080 and RTX 3090 reviews.

The 3070 should push performance per dollar to even higher levels, while drawing less power (and therefore generating less heat). It boasts 5344 CUDA cores compared to the 2304 in the RTX 2070, with roughly the same clock speeds and a similar GDDR6 memory interface rather than the more expensive GDDR6X VRAM we saw on the 3080 and 3090. Compared to the RTX 2080 Ti, the 3070 has more CUDA cores but less power and less memory bandwidth, so it'll be interesting to see how it compares in different titles.

RTX 3080 RTX 3070 2080 Ti RTX 2080 RTX 2070 GTX 1070
GPU GA102 GA104 TU102 TU104 TU106 GP104
CUDA cores 8704 5888 4352 2944 2304 1920
Memory Bus 320-bit 256-bit 352-bit 256-bit 256-bit 256-bit
Bandwidth 760GB/s 448GB/s 616GB/s 448GB/s 448GB/s 256GB/s
Base Clock 1440MHz 1500MHz 1350MHz 1515MHz 1410MHz 1506MHz
Boost Clock 1710MHz 1725MHz 1545MHz 1800MHz 1620MHz 1683MHz
TDP 320W 220W 250W 225W 175W 150W
Die Size 628mm2 392mm2 754mm2 545mm2 445mm2 314mm2
Transistors 28B 17.4B 18.6B 13.6B 10.8B 7.2B

For our review, we're testing the RTX 3070 Founders Edition, which retails at the same $500/£450 RRP as entry-level custom designs from Nvidia's partners. That's a change from last-generation RTX 20-series FE models, which came with a factory overclock and a mildly higher power target but also cost around $100 more than the cheapest third-party equivalents.

Like Nvidia's 3080 and 3090 FE models, the 3070 Founders Edition uses a dense 'flow-through' cooling solution, with a relatively small PCB and a lot of metal fins. Unlike the higher models though, the 3070 has both of its fans on the same side and the card is noticeably smaller - at around 24 by 11cm (roughly 9.5 by 4.5 inches), it's more or less the same height and just a centimetre longer than the outgoing RTX 2070 FE card despite promising significantly higher performance. It's also significantly lighter than other Ampere cards, weighing just 1035 grams compared to 1352 grams for the RTX 3080 FE and a monstrous 2157 grams for the 3090 FE.

In terms of I/O, we're looking at the same outfit as the RTX 3080 and 3090: three DisplayPort 1.4 ports and one HDMI 2.1 port capable of 4K 120Hz or 8K 60Hz with the right screen, but no USB-C VirtualLink (RIP). AV1 encoding is included here too, which should allow sites like Netflix and Twitch to stream high-quality, high frame-rate video content to you while requiring significantly lower bandwidth than current codecs. Finally, the card is PCIe 4.0 compatible but works just fine in PCIe 3.0 slots too so you aren't expected to rush out and buy a new motherboard.

The 3070 FE sips power compared to its bigger siblings, requiring only a single 8-pin power input - connected via the adapter in the box to the same 12-pin mini connector as we saw on Nvidia's RTX 3080 and 3090 cards. Those faster models required two 8-pin power inputs for their 12-pin adapters, so it's no surprise to learn that the 3070's rated TDP is 220W, significantly less than the 320W and 350W of the 3080 and 3090, respectively.

To put these power figures into perspective, we used Nvidia's PCAT system, a board that sits between the graphics card and its power inputs, to measure actual power consumption when gaming. When twinned with in-game performance metrics, we can use this to work out exactly how many watts are required to render each frame, and thereby get a measure of real-world efficiency of the RTX 3070 and its competitors from Nvidia and AMD.

Watts Per Frame RTX 3090 FE RTX 3080 FE RTX 3070 FE RTX 2080 Ti FE RTX 2070 RX 5700XT
Death Stranding 3.209 3.349 2.915 3.562 3.651 3.846
Percentage Diff 110.1% 114.9% 100.0% 122.2% 125.2% 131.9%
Watts Per Frame RTX 3090 FE RTX 3080 FE RTX 3070 FE RTX 2080 Ti FE RTX 2070 RX 5700XT
Gears 5 4.055 4.156 3.603 4.266 4.734 5.403
Percentage Diff 112.5% 115.3% 100% 118.4% 131.4% 150.0%

The 3070 gets off to an early lead in a run of Gears 5, with the RTX 2080 Ti delivering equivalent performance but requiring nearly 20 per cent more power to do so. The RTX 3080 and 3090 achieve higher frame-rates but use much more power, giving the 3070 the efficiency edge. AMD's latest high-end card at the time of this review, the RX 5700 XT, requires 50 per cent more power to render each frame.

We see a similar result in Death Stranding when it comes to the RTX 2080 Ti, which outperforms the RTX 3070 slightly but draws more than 20 per cent more power per frame. AMD cards are more efficient in Death Stranding than Gears 5, but the RX 5700 XT still require 32 per cent more watts per frame than the 3070. Of course, AMD's new Big Navi cards are potentially just weeks away and should come with efficiency improvements of their own. For the here and now though, the 3070 is our new efficiency champion.

With the power testing out of the way, it's time to move onto the game benchmarks. You'll see a few new titles here if you've not checked out our RTX 3080 and 3090 reviews yet, as well as a new test platform.

While AMD's Ryzen 3950X and its PCIe 4.0 compatibility were considered for our graphics test bed, in the end we went with the Intel Core i9 10900K, which outperformed the 3950X in our gaming benchmarks even with PCIe 4.0 cards like the RTX 30-series limited to PCIe 3.0 speeds. At least for now, raw CPU power seems to have a great influence on frame-rates than PCIe bandwidth - outside of a few specific scenarios.

The Core i9 10900K is locked to a 5GHz all-core frequency and backed by an Asus Maximus 12 Extreme Z490 motherboard and two 8GB sticks of G.Skill Trident Z Royal 3600MHz CL16. Cooling duties are handled by a 280mm AiO from Corsair. Our games are run from a capacious 2TB Samsung 970 Evo Plus NVMe drive provided by Box. The whole rig is powered by a 1000W Corsair power supply.

Now that you know the lay of the land, let's get into some gaming tests.

Nvidia GeForce RTX 3070 Analysis