Product Listing

NVIDIA GeForce RTX 3070 Founders Edition review: The end of the 1080p gaming era

By Aaron Yip - 31 Oct 2020

Benchmark Numbers

Test Setup

I am using the same test rig as the one used to test previous RTX 3080 and RTX 3090 graphics cards, and comes with the following specifications:

  • 10th Gen Intel Core i9-10900K (3.7GHz base clock, 20MB Intel Smart L3 cache)
  • ASUS ROG Maximus XII Formula (Intel Z490)
  • 32GB G.Skill DDR4 memory
  • WD Black SN750 2TB SSD
  • Windows 10 Home 64-bit
  • Acer Predator X27

I'll be stacking the RTX 3070 Founders Edition not only against the more powerful RTX 3080 and RTX 3090 Founders Edition cards, but also the full suite of the RTX 2080 Founders Edition cards, including the original 20802080 Super and RTX 2080 Ti. NVIDIA has claimed that the RTX 3070 offers similar performance to the RTX 2080 Ti, but at almost half the price. So let's pay extra attention to the comparisons between these two cards below.

Test cards compared
  NVIDIA GeForce RTX 3070 Founders Edition NVIDIA GeForce RTX 3080 Founders Edition NVIDIA GeForce RTX 3090 Founders Edition NVIDIA GeForce RTX 2080 Ti Founders Edition NVIDIA GeForce RTX 2080 Super Founders Edition NVIDIA GeForce RTX 2080
  NVIDIA GeForce RTX 3070 Founders Edition NVIDIA GeForce RTX 3080 Founders Edition NVIDIA GeForce RTX 3090 Founders Edition NVIDIA GeForce RTX 2080 Ti Founders Edition NVIDIA GeForce RTX 2080 Super Founders Edition NVIDIA GeForce RTX 2080
Core Code
  • GA104
  • GA102-200
  • GA102
  • TU102
  • TU104
  • TU104
GPU Transistor Count
  • 17.4 billion
  • 28.3 billion
  • 28,300 million
  • 18.9 billion
  • 13.6 billion
  • 13.6 billion
Manufacturing Process
  • 8nm (Samsung)
  • 8nm (Samsung)
  • 8nm (Samsung)
  • 12nm FinFET
  • 12nm FinFET
  • 12nm FinFET
Core Clock
  • Base: 1,500MHz
  • Boost: 1,730MHz
  • Base: 1,440MHz
  • Boost: 1,710MHz
  • Base: 1,395MHz
  • Boost Clock: 1,695MHz
  • Core: 1,350MHz, Boost: 1,635MHz
  • Core: 1,650MHz, Boost: 1,815MHz
  • Core: 1,515MHz, Boost: 1,800MHz
Stream Processors
  • 5,388
  • 8,704
  • 10,496
  • 4,352
  • 3,072
  • 2,944
Texture Mapping Units (TMUs)
  • 184
  • 272
  • 328
  • 272
  • 192
  • 184
Raster Operator units (ROP)
  • 64
  • 96
  • 112
  • 88
  • 64
  • 64
Memory Clock (DDR)
  • 8GB GDDR6
  • 14,000MHz
  • 10GB GDDR6X
  • 19,000MHz
  • 24GB GDDR6X
  • 19,500MHz
  • 11GB GDDR6 14,000MHz
  • 8GB GDDR6 15,500MHz
  • 8GB GDDR6 14,000MHz
Memory Bus width
  • 256-bit
  • 320-bit
  • 384-bit
  • 352-bit
  • 256-bit
  • 256-bit
Memory Bandwidth
  • 512GB/sec
  • 760GB/sec
  • 936.2GB/s
  • 616GB/s
  • 496.1GB/s
  • 448GB/s
PCI Express Interface
  • PCI 4.0 / 3.0
  • 4.0 / 3.0
  • PCI 4.0 / 3.0
  • 3.0
  • 3.0
  • 3.0
Power Connectors
  • 1 x 12p-pin
  • 1 x 12p-pin
  • 1 x 12p-pin
  • 2 x 8-pin
  • 1x 8-pin, 1x 6-pin
  • 1x 8-pin, 1x 6-pin
Multi GPU Technology
  • No
  • No
  • NVLink
  • Yes (2-way SLI)
  • Yes (2-way SLI)
  • Yes (2-way SLI)
DVI Outputs
  • No
  • No
  • No
  • No
  • No
  • No
HDMI Outputs
  • 1
  • 1
  • 1
  • 1x
  • 1x
  • 1x
DisplayPort Outputs
  • 3
  • 3
  • 3
  • 3x
  • 3x
  • 3x
HDCP Output Support
  • Yes
  • Yes
  • Yes
  • Yes
  • Yes
  • Yes
Stream Processor Clock
  • 1,650MHz

Benchmarks

As for benchmarks used, these are the following tools and games that I've drawn up to run the RTX 3070 Founders Edition on. 

  • 3DMark
  • Horizon Zero Dawn
  • Wolfenstein: Youngblood
  • Deus Ex: Mankind Divided
  • Shadow of the Tomb Raider
  • Metro: Exodus

For temperature, I ran "Time Spy Extreme" in Stress Mode to measure the card's peak heat level.

 

3DMark

3DMark is a synthetic benchmark that tests graphics and computational performance at different resolutions, starting at 1080p and going all the way up to 4K. A series of two graphics test, one physics test, and then a combined test stresses the hardware in turn to assess its performance.

The chart above shows performance normalcy with the RTX 3070 performing behind the quicker RTX 3080 and RTX 3090 cards. It's when we compare it with the RTX 2080 Ti that's where we see the RTX 3070 matching and even slightly edging the latter. It looks like NVIDIA claiming that the RTX 3070 is capable of matching and in certain scenarios even outperforming the older RTX 2080 Ti holds some truth to it.

Since the RTX cards support hardware ray tracing through its RT Cores, I’ve also used 3DMark’s Port Royal Ray Tracing benchmark to size up each card’s ray-tracing capabilities. While it's no surprise that the RTX 3080 and RTX 3090 cards left everyone else in smokes, what's eye-catching is how far the RTX 3070 lags behind both cards - considering the performance difference between the RTX 3080 and RTX 3090 isn't that wide a gap.

You'd also notice that the RTX 3070 actually performed poorly in ray-tracing even against the RTX 2080 Ti, but that's really not surprising. You see, the RTX 2080 Ti still has a higher RT Core count (68) over the RTX 3070's RT Core count of 46. For the uninitiated, RT Cores process all the BVH traversal and ray-triangle intersection testing, aka ray tracing. The more RT Cores a card have, the better at ray tracing performance it is. Going by pure numbers, the RTX 3070 should actually rack up much lower performance figures, but it didn't because the RT cores in the Ampere GPUs have been beefed up considerably. So while it still loses to an RTX 2080 Ti, the beefed-up 46 RT cores on the RTX 3070 managed a closer standing. We’ll dive deeper into this with real-world games benchmarks next page. In the meantime, let’s look at how the cards fared with DLSS turned on.

For the uninitiated, deep learning super sampling or DLSS uses artificial intelligence and machine learning to produce an image that looks like a higher-resolution image, without the rendering overhead. The idea is to make games rendered at 1440p look like they’re running at 4K resolution, or 1080p games to look like 1440p resolution. This is all possible thanks to NVIDIA’s Tensor cores, which are only available in RTX GPUs.

Although RTX 20 series GPUs have Tensor cores inside, the RTX 30 Series come with NVIDIA’s second-generation Tensor Cores, which offer greater per-core performance – and this makes for some interesting insights using the results above. You see, even though the RTX 3070 has a much lower Tensor Core count than the RTX 2080 Ti (184 versus 544), it actually performed better.

Now let’s look at how the card performs in games.

Join HWZ's Telegram channel here and catch all the latest tech news!
9.0
  • Performance 9
  • Features 8.5
  • Value 9
The Good
Excellent price point
Great performance for 1440p gaming
Solid build
Form factor suitable for powerful compact rigs
The Bad
Requires cable adapter
Supply expected to be very limited at launch
Average ray tracing performance
Not exactly future proofed
Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.