I know at least for my wife and myself, when shopping for our PC upgrades the highest-end cards are typically out of our budget and frankly that is fine because with our monitors we are also not normally gaming at 4k where those cards are needed. That’s why when new generations of cards start getting down into the middle of the product stack things start to get exciting and with tomorrow's launch of the Nvidia RTX 4070 we are starting to get into that range. So today I’m excited to check out the RTX 4070 Founders Edition to see how Nvidia’s new GPU performs. Before doing that I will dig into the specs to see what it has going on and then take a closer look at the card itself. So let’s dive in!

Product Name: Nvidia RTX 4070 Founders Edition

Review Sample Provided by: Nvidia

Written by: Wes Compton

Amazon Affiliate Link: HERE

 

RTX 4070 Details

The Nvidia RTX 4070 is the next addition to their 4000 Series of cards and like the others, it is based on the same Ada Lovelace architecture. This means that it does have the same full support for DLSS 3 and the other changes that the other cards have like the new fourth-generation tensor AI-focused cores, third-generation ray tracing cores, and the AV1 encoders which will be huge for streamers and also a way to lower streaming bandwidth while improving quality. The 4070 sits below the 4070 Ti in the lineup but is based on the same AD104 GPU as the 4070 Ti and 4080 which is built on the 4N custom process by TSMC. The RTX 4070 may use the same AD104 GPU but Nvidia has cut it down slightly with 4 GPCs or graphics processing clusters to the 4070 Ti’s 5. Translates to 23 TPCs or texture processing clusters and 46 streaming multiprocessors. For CUDA cores the RTX 4070 has 5888 down from the 7680 of the 4070 Ti which is a 23% drop. The tensor core count is 184, the ray tracing core count is 46, and the texture units are 184 as well which are all a 30% drop from the 240/60/240 on the RTX 4070 Ti. The RTX 4070 also has a lower boost clock with it running at 2475 MHz to 2610 MHz but the memory is clocked the same and has the same 21 Gbps data rate. They also have the same 12 GB of GDDR6X which is notably an improvement over the last generations RTX 3070 which has 8GB but the memory interface for the 4070 is 192-bit which is a downgrade but does match the 4070 Ti. The cut down GPU and lower clock speeds did help with the overall TDP though which is now 200 watts compared to the 285 watts on the RTX 4070 Ti.  

image 1

 

RTX 3070

RTX 4070

RTX 4070 Ti

Graphics Processing Clusters

6

4

5

Texture Processing Clusters

23

23

30

Streaming Multiprocessors

46

46

60

CUDA Cores

5888

5888

7680

Tensor Cores

184 (3rd Gen)

184 (4th Gen)

240 (4th Gen)

RT Cores

46 (2nd Gen)

46 (3rd Gen)

60 (3rd Gen)

Texture Units

184

184

240

ROPs

96

64

80

Boost Clock

1725 MHz

2475 MHz

2610 MHz

Memory Clock

1750 MHz

1313 MHz

1313 MHz

Memory Data Rate

14 Gbps

21 Gbps

21 Gbps

L2 Cache Size

4MB

36 MB

49 MB

Total Video Memory

8 GB GDDR6X

12 GB GDDR6X

12 GB GDDR6X

Memory Interface

256-bit

192-bit

192-bit

Total Memory Bandwidth

448.0 GB/sec

504 GB/sec

504 GB/sec

Texture Rate (Bilinear)

317.4 Gigatexels/sec

455.4 Gigatexels/sec

626 Gigatexels/sec

Fabrication Process

Samsung 8 nm 8N NVIDIA Custom Process

TSMC 4N NVIDIA Custom Process

TSMC 4N NVIDIA Custom Process

Transistor Count

17.4 Billion

35.8 billion

35.8 billion

Connectors

3 x DisplayPort

1 x HDMI

3 x DisplayPort

1 x HDMI

3 x DisplayPort

1 x HDMI

Power Connectors

1x12 pin

(Dongle to 1x 8-Pin)

2x PCIe 8-pin cables OR 300w or greater PCIe Gen 5 cable

2x PCIe 8-pin cables OR 300w or greater PCIe Gen 5 cable

Minimum Power Supply

650 Watts

650 Watts

700 Watts

Total Graphics Power (TGP)

220 Watts

200 Watts

285 Watts

Maximum GPU Temperature

93° C

90° C

90° C

PCI Express Interface

Gen 4

Gen 4

Gen 4

Launch MSRP

$499

$599

$799

 

With the RTX 4070 having support for DLSS 3 it supports DLSS Frame Generation which I’ve tested in the past and has been a huge performance improvement, especially in games that are CPU limited. The combination of DLSS Super Resolution and frame generation mean that when both are on the card renders 1 out of 8 pixels with traditional rendering. Super Resolution scales that up and then frame generation can use AI to render an entire frame in between the traditional frames. This is also why they can improve performance even when CPU-limited. DLSS 3 is being adopted quicker than DLSS 2 at a rate of 7 times faster according to Nvidia. They do have a good list of titles going and they finally have one website put together that shows you all of the games that support ray tracing, DLSS 3, DLSS 2, and AI processing which makes it easier to see if the games you want to play are included, here is a link.

image 6

image 7

On the call about the 4070 Nvidia also showed some of the numbers that they are seeing as far as ray tracing and DLSS use for 4000 series card owners as well as percentages of people who have 144hz or higher monitors and people who are gaming at 4k. At 83% that is a big portion of the people who are buying 4000 series cards who are using it and DLSS was a little lower but not far at 79%. They also compared the same numbers with 2000 Series owners which was the first generation with RTX support and those numbers are significantly lower. It makes sense though, with the original 2000 Series only the highest ends cards were able to handle RTX and DLSS has improved a lot. Having DLSS 3 on the new cards helps open up ray tracing performance in situations where you might otherwise skip it for the performance.

image 8

image 4

On that same note, they also showed just how much more RTX offers with today's offers whereas back in 2018 the cards were limited and games had to pick and choose which effects they wanted to utilize. This can really be seen when you see the change in ray tracing operations per pixel between Battlefield V and Cyberpunk 2077 with it going from 39 to 635. Nvidia also mapped out the number of transistors on their cards between then and now as well on the same graph. This does a great job of showing that while the cards have grown in processing power, the about of ray tracing operations per pixel has out scaled that by a huge amount.

image 3

image 5

Another big announcement, at least for me was that VLC is adding RTX Video Super Resolution. This one hits home because I watch a lot of TV both in the living room and on my PC. In the living room I have an Nvidia TV which has had AI upscaling for a while now and I recently noticed that some of the video files that I have on the network were a lot lower in resolution than I thought when I want to watch them on VLC on my PC. I hadn’t noticed it on the TV. Video Super Resolution could do the same on VLC, improving the resolution of all of your movies/videos as you watch them.

image 2

Before getting into testing I did also run GPUz to double-check that our clock speeds match up with the specifications. The RTX 4070 Founders Edition came in at 2475 MHz for the boost clock which matches the specifications and for the driver I used the 531.42 driver that Nvidia provided ahead of the launch. The board BIOS is also noted for future reference if needed.

image 41

 

 

Log in to comment

We have 1743 guests and no members online

supportus