We enter the 2nd quarter of 2023 with the announcement of the Nvidia Geforce Rtx 4070 12GB GDDR6X Graphics Card. The RTX 4070 is the successor of the fan-favorite RTX 3070 GPU. In typical Nvidia fashion and as seen with the recent RTX 4070 Ti, RTX 4080, and RTX 4090, the RTX 4070 comes with a significant price bump to $599 from the $499 launch price. Nvidia claims the RTX 4070’s performance to be on par with the $799 RTX 3080 on pure rasterization and up to 1.2x with DLSS 3 all while being almost 50% more power efficient. Let’s put the Nvidia GeForce RTX 4070 Founders Edition to the test.
Nvidia Ada Lovelace Architecture
Following the success of the RTX 30 series card and the Ampere architecture, Nvidia continues to double down on its proven and tested formula with the debut of the Nvidia Ada Lovelace architecture featuring 76 Billion Transistors. Nvidia has moved from Samsung 8N to TSMC’s 4N process, resulting in much bigger performance gains and slighter higher pricing as TSMC recently hiked up its prices. The new Ada Lovelace features 3rd Generation RT Cores at 200 RT TFLOPs and 4th Gen Tensor Cores for 1,4000 TFLOPs as well as a new streaming multiprocessor with up to 2x power efficiency. Moreover, Ada Lovelace also brings in new features exclusive to the RTX 40 series graphics card such as DLSS 3, AV1 Encoding, and Shader Execution Reordering.
Shader Execution Reordering
Shader Execution Reodering largely contributes to the performance gains of the RTX 40 series GPUs thanks to the more efficient usage of the shaders. SER is a scheduling system that reorders shading work on-the-fly for better execution efficiency, especially on ray-traced workloads. Think hyperthreading plus thread director but for GPUs particularly the RTX 40 series where the rays are reordered so that the ones that are running on the same program or similar are grouped together resulting in the shader working consistently instead of waiting on each other.
SER is an NVAPI extension meaning that developers would have to enable it on their games. We should expect SER to be heavily and quickly adopted especially in games where there are already Nvidia features enabled such as Reflex, RTX, and DLSS.
Nvidia DLSS 3
DLSS 3 uses the 4th Gen Tensor Cores and the Optical Flow Accelerator of the RTX 40 series to improve overall performance using AI-generated frames. In addition to already upscaling existing frames like DLSS 2, DLSS 3 generates intermediate frames in your game aka inserting AI-generated frames to improve performance. The RTX 40 series GPU will analyze two in-game frames and generate an optical flow field, closely looking at all elements in the game and using the information gathered to construct a frame in between. Additionally, DLSS 3 incorporates Nvidia Reflex ultra-low-latency solution making your games way more responsive and providing more frames. Games that support DLSS 3 will automatically support Nvidia Reflex.
DLSS 3 can be used by older Nvidia RTX graphics cards but won’t be able to fully maximize the feature as the biggest improvement of the technology, Frame Generation, is exclusive to the RTX 40 series graphics cards.
Nvidia RTX 40 Series Specifications
RTX 40 SERIES SPECS | RTX 4090 | RTX 4080 16GB | RTX 4070 Ti | RTX 4070 |
---|---|---|---|---|
Cores | 16,384 | 9,728 | 7,680 | 5,888 |
Tensor Cores | 184 | |||
RT Cores | 46 | |||
Boost Clock (GHz) | 2.52 | 2.51 | 2.61 | 2.47 |
Base Clock (GHz) | 2.23 | 2.21 | 2.31 | 1.92 |
Memory Config | 24GB GDDR6X | 16GB GDDR6X | 12GB GDDR6X | 12GB GDDR6X |
Memory Interface | 384-bit | 256-bit | 192-bit | 192-bit |
Graphics Power | 450W | 320W | 285W | 200W |
Required System power | 850W | 750W | 700W | 650W |
Price | Php 112,300 | Php 84,200 | Php 63,200 | Php 40,590 |
The Nvidia GeForce RTX 4070 sports an average of 24% fewer cores than the RTX 4070 Ti and with a significantly lower base clock of only 1,920 Mhz, making it the lowest-clocked RTX 40 GPU. It does, however, sport the same 12GB GDDR6X and 192-bit memory configuration. The 24% lower core count translates to a much lower TGP of only 200 watts and a 35 percent cheaper price tag of only $499 (Php 40,590) albeit it’s $100 more than the launch price of the GeForce RTX 3070.
Nvidia RTX 4070 Founders Edition
Like the previous generation RTX 3070, Nvidia has a Founders Edition variant which is quite surprising as the higher tier RTX 4070 Ti didn’t have an FE model.
The Nvidia GeForce RTX 4070 Founders Edition has similar packaging to the RTX 4080 and RTX 4090 Founders Edition variants. Apart from the GPU itself, the package contains a 12VHPWR adapter.
The RTX 4070 Founders Edition uses the same FE cooler design albeit in a much more compact form factor. Its aesthetic closely matches the RTX 3080 FE design but with the size measurements of an RTX 3070 FE.
The back part shows the stunning backplate with an exposed second fan completing the flow-through cooler design. The compact nature of the RTX 4070 FE makes the FE design much more appealing compared to the bulkier versions found on the RTX 4080 and RTX 4090.
While the RTX 4070 Founders Edition uses the controversial 12VHPWR connector, Nvidia allowed its board partners to use the standard 8-pin PCI-E connector as the RTX 4070 has a rated 200W TGP. The included 12VHPWR adapter uses two 8-pin PCI-E cables – quite an overkill for a 200W TGP card, but hey at least you could reuse the adapter on the Ti variant.
The RTX 4070 Founders Edition’s size sits in between the RTX 3080 FE and RTX 4070 FE – It’s basically what the RTX 3070 FE would look like if it used a shorter version of the RTX 3080 FE cooler. Its Dual-slot design paired with an efficient 200W TGP makes it an ideal graphics card for ITX and SFF builds.
The Nvidia GeForce RTX 4070 Founders Edition has the standard display output configuration with three DisplayPort 1.4a ports and one HDMI 2.1 port. The lower half of the IO bracket has large cutouts for ventilation.
Test Bench Setup and Benchmark Methodology
Gadget Pilipinas’ testing philosophy is to provide detail-oriented results as accurately as possible so that our readers can replicate our tests given that these conditions are met. Different benchmarking apps and sequences are used depending on the component or device being tested.
RTX 4070 Founders Edition Testbench Specs
CPU | Intel Core i9 13900K |
COOLER | ASUS ROG RYUJIN II 360mm V2 – Noctua NT-H2 Thermal Paste |
MOTHERBOARD | ASUS ROG Strix Z790-E Gaming |
MEMORY | Kingston Fury Beast RGB 32GB(2x16GB) 6000 MHz DDR5 |
GPU | Nvidia RTX 4090 FE | Colorful iGame RTX 4090 Vulcan OC-V | RTX 4080 FE | ASUS TUF RTX 4080 | 531.42 driver |
STORAGE | Kingston KC3000 1TB PCI-E Gen 4 SSD |
POWER SUPPLY | Gigabyte UD1000GM PG5 1000W |
OPERATING SYSTEM | Windows 11 Pro Build 22H2 |
DISPLAY | Lenovo Legion Y27Q Gaming Monitor |
We use CapFrameX 1.7.0 Beta as our primary FPS capture and analysis tool for all our gaming benchmarks. The latest build version of Windows 11 Pro and WHQL-certified drivers are used for our benchmarks. Readings such as temperatures and power draws are recorded using HWINFO64, and other relevant software for cross-checking.
RTX 4070 Founders Edition Benchmarks
COUNTER-STRIKE: GLOBAL OFFENSIVE
Counter-Strike: Global Offensive aka CSGO is an 11-year-old first-person shooter title that’s still widely popular despite its old Source Engine. CS:GO is a great game to test out CPU performance as the game is heavily CPU-dependent like any other FPS title. Our CS:GO benchmark process uses ULLETICAL’s FPS Benchmark workshop mod to measure the hardware’s performance.
DOTA 2
Most gamers play at least one of the following free-to-play titles: League of Legends, CS: GO, Dota 2, and/or Fortnite. Dota 2 is a good representation of the F2P titles as it is the most demanding game in the spectrum. The benchmark will give you an idea of the graphic card’s relative performance on other lesser demanding titles. Our benchmark sequence is based on a replay of OG vs. Liquid in the TI9 grand finals from the team fight that happened from 28:30 to 29:30.