Introduction

Gainward GeForce RTX 4070 Ghost is the company's most compact and affordable graphics card based on the new RTX 4070 Ada that NVIDIA is unveiling today, as it's priced bang on the NVIDIA-set MSRP of $600. You'll find that all the RTX 4070 graphics cards we're reviewing today (April 12) come at this price, as NVIDIA is giving them the special privilege of an earlier review publication date, to encourage board partners to come up with higher quality custom-designs at the baseline price. Reviews of premium, factory-overclocked custom-design RTX 4070 cards priced above the MSRP go live tomorrow (April 13), when these cards are available to purchase.



The new GeForce RTX 4070 is designed to consolidate NVIDIA's position in the performance-segment. This could be among the best-selling SKUs from this generation from the company, looking at how its predecessors such as the RTX 3070 fared. The RTX 4070 is designed for maxed out AAA gaming at resolutions up to 1440p, including real time ray tracing; or high refresh-rate competitive e-sports gameplay at 1080p and 1440p. 4K Ultra HD gaming is very much possible, but you'll need to know your way around your game settings, or use DLSS. Or you can just get GeForce Experience to find you the best settings.

The GeForce Ada graphics architecture powering the RTX 4070 in today's review, heralds the third generation of NVIDIA RTX, which has had a transformative impact on gaming graphics realism. NVIDIA figured that while fully ray traced 3D graphics may still be a distant technical reality for client applications, you could combine conventional raster 3D graphics with certain real time ray traced elements, such as reflections, lighting, shadows, global illumination, and motion blur; to significantly increase realism. Even this bit takes enormous compute power, and so NVIDIA loaded its silicon with fixed function hardware for the task, such as RT cores; while relying on an AI-based real time de-noising technology, which is where the acceleration from the Tensor cores helps. AI is also used in features such as DLSS, and more importantly, in the new DLSS 3 Frame Generation, which generates entire alternate frames using AI, without involving much of the graphics rendering machinery.

The GeForce RTX 4070 we're reviewing today is based on the same 5 nm AD104 silicon as the RTX 4070 Ti, but while the latter maxes out the silicon, the RTX 4070 is heavily cut down from it. This GPU is endowed with 5,888 CUDA cores, 46 RT cores, 184 Tensor cores, 184 TMUs, and 64 ROPs. It gets this shader count by enabling 46 out of the 60 streaming multiprocessors (SM) present on the silicon. Thankfully, the memory configuration is unchanged—you still get 12 GB of 21 Gbps GDDR6X memory across a 192-bit wide memory bus, with 504 GB/s of bandwidth on tap. The bandwidth is higher than the 448 GB/s that the previous-generation RTX 3070 contents with. The 192-bit memory bus may seem narrow, but is the result of NVIDIA trying to restructure the memory sub-system, with greater use of on-die caches on the GPU.

The Gainward GeForce RTX 4070 Ghost is designed for those who simply want a stock-priced RTX 4070 to install and forget about. It offers a no-frills product design, with its GPU sticking to NVIDIA reference clock speeds. The cooler design makes the Gainward Ghost among the most compact and lightweight custom-design RTX 4070 cards we've reviewed—which could be a good thing for SFF gaming PC builders. You also get the convenience of a single 8-pin PCIe power input. Yes—NVIDIA has allowed board partners to use the legacy connector, as the typical graphics power (TGP) of the RTX 4070 is just around 185 W.