Its specs have been carefully revealed to anyone who’s paying close attention, but today Nvidia is making it official: it’s launching its GeForce RTX 4070 graphics card at a starting price of $599. This GPU will be the cheapest way to buy Nvidia’s Ada Lovelace family of GPUs, which, in addition to better performance and power efficiency, gives you access to Nvidia DLSS 3 scaling and frame generation features, as well as hardware acceleration for AV1 video encoding. The cards will be available from Nvidia and other partners starting April 13th.
The Nvidia Founders Edition card is significantly smaller than its RTX 4090 and 4080 GPUs, with a relatively thin dual-slot cooler that will fit easily into most PC cases. GPUs made by Nvidia partners can (and will) be more expensive and physically larger, and some will also ship with 8-pin PCI Express power connectors rather than the newer but somewhat controversial 12VHPWR power connector that Nvidia uses.
The RTX 4070 is $100 more expensive than the RTX 3070 and RTX 2070’s $499 starting price, which was $120 more than the GTX 1070’s $379 starting price. Even adjusting for inflation since 2016 ($379 in 2016 is roughly $475 now), it continues the tradition of raising prices for GPUs with the same relative position in Nvidia’s lineup, which Nvidia partially (and not falsely) justifies by pointing out how much faster its modern graphics cards.
What does $599 get you?
| RTX 4090 | RTX 4080 | RTX 4070 Ti | RTX 4070 | RTX 3080 Ti | RTX 3080 10 GB | RTX 3070 Ti | RTX 3070 | RTX 3060 | |
|---|---|---|---|---|---|---|---|---|---|
| CUDA cores | 16 384 | 9728 | 7680 | 5888 | 10 240 | 8704 | 6144 | 5888 | 3584 |
| Clock boost | 2520 MHz | 2505 MHz | 2610 MHz | 2475 MHz | 1665 MHz | 1710 MHz | 1765 MHz | 1725 MHz | 1777 MHz |
| Memory bus width | 384-bit | 256-bit | 192 bit | 192 bit | 384-bit | 320 bit | 256-bit | 256-bit | 192 bit |
| Memory clock | 1313 MHz | 1400 MHz | 1313 MHz | 1313 MHz | 1188 MHz | 1188 MHz | 1188 MHz | 1750 MHz | 1875 MHz |
| Memory | 24GB GDDR6X | 16GB GDDR6X | 12GB GDDR6X | 12GB GDDR6X | 12GB GDDR6X | 10GB GDDR6X | 8GB GDDR6X | 8GB GDDR6 | 12GB GDDR6 |
| THP | 450 W | 320 W | 285 W | 200 W | 350 W | 320 W | 290 W | 220 W | 170 W |
The RTX 4070 has a lot in common with the 4070 Ti, a GPU originally billed as “RTX 4080 12GB”before it was “shut down”and restarted. They use the same AD104 GPU die, a smaller chip than the flagship AD102 used in the RTX 4090 or the AD103 used in the 4080 series. They also use 12GB of GDDR6X memory on a 192-bit interface. The 4070 has fewer CUDA cores (5888 compared to the 7680), but the 4070 and 4070 Ti are a lot more similar than the 16GB and 12GB RTX 4080 cards.
This 192-bit memory interface is narrower than the 256-bit interface used by the 3070 and 3070 Ti, which Nvidia made up for by adding 32MB of additional L2 cache to the 4070 (36MB in total). This cache, as well as the much higher base and boost clock speeds, are the 4070’s biggest improvements over the last generation cards.
On paper, there’s a lot to recommend for the 4070. It’s the first Ada Lovelace graphics card to launch for closer to $500 than $1,000, it’s the cheapest way to buy RTX 4000 series features, and at least you can buy versions that will physically fit in cases. smaller computers. It also promises to be significantly more power efficient than the 3070 and 3080 series cards it replaces (and competes with them given how widely the new 3000 series GPUs are available).
But in terms of raw rasterized game performance, it doesn’t look like as big of a flop as the 3070 did when it was released. This GPU could greatly outperform the more expensive RTX 2080 and 2080 Super of the previous generation for less money while trading blows with the RTX 2080 Ti. Nvidia’s performance numbers for the 4070 show it outperforming the 3070 Ti and either slightly outperforming or slightly behind the 10GB versions of the RTX 3080. and it still sells for at least the same price unless you’re buying used or refurbished, but previous mid-range GeForce cards have done a bit better at downgrading last-generation flagship performance to more reasonably priced PCs.
Nvidia is relying on DLSS 3 and AI-accelerated frame generation to make those numbers more impressive. This feature, as you may remember from our review of the RTX 4090, “doubles” the frame rate by producing one frame with AI interpolation for every frame rendered by the GPU. DLSS FG works in conjunction with the typical “super high resolution”scaling feature that adds pixels to the image your GPU is processing to make it look more detailed. The result can be a clean, high-resolution, high-frame-rate image rendered using only a fraction of the GPU power needed to originally render the same scene. Nvidia claims that the RTX 4070 can be up to 1.4x faster than the RTX 3080 when DLSS FG is enabled.
But as we found in our RTX 4090 review, it tends to perform best when your GPU is already putting out reasonable frame rates, making it hard to perceive flaws in AI-generated frames. In other words, it’s much better to make a 60 fps image look like a 120 fps image than it is to make a 15 fps image look like a 30 fps image. DLSS FG can also increase latency, making it harder to recommend for fast-paced competitive first-person shooters, and is compatible with a relatively narrow set of games (although Nvidia’s list of “50+ games”includes big names like Microsoft Flight Simulator, Diablo IV and Cyberpunk 2077).
As for AMD, we are still waiting for the RX 7000 series and RDNA3 architecture to move to cheaper GPUs, so AMD is fighting this battle with the latest generation cards. A recent email from AMD, which was likely sent ahead of the 4070’s release, beat the $579 and $649 MSRPs for the Radeon RX 6800 XT and 6950 XT. If you can’t find the 4070 in stock for $599 – and if you don’t mind the lack of DLSS support and poor ray tracing performance – either one is a plausible alternative to the 4070 for high quality 1440p for budget 4K. games.
We’ll take a closer look at the RTX 4070 in a full review shortly.