Nvidia GeForce RTX 4070 Ti Super review: More VRAM and bandwidth, slightly higher performance

The Nvidia RTX 4070 Ti Super marks the second salvo in Nvidia's mid-cycle 40-series Super refresh, taking over from the existing RTX 4070 Ti that will now be phased out. On paper, this looks like a nice upgrade, bumping from the AD104 to the AD103 GPU and, in the process, adding 4GB of memory and a 64-bit wider interface, plus 10% more shader cores, all at the same $799 MSRP. But therein lies the problem, as the Ada Lovelace architecture has always felt like too much of a generational increase in pricing. Still, the RTX 4070 Ti Super, with both of its suffixes, will go up against the best graphics cards, hoping to entice gamers who haven't snagged a GPU in a few years.

We noted at the announcement, as well as in the RTX 4070 Super review, that the cards appear to be launching in reverse order of excitement. We found plenty to like with the 4070 Super, and the additional memory finally gets Nvidia to 16GB of VRAM and a 256-bit interface for under $800. Prior to the 4070 Ti Super, you would have needed to spend 50% to get the RTX 4080 if you wanted that much memory and bandwidth, and at that point, why not just go all the way and buy an RTX 4090? It's not that $799 graphics cards are affordable, but they're certainly more affordable than what was previously available.

Realistically, no one who has already purchased an RTX 40-series GPU or one of AMD's RX 7000-series RDNA 3 GPU alternatives will be looking to buy one of the new Super models. Instead, it's those who haven't upgraded since the RTX 20-series and RX 5000-series (or earlier) that are the target. That means we're talking about people who waited an extra year for a bump in performance and specs and don't want to just give the current generation GPUs a pass and wait for RTX 50-series and RDNA 4. There are also people who might be building a new PC, of course, and if it's time to upgrade, then the Super cards all represent a move in the right direction: more performance, improved specs, and better values.

Here's the full rundown of the specifications for all the current generation Nvidia and AMD GPUs that have launched. Note that the RX 7600 XT and RTX 4080 Super are still missing, as those haven't quite launched yet — with the AMD part arriving tomorrow, alongside official retail availability for the 4070 Ti Super and the 4080 Super coming next week.

The RTX 4070 Ti Super officially launches on January 24, 2024, and we anticipate seeing plenty of retail inventory. The 4070 Super had a few of the base-priced models go in and out of stock over the past week, but overall availability has kept up with demand. Whether that's due to a large supply or a lack of demand is a different matter, of course.

Looking at the raw specs, the 4070 Ti Super has the same clocks as the 4070 Ti but bumps the SM count from 60 to 66, a 10% increase in theoretical compute. Of course, the bigger change comes in the memory subsystem and GPU, with the move to the larger AD103 die and a full 256-bit interface. Memory clocks remain at 21 Gbps, so you get 33% more memory and memory bandwidth. Total L2 cache remains the same as the 4070 Ti non-Super, as Nvidia limits the GPU to eight blocks of 6MB versus the maximum 8MB per 32-bit memory interface.

Based on the specifications, we expect to see around 10% higher performance in a lot of situations, with games and applications that are VRAM-limited potentially seeing a larger improvement. Spoiler: Most of the improvements track closer to the compute increase rather than the VRAM increase.

That might seem surprising, but the reality is that many games still don't need more than 12GB of VRAM, and the large L2 caches on Nvidia's Ada architecture really do a good job of alleviating raw memory bandwidth demands. There's also the noteworthy exception of being able to run AI and LLM workloads that need more than 12GB — and less than 16GB.

The remaining specs remain the same as the RTX 4070 Ti, with a 285W TGP (Total Graphics Power). For now, RTX 4070 Ti cards remain available for purchase, but Nvidia intends to phase those out, and we don't expect supply to last long. You might be able to score a sale in the short term, but that will also have to be balanced against the loss of VRAM.

Asus RTX 4070 Ti Super TUF Gaming

Image 1 of 10

Nvidia won't make a Founders Edition for the RTX 4070 Ti Super, leaving it to its add-in board partners. It did the same thing with the RTX 4070 Ti last year. Reviews of cards priced at the starting $799 MSRP can go up today, with reviews of higher-priced cards going live tomorrow — though both will have retail availability tomorrow. For this review, Nvidia shipped us the Asus RTX 4070 Ti Super TUF Gaming. This is a reference clocked card, and we would expect factory overclocked models to perform perhaps 5% better overall.

Our understanding is that all RTX 4070 Ti Super cards will come with the new 16-pin 12V-2x6 power connector. That means they'll also include a dual 8-pin to 16-pin adapter, which we still think is an ugly solution. If you haven't upgraded to a new ATX 3.0 power supply, that might be something to consider adding to the shopping cart to go with your graphics card upgrade.

The Asus card's design is something we've seen plenty of times already. You get triple 89mm fans, and the middle one spins clockwise while the outside fans spin counterclockwise, supposedly reducing turbulence and improving cooling. The card measures 301x137x64 mm (our measurements — and most of the card is 129mm tall, except for the last 5cm) and weighs 1323g without the 16-pin adapter, or 1354g, including the adapter. Incidentally, that's slightly wider than a triple-slot card, so plan on not having access to four slots, even though the IO bracket is only a 2-slot width.

Image 1 of 5

Asus has several tiers of graphics cards, with the TUF Gaming brand occupying the middle range. ROG Strix cards carry a significant price premium and come with more RGB lighting, improved power circuitry, and a few other potential extras. As such, the only RGB lighting on the 4070 Ti Super we're looking at today is the TUF logo and strip at the top-rear edge of the card.

As we've seen with a lot of higher-end Asus cards, you get a slight upgrade in display outputs: three DP1.4a ports and two HDMI 2.1 ports. Our understanding is that only four can be used simultaneously, but since HDMI 2.1 offers higher bandwidth, it's a more flexible configuration than the typical triple DP and single HDMI we commonly see.

There aren't many extras in the package, but one nice feature is the small magnetic support stand that Asus includes. It has a limited height range and may not work with all cases and graphics cards, but with the weight of many modern GPUs, it's helpful to have support at the opposite end of the card, away from the IO ports. We've heard about cracking graphics card PCBs, so if you have a large and heavy graphics card, such supports are basically required for long-term functionality in our book.

Nvidia RTX 4070 Ti Super Test Setup

Our current graphics card testbed has been in use for over a year now, and so far we haven't seen any pressing need to upgrade. The Core i9-13900K is still holding its own, and while the i9-14900K or Ryzen 9 7950X3D can improve performance slightly, at higher resolutions and settings we're still almost entirely GPU limited — though perhaps not when the future RTX 50-series and RX 8000-series arrive. We also conduct professional and AI benchmarks on our Core i9-12900K PC, which is also used for our GPU benchmarks hierarchy.

We're using updated preview drivers from Nvidia, version 551.15. In our 4070 Super testing, we noticed some changes in several of the benchmarks and retested the Nvidia cards in the affected games. We've also retested AMD's GPUs using the 23.12.1 drivers. The most impacted games are Borderlands 3, Far Cry 6, Forza Horizon 5, Microsoft Flight Simulator, and Spider-Man: Miles Morales. We've also fully retested the RTX 4070 and RTX 4070 Ti.

There's not really a direct previous generation part for the RTX 4070 Ti Super, but we've tossed in results from the previous generation RTX 3070 Ti and the four-year-old RTX 2070 Super as well. Both of those were recently retested, so their results should be fully up to date.

Otherwise, as we're nearly finished with the current generation GPUs, we've opted to include all the latest AMD and Nvidia GPUs in our charts, minus the RTX 4090 — because it costs over twice as much and is becoming difficult to find, plus we need to save something for next week's 4080 Super launch. The Arc A770 16GB is also representing Team Blue, obviously at a much lower performance bracket.

Our current test suite consists of 15 games. Of these, nine support DirectX Raytracing (DXR), but we only enable the DXR features in six games. The remaining nine games are tested in pure rasterization mode. While many of the games in our test suite support upscaling — twelve support DLSS 2, five support DLSS 3, five support FSR 2, and four support XeSS — our focus is primarily on native resolution performance. GPUs generally show similar scaling if you enable DLSS/FSR2/XeSS, and we do have some additional 'bonus' test results that show how the 4070 Ti Super and other GPUs perform with Quality mode upscaling.

We tested at 1080p (medium and ultra), 1440p ultra, and 4K ultra for this review — ultra being the highest supported preset if there is one, and in some cases maxing out all the other settings for good measure (except for MSAA or super sampling). Like the RTX 4070 Super, we still feel 1440p native resolution gaming is the optimal target for the 4070 Ti Super. It's also a good stand-in for the kind of performance you'll see at 4K with Quality mode DLSS enabled.

Our PC is hooked up to a Samsung Odyssey Neo G8 32, one of the best gaming monitors around, allowing us to fully experience the higher frame rates that might be available. G-Sync and FreeSync were enabled, as appropriate. As we're not testing with esports games, most of our performance results are nowhere near the 240 Hz limit, or even the 144 Hz limit of our secondary test PC.

Image 1 of 8

We've installed Windows 11 22H2 and used InControl to lock our test PC to that major release for the foreseeable future (though critical security updates still get installed monthly — and one of those probably caused the drop in performance that necessitated retesting a third of the games in our suite).

Our test PC includes Nvidia's PCAT v2 (Power Capture and Analysis Tool) hardware, which means we can grab real power use, GPU clocks, and more during all of our gaming benchmarks. We'll cover those results on our page on power use.

Finally, because GPUs aren't purely for gaming these days, we've run some professional content creation application tests, and we also ran some Stable Diffusion benchmarks to see how AI workloads scale on the various GPUs.

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content