The GeForce RTX 4060 Ti was arguably Nvidia's least-appealing last-generation graphics card. It shipped for $399 with just 8GB of GDDR6 memory and a crippling 128-bit memory interface, which combined to make it a deeply unappealing option compared with the standard RTX 4060, the RTX 4070, or AMD’s Radeon RX 7700 XT. (Nvidia also released a 16GB version for $499, an even tougher sell.) Today's release of the Nvidia GeForce RTX 5060 Ti, available with either 8GB ($379) or 16GB ($429) of GDDR7 video memory, is poised to right some of these wrongs. We've put PNY’s $429 GeForce RTX 5060 Ti 16GB Overclocked Dual Fan model through our rigorous tests to see how it fares.

While Nvidia, bafflingly, retained its narrow 128-bit memory interface, the RTX 5060 Ti is far more appealing than the RTX 4060 Ti ever was, with GDDR7 offsetting this shortcoming quite well. This point is particularly true when considering the RTX 4060 Ti 16GB launched at $499, with the RTX 5060 Ti 16GB costing a surprising $70 less. The RTX 5060 Ti is ideal for gamers who want to play at 1440p or 1600p, where it can run current games with near-maximum settings while maintaining a smooth refresh rate. That earns it our Editors’ Choice award for mainstream 1440p play. (Gamers looking to run games at 1080p with a high refresh rate would also get a lot of joy from the RTX 5060 Ti, but it's edging on overkill for that resolution.) All this is contingent, of course, on the supply of the GPUs and how the ultimate pricing shakes out for US shoppers, with tariffs and possible shortages looming.

Configuration: New and Improved, But Why Still 128-Bit?

The GeForce RTX 5060 Ti houses a GB206 graphics processor based on Nvidia’s "Blackwell" graphics architecture. This is the first desktop graphics card we’ve seen to use this graphics chip, which is similar to the AD106 chip inside the Nvidia GeForce RTX 4060 Ti but with some enhancements.

In addition to the performance boost the Blackwell architecture provides, the RTX 5060 Ti also has a higher core count than the RTX 4060 Ti. It has 36 streaming multiprocessors (SMs) compared with the RTX 4060 Ti’s total of 34. Most of the hardware resources of an Nvidia graphics chip sit inside these SMs, and so this increase brings with it a higher number of CUDA cores, texture mapping units (TMUs), tensor cores, and ray-tracing cores, all of which you can see in the spec chart.

Nvidia’s reference design for the RTX 5060 Ti calls for it to be clocked slightly higher than the RTX 4060 Ti. (Note: There is no Founders Edition version of this card.) This increase in clock speed, combined with the bump up in hardware resources, should provide around a 10% increase in performance, in addition to any gains derived from the improved Blackwell architecture.

Unfortunately, the RTX 4060 Ti's least appealing feature, its narrow 128-bit memory interface, lives on in the RTX 5060 Ti. Memory bandwidth is essential for driving fast graphics performance, and this design choice hampered the RTX 4060 Ti's performance, particularly at higher resolutions. This 128-bit-wide memory interface is one of the key reasons the RTX 4060 Ti was a bummer of a card.

(Credit: Joseph Maldonado)

This interface remains a limiting factor on the new Nvidia GeForce RTX 5060 Ti, but it is far less significant than it was on the 4060 Ti. Nvidia equipped the RTX 5060 Ti with 28Gbps GDDR7 memory, which is far faster than the 18Gbps GDDR6 memory that the 4060 Ti shipped with. This resulted in a greater than 50% increase in overall memory bandwidth (to 448GBps, from 288GBps).

(Credit: Joseph Maldonado)

Assuming you can find a card at the actual MSRP, you'll also find improvement in this generation in terms of pricing, particularly on the 16GB models. The RTX 4060 Ti launched at $399 for the 8GB model and a prohibitively high price of $499 for the 16GB model, which I genuinely hope no one paid full price for. The 8GB RTX 5060 Ti is a little more affordable than its predecessor at $379, and the 16GB model is much more enticing at $429. I would have considered this a bit high for the RTX 4060 Ti in retrospect due to its memory issues, but as you are about to see, GDDR7 goes a long way to address those limitations.

A 'Quick' Aside About Memory Bandwidth

The memory interface on a component like the graphics card essentially works like pipes; think of data as the water. The wider the interface, the more pipes you have, and the more data can travel to or from the graphics chip at any given time. Naturally, if we increase the number of pipes by widening the interface from, say, 128 bits wide (like on the Nvidia GeForce RTX 5060 Ti) to 192 bits wide (like on the Nvidia GeForce RTX 5070), we enable more data to be in motion between the graphics chip and the VRAM at any given time. We typically measure the bandwidth of how much data can pass through this connection in terms of gigabytes per second, or GBps for short.

Widening the memory interface is an effective way to increase the amount of bandwidth a graphics card has. But it comes at a cost: You will need more memory chips, a more complicated PCB, and additional circuitry on the graphics chip itself to support this. All of that naturally will impact the price, but there's an easier way to increase bandwidth: by simply opting for VRAM that runs at a higher speed. Going back to our pipes analogy, this effectively works like increasing the water pressure, forcing more through each second.

This is why the adoption of GDDR7 memory for the Nvidia GeForce RTX 5060 Ti is so impactful. This GDDR7 is rated to push 28Gbps (gigabits per second) as opposed to the 18Gbps of the GDDR6 used on the Nvidia GeForce RTX 4060 Ti. The math for calculating bandwidth works like this:

Memory interface width in bits x RAM speed = Bandwidth in Gbps

From here, all we have to do in order to get the bandwidth in GBps is to divide by eight. Plugging in those numbers for the RTX 4060 Ti, we get this this:

128 x 18 = 2,304Gbps

Dividing that by eight gives us a final bandwidth for the RTX 4060 Ti of 288GBps.

If we do the same for the Nvidia GeForce RTX 5060 Ti, we have the following equation...

128 x 28 = 3,584Gbps

...which then gives us a total bandwidth of 448GBps after dividing by eight.

Effectively, this means that the Nvidia GeForce RTX 5060 Ti has roughly 56% more bandwidth than the RTX 4060 Ti while maintaining the same 128-bit interface. This goes a long way to reducing the negative impacts of having a 128-bit bus.

PNY's Card Design: Two Fans and No RGB

Before delving into the benchmark tests, let’s look at the specific RTX 5060 Ti model we received for review.

PNY’s GeForce RTX 5060 Ti 16GB Overclocked Dual Fan graphics card is a relatively plain model in the series. It lacks RGB lighting and uses a simple dual-fan thermal solution. It also has a metal backplate and occupies only two open expansion slots.

(Credit: Joseph Maldonado)

This card's rear I/O bracket features three DisplayPort connections and a single HDMI port.

For power delivery, this particular PNY card requires just a single 8-pin power connection, which is a refreshing change from the banks of multiple 6-pin or 8-pin connectors or the polarizing, Nvidia-specific 12VHPWR connector that has dominated the higher-end RTX 50-series cards we have tested. PNY suggests a minimum 600-watt PSU for use with this card.

(Credit: Joseph Maldonado)

Test Setup and Competition

We tested the PNY GeForce RTX 5060 Ti 16GB Overclocked Dual Fan graphics card on our recently rebuilt graphics card testbed, which features an AMD Ryzen 9 9950X CPU. This CPU is installed on a Gigabyte X870E Aorus Master motherboard with two 16GB sticks of DDR5 RAM in a dual-channel configuration, using an AMD EXPO memory profile to set the RAM to 6,000MHz. The CPU is also cooled by a 360mm Cooler Master liquid cooler.

This system is configured with two 2TB Crucial PCIe 4.0 M.2 SSDs, one dedicated to holding games and the other to holding Windows 11 and everything else. It is powered by a 1,500-watt Corsair power supply.

As the name indicates, our PNY card has a slight overclock over stock speeds, as indicated by the base/boost clocks in the chart above.

Upon launch, the Nvidia GeForce RTX 5060 Ti meets an open playing field in its price range. The AMD Radeon RX 9070 and the Nvidia GeForce RTX 5070 are both $549 MSRP, creating substantial space between them in terms of cost and direct competition. This gap leaves only last-generation graphics cards, like the AMD Radeon RX 7800 XT and Nvidia GeForce RTX 4060 Ti, to compete with the RTX 5060 Ti near its price point. These older cards can still put up a fight, but as they use older technology, the RTX 5060 Ti has the advantage.

Synthetic Benchmarks

The synthetic test results suggest a substantial performance gain when moving from the RTX 4060 Ti to the RTX 5060 Ti across the board. Results against AMD’s Radeon RX 7800 XT were more mixed, but the RX 7800 XT came out ahead more often than not.

These test results also clearly show that the RTX 5070 and AMD’s Radeon RX 9070 are far ahead of the RTX 5060 Ti, which is reflected in their shared starting price.

Procyon AI Benchmarks

UL’s Procyon AI text generation is one of the first reliable tests we have to gauge AI performance. (Testing AI performance in any holistic sense is difficult, particularly as on-card AI muscle addresses a wide range of tasks—training versus inferencing, for one thing. It is impossible with current tools to test a very broad set of AI local-processing scenarios in an all-encompassing way.)

This test presses the GPU on a series of text-creation inferencing tasks via four popular AI large language models (LLMs) run locally, and produces an overall score in each case. (Here, Procyon uses versions of Mistral, Microsoft's PHI, and two flavors of Meta AI's Llama.) It also reports how many tokens (or discrete units of text, in the context of text generation) the hardware can produce per second with each model. It also measures the time it takes to produce the first token.

Regarding AI performance, the RTX 5060 Ti was once again a notable step ahead of the RTX 4060 Ti. (Note that our RTX 4060 Ti test sample was an 8GB card.) It was substantially faster in all of the tests, and the RTX 5060 Ti ran the Llama 2 tests without issue, whereas the RTX 4060 Ti failed to run those benchmarks.

AMD didn't stand much chance here, with the RTX 5060 Ti beating even the Radeon RX 9070 XT in this area in all but three tests. That’s a big deal if you care about AI performance and plan to heavily use local-processed AI in your work or leisure time, but for most people, this likely isn’t important right now. And AI pros aren't buying mainstream consumer cards like these.

Content Creation Benchmarks

Using graphics cards for GPU-accelerated content creation tasks is common today, and here, Nvidia’s GeForce RTX 5060 Ti scored a more substantial win over AMD's cards. The RTX 5060 Ti effectively tied with the Radeon RX 9070 XT in Adobe Premiere, and it was far faster than the Radeon RX 7900 GRE in Blender. The RX 9070 XT failed to run the Blender benchmark, but unless it would have nearly doubled the performance of the RX 7900 GRE, it will be slower than the RTX 5060 Ti in this area, too.

The RTX 5060 Ti also significantly outpaced the RTX 4060 Ti here. However, I wouldn’t recommend anyone upgrade from the RTX 4060 Ti to the RTX 5060 Ti, as I don’t consider these gains enough to justify the cost of replacing such a recent card. Regardless, you would see a noticeable uplift in many content creation tasks.

Screen Optimization Benchmarks

AMD’s FSR, Intel’s XeSS, and Nvidia’s DLSS technologies work in two ways to algorithmically enhance performance. First, they use algorithms crunched on GPU silicon to upscale images rendered at lower resolutions into sharper images. Second, they similarly use algorithms to generate and insert additional frames between originally rendered frames to boost frame rates. DLSS is the most advanced in this second technique, generating up to three additional frames per raw frame render via what Nvidia calls Multi-Frame Generation in its 50-series cards.

Regardless of version, each of these techniques works differently from each other to achieve similar goals, so you should be careful when considering results using these technologies against each other. However, due to the popularity of these techniques, it’s helpful to know how they compare in terms of performance.

In these tests, the RTX 5060 Ti outpaced the RTX 4060 Ti again. The RTX 5060 Ti wasn’t a match for the RTX 5070 or the AMD Radeon RX 9070, but it performed similarly to the Radeon RX 7800 XT. With frame generation off, the RX 7800 XT lagged behind the RTX 5060 Ti, but it pulled ahead when this feature was enabled.

Ray-Traced Gaming Benchmarks

I do not doubt that the Nvidia GeForce RTX 5060 Ti's GDDR7 memory helps it overcome the limitations of its narrow 128-bit memory interface. Still, as I said earlier, it doesn’t entirely fix the issue. In-game benchmarks give us the best idea of how much of a problem the 128-bit memory interface is, and they clearly show that the issue was much improved.

A technical glitch prevented us from testing the RTX 5060 Ti at 1080p in Cyberpunk 2077, but with the resolution set to 1440p and 4K, the RTX 5060 Ti outpaced the RX 7900 GRE and the RX 7800 XT, both of which have wider 256-bit memory interfaces and more memory bandwidth. The RTX 5060 Ti’s improved graphics cores and faster GDDR7 memory speeds enabled it to pull ahead despite this disadvantage in Cyberpunk 2077. It also trounced the RTX 4060 Ti.

At 1080p in F1 2024, the RTX 5060 Ti was slightly ahead of AMD’s Radeon RX 7900 GRE before it effectively tied at 1440p and 4K. The RTX 5060 Ti still holds the overall advantage, but it notably dropped further than the RX 7900 GRE when transitioning from 1080p to 1440p. This is likely due to the RTX 5060 Ti having less bandwidth than the RX 7900 GRE. The impact here is slight, but the issues that the narrow 128-bit interface can cause are more evident in Far Cry 6.

Typically, if memory bandwidth isn’t a problem, performance tends to scale linearly with resolution. This means most graphics cards tend to have their frame rates drop a similar proportional amount when moving from, say, 1080p to 1440p. When we see one card dropping faster than the others, it's typically a sign of a bottleneck somewhere, and for the RTX 5060 Ti, that’s almost certainly its narrow memory interface.

In F1 2024, the impact of this was negligible. The RTX 5060 Ti lost 36% of its frame rate when transitioning from 1080p to 1440p, whereas the RX 7900 GRE dropped by 33%. This is enough difference to close the performance gap between the two, but not much else.

This is far more noticeable in Far Cry 6, where the RTX 5060 Ti went from having a significant lead over the Radeon RX 7900 GRE and the Radeon RX 7800 XT at 1080p to being slower than both at 1440p and 4K. From 1080p to 1440p, the RTX 5060 Ti dropped 22% of its performance while the RX 7900 GRE only lost 8%. Both dropped a fair bit when cranking the resolution up to 4K, but the RTX 5060 Ti remained behind.

A similar situation is observable in Returnal, where the RX 7900 GRE tied with the RTX 5060 Ti at 1080p, but the RX 7900 GRE came out on top at 1440p and 4K. This is again likely due to the RTX 5060 Ti's memory-bandwidth limitations brought about by the card's narrow 128-bit interface.

The RTX 5060 Ti was behind the RX 7900 GRE at all resolutions in Avatar: Frontiers of Pandora and Call of Duty: Modern Warfare III, and this likely wasn't a result of the limited memory bandwidth. The bandwidth constraints of the RTX 5060 Ti aren't likely to impede 1080p resolutions, which the performance in the other games we tested supports—the RTX 5060 Ti is generally faster at 1080p. Instead, AMD's win here is likely due to better optimization for AMD hardware in these two games.

Raster Gaming Benchmarks

In games that don’t support ray tracing, like Total War: Three Kingdoms and Shadow of the Tomb Raider, AMD’s previous-generation Radeon RX 7900 GRE and Radeon RX 7800 XT held a sizable lead over the RTX 5060 Ti. Those cards struggled to compete with the RTX 5060 Ti in games with ray-tracing support as their ray-tracing hardware is less robust, but if ray tracing is out of the equation, they can handily beat the RTX 5060 Ti.

The Radeon RX 7900 GRE's and RX 7800 XT’s performance triumph in Shadow of the Tomb Raider, in particular, was helped by the RTX 5060 Ti’s narrow memory interface, which again held it back at 1440p and 4K.

Power Consumption and Thermal Tests

We use a Kill-A-Watt wall meter to measure the power consumption of our graphics card testbed as a whole during select tests to gauge the power draw of each tested graphics card.

Power consumption in gaming increased a fair bit from the RTX 4060 Ti to the RTX 5060 Ti. The performance increase outweighs this increase in power draw, but drawing more power is never ideal. However, the RTX 5060 Ti’s power consumption was considerably lower than that of the AMD competition, except for the faster Radeon RX 9070, which is more efficient overall.

The gaming thermals we recorded from the PNY GeForce RTX 5060 Ti weren’t terrible, but they certainly could have been better. Most of the other graphics cards in the chart stayed cooler during the testing process. Different models of the RTX 5060 Ti with more robust thermal solutions could perform better here. However, it isn’t strictly necessary, as 70 degrees C is within the safe operating range for a modern graphics card.