MSI GeForce RTX 5060 Gaming OC review: you’re gonna need a bigger buffer

With Computex 2025 now in the rear-view mirror, I finally have the chance to sit down with Nvidia’s new GeForce RTX 5060. This is as budget as Blackwell gets for the foreseeable, and though the graphics card thankfully offers a good amount of value, it’s often fighting against its technical makeup to deliver the requisite punch.

Starting at £279 / $299 with plentiful retail availability, RTX 5060 matches the cost of its predecessor, RTX 4060. However, competition is about to heat up with rival Radeon RX 9060 XT launching in just a few days at the same price point. That’s not forgetting the more affordable Arc B580, too.

Nvidia’s mid-range arrival effectively signals the start of a new round in the bout for the title of best graphics card for budget gamers. RTX 5060 certainly appears to be a safe competitor in many respects owing to its Blackwell roots and support for DLSS 4, but let’s see how it shapes up once it enters the ring.

Specifications

RTX 5060 uses the same GB206 GPU die as RTX 5060 Ti albeit with some expected cutbacks to core counts and clock speeds. Where it separates itself most from existing siblings is in memory configuration; there’s only an 8GB VRAM capacity to choose from as Nvidia forgoes providing a 16GB option.

	RTX 5060	RTX 5060 Ti
Released	May 2025	Apr 2025
Codename	Blackwell	Blackwell
GPU	GB206	GB206
Process	TSMC 4N (4nm)	TSMC 4N (4nm)
Transistors	21.9bn	21.9bn
Die size	181mm²	181mm²
CUDA cores	3,840	4,608
Boost clock	2,497MHz	2,572MHz
FP32 Boost TFLOPS	19.2	23.7
SM count	30 of 36	36 of 36
RT cores	30 (4th Gen)	36 (4th Gen)
RT TFLOPS	58	72
Tensor cores	120 (5th Gen)	144 (5th Gen)
ROPs	48	48
Memory	8GB	8GB / 16GB
Memory type	GDDR7	GDDR7
Mem. clock	28Gb/s	28Gb/s
Mem. interface	128-bits (PCIe 5.0 x8)	128-bits (PCIe 5.0 x8)
Mem. bandwidth	448GB/s	448GB/s
Board power	145W	180W
Launch MSRP	$299	$379 / $429

Nvidia trims back six Streaming Multiprocessors (SMs) from the GB206 die in crafting RTX 5060. This cut results in a 17% decline in core counts across the board relative to the full-fat configuration which RTX 5060 Ti enjoys. Both low-end and high-end cards are receiving the same treatment as this same percentage coincidentally also separates the GB203 GPUs of RTX 5080 and RTX 5070 Ti Super.

RTX 5060 clock speeds thankfully haven’t fallen much from RTX 5060 Ti levels, settling at 2,497MHz. Of course, this only refers to Nvidia’s reference specification and board partners will provide overclocks of varying degrees on many of their models, so expect some variance. Running at this frequency, the GPU can manage up to 19.2 FP32 boost TFLOPs, while 30 RT cores provide up to 58 RT TFLOPs.

Like its other Blackwell siblings, RTX 5060 benefits from a switch to GDDR7 VRAM. Running at 28Gb/s on a 128-bit bus, its memory bandwidth is identical to RTX 5060 Ti at 448GB/s. However, in lieu of 3GB modules, Nvidia can only equip the GPU with either 8GB of 16GB total capacity due to the width of its interface, opting for the former. An unfortunate turn of events in my eyes, as this buffer size can limit the capabilities of the silicon, as will become clear in the performance section of this review.

Rocking a TGP (Total Graphics Power) of 145W, it should come as no surprise that you don’t need a beefy power supply to drive RTX 5060. Nvidia specifically recommends running a 550W PSU with the pixel pusher but this is while paired with a Ryzen 9 9950X. Less-powerful (and arguably more appropriate) processors like Ryzen 5 9600X should provide more wiggle room if you have fewer watts at your disposal.

Before moving on to inter-generational comparisons, I should highlight that RTX 5060 comes with a PCIe 5.0 x8 interface. This shouldn’t cause any problems for owners of motherboards with PCIe 5.0 or 4.0 primary expansion slots although performance can drop on older PCIe 3.0 boards in some games owing to constrained bandwidth. This is only an issue for ageing systems but one to bear in mind if you’re in such a boat.

	RTX 5060	RTX 4060	RTX 3060
Released	May 2025	Jun 2023	Feb 2021
Codename	Blackwell	Lovelace	Ampere
GPU	GB206	AD107	GA106
Process	TSMC 4N (4nm)	TSMC 4N (4nm)	Samsung 8N (8nm)
Transistors	21.9bn	18.9bn	12.0bn
Die size	181mm²	159mm²	276mm²
CUDA cores	3,840	3,072	3,584
Boost clock	2,497MHz	2,460MHz	1,777MHz
FP32 Boost TFLOPS	19.2	15.1	12.7
SM count	30 of 36	24 of 24	28 of 30
RT cores	30 (4th Gen)	24 (3rd Gen)	28 (2nd Gen)
RT TFLOPS	58	35	25
Tensor cores	120 (5th Gen)	96 (4th Gen)	112 (3rd Gen)
ROPs	48	48	48
Memory	8GB	8GB	12GB
Memory type	GDDR7	GDDR6	GDDR6
Mem. clock	28Gb/s	17Gb/s	15Gb/s
Mem. interface	128-bits	128-bits	192-bits
Mem. bandwidth	448GB/s	272GB/s	360GB/s
Board power	150W	115W	170W
Launch MSRP	$299	$299	$329

Looking back at prior 60 class graphics cards, it’s clear that while RTX 5060 shares much in common with RTX 4060 it separates itself with several generational upgrades. Depending on which way you look at these improvements, though, you could argue this is in some parts the proper successor to RTX 3060 we’ve been waiting for.

While a shift in architecture naturally creates differences in GPU between RTX 5060 and 4060, Nvidia gives its new offering a boost by providing a higher class of die (GB206 vs. AD107). This marks a return to the stylings of RTX 3060 with its GA106 chip, and affords the budget Blackwell card a welcome helping of cores and transistors that exceed its forebear in both quantity and quality.

RTX 5060 does demand more power than RTX 4060 to feed its larger GPU, but an additional 35W for a 150W total isn’t worth splitting hairs over. Besides, it’s downright impressive how much more efficient the card is relative to RTX 3060.

Following the generational regression in memory bandwidth between RTX 3060 (360GB/s) and RTX 4060 (272GB/s), I’m happy to see RTX 5060 trump both of its predecessors. Running at up 448GB/s its GDDR7 modules make for a 65% improvement over the prior generation’s GDDR6 chips, but a more humble 24% looking a further generation back.

The real kicker here, of course, is that RTX 3060 boasts a 12GB buffer thanks to its 192-bit bus. RTX 5060’s memory is faster but capacity is king in many circumstances.

Design

Like RTX 5060 Ti, there’s no reference design to point to for RTX 5060. In lieu of a Founders Edition, it’s up to board partners to provide models that hit or come as close as possible to Nvidia’s suggested £279 / $299 pricing.

For this review, I’m taking a look at a more premium option: MSI GeForce RTX 5060 Gaming OC. It’ll set you back around £335 / $370, and such pricing pushes it dangerously close to RTX 5060 Ti 8GB territory. There are elements here that those in search of a higher-quality design will appreciate, and in fairness, budget-constrained buyers would do well to seek out Ventus 2X from MSI’s arsenal.

Gaming OC isn’t especially large nor heavy, rocking a dual-slot design and weighing in at 649g. This isn’t surprising given the power draw of RTX 5060 doesn’t necessitate a beefy cooler. Regardless, the card looks and feels the part of a premium model with rock-solid build quality and materials.

The all-black appearance of Gaming OC may feel a little non-descript for some but I appreciate stealthy styling. The proverbial block of obsidian does feature a splash of colour once powered on, though, with RGB LEDs running through the side-facing MSI logo. The silver dragon on the metal backplate also nicely reflects the surrounding light of your system.

MSI pushes power consumption on Gaming OC up to 155W (+5W) out of the box, in service of its built-in overclock. It runs at 2,625MHz by default, netting an additional 128MHz versus stock but you can gain an additional 15MHz via MSI Center to a 2,640MHz total. This won’t make an enormous difference to performance in applications and games but any increase in frequency is welcome.

You won’t need to reach for a 12V-2×6 cable or dongle to power Gaming OC as a single 8-pin PCIe connector handles power delivery duties. Like RTX 5060 Ti, Nvidia isn’t mandating the use of the newer port with RTX 5060.

Speaking of cooling, MSI equips Gaming OC with two of its flagship Stormforce fans that you’ll find on its most-premium RTX 5090 models including the flagship Suprim X design. The seven blades on each blower feature a claw texture which apparently enhances air pressure for enhanced thermal management. Returning to aesthetics a moment, I’m very keen on the reflective holographic dragon logos in the centre of each fan.

Underneath the fans, MSI’s wave-curved heatsink design helps improve cooling efficiency by widening the size of wave edges where there’s less airflow. This setup works in tandem with a nickel-plated copper baseplate, which transfers heat to square-shaped core pipes that capture heat from the GPU baseplate more effectively than traditional heatpipes.

Running RTX 5060 Gaming OC in the Club386 test system, power consumption is nice and low at 259W under load. It expectedly pulls more watts from the wall than RTX 4060, but what it does with those additional 21W is nothing short of transformative in practice.

More power means higher temperatures but Gaming OC does an admirable job of keeping RTX 5060 cool and quiet. In my testing, the card’s thermals averaged out to a lowly 62°C under load peaking at 65°C, all while running at a practically inaudible 36.4db. Custom cooling on this calibre of card doesn’t come much better.

Performance

It’s time to fire up the Club386 test bench once again as I put RTX 5060 through its paces. The system is certainly overkill for a mainstream graphics card, yet its Ryzen 9 7950X3D CPU will help keep any bottlenecks at bay.

I’m benchmarking RTX 5060 at both FHD (1080p) and QHD (1440p). While its VRAM capacity hampers its abilities to play at the more-demanding resolution without a second thought, the card can rise above the limits of its capacity in some games. Moreover, it’s important to understand how and when this buffer size can be problematic.

Application & AI

In 3DMark, RTX 5060 creates a great deal of distance between itself and RTX 4060. Speed Way illustrates a 31% difference in ray tracing performance, but it’s the rasterised battleground of Steel Nomad that forms the largest gap at 40%.

Stepping up to RTX 5060 Ti 16GB provides a small leg up of 18% in Speed Way and 13% in Steel Nomad. Considering this comes with at least a 27% price bump for the 8GB variant, RTX 5060 emerges the better-value buy on paper. However, it’s worth nothing that these 3DMark benchmarks operate comfortably within an 8GB buffer.

Radeon RX 9060 XT 16GB will serve as the closest competitor to RTX 5060 come June 5. For the moment, neither RX 7600 nor 7600 XT come close to matching Nvidia’s offering in either workload. It will take a herculean effort to catch up, but AMD has proven RDNA 4 can compete with Blackwell via RX 9070.

Meanwhile, Arc B580 puts in an uncomfortably strong showing in Steel Nomad, as RTX 5060 pulls ahead by a mere 5%. Nvidia more decidedly pushes Intel back in Speed Way, though, with a 41% advantage.

It’s no secret that Blender loves CUDA allowing every GeForce graphics card to expectedly zip past Radeon and Arc alike.

Accruing 3,691 samples per minute, RTX 5060 finds itself ahead of RTX 3070 and nipping at the heels of RTX 4060 Ti 8GB and RTX 3070. You only gain an additional 19% worth of performance jumping up to RTX 5060 Ti 16GB, as value once again favours the more-affordable of the two.

Turning attentions to AI, RTX 5060 absolutely creams RTX 4060 in Geekbench AI by a whopping 26%. Its fifth-generation Tensor cores give the graphics card the gusto to pip past RTX 4060 Ti 8GB too by a cool 9%.

RTX 5060 Ti 16GB claws a 13% lead over RTX 5060, the smallest increase of the whole suite. As much as its beefier GPU doesn’t result in enormous gains to half precision (FP16) performance, its memory capacity can balance the scales elsewhere in the realms of deep learning.

Running an LLM (Large Language Model) with a Llama 3.1 base is smoother on RTX 5060 than prior generation graphics cards in the same class. In Procyon AI Text Generation, it enjoys a 38% lead over its predecessor.

However, it’s Arc B580 that takes the value crown in the competition. Its 12GB buffer is undoubtedly helping it punch above its weight, earning second place on the chart behind RTX 5070, but that capacity would be for naught if not for Intel’s surprisingly stout XMX Engines.

Gaming

Starting the suite of gaming benchmarks with Assassin’s Creed Mirage, RTX 5060 delivers frame rates that keep Basim’s parkour looking smooth. At FHD the graphics card is a dead ringer for RTX 4060 Ti 8GB, averaging 119fps, but it pulls ahead of its Lovelace bedfellow by two frames at QHD with an 87fps result. Impressively, it maintains a healthy minimum frame rate north of 60fps, too.

Looking back at its predecessor, RTX 5060 outpaces RTX 4060 by 23-24% across the two resolutions. Keeping analysis in the family, RTX 5060 Ti 16GB enjoys a small 10% lead at FHD, increasing to 18% at QHD.

According to the game’s VRAM usage tracker, 8GB is more than enough to handle Baghdad with all its bells and whistles so these gains are all down to silicon. Case in point, neither RX 7600 XT nor Arc B580 can keep up with RTX 5060’s raster chops despite their larger buffers.

Logging into Final Fantasy XIV: Dawntrail, RTX 5060 continues its reign of relative supremacy over RX 7600 XT and Arc B580 in the world of rasterisation. The MMORPG practically skips along at 145fps on average at FHD, continuing steady at 94fps at QHD albeit with minimums falling just below the 60fps threshold.

Once again, RTX 5060 finds itself squeaking past RTX 4060 Ti 8GB in the charts with just 2-3fps separating the two cards. Moving down the Lovelace stack, a smaller 16-18% generational gap keeps the budding Blackwell from RTX 4060.

RTX 5060 Ti 16GB doesn’t shift the needle in a significant manner, offering up a small 11-12% uplift over RTX 5060. Results like this further cement that there is more value to extract from the less-powerful SKU in terms of raw performance, providing you can work within the limitations of its memory. Hold that thought.

It’s on the tracks of Forza Motorsport that RTX 5060 and all other graphics cards with 8GB of VRAM hit the brakes. Where once this seemingly budget darling occupied the higher rungs of the leaderboard, it now sputters along in the pits with unplayable frame rates.

Make no mistake, this is a GPU buckling under the constraints of its buffer. Take a look at how much better RTX 5060 Ti 16GB fares in this benchmark, outright lapping RTX 5060 as its better by 165% at FHD and barely slowing down at QHD with a 161% advantage. As a reminder, all that separates these two cards outside of VRAM capacity is 17% worth of cores which is not nearly enough to create these kind of performance differences.

If the advantages of a bigger buffer weren’t already obvious, look at RX 7600 XT. Despite its weaker ray tracing capabilities, AMD’s prior generation pixel pusher is delivering playable if less-than-desirable frame rates while RTX 5060 embarrassingly can’t even muster 30fps on average using the same settings.

Studying results like this makes me wish Nvidia offered the likes of RTX 5060 with 16GB of VRAM for a small premium. I understand that there are markets with weaker currencies where this 8GB model makes the most sense, but to leave this silicon hamstrung worldwide feels like missing an obvious win.

Bruised but not broken, RTX 5060 finds its footing in the rasterised stirrups of Mount & Blade II: Bannerlord. It’s difficult to go wrong with any modern graphics card in this game, but this one proves its worth with 194fps on average at FHD and 138fps at QHD.

Relative to RTX 4060, its current generation replacement rides the better of 25% at FHD and jumps up to 31% at QHD. RTX 5060 once again largely mirrors RTX 4060 Ti 8GB, with 3-9fps in favour of the newer card staving off any chance of mistaken identity.

Bannerlord is a rosier battleground for RTX 5060 Ti too, lording over RTX 5060 with a 21% performance advantage at FHD but falling to 16% at QHD. These leads aren’t quite enough to usurp its less-expensive sibling in terms of value, but it’s a stronger showing compared to most of the suite.

Pumping out over 200fps in Tom Clancy’s Rainbow Six Extraction at FHD, there’s little doubt that RTX 5060 has the credentials of an inexpensive esports graphics card. That’s not forgetting its support for Nvidia Reflex, the most prevalent latency reduction feature on the scene. As an RTX 50 Series GPU, it’ll eventually be able to make use of Nvidia Reflex 2 too once it emerges from the GeForce labs.

It’s a good thing RTX 5060 has Nvidia Reflex under its belt, as Arc B580 gives it a stonking good run for its money. There’s just four frames per second at most keeping these two competitors apart, which paints Intel’s card favourably considering it’s $50 cheaper.

DLSS 4

Outside of under-the-hood improvements, Blackwell’s most attractive selling point is DLSS 4. As much as Nvidia seems intent on making Multi Frame Generation (MFG) the star of the show, the brand’s also made improvements to the quality of DLSS Super Resolution too via a new ‘Transformer’ model. While every member of the RTX family benefits from these upscaling upgrades, only 50 Series models like RTX 5060 can do so without any cost to performance.

In practice, the latest version of DLSS Super Resolution is king of the hill when it comes to upscaling despite growing competition from FSR 4 and XeSS 2. Even at 1080p, it’s a technology I recommend using as it brings a welcome performance uplift with minimal impact to image quality. It’s as close to a free frame rate boost with no strings attached as you could hope to get, raising RTX 5060 frame rates from anywhere between 9-76% percent using the highest quality preset in my tests.

Both Single and Multi Frame Generation are welcome enhancements albeit less universally useful. These motion smoothing features outright require rock-solid base frame rates to operate as intended, and this can be a tall order for cards like RTX 5060 in more-demanding games. With some small amount of forethought they’re tools worth using to improve performance but it’s easy to mishandle them if you forget their limitations.

RTX 5060 makes short work of Call of Duty: Black Ops 6 sans any performance enhancements, running the game at 60fps+ at both FHD and QHD. However, it isn’t alone in this feat as both RTX 4060 and RTX 3060 Ti both hit this threshold. Engaging DLSS Super Resolution gives all three graphics cards a leg up. Running the ‘Quality’ upscaling preset pushes RTX 5060 up in particular by 9% at both resolutions, netting it a triple digit average frame rate at FHD.

Comfortably north of 60fps, there’s nothing stopping RTX 5060 from enjoying the fruits of DLSS Frame Generation. Enabling the feature lifts the card up by 43% at FHD and 34% at QHD. This results in a noticeably smoother looking gameplay experience but one that doesn’t feel as responsive as frame rates accrued through traditional means would. Interpolating generated frames into the render pipeline unavoidably comes at the cost of latency after all.

Despite Call of Duty: Black Ops 6 not sporting DLSS 4 out of the box, it is possible to inject it into the game via Nvidia app overrides. This is true of almost every game rocking older versions of Super Resolution and Frame Generation. This method can welcomely provide higher upscaled image quality and unlock MFG in a few clicks.

Without DLSS Super Resolution under its wings, the only chance RTX 5060 would have of playing Cyberpunk 2077 with path tracing would be through the make believe world of a braindance. A 44fps average frame rate at 1080p is by no means ideal, but it’s playable nonetheless. It’s also still bonkers to me that $300 graphics cards can handle this kind of workload with a small dose of upscaling in tow.

RTX 5060 is well short of the 60fps average necessary for the optimal DLSS Frame Generation experience. More importantly, the card’s 8GB buffer is also under a huge amount of strain owing to the demands of a path traced Night City. These two factors combined stop the feature dead in its tracks when enabled, resulting in repeatable performance regressions.

As with Forza Motorsport, RTX 5060 is crying out for more room to breathe. Looking at RTX 5060 Ti 16GB’s results in the same scenario, it suffers no such setback. This is undoubtedly in part thanks to its more powerful GPU but its VRAM capacity is playing a far larger part in this case.

From one path traced game to another, the otherworldly Alan Wake 2 shines a more favourable light on RTX 5060 than the neon glow of Night City does. It’s 18-20% faster than RTX 4060, though this only translates to an additional 8fps at most. The game is playable on the graphics card with 30fps+ average frame rates, but sadly falls short of the 60fps threshold for Frame Generation.

Unlike Cyberpunk 2077, enabling the feature regardless doesn’t result in performance regressions. Instead, Frame Generation raises RTX 5060 frame rates by 75%-217% at FHD 69-200% at QHD. The motion smoothing effect this creates is impressive to look at from a distance but generally doesn’t feel fantastic to play owing to that suboptimal base frame rate.

Resolution	DLSS	DLSS+FG	DLSS+FG 3x	DLSS+FG 4x
FHD	54.3ms	65.1ms	70.6ms	73.0ms
QHD	81.0ms	100.0ms	105.8ms	112.4ms

Examining the average latency values above helps to better explain this phenomena. The lower the base frame rate, in this instance with DLSS enabled, the higher your starting latency value and the more sluggish the game feels.

Enabling Frame Generation in Alan Wake 2 on RTX 5060 increases latency by at least 20-23% depending on the resolution in question. This is less detrimental at FHD as the result is still below the 100ms mark, but nudges QHD into a even more sluggish territory despite the benefits to motion smoothness.

Pushing on with MFG in a 3x or 4x configuration doesn’t increase latency as severely as just enabling Single Frame Generation but does come at further cost. For RTX 5060, the end result is a 35-38% rise. For me, the latency values are simply too high at QHD and are borderline passable at FHD.

Your sensitivity to these effects may vary so it’s at the very least worth trying but I’d advise caution, particularly in games that are already saturating RTX 5060’s VRAM capacity. Regardless of how you respond to Frame Generation, though, the rule of thumb of getting as high a base frame rate as you can before turning it on remains the same.

Conclusion

There’s a lot riding on the success of RTX 5060 as it should prove to be the most popular entry in the GeForce RTX 50 Series, following in the footsteps of its predecessors. I like a lot of what this graphics card offers, particularly how much of a performance improvement it is over RTX 4060. Still, it’s difficult to shake the concerns I have over its piddly VRAM capacity.

DLSS Super Resolution does help lighten the load on RTX 5060’s 8GB buffer, but there’s only so much it can reasonably accomplish. This is providing that the game you’re playing supports the feature, of course. In some cases you have no other choice but to reduce settings despite the GA106 GPU having more to give as RTX 5060 Ti 16GB demonstrates.

Meanwhile, RTX 5060 exemplifies the caveats that come part and parcel with DLSS Multi Frame Generation. It’s better to think of the feature as a motion smoothing technology than it is a performance enhancement, despite it lifting frame counters upwards. Through this mindset, it’s easier to accept its limitations and forgo using it should latency prove too high.

My advice to prospective buyers of RTX 5060 is to wait and see how RX 9060 XT stacks up, mainly its 16GB variant. For an additional $50, having so much more VRAM at your disposal will be transformative to frame rates in some cases. A bigger buffer will also generally improve the longevity of your card as ray tracing becomes more commonplace and demands more memory capacity. It’s for this reasons I gravitate more to RTX 5060 Ti 16GB despite its higher price.

As far as partner cards go, there’s little more MSI can do to elevate this budget GPU to higher echelons. Overclocked frequencies are extremely stable, temperatures barely warm, noise practically non-existent, plus it’s a looker, too.