AMD Radeon RX 9060 XT PCIe Comparison: 8GB vs. 16GB

On store shelves, AMD’s Radeon RX 9060 XT looks like a compelling option for gamers. It’s affordable, backed by a trusted brand, and appears well-equipped for modern gaming. But there’s a catch: the 8GB version is severely constrained by its limited video memory – a technical bottleneck that may not show up in standard benchmarks. That missing memory turns the 8GB model into a ticking time bomb for gamers hoping to future-proof their rigs.

The 8GB RX 9060 XT is marketed under the same name as the 16GB model, yet performance-wise, they tell very different stories. In our review, we highlighted several modern games with rising VRAM demands where the 8GB version didn’t just struggle – it broke.

Frame rates tanked. Texture streaming stuttered. What should have been smooth gameplay at reasonable settings became a frustrating experience. This is problematic for several reasons, but most importantly, a brand-new graphics card should not be bottlenecked by limited VRAM in today’s titles – and yet, here we are.

This becomes a major concern when you consider that most buyers expect to keep their GPU for at least three to four years. And if 8GB is already falling short in 2025, the road ahead looks even rougher.

The Future of VRAM Usage

In those 3 to 4 years, games will almost certainly become more demanding, requiring significantly more VRAM than they do today. Developers are already sounding the alarm. Even now, optimizing games for 8GB GPUs is becoming a time-consuming hurdle – a problem many in the gaming community flagged well before it showed up in review metrics. Some developers have spent weeks, even months, building workarounds just to make games functional on 8GB cards, only to see players expect next-gen experiences on last-gen memory footprints.

No VRAM? No problem. I’ll just turn up ugly graphics.

Testing VRAM limitations can be difficult, and in many cases, a quick 30- to 60-second benchmark run won’t reveal the performance issues players encounter during extended gaming sessions.

For example, to see VRAM-related performance issues in Halo Infinite on an 8GB GPU using ultra settings at 1080p or 1440p, you generally have to play the game for a prolonged period. It takes time to overwhelm the VRAM buffer. Texture issues, however, tend to appear more quickly, but you have to be looking for them – and they won’t show up in benchmark graphs.

A second Halo Infinite example: 16GB vs. 8GB VRAM “scaling”

Another factor reviewers often overlook is hardware configuration. When testing GPUs, it’s standard practice to pair them with the latest and fastest CPUs to showcase their full potential, and that’s exactly what we do.

But when it comes to VRAM limitations, testing on a high-end CPU with fast memory and a modern platform represents a best-case scenario for pushing the VRAM buffer. We briefly demonstrated this in our RX 9060 XT 8GB review by running tests on a Core i7-8700K, a processor limited to PCI Express 3.0 and slower DDR4 memory – and the results were shocking.

Testing Method

For this test, we’re using our high-end 9800X3D test system to evaluate PCIe performance. This means we’re isolating the PCIe bus and testing it specifically, giving us a clear understanding of how PCIe bandwidth influences performance – particularly when dipping into system memory.

We had considered using older CPUs to represent PCIe 3.0 and 4.0 configurations, but that would introduce too many variables. Instead, we’re sticking with the 9800X3D and manually adjusting the PCIe mode in the BIOS.

Also check out: The Inner Workings of PCI Express, How Hardware Works

This allows us to measure the performance of the 8GB RX 9060 XT using PCIe 3.0, 4.0, and 5.0, and compare that data to the 16GB RX 9060 XT using PCIe 3.0. These results are especially useful for those without PCIe 5.0-capable systems, as many platforms are still limited to PCIe 4.0 or even 3.0.

Let’s dive into the testing.

Benchmarks

F1 25 – Very High

Starting with F1 25 using the very high preset – the second highest quality setting in the game – we also checked performance using the high preset, which we’ll cover shortly. The 16GB RX 9060 XT averaged 73 fps in this test, with 1% lows at 51 fps. That’s solid, very playable performance.

With PCIe 5.0, the 8GB version was 18% slower, dropping to 60 fps, with 1% lows down by 22%. Remember, this is the 8GB model on PCIe 5.0 versus the 16GB model on PCIe 3.0.

When downgrading the 8GB card to PCIe 4.0, performance dropped by 30% to just 42 fps.

On PCIe 3.0 – the same interface used by the 16GB model in this test – performance fell further to just 35 fps.

That’s a 17% drop compared to PCIe 4.0, and 42% slower than PCIe 5.0. Worse still, on PCIe 3.0, the 8GB RX 9060 XT was 53% slower than the 16GB version on the same interface – or to put it another way, the 16GB model was 111% faster.

F1 25 – High

Here’s F1 25 again, this time using the high preset at 1440p with quality upscaling. The 16GB RX 9060 XT delivered 190 fps on average with 1% lows at 156 fps, all on PCIe 3.0. By comparison, the 8GB model was 11% slower when using PCIe 5.0.

Switching the 8GB card to PCIe 4.0 led to a further 12% drop to 149 fps, and another 13% drop using PCIe 3.0. With both GPUs on PCIe 3.0, the 16GB model provided nearly 50% higher average frame rates and 63% higher 1% lows. Again, these are very reasonable settings – high preset at 1440p with quality upscaling – so if you’re using an older PC, it’s essential to opt for the 16GB model.

Monster Hunter Wilds

In Monster Hunter Wilds, the game simply doesn’t run well on 8GB GPUs using the ultra preset at 1440p with quality upscaling. The 16GB model averaged over 60 fps on PCIe 3.0 – twice the performance of the 8GB card under the same conditions.

Even with PCIe 5.0, the 8GB version struggled, delivering terrible frame time performance and 1% lows of just 2 fps.

Interestingly, increased PCIe bandwidth did improve average frame rates for the 8GB model, but it actually worsened the 1% lows, creating more stuttering. This seems to be a result of rendering more frames overall while still running out of VRAM. The takeaway here is that PCIe bandwidth alone cannot solve the VRAM bottleneck of the 8GB RX 9060 XT in this title – especially with only 8 lanes available.

Marvel’s Spider-Man 2

Spider-Man 2 runs exceptionally well at 1440p with upscaling on the 16GB RX 9060 XT, averaging 112 fps with 1% lows of 66 fps. That’s 26% faster than the 8GB model using PCIe 5.0, and 38% faster for 1% lows.

Switching the 8GB card from PCIe 5.0 to 4.0 caused a 20% performance drop. Moving to PCIe 3.0 resulted in an even larger 27% decline, dropping the average frame rate to just 52 fps.

This means in a PCIe 3.0 system, using the settings shown here, the 16GB version of the RX 9060 XT delivers at least 115% higher average frame rates and 175% higher 1% lows. Even in PCIe 4.0 systems, the 16GB model is still 58% faster on average and 74% faster for 1% lows.

Ratchet & Clank: Rift Apart

Moving on to Ratchet & Clank: Rift Apart at 1440p with quality upscaling using the very high preset, the 16GB RX 9060 XT on PCIe 3.0 averaged 128 fps, with 1% lows of 97 fps – excellent performance overall.

Under the same conditions with PCIe 3.0, the 8GB model was 103% slower on average and 155% slower in 1% lows. Upgrading to PCIe 4.0 improved performance for the 8GB model by 30%, but the 16GB card was still 56% faster. Moving from PCIe 4.0 to 5.0 gave the 8GB model another 32% boost, making it usable, although the 16GB model still maintained a 20% performance advantage.

Indiana Jones and the Great Circle

Next up is Indiana Jones and the Great Circle. Here, the 16GB RX 9060 XT using PCI Express 3.0 rendered an average of 135 fps, with 1% lows at 113 fps – very solid performance overall. Under the same conditions, the 8GB model using PCIe 3.0 managed just 45 fps, making the 16GB version 200% faster, and 223% faster when comparing 1% lows.

Increasing the PCIe bandwidth to PCIe 4.0 improved the 8GB model’s average frame rate by a substantial 44%, reaching 65 fps. However, this was still less than half the performance of the 16GB model. Switching to PCIe 5.0 delivered a further 42% performance boost, and overall performance became quite good. Even so, the 16GB card remained 47% faster – while still running on PCIe 3.0.

Dragon Age: The Veilguard

Dragon Age: The Veilguard does not run well on 8GB graphics cards using the ultra preset at 1440p with upscaling, regardless of the PCI Express mode. On PCIe 3.0, the 16GB model was 174% faster.

Even after switching the 8GB card to PCIe 4.0, the 16GB version remained 70% faster, and still held a 31% lead when the 8GB card used PCIe 5.0. Despite these gains, 1% lows on the 8GB card were poor across all PCIe configurations.

Bonus: 5800X3D Testing

F1 25 (5800X3D)

To wrap up testing, we installed both GPUs into our Ryzen 7 5800X3D system, which is limited to PCIe 4.0 – the highest interface supported by the AM4 platform.

Interestingly, the 16GB RX 9060 XT delivered nearly identical performance on both the 5800X3D and 9800X3D systems, as GPU compute performance was the limiting factor. So, regardless of the CPU used, the full potential of the 16GB model was unlocked with these quality settings.

The 8GB model, however, continued to show inconsistencies. Using PCIe 4.0 with the 9800X3D, the 16GB version was already 27% faster than the 8GB model. But on the 5800X3D, the performance gap widened to 44%, due to an additional 11% performance drop – likely caused by the slower DDR4-3600 memory used on AM4.

The takeaway: at best, you can expect performance similar to the 9800X3D when using an older or slower CPU, but more often than not, the gap between the 8GB and 16GB models will be even larger.

Ratchet & Clank (5800X3D)

It’s important to note that this performance gap isn’t consistent across all titles.

Take Ratchet & Clank: Rift Apart, for example. Here, the 8GB RX 9060 XT using PCIe 4.0 delivered nearly identical average frame rates when paired with either the 5800X3D or the 9800X3D. However, 1% lows – which were already poor on the 9800X3D – were even worse on the 5800X3D.

Making Sense of the Numbers

So there you have it. As shown earlier with the Core i7-8700K, you absolutely do not want to run out of VRAM on a PCI Express 3.0 system, even when all 16 lanes are available. Those using PCIe 4.0 systems will fare much better, but as we also saw with PCIe 5.0, a significant performance hit isn’t always avoidable.

In nearly all of the games tested, the 8GB RX 9060 XT was effectively unusable when limited to PCIe 3.0 – F1 25 with the high preset being the only real exception, though even then, performance lagged well behind the 16GB model.

This highlights that even with more modest settings, performance can drop sharply on older platforms. You only need to exceed the local video memory limit slightly to encounter serious performance issues… problems that would often go unnoticed on modern systems with sufficient VRAM and bandwidth.

PCIe bandwidth should never be a performance-limiting factor, and the only reliable way to ensure that is by having enough VRAM. At this point, 16GB is sufficient, 8GB is not.

Another important factor to consider is motherboard configuration. Many motherboards, especially budget-oriented models, will reduce the PCIe lane count to the primary GPU slot when the secondary PCIe slot is populated. So, if anything is installed in the secondary PCIe x16 slot, both the primary and secondary slots may operate in an x8/x8 configuration, something that is extremely common.

This effectively halves the available PCIe bandwidth for the graphics card. For example, on a PCIe 4.0 system with all 16 lanes active, there’s 64 GB/s of bandwidth. But with only 8 lanes, that drops to 32 GB/s, or the same bandwidth you’d get from PCIe 3.0 x16.

Ultimately, PCIe bandwidth should never be a performance-limiting factor, and the only reliable way to ensure that is by having enough VRAM. At this point, 16GB is sufficient, 8GB is not. That said, having the full 16 lanes available still benefits the 8GB RX 9060 XT significantly. We’ll revisit this and some new testing in a follow-up benchmark session with different GPUs and PCIe configurations.