Learn extra at:
Lately we examined how PCI Categorical bandwidth influences the efficiency of the 8 GB Radeon RX 9060 XT when native video reminiscence (VRAM) is exceeded. All the goal of that testing was to push previous the VRAM restrict, which, sadly for 8 GB graphics playing cards, is a comparatively straightforward job in 2025. This may occur even when utilizing settings that may in any other case be extremely playable, as demonstrated by the 16 GB mannequin.
That is an fascinating take a look at for a number of causes, probably the most notable being that PCIe bandwidth turns into a main bottleneck when required VRAM is exceeded. Rising bandwidth may help scale back the efficiency penalty of shifting recreation property into native system reminiscence (RAM).
Understanding VRAM and PCIe Bandwidth
To briefly clarify, when VRAM runs out, recreation property are shifted to system RAM. The info should go by means of the PCI Categorical bus to the CPU’s reminiscence controller after which out to RAM, the place it may be saved.
The explanation why that is typically catastrophic for efficiency is bandwidth. Within the case of the GeForce RTX 5060 Ti, the VRAM has a peak bandwidth of 448 GB/s, which is rather a lot. In distinction, the theoretical peak bandwidth for system reminiscence is way decrease – round 96 GB/s for dual-channel DDR5-6000 and roughly 50 GB/s for dual-channel DDR4-3600, relying on a number of elements.
| RTX 5060 | RTX 5060 Ti | RTX 5070 | RTX 5070 Ti | RTX 5080 | |
|---|---|---|---|---|---|
| Worth MSRP | $300 | $380 / $430 | $550 | $750 | $1,000 |
| Launch Date | Could 2025 | April 2025 | Mar 2025 | Feb 2025 | Jan 2025 |
| Course of | TSMC 4N | ||||
| Die Dimension (mm²) | 181 mm² | 263 mm² | 378 mm² | ||
| Core Config | 3840 : 120 : 48 | 4608 : 144 : 48 | 6144 : 192 : 80 | 8960 : 280 : 96 | 10752 : 336 : 112 |
| L2 Cache (MB) | 32 MB | 48 MB | 64 MB | ||
| GPU Enhance Clock | 2497 MHz | 2572 MHz | 2512 MHz | 2452 MHz | 2617 MHz |
| Reminiscence Capability | 8 GB | 8 GB / 16 GB | 16 GB | ||
| Reminiscence Velocity | 28 Gbps | 30 Gbps | |||
| Reminiscence Sort | GDDR7 | ||||
| Bus Sort / Bandwidth | 128-bit, 448 GB/s | 192-bit, 672 GB/s | 256-bit, 896 GB/s | 256-bit, 960 GB/s | |
| PCIe Bus Interface | PCIe 5.0 x8 | PCIe 5.0 x16 | |||
| Complete Board Energy | 145 W | 180 W | 250 W | 300 W | 360 W |
The important thing takeaway is that fifty to 96 GB/s is significantly lower than 448 GB/s, and this does not even account for elevated latency. These figures additionally ignore the PCIe interface as a possible bottleneck, which may additional scale back efficient bandwidth relying on the setup.
PCI Categorical 5.0, for instance, offers a bi-directional bandwidth of 128 GB/s with all 16 lanes accessible. That drops to 64 GB/s for PCIe 4.0 and 32 GB/s for PCIe 3.0. Nevertheless, the RTX 5060 Ti is proscribed to simply 8 lanes, so these figures are halved once more: 64 GB/s for PCIe 5.0, 32 GB/s for PCIe 4.0, and solely 16 GB/s for PCIe 3.0.
Because of this when the 8 GB VRAM restrict is exceeded, efficiency may very well be considerably higher on PCIe 5.0 in comparison with 4.0 or 3.0. And that is precisely what we discover on this overview.
For testing the RTX 5060 GPUs, we are going to give attention to upscaled 1440p – so not native 1440p, but in addition not 1080p. As we have demonstrated up to now, it is very straightforward to overwhelm 8 GB GPUs at 1080p in trendy video games utilizing settings which might be nonetheless very playable on a 16 GB mannequin. Moreover, these 8 GB GPUs weren’t ‘supposed’ or ‘designed’ particularly for 1080p gaming. That declare is just unfaithful and never mirrored in how these merchandise have been marketed.
AMD, for instance, has constantly promoted the 9060 XT as providing ultra-fast 1440p gaming. Nvidia has additionally formally showcased the 5060 Ti at 1440p utilizing most high quality settings, albeit with 4x body technology enabled. The purpose is, neither firm is advertising and marketing these $300+ GPUs as 1080p-only options, and rightly so. At this value level, consumers ought to count on succesful 1440p efficiency, which occurs to be what you obtain so long as you buy the 16GB mannequin.
Our take a look at mattress shall be our normal high-end 9800X3D take a look at system to evaluate PCIe efficiency particularly. This permits us to isolate the PCIe bus and consider how a lot PCIe bandwidth influences efficiency when the GPU accesses system reminiscence.
We measured the 8 GB 5060 Ti‘s efficiency utilizing PCIe 3.0, 4.0, and 5.0, and in contrast it to the 16 GB 5060 Ti working on PCIe 3.0. This information is helpful for customers who haven’t got PCIe 5.0 programs, as many CPUs and platforms nonetheless depend on PCIe 4.0 and even 3.0.
Let’s dive into the benchmarks.
Benchmarks
Dragon Age: The Veilguard
Beginning with Dragon Age: The Veilguard, efficiency goes from dangerous to worse for the 8 GB mannequin. However first, let’s focus on the 16 GB card, which we examined utilizing PCIe 3.0. When VRAM capability is just not exceeded, the PCIe mode has little influence on efficiency. On this case, the 16 GB card averaged 66 fps with a 1% low of 47 fps, delivering a really playable expertise.
Underneath the identical circumstances utilizing PCIe 3.0, the 8 GB mannequin averaged simply 27 fps with a 1% low of 9 fps. That made the 16 GB card 130% sooner by way of common body charge and 411% sooner for the 1% lows. Rising the PCIe bandwidth to 4.0 improved the 8 GB card’s common body charge by 22% and its 1% lows by 111%. Shifting as much as PCIe 5.0 introduced an extra 12% achieve in common body charge, with an analogous enchancment in 1% lows.
Even with PCIe 5.0, the 8 GB mannequin was unable to ship playable efficiency. The 16 GB card utilizing PCIe 3.0 was nonetheless 68% sooner on common and 119% sooner for the 1% lows. Whereas elevated PCIe bandwidth helped considerably, the 8 GB 5060 Ti carried out so poorly on this take a look at that the PCIe specification made little distinction. Playable efficiency would nonetheless require reducing high quality settings.
F1 25
Subsequent is F1 25 utilizing the excessive preset. You will need to be aware that 8 GB GPUs carry out very poorly with the ‘very excessive’ preset, whereas 16 GB fashions nonetheless present easy efficiency.
Even on the excessive preset, the 8 GB mannequin confirmed noticeable decline, with 1% lows roughly 30% decrease than the 16 GB mannequin, even when utilizing PCIe 4.0 or 5.0. When restricted to PCIe 3.0, the 16 GB card ended up 114% sooner in common body charge and 235% sooner in 1% lows.
Indiana Jones and the Nice Circle
In Indiana Jones, the 16 GB 5060 Ti averaged 118 fps with a 1% low of 96 fps. This supplied a extremely playable and easy high-refresh-rate expertise utilizing PCIe 3.0, for the reason that VRAM capability was not exceeded.
The identical can’t be mentioned for the 8 GB mannequin. It’s obscure how these two variations of the 5060 Ti share the identical title, given the huge efficiency distinction. Positive, decreasing visible high quality can match the efficiency, however the visible downgrade on this recreation is critical and really noticeable.
Frankly, it’s astonishing that this example is appropriate to some. The 16 GB mannequin is just not solely delivering a easy expertise, however it is usually as much as 883% sooner utilizing the identical PCIe 3.0 configuration.
Sadly, switching the 8 GB card to PCIe 4.0 or 5.0 makes little distinction. With these settings – which run completely nicely on the 16 GB mannequin – the 8 GB card is totally overwhelmed. It might be unwise to not see this as a transparent indication of the place issues are heading.
Marvel’s Spider-Man 2
We examined Marvel’s Spider-Man 2 utilizing the very excessive preset with high quality upscaling at 1440p. This isn’t native 1440p – render decision is definitely nearer to 1080p. Underneath these circumstances, the 16 GB 5060 Ti achieved a formidable 90 fps with 1% lows of 56 fps, delivering extremely playable efficiency.
Utilizing PCIe 3.0, the 16 GB mannequin was 592% sooner than the 8 GB mannequin. Even when the 8 GB card was moved to PCIe 4.0, the 16 GB model remained 210% sooner, and nonetheless 120% sooner when utilizing PCIe 5.0. Attributable to its 8-lane PCIe limitation, the 8 GB mannequin turns into virtually unusable as soon as VRAM capability is exceeded, which is precisely the state of affairs examined right here.
Monster Hunter Wilds
Monster Hunter Wilds is one other instance the place PCI Categorical specification makes little distinction when VRAM is the limiting issue. If you wish to use the extremely high quality preset and see all textures rendered correctly, greater than 8 GB of VRAM is completely required – even when gaming at 1080p.
Whereas the 16 GB mannequin averaged 84 fps underneath these circumstances, the 8 GB model managed simply 34 fps, with extraordinarily poor 1% lows of 16 fps.
The Final of Us Half II
The Final of Us Half II is technically playable on an 8 GB graphics card, although you’ll encounter frequent body time inconsistencies. On this take a look at, the 16 GB mannequin averaged 91 fps, making it 40% sooner than the 8 GB model underneath similar settings. Switching the 8 GB mannequin to PCIe 4.0 boosted its efficiency by 18%, decreasing the efficiency hole to 18%. Lastly, utilizing PCIe 5.0, the 8 GB card reached 92 fps – matching the 16 GB card’s common.
Ratchet & Clank: Rift Aside
Lastly, Ratchet & Clank: Rift Aside runs exceptionally nicely on the 16 GB mannequin, averaging 133 fps. That makes it 110% sooner than the 8 GB card utilizing PCIe 3.0. The 8 GB card struggled considerably on PCIe 3.0, with very poor body time efficiency. Shifting to PCIe 4.0 resolved a lot of the points, and the cardboard reached 95 fps. Nevertheless, the 16 GB mannequin was nonetheless 37% sooner. On PCIe 5.0, the 8 GB 5060 Ti carried out nicely, providing comparable 1% lows to the 16 GB mannequin, though the 16 GB model remained 11% sooner in common body charge.
What We Discovered
As we had come to count on, based mostly on earlier assessments, exceeding VRAM capability ends in considerably worse efficiency when utilizing PCI Categorical 4.0 or 3.0 in comparison with PCIe 5.0. Whereas most eventualities led to utterly unplayable efficiency, even the most effective circumstances confirmed a considerable efficiency hit.
For probably the most half, the 8 GB 5060 Ti is sort of unusable when VRAM is exceeded, largely because of its 8-lane configuration. Even PCIe 5.0 can not overcome this limitation. The 9060 XT does carry out higher on PCIe 4.0 and 5.0, due to its full 16-lane interface, which offers twice the bandwidth underneath the identical PCIe model.
The lesson may be very clear: in case you intend to do something past enjoying esports titles at 1080p, you will need to keep away from the 8GB mannequin just like the plague. Even for esports gamers, we’d suggest steering clear, if just for higher resale worth down the highway.
Whatever the settings used, in case you’re excited about enjoying trendy AAA titles, want to discover increased resolutions sooner or later, or plan to maintain your graphics card for a minimum of the following 2 years, not at all will you need to buy the 8GB variations of both the 9060 XT or 5060 Ti. Selecting considered one of these fashions will seemingly power you to improve ahead of anticipated, and within the meantime, you will typically be caught tweaking settings to keep away from body time spikes and subpar efficiency.
What we discover most complicated concerning the 8 GB vs 16 GB debate – regardless of which aspect of the fence you are on – is how anybody can defend a product utilizing the very same title for 2 very totally different configurations, a selection that may end up in radically totally different efficiency.
Even if you’re of the opinion that 8GB continues to be fantastic and it is all you want … how do you justify two merchandise with the very same title, delivering radically totally different efficiency, utilizing settings that proved extremely playable for the 16GB mannequin?
Take F1 25 for instance. Utilizing the excessive preset (not even the utmost setting), the 8 GB mannequin delivered playable efficiency, which is a optimistic. Nevertheless, the 16 GB model was 114% sooner on PCIe 3.0, and as much as 30% sooner when the 8 GB card used PCIe 4.0 or 5.0. How is it acceptable for a similar product to carry out 30% higher just because it has extra VRAM?
We can not recall a time when it was thought of acceptable for 2 configurations of the identical product to vary by 30% in real looking circumstances. The same sample was noticed in Ratchet & Clank. On PCIe 4.0, the 8 GB mannequin averaged 96 fps, which is respectable. However even then, the 16 GB card on PCIe 3.0 was practically 40% sooner. That could be a large hole for 2 playing cards marketed underneath the identical mannequin… and these are among the higher outcomes for the 8 GB mannequin.
We additionally don’t perceive the argument that the 8 GB model is meant for “low” or “medium” settings in video games like Indiana Jones and the Nice Circle. That logic merely doesn’t maintain up. If it had a unique product title and was marketed as a lower-tier providing for modest settings, maybe that may make sense. However that isn’t the case. Actually, the alternative is true – each AMD and Nvidia promoted these merchandise based mostly on the efficiency capabilities of the 16 GB variations.
