• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

NVIDIA GeForce RTX 5060 Ti PCI-Express x8 Scaling

W1zzard

Administrator
Staff member
Joined
May 14, 2004
Messages
28,804 (3.74/day)
Processor Ryzen 7 5700X
Memory 48 GB
Video Card(s) RTX 4080
Storage 2x HDD RAID 1, 3x M.2 NVMe
Display(s) 30" 2560x1600 + 19" 1280x1024
Software Windows 10 64-bit
The new NVIDIA GeForce RTX 5060 Ti uses the modern PCI-Express 5.0 bus interface, but supports only an 8-lane configuration. This is just half the bandwidth of the higher-end models, which run at x16. Are you in trouble with an older system, which only supports running at PCIe 4.0?

Show full review
 
So pretty much anyone with a Gen 4 board should be fine and even Gen 3 is running at such a small loss that it is not an issue considering the CPUs on those platforms. As expected.
@W1zzard
The Relative Performance chapter is showing up two times in the article, just as a heads up.
 
The Relative Performance chapter is showing up two times in the article, just as a heads up.
fixed

This test would make a lot more sense for the 8GB version where PCIE speed will be way more impactful.
No worries, I just ordered a 8 GB card
 
I was really hoping to see the compute figures here even more than the gaming numbers, which I wasn't even expecting to suffer much. Compute tasks, like LLM performance is where I would expect to see a performance impact.
 
from the drop down menu Conclusion takes to page 33 which doesnt exist and its actually page 32.
 
from the drop down menu Conclusion takes to page 33 which doesnt exist and its actually page 32.
Yeah that's because I removed a duplicate page. If you load the review now the dropdown will be correct
 
No worries, I just ordered a 8 GB card
Is it possible to add Shared memory to benchmmarks? It's really interesting, cause Vram can be 6-7 out of 8 GB, but also 2-3 GB shared memory.
There are few ways to see that value. PCI-e load is also useful, but RTX 5000 shows incorrect values.
 
5060Ti: "I got 99 problems, but the PCIe ain't one."
 
No worries, I just ordered a 8 GB card

Interesting enough review, that I wanted to register to say thank you for taking the time to do this!

Throwing another corner case out there alongside the 8gb pass I'll be keenly watching for...

...on some motherboards/BIOS configs, it's possible to not plug your monitor into the video card at all, but instead run it off the onboard GPU output. Windows 11 handles this cleanly, giving you options for which GPU (onboard or dedicated) you want to actually do the work of rendering any given app or game...but whichever choice you make, obviously the output is through the onboard HDMI or DP or whatever. So the dedicated card must be pushing those 'finished frames' out to the onboard GPU via that PCI-E bus.

Figuring a 4k image at 32bpp * 60 fps = works out to something like 2 GB/s bandwidth demand, right? Seems that would eat a big chunk out of a PCI-E 8x bus...

I'm wondering if that would be a benchmark possible to run?
 
Good idea, judging by this video the 8GB card's performance regularly collapses when VRAM is exceeded, which is more pronounced at lower PCIe speeds. We aren't just talking about a few percent. The 5060 Ti will presumably be affected in just the same way.
All 8GB cards will be bad in 2025 with more demanding and popular games. There's no question about it. That's why Nvidia banned AIBs from sending any 8GB cards to reviewers... Nvidia knows very well that 8GB cards are bad in this regard. Anyone who purchases those cards with low VRAM will have to dial down several settings and live with heavy compromises.
 
All 8GB cards will be bad in 2025 with more demanding and popular games. There's no question about it. That's why Nvidia banned AIBs from sending any 8GB cards to reviewers... Nvidia knows very well that 8GB cards are bad in this regard. Anyone who purchases those cards with low VRAM will have to dial down several settings and live with heavy compromises.
They never should have been made.
 
PCI-Express 5, 4 and even 3, do not have that big of a effect on GPU's. Its the industry who needs faster and more lanes to support the ever growing need of faster storage, NIC's and such.

Obviously making these things universal, i.e no AGP anymore, would make all sense in the world.
 
So that mean the PCIe x16 3.0 results are the same as PCIe x8 4.0, correct? I mean, why would anyone run his GPU at x8 instead of x16 ?
 
So that mean the PCIe x16 3.0 results are the same as PCIe x8 4.0, correct? I mean, why would anyone run his GPU at x8 instead of x16 ?

Cheaper motherboards do skimp out on the amount of available lanes. But 4.0 x8 is the same as 3.0 X16.
 
I'm asking since my mobo only has PCIe 3.0 slots, so I'm guessing this card will use x16 or x8 lines for PCIe 3.0 ??? Or the card's layout is only made for x8 lines no matter the port speed?
 
I'm asking since my mobo only has PCIe 3.0 slots, so I'm guessing this card will use x16 or x8 lines for PCIe 3.0 ??? Or the card's layout is only made for x8 lines no matter the port speed?

Yes.

and chose to give the RTX 5060 Ti and the GB206 silicon driving it a PCI-Express 5.0 x8 host interface

But you should not lose a lot in regards of performance.
 
But you should not lose a lot in regards of performance.

There are caveats to that, evidently it CAN make a huge difference under some circumstances. When HUB deliberately used settings to find the 8GB card's limits it absolutely choked in situations where the 16GB card was unaffected. This was both at PCI-E 3.0 and 4.0. That was the 4060 Ti but I can't see why the 5060 would be any better. Especially on 3.0 motherboards where it will only run at 3.0 x 8.
 
Back
Top