NVIDIA GeForce RTX 4060 Ti to Feature a PCI-Express 4.0 x8 Bus Interface

erek

[H]F Junkie
Joined
Dec 19, 2005
Messages
10,897
Cute lil card, I like

“While this is still plenty of interface bandwidth for a GPU of this market segment, with bandwidth comparable to that of PCI-Express 3.0 x16, using the RTX 4060 Ti on older platforms, such as 10th Gen Intel Core "Comet Lake," or even much newer processors such as the AMD Ryzen 5700G "Cezanne," would run the GPU at PCI-Express 3.0 x8, as the GPU physically lacks the remaining 8 lanes. The lower PCIe lane count should simplify board design for AIC partners, as it reduces the PCB traces and SMDs associated with each individual PCIe lane. Much like DRAM chip traces, PCIe traces are meticulously designed by EDA software (and later validated), to be of equal length across all lanes, for signal integrity.”

1684120778994.png


Source: https://www.techpowerup.com/308599/...to-feature-a-pci-express-4-0-x8-bus-interface
 
It has, of course, absolutely nothing to do with artificially restricting performance for product segmentation.

It's not as if there's any $120 16-lane cards being produced at this very moment that proves the sillyness of the claimed motivation for this.
 
It’s not going to hurt performance.
At PCIe 3.0 8x it will hurt performance, not a lot, but it will.
This is a cost-savings and product segmentation by NVIDIA that is completely unnecessary as a product sold directly to customers, but for OEMs and pre-builts it might make a bit more sense.
 
it reduces the PCB traces and SMDs associated with each individual PCIe lane
Anyone who can't figure this out really should not be allowed have/use a computer....

It's yet ANUTHA money-grubbin cash grab from nGreediya, because it lowers their BOM/design costs, AND screws the consumer all in 1 slick-dick move, just like they always do !
 
Anyone who can't figure this out really should not be allowed have/use a computer....

It's yet ANUTHA money-grubbin cash grab from nGreediya, because it lowers their BOM/design costs, AND screws the consumer all in 1 slick-dick move, just like they always do !

It’s not a cash grab if no one buys it, just sayin’…
 
Lame.

It may not need more than 8x Gen4, but what happens in an older system with 8x Gen3?

It’s not going to hurt performance.

At PCIe 3.0 8x it will hurt performance, not a lot, but it will.
This is a cost-savings and product segmentation by NVIDIA that is completely unnecessary as a product sold directly to customers, but for OEMs and pre-builts it might make a bit more sense.
It will hurt performance, but it's only maybe 2%. Someone will do more tests running in the various pcie gens and compare performance. Likely 1% difference.

"But what about 4k!!!" you say. Not a 4k card, expecting that would be stupidity.
 
I don't see it as a big deal. If you still using 3.0 boards the cpu might be bottlenecking the 4060 anyway.
 
  • Like
Reactions: erek
like this
I don't see it as a big deal. If you still using 3.0 boards the cpu might be bottlenecking the 4060 anyway.

5800x3D will still smoke most processors in gaming even at PCIe3.0, with the competing processors using PCIe4.0.

By the way, the difference between PCIe3.0 and 4.0 for a GPU in gaming, all else considered equal, is negligible:

https://www.techpowerup.com/review/nvidia-geforce-rtx-4090-pci-express-scaling/

PCIe3.0 still has plenty of bandwidth for gaming, but not when you cut off half of the lanes. Not to mention that people likely still on PCIe3.0 platforms are exactly the target market for a card like the 4060 or 4060Ti, because people buying that card are likely on a budget. The RX 6500 demonstrated why this is a terrible design choice.
 
So the 8 lane pci express matters only if the 16gb vram gets full right ??
 
So the 8 lane pci express matters only if the 16gb vram gets full right ??
I imagine it could be a bit more complicated than that, a 4090 24gb will rarely get full and will still show a small drop in performance if you go from x16 4.0 to x16 2.0 and a big one if you do down to x16 1.x, while a 3080 10gb does not seem to show much of a difference between 4.0 and 3.0 (or x16 and x8 in 4.0):
https://www.gamersnexus.net/guides/2488-pci-e-3-x8-vs-x16-performance-impact-on-gpus, but can show a small one versus 4.0x16 and 3.0x8

And for both a 3080 and 4090 the impact of going with small bandwith does not seem to get worst passing from 1080 to 4k, even it is often less (I imagine higher the FPS, less time you have to interact with the gpu), as there will be more interaction with a GPU than filling the memory of texture
 
So the 8 lane pci express matters only if the 16gb vram gets full right ??
Well from all the test I’ve seen a 4090 going from 4.0x16 to 3.0x8 saw at worst a 10% reduction. I doubt that even at 16gb a 3060ti will have many issues there as there will be other bottlenecks in the way.
 
Back
Top