[ANSWERED] Iray - Do PICe lanes matter once the scene is loaded into a graphics card?
![Joe827](https://secure.gravatar.com/avatar/88b983aa04098c69b204a121d3c37c44?&r=pg&s=100&d=https%3A%2F%2Fvanillicon.com%2F88b983aa04098c69b204a121d3c37c44_100.png)
Gut check on my understanding...
During an Iray render, a "scene" is sent to the graphics card for processing by the GPU. The scene consists of geometries, textures etc. I will be replacing a 10 year old mother board in the near future. I currently use 2 Evga 1080TI's for rendering. Initially, I assumed that all LGA 1151 boards operate in a x16/x8 or x8/x8 configuration when two cards are installed. However, I did find exactly one LGA 1151 board that operates in a x16/x16 configuration with two cards.
So nuts n bolts question, outside of scene transfer time, is anyone able to speak for the impact of available PCIe lanes on render time? My assumption is that the difference isn't big since the bulk of the data is transfered when the scene is sent to the card, but that's just my gut and realize there may be more to this than meets the eye.
Comments
There are people who render on x1 slots and have reported just fine results.
You might like to check this link:
https://www.pugetsystems.com/labs/articles/NVIDIA-GeForce-RTX-2080-Ti-PCI-Express-Lane-Scaling-in-OctaneRender-and-Redshift-1259/#DoPCI-Expressx1,x4,x8,vsx16LanesMatterforGPURendering?
In short: The RTX 2080 ti and above cards are the first time cards that have a very minor peformance advantage with X16. But it is not relevant for real world speed. I remember there were some other benchmarks (not linked) concluding that the 1080 ti is not limited by x8.
More interesting is the next card generation. As - at least with AMD CPUs and boards - we now have PCI4 and if Nvidia is making a PCI4 card then there might be relevant speed improvment in the future (not discussed in the linked article)..
I saw some calculations on that, based on current games. It seems unlikely that even a game trying to push 90fps at 4k would saturate a PCIE gen 3 x16 connection. PCIE gen 4 (and gen 5 which is coming soonish) are going to be most relevant, at least for the next 3 or 4 years, to high speed storage. A PCIE gen 4 m.2 drive on a x4 connection has a maximum theoretical bandwidth of 64 GT/s (roughly 8Gb/s)
Quick summary, the difference between 8 and 16 lanes is
Thank you. That's very helpful.