PCI Express scaling with Geforce RTX 4090 and Core i9-13900K

PCI Express scaling with Geforce RTX 4090 and Core i9-13900K

from Andrew Link
What is the status of scaling the PCI Express lanes between processor and graphics card? Techpowerup asked a Geforce RTX 4090 and a Core i9-13900K to dance. The case is clear, but not applicable to every scenario.

Techpowerup carried out the scaling test for PCI Express with current hardware and determined whether anything had changed since the last tests – for example those of PC Games Hardware (see list of links). So does it make a difference whether the graphics card is connected to the CPU with 16 or 8 lanes? Beware of spoilers: still not worth mentioning in this constellation.

Accompanying material:

The scaling test was carried out in three typical resolutions with a mix of games and the result clearly shows that the bandwidth of PCI Express 4.0 with 8 lanes (or Gen 3 with 16 lanes) is completely sufficient at the moment. Anything above that brings no real added value. Accordingly, AMD and Nvidia are also reluctant to switch graphics cards to Gen 5, because this is also associated with additional costs that either have to be passed on to the customer or, alternatively, squeeze the margin.

With 4 lanes, the loss of performance becomes more obvious. You wouldn’t recommend it if you didn’t have to, but if you have to, it’s still bearable. However, this should primarily be an issue for laptops with – potentially external – graphics cards.

However, one must not forget that there is movement in the market. Just because the combination of Geforce RTX 4090 and Core i9-13900K doesn’t benefit doesn’t mean there can’t be differences. PCGH had taken on the cause with the Radeon RX 5500 XT, among other things. With 8 lanes PCI Express 3.0, it is noticeably slower than with PCI Express 4.0, because in absolute terms a few frames count with slow graphics cards and if the cheap models are still running out of memory, the problem worsens due to access to the main memory. Speaking of main memory: Direct Storage also generally benefits from higher bandwidths, but is used sparingly at the moment. And the measurement results from Techpowerup also show that ray tracing can potentially benefit from more bandwidth. But the same applies here: the difference isn’t dramatic, and it doesn’t affect a large number of games.

See also  Fortnite: Is there a Season 3 Live Event coming?

Source: tech powerup