PCIe lanes are not the new IRQ.
PCIe lanes are not the new IRQ.
I recently assembled a new PC and encountered an unexpected issue: sufficient PCIe lanes. While I received some feedback here and ASUS provided a general response, there hasn't been a detailed analysis in any articles or videos about how PCIe sharing affects performance. Just because the motherboard offers two Gen 4 M.2 slots doesn’t guarantee each slot has its own dedicated lanes. Adding a non-GPU component that shares lanes with another SSD could impact overall speed. Similarly, SATA drives would also be affected by lane sharing. A system with many components tends to lag compared to a minimal setup with just a GPU and one NVMe drive, which gets dedicated bandwidth. This situation reminds me of the days when you had to monitor interrupt queue resources closely.
It's the same pace. PCIe lanes aren't divided by bandwidth but by how they're assigned. For example, a M.2 port might get four lanes or the PCIe slot could handle four; you either receive M.2 PCIe or two additional SATA connections. Getting this right requires careful planning to avoid removing something essential by inserting another component.
I possess an ASUS TUF x570-Plus, R5-3600, with an NVMe SSD in the first M.2 slot and another in the second slot. There are also two mirrored Iron Wolf Pro HDDs for long-term, bulk storage. The setup aimed to use the second drive mainly as a cache or scratch disk. It functions well, but I'm uncertain about its optimal configuration. Asus tech support explained that x16 is reserved for the GPU, x4 for M.2 #1, and M.2 #2 uses lanes from the four chipset lanes. Isn't this naturally slower? Also, does this mean the second drive shares the chipset's x4 with all other parts?
You choose whether the speed benefit outweighs the need for extra storage.
All the content focused on maximizing speed seems to overlook this aspect. In my situation, selecting the second Gen 4 M.2 was aimed at creating a highly efficient cache or scratch disk, which offers quicker performance than a standard SSD—though even with high-end processors like i9 or Ryzen 9, you're limited to around 24 lanes. Argh!
The setup will vary based on how ASUS configures the X570 PCI-E lanes. Generally, AMD indicates the X570 chipset offers a flexible configuration—allowing you to utilize all available lanes or select specific ones. This impacts M/2 slots, PCI-E ports, and SATA connections, depending on whether you customize them. The second M.2 port may not be faster but will experience greater latency because it must pass through the X570 chipset before reaching the CPU. In contrast, the first M/2 slot connects directly to the CPU. With the ASUS TUF X570-PLUS, the M.2 and SATA ports remain functional even when the second M.2 slot is occupied, as confirmed by the motherboard manual. The first M.2 slot travels through the CPU, while the 4.0 X16 slot goes directly to the CPU. The 4.0 X4 2nd PCI-E X16 slot connects via the X570 chipset, and the 4.0 X4 second M.2 port passes through the X470 chipset. All PCI-E X1 slots stay on the X570 chipset; Intel’s i9 or Ryzen 9 chips are considered mainstream consumer options. It took some time for Intel to adapt from being limited to 16 PCI-E lanes until the 11th generation, but it now focuses on one GPU and one M.2 slot. Ultimately, the board manufacturer’s choices play a significant role.
You're asking if the first M.2 is actually running at 4.0 x4 speed instead of 4.0 x4 performance. The X16 slot is for the GPU, and you're wondering why it might not match the performance of the first slot. The new Asus ProArt x670 motherboard includes four M.2 ports—two at 5.0Gb and two at 4.0Gb—and two X16 slots. Despite this, a 7000 series Ryzen CPU still has only 24 lanes, which raises questions about the actual performance differences between slots.
It's a setup with 24 lanes in total. The configuration includes 16 PCI-E 5.0 lanes for the GPU, plus 4 PCI-E 5.0 lanes for the first M.2 slot and another 4 PCI-E 5.0 lanes connecting the GPU to the X670 chipset. The second M.2 slot is likely routed through the X670 chipset as well.
But isn't it forming a slow point for storage—specifically for that final x4 connection to the chipset? I've noticed many discussions before about how helpful the dedicated x4 in the first M.2 is. The initial GPU handles itself with 5.0 x16, just like the first M.2 with its own x4. Yet all the other components on that x670 board—especially the extra three M.2s—are limited by the chipset's small x4 supply. It makes sense to think that if there are four M.2 slots, they should all work fine. Where does the second set of 5.0 x16s get their bandwidth from? That’s why I’m curious about videos or articles checking how much performance or stability gets affected by using the chipset’s x4.