PCIe Bandwidth Calculator

A PCIe Gen4 x16 slot delivers a theoretical maximum of 31.5 GB/s (gigabytes per second) in each direction. This is calculated as 1.969 GB/s per lane times 16 lanes. It's the standard slot for high-end graphics cards like the RTX 4090 or Radeon RX 7900 XTX. Actual throughput may be slightly lower due to protocol overhead.

No, a PCIe Gen4 SSD will operate at Gen3 speeds if installed in a Gen3 slot. For example, a Gen4 x4 SSD rated for 7 GB/s will be limited to about 3.94 GB/s in a Gen3 x4 slot. Both the motherboard and CPU must support Gen4 for full performance.

Yes, PCIe is physically and electrically backward compatible. You can use a PCIe x1 card in a x16 slot, but the card will only use one lane, so bandwidth will be limited to the x1 maximum for that generation. The extra lanes go unused.

Some motherboard slots are full-length (x16) for mechanical compatibility but have only x4 or x8 electrical lanes. In this case, any installed card is limited to the number of active lanes. For example, a Gen3 x16-sized slot with only x4 lanes provides 3.94 GB/s, not the 15.76 GB/s of a true x16 slot.

You can use software tools like GPU-Z, HWiNFO, or CrystalDiskInfo to check the current PCIe generation and lane width for GPUs and SSDs. These tools report the negotiated link speed and width in real time, which is helpful for verifying slot wiring and device compatibility.

For most gaming workloads, PCIe Gen3 x16 (15.76 GB/s) is enough for current GPUs like the RTX 4070 or RX 6800. However, some AI, compute, or workstation tasks can benefit from Gen4 or Gen5 bandwidth, and using a card in a lower-bandwidth slot (x8 or Gen3 x8) can reduce performance in data-intensive scenarios.

Each PCIe generation doubles the per-lane bandwidth of the previous one. Gen3 offers 0.985 GB/s per lane, Gen4 provides 1.969 GB/s per lane, and Gen5 jumps to 3.938 GB/s per lane. This means faster SSDs, GPUs, and network cards can make full use of the latest platforms.

Common reasons: the SSD is in a lower-generation slot (Gen3 instead of Gen4), lane count is limited (x2 or x1 instead of x4), or the system’s CPU/motherboard does not fully support the required PCIe standard. Always check your board’s specs and use the calculator to confirm slot bandwidth.

Yes, especially on consumer motherboards where chipset lanes are shared among multiple devices. Installing multiple NVMe SSDs or PCIe cards can cut available lanes for each slot, lowering bandwidth. Always check your motherboard manual and CPU lane allocation.

Check your motherboard’s official specifications and CPU support list. Gen5 support is currently found on Intel 12th/13th/14th Gen (Z690, Z790) and AMD Ryzen 7000 (X670, B650) platforms. Even if the board has a Gen5 slot, your CPU must also support it.

PCIe protocol overhead (encoding, error correction, and packet headers) reduces usable bandwidth by 2-5 percent compared to the theoretical maximum. For example, a Gen4 x4 slot with 7.876 GB/s raw bandwidth may deliver about 7.5 GB/s in practice. Device controller efficiency and firmware can further reduce real-world speeds.

Thunderbolt 3 and 4 connections provide a PCIe Gen3 x4 link, which means 3.94 GB/s of theoretical bandwidth. This is much lower than a desktop x16 slot, so high-end GPUs in external enclosures will see reduced performance compared to native installs.

Yes, but the SSD will run at Gen4 speeds. For example, a Gen5 x4 SSD capable of up to 15 GB/s will be limited to about 7.8 GB/s in a Gen4 x4 slot. Both the motherboard and CPU must support Gen5 for full performance.

A 100GbE NIC requires at least Gen3 x16 or Gen4 x8 for full throughput without bottleneck. For 200GbE, Gen4 x16 or Gen5 x8 is recommended. Always check your NIC and motherboard documentation to ensure sufficient PCIe bandwidth.

PCIe bifurcation splits a single slot’s lanes between multiple devices (often for RAID or multiple NVMe SSDs). The total bandwidth is shared across connected devices. For example, bifurcating a Gen4 x16 slot into four x4 segments gives each device 7.876 GB/s, but the total bandwidth remains 31.5 GB/s.

Yes. The calculator shows the raw, theoretical maximum (excluding overhead). In reality, protocol overhead, device controller efficiency, and system architecture reduce usable throughput by 2-10 percent. Always expect slightly lower actual speeds when benchmarking.