PCI Express (PCIe) 5.0 is the latest generation of the PCI Express interface standard that is used to connect high-speed peripherals to a computer. Some key uses and improvements of PCIe 5.0 include:
Faster Data Transfer Speeds
PCIe 5.0 offers significantly higher maximum per-lane data bandwidth compared to previous generations of PCIe:
| PCIe Generation | Max Bandwidth Per Lane |
|---|---|
| PCIe 1.0 | 250 MB/s |
| PCIe 2.0 | 500 MB/s |
| PCIe 3.0 | 985 MB/s |
| PCIe 4.0 | 1969 MB/s |
| PCIe 5.0 | 3938 MB/s |
This nearly doubles the bandwidth per lane compared to PCIe 4.0. For a typical x16 slot, this translates to a maximum possible bandwidth of 63 GB/s with PCIe 5.0, up from 31 GB/s with PCIe 4.0.
Lower Latency
In addition to higher bandwidth, PCIe 5.0 also reduces latency compared to previous generations. The specification quotes a minimum payload round trip latency of 21ns for PCIe 5.0, down from 29ns for PCIe 4.0. This lower latency benefits time-sensitive applications that require frequent small data transfers, such as real-time media processing.
Increased Power Efficiency
PCIe 5.0 inherits improved power efficiencies from the PCIe 4.0 specification but takes it even further. The PCI-SIG claims PCIe 5.0 offers up to 36% better power efficiency per bit compared to PCIe 4.0. This extends the battery life of mobile devices and reduces power and cooling requirements for high-performance computing applications.
Compatibility and Backwards Compatibility
PCIe 5.0 is designed to be fully backwards and forwards compatible with previous PCIe generations. A PCIe 5.0 device can slot into a PCIe 4.0 or 3.0 slot and automatically negotiate to run at the lower generation’s bandwidth. Similarly, a PCIe 3.0 device can function in a PCIe 5.0 slot. The different PCIe generations are interoperable through standardized electrical channels, connectors, software protocols, and programming interfaces.
Use Cases
Some example use cases and applications that can benefit from PCIe 5.0’s improvements are:
- Gaming: High-end graphics cards for gaming PCs to drive higher resolutions and framerates for immersive experiences.
- Creative Professionals: Video production and editing, 3D modeling and rendering, that require moving large files quickly.
- Data Centers: AI processing, simulation and modeling, rapid database access.
- Networking: 100GbE network interfaces, routers, switches and infrastructure.
- Storage: Fast NVMe SSDs, RAID arrays, and interconnects to provide real-time access to vast datasets.
Graphics Cards
One major application of PCIe 5.0 will be supporting cutting-edge graphics cards and GPUs. Modern graphics cards are incredibly data-hungry – high resolutions, frame rates, and graphics effects demand tons of bandwidth. Previous generation cards have already begun to saturate PCIe 4.0’s limits.
AMD and Nvidia’s top consumer cards in 2022 include:
- Nvidia RTX 3090 Ti – 24GB RAM, 450W TDP, Bandwidth needs – Over 60 GB/s
- AMD Radeon RX 6900 XT – 16GB RAM, 330W TDP, Bandwidth needs – Over 50 GB/s
These bandwidths already brush up against the ~32 GB/s limit of a PCIe 4.0 x16 slot. By providing 64 GB/s on an x16 slot, PCIe 5.0 offers plenty of headroom for future graphics cards to push limits even further with higher resolutions, refresh rates, HDR quality, ray tracing effects, and game physics – enabling deeply immersive gaming experiences.
NVIDIA’s RTX 4090
Nvidia’s new flagship GeForce RTX 4090 GPU, released in October 2022, is one of the first consumer products to utilize a PCIe 5.0 connection. It achieves nearly 2x the performance of the previous generation RTX 3090 Ti.
Compared to its predecessor, key specs include:
| RTX 3090 Ti | RTX 4090 | |
|---|---|---|
| PCIe Generation | 4.0 | 5.0 |
| GPU Bandwidth | 21 Gbps | 23.75 Gbps |
| RT Cores | 84 | 128 |
| Tensor Cores | 336 | 512 |
| TDP | 450W | 450W |
Despite having the same power draw, the 4090 outperforms the 3090 Ti significantly thanks to its efficient new architecture, faster GPU, and critically – access to much more bandwidth from its PCIe 5.0 x16 slot.
Radeon RX 7900 XTX
AMD’s newest flagship GPU, the Radeon RX 7900 XTX launching December 2022, will also take advantage of PCIe 5.0. It achieves up to 1.7x higher performance than previous generation cards in the same power bracket. AMD claims it is the “fastest gaming GPU in the world”.
Key specs compared to previous generation include:
| 6900 XT | 7900 XTX | |
|---|---|---|
| PCIe Generation | 4.0 | 5.0 |
| Stream Processors | 5120 | 6144 |
| Game Clock (GHz) | 2.015 | 2.3 |
| Memory Size | 16GB | 24GB |
| Infinity Cache (MB) | 128 | 192 |
| TDP | 300W | 355W |
The 7900 XTX takes advantage of an enhanced GPU architecture, more stream processors, faster clocks, and crucially – PCIe 5.0 support – to achieve its huge generational performance gains while still using a similar power envelope. The extra bandwidth from PCIe 5.0 is critical in feeding the beastly 7900 XTX.
NVMe SSD Storage
PCIe 5.0 will also drive the next generation of blazing fast NVMe solid state drives for storage. NVMe SSDs directly connect to PCIe lanes to provide orders of magnitude higher performance compared to traditional SATA SSDs and hard drives.
Already, the fastest PCIe 4.0 NVMe drives on the market like the Samsung 980 Pro are brushing up against the limits of PCIe 4.0’s ~4GB/s bandwidth. Native PCIe 5.0 SSDs will smash through that barrier with bandwidths over 7GB/s.
Some models under development include:
- Seagate FireCuda 530 PCIe Gen5 NVMe SSD – Up to 7300 MB/s seq. reads
- ADATA Project Nighthawk PCIe 5.0 SSD – Over 7000 MB/s seq. reads
- Gigabyte PCIe 5.0 SSD – Up to 14000 MB/s seq. reads
These blazing SSDs will provide load times under 5 seconds for even the biggest games, nearly instantaneous boot times, and the ability to copy 100GB of files in under 10 seconds. For professional media workflows, PCIe 5.0 SSDs remove storage bottlenecks and allow nearly real-time editing of 8K RAW video footage.
AMD and Intel Platforms
To take advantage of PCIe 5.0, both a compatible CPU and motherboard are required. AMD and Intel have both introduced platforms that support PCIe 5.0.
AMD
AMD Ryzen 7000-series processors launched in Fall 2022 are the first from AMD with native PCIe 5.0 support. Paired with an AMD X670 or X670E chipset motherboard, they offer full PCIe 5.0 connectivity for graphics cards and storage.
Key processors include:
- Ryzen 9 7950X – 16 cores, up to 5.7 GHz, TDP 170W
- Ryzen 9 7900X – 12 cores, up to 5.6 GHz, TDP 170W
- Ryzen 7 7700X – 8 cores, up to 5.4 GHz, TDP 105W
These processors use the new AMD AM5 socket and also deliver enhanced performance, efficiency, overclocking capability and DDR5 memory support along with PCIe 5.0.
Intel
Intel’s upcoming 13th Gen Core processors, expected to launch in Fall 2022 on the new Z790 chipset, will introduce support for PCIe Gen 5.0. They will provide up to 24 PCIe 5.0 lanes off the processor. Paired with a Z790 motherboard, this platform will enable PCIe 5.0 support for graphics cards and storage.
Intel is touting up to a 15% single-threaded and 41% multi-threaded performance improvement over previous generation CPUs with the Gen13 family. Their flagship 13600K is rumored to offer:
- 16 cores / 24 threads
- Up to 5.5 GHz frequencies
- 175W TDP
Both Intel and AMD’s new platforms will drive widespread consumer adoption of PCIe 5.0 starting in 2023 and pave the way for rollouts in enterprise data centers.
Networking
Faster networking technologies are also taking advantage of PCIe 5.0 connectivity. Terabit network interface cards (NICs) that can hit speeds up to 400Gbps are coming to market using PCIe 5.0 x16 slots. Major vendors include:
- Marvell: Offering 400GbE adapters using PCIe 5.0 and 50G PAM4 SerDes technology.
- Nvidia: Released the Nvidia Quantum-2 series adapters with 400Gbps speed based on PCIe 5.0.
- Broadcom: Announced 200Gb and 400Gb Ethernet adapter prototypes leveraging PCIe 5.0.
These ultra-fast NICs will accelerate server and network infrastructure connectivity in cloud data centers. Cisco estimates 400GbE adoption will grow 10x faster than 100GbE for data centers.
AI and GPGPU Acceleration
AI acceleration, high performance computing (HPC), and GPGPU applications will also benefit enormously from PCIe 5.0. Training deep neural networks requires massive compute capability and bandwidth to shuttle data between GPUs and host system memory or storage. PCIe 5.0 provides a 2x increase in bi-directional bandwidth over PCIe 4.0, accelerating model training times.
NVIDIA’s latest H100 GPU for data centers is designed to leverage PCIe 5.0. Four H100 GPUs can achieve an aggregate bandwidth over 1 Terabit/s through PCIe 5.0. This allows assembling exascale-class supercomputers focused on AI workloads.
Challenges With Adoption
While PCIe 5.0 brings tantalizing performance improvements on paper, there are some challenges to its widespread adoption:
- Costs: The transition to PCIe 5.0 requires purchasing new CPUs, motherboards, NICs and risers compatible with the standard. This represents a significant upgrade investment.
- Immature Ecosystem: CPUs, motherboards, GPUs, storage and network devices built for PCIe 5.0 are just emerging. The ecosystem needs time to mature.
- Limited Use Cases Initially: The bleeding edge performance of PCIe 5.0 may be overkill for common workstation and PC needs today. Power users like gamers and content creators will drive early adoption.
- Heating and Power Delivery: PCIe 5.0’s increased speeds require more robust power delivery and cooling solutions, increasing costs and complexity.
Therefore, while compatible hardware is now releasing, widespread adoption of PCIe 5.0 may take 2-3 years. During this time, costs will come down, the ecosystem will expand, and use cases requiring these extreme speeds will compel upgrades.
The Future – PCIe 6.0
Even as PCIe 5.0 is beginning its rollout, the PCI-SIG is already developing the next generation PCIe 6.0 specification, expected to be finalized in 2021. PCIe 6.0 will again double bandwidths to provide speeds up to 256 GB/s over a x16 slot.
Some headline numbers for PCIe 6.0 include:
- Up to 256 GB/s per direction (x16 slots)
- Up to 64 GB/s per direction (x1 slots)
- Sub-10ns latency
- Pulse amplitude modulation (PAM4) signaling
This could enable truculent PCIe bandwidths of over 1 Terabyte/s in a dual x16 slot configuration. Applications like machine learning, network infrastructure, HPC, and cloud computing will surely take advantage.
However, PCIe 6.0 will require even more advanced controllers, connectors, cabling and power delivery to handle these speeds while maintaining signal integrity. Widespread PCIe 6.0 adoption may only come around 2026 and beyond.
Conclusion
PCI Express 5.0 represents a significant leap forward in interface bandwidth and latency. It provides major headroom for bleeding-edge graphics cards, hyperfast storage, networking, AI acceleration and more. While just entering the market, compatible hardware will drive rapid development of the PCIe 5.0 ecosystem over the next 2-3 years. Adoption will be led by early adopter power users at first before hitting mainstream. Looking even further ahead, PCIe 6.0 on the horizon will again double theoretical speeds to 256 GB/s per lane. The future of high performance computing will undoubtedly ride the PCIe express.