Mon Jul 08 2024
PCIe 5.0 is the latest generation of PCIe, or peripheral component interconnect express, adopted by mainstream hardware. Developed in 2019, it has only recently shown up in motherboards in the M.2 slot and the x16 GPU slot.
While there have been no mainstream GPUs supporting the PCIe 5.0 standard as of now, a lot of storage devices and M.2 NVMe drives are on the market utilizing PCIe 5.0. Since the standard is backward compatible, so slotting in PCIe 4.0 or the older PCIe 3.0 devices in these slots is okay. So, what are the differences in PCIe 5.0? Does it warrant any consideration when configuring your next system?
PCIe 5.0, the latest PCI Express standard, represents a doubling of speed over its predecessor PCIe 4.0, with giga transfers increasing from 16GT/s to 32GT/s per lane. The increased speed per lane doubles from approximately 2GB/s to 4 GB/s, an aggregate bandwidth of about 64 GB/s. PCIe 5.0 is the fastest PCIe standard on the market.
| Lanes | PCIe 5.0 (32 GT/s) | PCIe 4.0 (16 GT/s) |
|---|---|---|
| x1 | 4 GB/s | 2 GB/s |
| x2 | 8 GB/s | 4 GB/s |
| x4 | 16 GB/s | 8 GB/s |
| x8 | 32 GB/s | 16 GB/s |
| x16 | 64 GB/s | 32 GB/s |
PCIe 5.0 rapidly moves data between CPUs, GPUs, storage devices, networking, and any other PCIe peripherals connected. These doubling in speeds hinge on the devices, the lane standard, and the application to go hand in hand in optimizing data transfer at the highest throughput.
CPUs that support PCIe 5.0 include
Keep in mind the lane splitting and where PCIe 5.0 is routed to on each motherboard to ensure you know which slot is PCIe 5.0-enabled. We will go over the different PCIe 5.0 devices and if they make a meaningful difference in various workloads.
As of June 2024, there are no Gen 5 GPUs in the NVIDIA RTX line (up to the 40-series) or the AMD Radeon RX line (up to the 7000-series). However, it is expected that the next generation of GPUs will support PCIe 5.0, following the increased adoption of modern motherboards and CPUs supporting PCIe 5.0.
But that begs the question, will PCIe 5.0 make a huge difference in GPU performance?
Judging by past generational leaps, we can say that PCIe 5.0 will likely result in little to no performance increase attributable solely to GPUs. When the standard evolved from PCIe 3.0 to PCIe 4.0, GPU performance did not significantly improve; modern mid-tier GPUs (4060 to 4070 Ti) can still run on PCIe 3.0 with minimal performance loss (within a 5% deviation). Therefore, the speedups gained from gaming on next-gen PCIe 5.0 GPUs will probably not result from the increased bandwidth alone.
While PCIe 5.0 may not make a huge difference for mainstream gaming, it will become more impactful as peripherals and applications fully optimize and utilize the increased performance.
The increased bandwidth can benefit scenarios where GPUs need to transfer large amounts of data quickly. Productivity applications can push even the highest-tier GPUs to their limits, and PCIe 5.0 can significantly enhance workloads involving multiple GPUs. When multiple GPUs execute calculations in parallel, seamless and fast data transfer and communication are essential, necessitating high GPU bandwidth.
Workloads such as video editing, high-performance computing (HPC), AI training, and large-scale simulations benefit from fast access to storage devices. When file sizes exceed GPU memory capacity, quick access to storage is crucial to minimize wait times.
After PCIe 5.0 was introduced on motherboards, manufacturers quickly released their own PCIe 5.0-enabled SSDs. While PCIe 4.0 SSDs reach a limit of about 7000 MB/s, current PCIe 5.0 SSDs on the market can achieve speeds over 10,000 MB/s, with a theoretical limit of 14,000 MB/s.
The increased performance and speed are beneficial for tasks involving large file and data transfers, such as loading applications, reading and writing data directly off storage, and other data-intensive workloads.
Data transfer speeds are simple and easy to understand – more lanes mean more speed. Gen 5 storage devices exemplify this. NVMe SSDs are now the gold standard in data storage due to their high speed, multitasking capability, energy efficiency, and reliability.
NVMe SSDs in the gumstick design M.2 form factor are popular for consumer use as primary storage. Fast storage allows mainstream users to access large files quicker, upload and download files to storage faster (if networking permits), and load games faster. Gaming titles featuring Direct Storage capabilities let the GPU access game files directly from the NVMe SSD without first transferring data to the CPU, reducing CPU overhead and thus speeding up load times.
In enterprise settings, M.2 SSDs are often used as the operating system boot drive, while PCIe 5.0 is found on the 2.5” form factor U.2 hot-swappable drive. Data-intensive workloads constantly pull data, especially in real-time simulation or data analytics, where the speed at which data is written and read directly correlates to time to completion. Real-time data analytics just got even closer to real-time.
Artificial intelligence and machine learning, which constantly pull data to make predictions, can execute faster due to the increased bandwidth offered by PCIe 5.0. Modern enterprise CPUs house ample full-speed PCIe 5.0 lanes, and servers dedicate an entire backplane of PCIe-enabled NVMe slots directly connected to the CPU to minimize performance bottlenecks and allow for easy RAID configurations for peak performance.
As PCIe 5.0 continues to mature, its promise of enhanced bandwidth and speed holds considerable potential for both gaming enthusiasts and professionals alike. While the immediate impact on gaming performance may be modest, the future readiness it brings to handle increasingly data-intensive applications will continue to become more and more optimized in using faster data transfer technology.
As PCIe 5.0-enabled hardware becomes more prevalent, its ability to handle large-scale data transfers swiftly and efficiently will become increasingly valuable. This technology promises to streamline workflows and reduce latency in demanding tasks, enhancing productivity and performance across various sectors. And sooner or later PCIe 6.0 (the newest standard that has already been developed and defined) will bring even more improvements.
In essence, while PCIe 5.0 may not revolutionize gaming overnight, its foundational role in enhancing data throughput and accessibility ensures that the technology remains a cornerstone of future-proof computing infrastructure.
If you're looking to configure the very best workstations to future proof your computing infrastructure and deployment, contact SabrePC today!
Read More
Mon Jul 08 2024
Computex 2024 AMD formally announces AMD Ryzen 9000, their 5th generation consumer desktop processor, and delivers a teaser for the data center. While these AMD Ryzen CPUs are geared towards peak gaming performance, that shouldn’t detract from the productivity they can deliver in certain applications. Let's go over the newly announced Zen 5 Ryzen offerings:
5th Gen AMD Ryzen is powered by the AMD Zen 5 core built on the same AM5 platform which means that 5th Gen Ryzen can be used in last gen motherboards. AMD has also committed to deliver CPUs for the AM5 platform to 2027.
Zen 5 is an engineering marvel, featuring more performance and more efficiency.
AMD Zen 5 delivers up to a 2x increase in instruction bandwidth, data bandwidth, and AI performance compared to Zen 4 with an average 16% average IPC uplift. Zen 5 seen in consumer and desktop processors will also power the data center.
A very brief but very important announcement was 5th Gen AMD EPYC Turing. Still using the SP5 socket, the newest EPYC announced is likely a successor to Bergamo. EPYC Turin will utilize Zen 5c a cut-down version of Zen 5 cores where the cache and size are significantly reduced to fit additional cores on the CPU.
5th Gen AMD EPYC Turin delivers 192 Zen 5C cores, trumping the EPYC Bergamo 9754 (128 cores) by 64 entire cores! AMD already has a stronghold in high-core density cores with 4th Gen Bergamo, but 5th Gen AMD EPYC Turin further extends their lead. A dual AMD EPYC Turin deployment delivers 384 Cores and 786 threads, an unheard-of per-node density.
These high-core count data center processors are perfect for powering enterprise cloud, where cores can be allocated and split up for different clients. Cloud-based workloads, virtualization servers, and other non-highly intensive applications can take advantage of the ample number of efficient cores in Zen 5c. Specifications and other AMD EPYC Turin CPU models have not yet been released, but we can expect 5th Gen EPYC to be delivered in 2H of 2024.
If you have any questions on sourcing AMD Ryzen or AMD EPYC, contact us today and we can deliver official quotes or official availability.
Read More
