Intel claims PCIe 4.0 Storage Leadership with their Rocket Lake processors | OC3D News (overclock3d.net) Intel's Ryan Shrout has released some benchmarking data for the company's upcoming Rocket Lake powered i9-11900K processor, highlighting the drive's PCIe 4.0 storage performance using PCMARK 10's Quick System Drive Benchmark. Compared to AMD's Ryzen 9 5950X processor, a 16-core Zen 3 CPU, Intel's 8-core i9-11900K achieved a reported 11% performance lead over AMD's current desktop flagship. These tests were conducted using comparable Ryzen and Intel systems using the same memory, SSDs GPUs and drivers with comparable motherboards.
Strangely, Intel lists their AM4 motherboard as ASUS' X570 ROG Rampage VIII Hero in their
benchmark notes, which is a motherboard that doesn't exist. Intel should not have made this mistake, but let's assume that Intel meant to list ASUS' X570 ROG CrossHair VIII Hero motherboard. Intel's Rocket Lake CPU architecture is the company's first to support PCIe 4.0 on desktop platforms, allowing Intel to access SSD performance levels which can push past the limits of PCIe 3.0. This will allow Intel systems to utilise faster drives and benefit from faster read/write speeds.
What exactly was Intel testing? In
UL Banchmarks' technical guide for PCMARK 10, the company stated that their Quick System Drive Benchmark comprises six stages, which are listed below.
- cps1 - Copying 339 JPEG files, 2.37 GB in total, into the target drive (write test)
- cps2 - Making a copy of the JPEG files (read-write test)
- cps3 - Copying the JPEG files to another drive (read test)
- exc - Using Microsoft Excel ill Using Adobe Illustrator
- psl - Using Adobe Photoshop (light use)
Please note that this test is not as extensive as PCMARK 10's Full System Drive Benchmark, a lengthier test that includes more stages and more varied testing. Another thing that is worth mentioning is that the "Quick" storage benchmark that Intel used is designed to "measure the performance of small system drives from traditional spinning drives (HDDs) at the low end and entry-level PCI Express SSDs at the high end". This benchmark is NOT designed for high-end PCIe 4.0 SSDs.
While the benchmarks that Intel have shared are legitimate, we are baffled that Intel refused to share data from PCMARK 10's Full System Drive Benchmark. UL Benchmarks' own documentation says that their Quick System Drive Benchmark is "better suited for testing entry performance level and lower capacity drives", making it a strange choice for a high-end PCIe 4.0 benchmark comparison.
What we fear is that this is a CPU performance benchmark masquerading as a pure storage benchmark. Intel's results imply that AMD's processors are not making the best use of PCIe 4.0 storage drives, though it is likely that AMD's Ryzen processors are facing a CPU-based performance limit that Intel's system is not.
High-end storage benchmarks are inherently difficult to produce. Most write tests require benchmarking tools to rapidly create new files to write to SSDs at rates that are now measured in gigabytes per second. Many outdated SSD benchmarking tools have now effectively become CPU benchmarks, as performance can be limited by single-threaded CPU processing power and not raw SSD performance.
Are Intel's benchmarks misleading? Yes and No. While Intel is using UL Benchmark's PCMARK 10 Quick System Drive test outside of its design intent, as it is a test that's designed for "small system drives from traditional spinning drives (HDDs) at the low end and entry-level PCI Express SSDs", Intel's results are the result of legitimate testing between two systems.
While Intel has used PCMARK 10's Quick Storage Benchmark outside of its original remit, it does highlight how storage performance and CPU performance interact. Fast storage solutions require processors that are fast enough to keep up with their data output. In PCMARK's Quick Storage test, Intel has showcased how their i9-11900K can deliver more performance than AMD's Ryzen 9 5950X.
Regardless of how you look at this testing, it is hard to deny that the results are fascinating. Regardless, it would have been nice to have also seen results from PCMARK 10's more extensive Full System Drive Benchmark, or at least hear a more in-depth explanation for why Intel's i9-11900K bests AMD's 16-core Ryzen 9 5950X. It would be nice to have some more context with these benchmarks. When speaking to a technical audience, a simple "here is the result, we win!" is not good enough.
Looks like Intel was tired of hearing about AMD's PCI-E 4.0 and finally decided to innovate in my personal opinion.