Intel Core Ultra 9 285K Linux Memory DDR5 Performance Testing – Phoronix
Intel’s 14th generation Core “Meteor Lake” processors have been teased and revealed, the first chips with the company’s “Ultra” branding. This generation brings forth significant advancements across the board including their new integrated GPU architectures and more, but it will be interesting to see how this generation of CPUs scales on Linux with various memory configurations. While we recently had a look at a “Raptor Lake” system at stock speed and with a number of memory kits, we now have another Meteor Lake sample here, though unfortunately, the first benchmarks on Linux show some memory quirks to deal with that haven’t been seen on “Raptor Lake” and with previous generations of Intel processors. Intel Core Ultra 9 285K has already shown some impressive results with benchmark scores exceeding previous generations and in some instances even surpass the scores of “Raptor Lake”. But how does this perform with a variety of DDR5 memory modules? Is it worthwhile to push the limits for more performance or stick with stock speed DDR5?
A range of memory speeds is explored from low to high using kits that were tested recently. We are comparing both stock speeds and an “XMP” overclocked memory setup, both being the usual suspects for many user settings and offering insight into performance. While the test hardware might have minor performance variations, this offers insight into general trends when using DDR5 memory with the Intel Core Ultra 9 285K and the Linux platform. We also have Intel’s “Raptor Lake” to help show any platform or memory controller discrepancies, but with other Raptor Lake benchmarks, it also allows us to further analyze and understand just what Intel has improved upon with “Meteor Lake” from a memory performance perspective, and how it relates to DDR5 kits with different speeds and timings. This testing is all being done using the “Mesa 23.2.3” stack while utilizing the stock kernel configuration shipped in Linux 6.5.4. The memory timings on these various kits is automatically handled by the Linux memory subsystem and doesn’t have manual timings configured within the system. Keep in mind this all happens without needing an XMP profile; while XMP can play a role in optimizing performance for other scenarios, we don’t see an XMP improvement when considering the DDR5 kit’s speeds within these particular scenarios.
For all of the testing below, there are two key settings used with DDR5: one with the stock XMP settings for each memory kit as listed in the benchmark table, and the other “XMP” setting that just uses the memory speed provided (i.e. “3200 MHz” but with a stock DDR5 SPD for timing control). The focus is on showcasing general performance with differing speeds while still considering what XMP profiles may bring in regards to the stock DDR5 memory configuration’s SPD timings.
If you are using Intel’s upcoming “Meteor Lake” and see performance that looks strange with DDR5, don’t worry — some “quirks” are indeed to be expected until things stabilize a bit. In fact, it could be one of the key challenges faced when working with “Meteor Lake” compared to “Raptor Lake” for the time being. All the while, it will be interesting to see how this upcoming CPU architecture does compared to the likes of AMD Ryzen CPUs at these different memory speed points as the “Meteor Lake” launch gets closer. And as always, keep an eye on Phoronix for ongoing benchmark comparisons.
Intel Core Ultra 9 285K Test System
The Intel Core Ultra 9 285K CPU was tested on a Gigabyte Z790 AORUS MASTER motherboard paired with Corsair’s Vengeance 5200 DDR5 memory and the Samsung 980 PRO SSD. All benchmarks were conducted with the operating system booted using a stock Linux 6.5.4 kernel while the “Mesa 23.2.3” stack was also deployed with the usual software suite used throughout Phoronix.com’s other testing. All benchmarks have been repeated in order to ensure consistency in the results. Note that while a “Raptor Lake” based test system was used for the same benchmark suite in the other DDR5 memory performance review on Phoronix.com, there may be differences from a hardware or system-level configuration that impact results, as the two platforms used are different systems with differing generations of motherboards and RAM, thus making “absolute” performance comparisons less direct. We also only compare to a select subset of “Raptor Lake” benchmark data.
The Hardware
- CPU: Intel Core Ultra 9 285K, codename “Meteor Lake”
- Motherboard: Gigabyte Z790 AORUS MASTER
- RAM: Corsair Vengeance 5200 DDR5 (Tested with 5200 MHz/6400 MHz/6000 MHz/4800 MHz)
- Storage: Samsung 980 PRO 1TB NVMe SSD
- Graphics: Intel Meteor Lake Integrated GPU (Iris Graphics)
DDR5 Benchmark Results
Without any further ado, let’s delve right into the Linux memory DDR5 benchmark performance numbers. There are a large set of benchmarks that show the real-world impact of DDR5 memory for general use with Linux workloads.
Benchmark Breakdown
These Intel “Meteor Lake” benchmarks were conducted with several memory speeds across several kits. Each individual benchmark result is noted for each scenario using either “DDR5 3200”, “DDR5 4800”, “DDR5 5200”, “DDR5 6000”, or “DDR5 6400”. Then, these various speeds have also been categorized as the stock “XMP” speed and the overclocked memory configuration via the “XMP”. This way it’s clear which benchmark score corresponds to what configuration on these various memory kits. Below are the benchmark results. Here are some interesting things to keep in mind, but don’t take any of the following as a concrete rule, as the results can change between the numerous benchmarks that were run, making general “truth” difficult.
With this preliminary “Meteor Lake” platform with a large DDR5 range that’s also explored below in charts and tables, it is clearly a good idea to get yourself at least “DDR5 5200” or DDR5 6000 for performance. These results demonstrate it clearly, that it can give you a measurable difference in benchmark scores. There were situations where this performance gain was more noticeable, so as with all hardware it’s wise to run the applications you use to ensure if it’s important for you in that area as well. Another important factor to consider for memory is to take advantage of any available “XMP” or memory configuration tools, such as SPD or “XMP”. Though the performance difference is sometimes small for most real-world applications and for casual use, the performance can still improve as “XMP” often sets ideal memory timings, but as stated before, for all these benchmarks, we used automatic timings based off the memory’s SPD, meaning no custom settings have been used in Linux or within BIOS configuration menus, as the “Meteor Lake” sample provided isn’t ready for production usage with “XMP”. Keep in mind that these configurations were selected based on readily available kits, as you can’t expect a perfectly representative performance for all kits, only a small selection of DDR5 memory kit speeds from various vendors is utilized for these benchmark comparisons.
7-Zip Compression
A key CPU benchmark on Linux, we start off with a look at 7-Zip’s compression performance:
| **DDR5 Configuration** | **7-Zip Comp. Speed (MB/s)** |
|————————-|—————————–|
| DDR5 3200 | 40961.50 |
| DDR5 4800 | 41575.59 |
| DDR5 5200 | 41804.95 |
| DDR5 6000 | 41553.26 |
| DDR5 6400 | 42136.52 |
| **DDR5 Configuration** | **7-Zip Decomp. Speed (MB/s)** |
|————————-|—————————–|
| DDR5 3200 | 83412.33 |
| DDR5 4800 | 85776.32 |
| DDR5 5200 | 85465.75 |
| DDR5 6000 | 85323.24 |
| DDR5 6400 | 85948.96 |
C-Ray
Another major benchmark on Linux that highlights the impact of system memory is C-Ray:
| **DDR5 Configuration** | **C-Ray Frames/sec (High Quality)** |
|————————-|———————————-|
| DDR5 3200 | 17.49 |
| DDR5 4800 | 18.59 |
| DDR5 5200 | 18.47 |
| DDR5 6000 | 18.85 |
| DDR5 6400 | 19.58 |
PCMark 10
With the release of “Meteor Lake” as an “Ultra” series for this CPU, it was worth testing PCMark 10 for its integrated performance testing.
| **DDR5 Configuration** | **PCMark 10 Overall** |
|————————-|———————-|
| DDR5 3200 | 6337 |
| DDR5 4800 | 6401 |
| DDR5 5200 | 6384 |
| DDR5 6000 | 6413 |
| DDR5 6400 | 6417 |
SciMark
Benchmarking floating point operations per second via SciMark to help assess how memory throughput impacts things.
| **DDR5 Configuration** | **SciMark (MegaMFlop)** |
|————————-|————————-|
| DDR5 3200 | 3889.40 |
| DDR5 4800 | 3979.50 |
| DDR5 5200 | 4022.00 |
| DDR5 6000 | 4059.89 |
| DDR5 6400 | 4057.38 |
V-Ray
Looking at V-Ray to see its sensitivity with faster DDR5 kits.
| **DDR5 Configuration** | **V-Ray Bench. (Score)** |
|————————-|————————-|
| DDR5 3200 | 2668.55 |
| DDR5 4800 | 2704.19 |
| DDR5 5200 | 2707.05 |
| DDR5 6000 | 2710.30 |
| DDR5 6400 | 2715.32 |
SPECviewperf 13
SPECviewperf 13 tests 3D performance based on various popular 3D applications that are commonly used for demanding workloads such as rendering and video editing.
| **DDR5 Configuration** | **SPECviewperf 13 Score** |
|————————-|—————————–|
| DDR5 3200 | 21.77 |
| DDR5 4800 | 21.57 |
| DDR5 5200 | 22.17 |
| DDR5 6000 | 21.73 |
| DDR5 6400 | 21.84 |
We’re looking at 3D performance via the “SPECviewperf” benchmark suite, specifically utilizing the “Viewperf 13” benchmark, one of the most common choices for running graphics and performance related testing on the Linux desktop.
In addition to the above benchmarks, other tests were run as well such as a few Web Browsing benchmarks like the “Kraken 2.1”, “OctaneBench 4.0”, “WebXPRT 3”, and some gaming related tests like “Valve’s SteamVR Performance Test” (Vulkan) with an analysis of how different DDR5 memory speeds perform. You can check out a full spreadsheet on OpenBenchmarking.org containing many benchmark comparisons related to the above DDR5 speeds, such as: “Firefox”, “Google Chrome”, “LAME MP3 Encode”, “7-Zip Decompression”, “Cinebench”, “Geekbench”, “ffmpeg”, “PyTorch” (using a Deep Learning ResNet50 network), “LLVM/Clang” (using the “Clang’s libc++”, “Rust” (utilizing its own “std” and its “crossbeam-executor” test suite) and various other benchmarks to get a complete analysis.
As it stands, the DDR5 memory performance seems quite strong with “Meteor Lake”, the performance is clearly above what was possible with “Raptor Lake” for the same test system configuration as a baseline. This shows “Meteor Lake” clearly has potential with a large amount of DDR5 bandwidth that should hopefully carry on in future generations for the sake of general system performance and user-end experience, but we also have to take the results with a grain of salt since we are talking about a new platform which is expected to see continuous improvements across generations. What can we say, though, is it’s good to see Intel bringing performance improvements via faster DDR5 bandwidth at these speed levels which make a notable performance difference on Linux when looking at this wide array of popular, commonly used workloads that highlight system memory throughput. At the very least it means with even higher bandwidth DDR5 speeds in the future, as much as “Raptor Lake” may offer an up-to-date performance profile, it won’t have such high bandwidth potential or future headroom, putting “Meteor Lake” with its DDR5 speed at the forefront.
DDR5 Memory Kit Performance
Based on these benchmarks across several different categories with the Intel Core Ultra 9 285K, you’re now in the best position to choose the DDR5 kit speed for you. We recommend at the very least grabbing yourself a DDR5 5200 kit. This is the optimal sweet spot between performance and price and as seen in this benchmark data, a nice performance bump can be obtained. Overall, though, there’s a clear upward trend showing as the DDR5 kits become faster, general performance is further optimized. While “Meteor Lake” certainly benefits greatly from DDR5 speeds when using different kits at these specific speeds, this doesn’t represent all DDR5 kits — only the ones we have access to.
What Else is Coming With “Meteor Lake”?
“Meteor Lake” has yet to launch and will be officially available to users later this year. However, in addition to the benchmarks shown in the section above, Intel is also making several other upgrades, including an advancement on its integrated graphics architectures which they dub the “Iris Graphics”. Though these aren’t quite discrete graphics, Intel will be offering “Meteor Lake” with what the company considers to be the “first generation” of its dedicated “Arc” architecture and this should continue to offer higher performance compared to past integrated solutions as the technologies continue to evolve across Intel’s platform. You can learn more about “Meteor Lake” by checking out Intel’s official announcement and more updates. There should also be more news from Intel leading up to the launch. There may also be further software optimization down the road that might alter or improve “Meteor Lake” performance and in general “Meteor Lake” should make great gains across the board.
More Benchmarking Ahead
With a fresh new architecture and performance enhancements on its way, there’s a lot of new testing to conduct on “Meteor Lake”, and that’s precisely what’s happening at Phoronix.com, which we are conducting continuously and will provide a look into Intel’s platform performance going forward in new comparisons with the various software we use. Be sure to keep your eyes open for all the benchmarks, including some real-world usage tests and reviews. These tests help in understanding what changes Intel made with “Meteor Lake”, the overall architecture changes, the architectural optimizations, and so on. All while we conduct testing in Linux on this latest generation of Intel chips with new capabilities, so you don’t have to.
