Memory speed is set in several layers, and the number on a kit’s box is only one of them. This guide separates those layers, reports what Intel and AMD say about memory profiles, and summarizes two independent reviews together with their test conditions. Gimaxo has not tested any memory, so no result below is ours. Where a sentence starts with “Our reading:”, it is this guide’s own conclusion, drawn from the source named just before it.

Photo: Onur Binay, Unsplash, published October 31, 2021, Unsplash License. A memory module installed in a PC, shown as an illustration. It shows no setting and no result. Not tested or chosen by Gimaxo.

Four settings that are easy to confuse

A memory kit can be described in four different ways, and each answers a different question.

• The kit’s stored default profile. Memory modules carry stored profiles, known as SPD data. Intel says its XMP modules first boot with the default JEDEC settings, so that the first start works even on a system that does not support overclocking. JEDEC is the memory industry’s standards body. CPUID describes its CPU-Z tool as showing memory type, size, timings and module specifications (SPD).
• The speed your processor officially supports. Processor makers publish it, and it can depend on how many modules you install. AMD’s page for the Ryzen 7 9800X3D lists a maximum of DDR5-5600 with two modules and DDR5-3600 with four, in the configurations it cites. Intel’s page for the Core Ultra 9 285K lists “Up to DDR5 6400 MT/s”.
• The setting your board actually applies. The BIOS, the firmware menu you open at startup, decides it. Neither the kit’s label nor a processor page states it.
• An XMP or EXPO profile. Intel says XMP overclocks compatible memory at higher and optimized speeds, beyond the standard settings, and that XMP 3.0 for DDR5 supports up to five profiles, three from the vendor and two rewritable. AMD describes EXPO as user-friendly memory overclocking for DDR5 on processors with the AM5 socket.

Our reading: the makers’ figures cannot tell you the speed your PC starts at. That depends on the kit’s stored profile, your processor’s limit for your configuration and what your board applies. Check your BIOS memory page, or a tool that reports memory frequency, instead of assuming a number.

What the makers claim

Intel’s XMP page lists “Higher FPS” as a benefit, described as better performance in demanding open-world and fast-paced games. Its footnotes say performance varies with system configuration, and that altering clock frequency or voltage may void any product warranties and reduce stability, security, performance and the life of the processor and other components.

AMD’s EXPO page says high memory frequencies and aggressive settings can unlock higher and smoother frame rates, and it carries a warranty footnote covered below. Another footnote describes AMD’s own tests of March 2026, on a Ryzen 7 9700X with a GeForce RTX 5090. Those tests compare two kinds of EXPO kit, not default memory against a profile.

These are claims, not measurements. Neither page gives a game-by-game figure for enabling a profile.

What two reviewers measured

Both reviews below used AMD processors. Gimaxo did not read a report with a readable protocol for an Intel platform in this pass, so nothing here describes one.

TechPowerUp, “DDR5 Memory Performance Scaling with AMD Zen 5”, by W1zzard, September 27, 2024. The test used a Ryzen 9 9950X on an ASUS X670E Crosshair Hero (BIOS 2201), a GeForce RTX 4090 at Ultra settings and Windows 11 23H2. One memory kit family was run at six listed settings. They included DDR5-4800 CL40 and DDR5-5600 CL40 at 1.1 V as its JEDEC baselines, DDR5-6000 CL28, DDR5-6400 CL32 at a 1:1 controller ratio, DDR5-7200 CL34 and DDR5-8000 CL38.

The review charts relative game performance across its game suite at 1080p, 1440p and 4K. Its text does not say what reference the percentages use. In all three charts the DDR5-8000 CL38 setting reads exactly 100.0%, so this guide reads the figures as relative to that setting. That reading is Gimaxo’s, not TechPowerUp’s. In those charts DDR5-4800 scored 92.6% at 1080p, 95.5% at 1440p and 96.7% at 4K. DDR5-5600 scored 95.8%, 97.2% and 98.3%. DDR5-6000 CL28 scored 99.7%, 99.8% and 100.2%.

These are scores for one processor, one graphics card and one game suite. They are not a loss to expect on other PCs. The reviewer’s own conclusion calls DDR5-8000 the fastest for gaming at typical resolutions by a wafer-thin margin, and says some benchmarks show only marginal differences because the CPU cores often work on data held in the processor’s caches. Limits: one non-X3D processor, one graphics card, a game list visible only in the report’s charts, and a single kit family at listed settings. The report lists some settings twice with small differences. The figures here use the entry labelled 1.1 V for DDR5-4800 and DDR5-5600, and the entry without a voltage label for DDR5-6000.

Phoronix, “DDR5-4800 vs. DDR5-6000 Performance With The AMD Ryzen 7 9850X3D In 300+ Benchmarks”, by Michael Larabel, January 29, 2026. The test used a Ryzen 7 9850X3D, supplied by AMD, on an ASRock X870E Taichi under Ubuntu 25.10 with Linux 6.17, and a GeForce RTX 5090 for the gaming tests. It compared G.Skill DDR5-6000 EXPO memory with DDR5-4800 across more than 300 benchmarks.

The gaming results come first. The gaming page includes Cyberpunk 2077 at 1080p and 1440p, DiRT Rally 2.0 at 1080p, and Xonotic and Unvanquished at 4K. For most gaming tests, DDR5-4800 gave about 99% of the DDR5-6000 result, with a slight benefit for DDR5-6000 in some games.

The average over all 300+ benchmarks is a different figure, because those benchmarks are not only games. Across all of them the geometric mean favored DDR5-6000 by just over 1%, and some non-gaming workloads, such as code compilation, database serving and SVT-AV1 video encoding, ran 5 to 10% faster. The 1% figure is an average over all tests, not a gaming result. Limits: Linux, not Windows, one X3D processor, two memory settings, and a review that treats the value question as depending on local memory prices.

Our reading: the two reviews differ in processor, graphics card, operating system and memory, and neither tested your PC. In Phoronix’s gaming tests the gap between the two settings was small. In TechPowerUp’s charts DDR5-4800 scored lower, by more at 1080p than at 4K. A result for your processor, games and resolution needs a report that matches them.

Check, enable and verify a profile

1. Read your board maker’s documentation for its memory settings. Intel says Z and B series motherboards have XMP capability.
2. Enable the profile in the BIOS. Intel says to enable XMP in the BIOS, and that the system then applies the best settings for your memory and motherboard. AMD publishes a memory compatibility list for EXPO kits.
3. Confirm the speed that was applied, in the BIOS or in a tool that reports memory frequency. The label does not show what your board chose.
4. Test stability before you rely on it. Intel says altering frequency or voltage may reduce stability. Microsoft documents a built-in Windows Memory Diagnostic Tool in a troubleshooting article: run mdsched, restart, and read the result. Standard tests run by default, and extended tests are more thorough but slower. It is a diagnostic tool, and the Microsoft page does not describe it as proof of stability. Our reading: a clean result is a good sign, not an exhaustive proof that every game and workload will run without errors.
5. To go back, Intel says reselecting the default profile resets the platform to its default settings.

Warranty conditions

Intel’s FAQ says that altering frequency or voltage outside Intel specifications may void the processor warranty. Its examples include overclocking and enabling XMP, which it calls a type of memory overclocking. AMD’s EXPO footnote says that operating outside AMD’s published specifications, including altering memory timing or voltage, will void any applicable AMD product warranty, even when enabled through AMD hardware or software. It adds that this may also void warranties from the system manufacturer or retailer.

These are the makers’ statements about their own product warranties, in the contexts quoted. They are not a conclusion about all your warranty rights, which also depend on your country, your seller and the terms of your other parts’ makers. Read the terms from your processor, memory, motherboard and PC makers before you enable a profile. This guide gives no legal advice. It does not cover manual timing or voltage tuning, which none of the pages above describe.

When it is not worth the effort

• Your frame rate is already where you want it. The reviews above measured small gaming gaps in some tests, so a profile may add little.
• Your processor is not the one that was tested. Phoronix measured about 99% of the DDR5-6000 result for DDR5-4800 in most gaming tests on an X3D processor, and TechPowerUp’s DDR5-4800 scores were lower on a different processor. Look for a review of your own processor.
• The limit is somewhere else. To find what limits your frame rate before you change memory, start with our gaming PC performance guide.
• The warranty matters more to you than a few percent. The makers’ own pages say the profile is overclocking.

Sources

Named here as plain text. Pages read on November 6, 2026 unless stated.

• Intel, Extreme Memory Profile (XMP) page and FAQ
• Intel, Core Ultra 9 processor 285K specifications page
• AMD, EXPO Technology for Socket AM5 page, including its footnotes
• AMD, Ryzen 7 9800X3D desktop processor page (read November 4, 2026)
• TechPowerUp, “DDR5 Memory Performance Scaling with AMD Zen 5”, W1zzard, 27 September 2024, including its relative performance charts
• Phoronix, “DDR5-4800 vs. DDR5-6000 Performance With The AMD Ryzen 7 9850X3D In 300+ Benchmarks”, Michael Larabel, 29 January 2026
• CPUID, CPU-Z page
• Microsoft Support, Windows Memory Diagnostic Tool instructions, in the article “How to fix Error 0xA: IRQL_not_less_or_equal”