How to Overclock DDR5 RAM
Overclocking is an art that has existed throughout the history of the PC. Who doesn’t want to get additional performance from their purchased hardware? Overclocking your CPU or graphics card usually gets all the attention because overclocking can significantly improve performance. However, memory overclocking can also have significant benefits if you are willing to work.
When DDR5 debuted at Intel’s Alder Lake, it brought many exciting improvements over DDR4, including performance, bandwidth, and capacity improvements. As you can imagine, next-generation chips such as Raptor Lake and AMD’s Ryzen 7000 will also support new memory, and many new options will soon be on the market.
The only caveat is that DDR5 is in the early stages of its life cycle and is not exactly at the same level as high performance DDR4 memory. That makes sense as DDR4 has been released for some time. However, that doesn’t mean you can’t overclock DDR5 to get more performance out of your memory kit. There is always a risk when running hardware that is out of the manufacturer’s specifications, and the product warranty does not cover overclocking. However, overclocking does not affect the state of the hardware, as long as it runs correctly and is within reasonable limits.
Many enthusiasts and casual users simply use the XMP profile built into the memory kit, which is an automated way to overclock memory. If you are interested in using this technique, see the article How to enable XMP. If you want to get even better performance, check out our guide to getting the best performance out of your DDR5 memory kit.
Get to know DDR5 IC
Different integrated circuits (ICs) behave differently, and DDR5 is in the same situation. Some ICs may be more noticeable because they hit higher frequencies, while others are better when aiming for lower timing. Leading manufacturers such as Micron, Samsung and SK Hynix have released early ICs for DDR5. However, DDR5 is still wet behind the ears, so there is no wide choice of ICs.
Even if the brand and product lineup are the same, the two ICs will not be the same. So, for example, if a particular “X” IC can be overclocked beyond DDR5-6000, not all “X” ICs may be overclocked. So, in the end, there’s still a bit of a silicon lottery involved.
Thaiphoon Burner was a great utility for reading the SPD of a DDR4 memory module to detect which IC was installed without disassembling the heat spreader. Unfortunately, due to the lack of documentation from JEDEC, Thaiphoon Burner developers were unable to add DDR5 support.
Fortunately, CPU-Z can pick up the IC vendor of the DDR5 memory module even if the exact IC model cannot be identified. So far, there aren’t many ICs from the same company, so this feature is useful. However, when a manufacturer puts multiple ICs on the market, CPU-Z is completely useless unless the developer finds a way to detect a particular IC.
IC vendor | IC part number | Die revision |
---|---|---|
micron | MT60B2G8HB-48B, MT60B1G16HC-48B | A-Die |
Samsung | K4RAH086VB-BCQK | B die |
SK Hynix | H5CG48MEBDX014, H5CG46MEBDX015 | M-die |
In the era of DDR4, Samsung B-die IC was the golden standard for tight timing. With DDR5, you can see that SKhynix M-die can be executed at the same timing as B-die. Based on experience with various DDR5 memory kits, SK hynix’s M-die IC has the best overclocking potential. For example, I was able to raise the M-die to DDR5-6600, overclock the DDR5-4800 M-die memory kit equally, and find stability with DDR5-6000.
Again, it should be emphasized that these are only preliminary observations of the current generation of ICs. Things can change over time as ICs mature and new ICs hit the market. However, at least for now, rank the ICs in overclocking order: SK hynix M-die, Samsung B-die, Micron A-die.
When overclocking memory, it first tries to determine the maximum safe voltage that can be used for the various ICs. Samsung said it doesn’t guarantee overclocking voltage for its ICs, but the absolute maximum DC drain voltage is 1.4V. He contacted SK Hynix and Micron representatives to inquire about their respective ICs. Unfortunately, there is no feedback on this subject. Nevertheless, we have talked to various memory vendors, but the consensus is that 1.4V is the maximum voltage for daily use.
Overclocking process
1. Set the DRAM target data rate. The first step is to decide how far to push the DDR5 memory. There are two approaches. You can choose a higher data rate and work from there, or you can choose a more conservative data rate and work to find the memory kit limits.
For reference, the default ratio for DDR5 is 100: 100. Still, certain data rates are only available in 100: 133 ratios such as DDR5-5333 and DDR5-6666.
Like Rocket Lake, Alder Lake offers two BCLK / DRAM ratios of 100: 100 and 100: 133. Rocket Lake improves performance in a 100: 133 ratio. However, Alder Lake gives similar performance at a 100: 100 ratio.
Alder Lake has three gear modes for memory, but Gear 2 is the only setting that is really important for mania level DDR5 overclocking. In Gear 1 (1: 1), the memory and the memory controller operate at the same frequency, but in Gear 2, the memory operates twice as fast as the memory controller (2: 1). Finally, Gear 4 runs memory four times faster than the memory controller (4: 1).
Gear1 does not work with DDR5 because the default setting is Gear2. Gear4 probably won’t work until it reaches a data rate of DDR5-8000 or higher. So while extreme overclockers using liquid nitrogen may need to step up to Gear 4, most of the DDR5 memory overclocking should only use Gear 2.
2. Fine-tune the memory timing. This step is a trial and error part of the overclocking process and can be time consuming. DDR5 has the same four primary memory timings as DDR4 (CAS latency, tRCD, tRP, and tRAS). The trick is to use Baby Steps, increment individual timings in 1 or 2 clock cycle increments, and then test for stability.
It’s a good idea to start with the advertised timing of your memory kit and proceed from there. If overclocking is unstable, add one clock cycle to the timing. If it is stable, try shortening the timing by one clock cycle to optimize overclocking.
If you’re adventurous, you can even play at the command rate. For DDR5, the default command rate is 2T. However, 1T is possible depending on the IC. I was able to run Samsung and SK hynix IC on 1T, but it didn’t work very well on Micron IC.
Memory overclocking doesn’t just play at the four main timings. You can significantly improve performance by fine-tuning the 2nd and 3rd timing. However, this is beyond the scope of this article, so I won’t delve into that rabbit’s hole.
3. Increase the voltage. There are quite a few voltage settings that affect memory overclocking, but the average user can get by fine-tuning four of them. Other voltages can be left at the discretion of the motherboard.
DRAM VDD: This voltage is one of the most important because it powers the memory chip. For daily use, it is recommended to keep it below 1.4V as a safety measure.
DRAM VDDQ: The voltage is supplied to the I / O of the memory chip. VDD and VDDQ are usually closely related. However, in some cases, desynchronizing the voltage and running higher VDDQ (50 mV to 100 mV) can help stabilize memory overclocking.
CPU VDDQ: The voltage flowing through the memory controller of the processor. 1.2V is sufficient for DDR5-4800 to DDR5-6000, but 1.4V is sufficient for DDR5-6200 and later.
VCCSA: This is the system agent voltage, but it helps with memory overclocking. This is one of the more sensitive voltages, so increase it little by little. For example, 1.25V to 1.35V is sufficient for most memory overclocking. It is best to keep VCCSA below 1.4V.
4. Test the stability of overclocking. Some users run stress tests for hours, while others run them overnight. You need to decide how important memory stability is to your system and how much time you spend testing. In addition, different programs use different algorithms to catch the error. We recommend that you test stability using a combination of two or more, rather than relying on a single program.
Anyway, don’t stick to punishing DDR5 memory. Once you’re happy with overclocking, daily use is ultimately the best stability test.
5. Save the overclock. In the old days, you had to remember the memory overclocking settings or write them down on paper. DDR5 comes with a new XMP 3.0 standard, bringing the total number of memory profiles to five. We recommend that you use two of these profiles as they are available to your users. If you find a stable memory overclock, don’t forget to save it to the DDR5 memory module. Options vary from motherboard to motherboard, but you can usually find features in the DRAM section.
Stability test software
There are many options for testing memory overclocking, but we’ve put together a small list of the most common and practical tools. Many of them are free, but some require a license.
- Memtest86: This is old-fashioned software for testing memory, but it’s still one of the best options. The program boots from a USB pen drive, so you can test overclocking stability without accessing Windows.
- HCI MemTest: The software detects memory errors very quickly. The best way is to open one instance per CPU thread and divide the memory evenly for testing.
- Google Stressful Application Test (GSAT): If you’re using Linux, no utility is better suited for memory testing than GSAT. Google is using this software to test machines.
- Karhu RAM test: The software excels in its capabilities and supports up to 8TB of memory. Unfortunately, it’s not free and requires a $ 10 license.
- TestMem 5: This lightweight and robust software supports custom configurations. The Anta777 Extreme configuration is currently the most popular and widely used preset.
Optimized DRAM timing
There is no shortcut to achieving maximum memory overclocking. Still, I’ve put together a small table with the optimized configuration possible with existing DDR5 memory kits. However, these are not magical formulas that can be connected to the motherboard without testing. Keep in mind that all memory kits are different, whether they have the same IC or not. However, these basic configurations can act as stepping stones and you can proceed from there. Don’t be afraid to try other combinations.
Stick to more standard data rates available on the market, such as DDR5-4800, DDR5-5200. However, you can also try other uncommon data rates such as DDR5-5066.
Data rate | timing | Voltage |
---|---|---|
DDR5-6400 | 40-40-40-90, 42-42-42-90, 44-44-44-90 | 1.35V-1.40V |
DDR5-6200 | 32-38-38-77, 36-36-36-77, 36-38-38-77 | 1.35V-1.40V |
DDR5-6000 | 36-36-36-77, 36-38-38-77, 38-38-38-77 | 1.35V-1.40V |
DDR5-5600 | 36-36-36-77, 36-38-38-77, 38-38-38-77 | 1.30V-1.35V |
DDR5-5200 | 36-36-36-77, 36-38-38-77, 38-38-38-77 | 1.20V-1.30V |
DDR5-4800 | 34-36-36-77, 36-36-36-77, 36-38-38-77 | 1.10V-1.20V |