Trouble Enabling XMP
Trouble Enabling XMP
Hello
Here’s a revised version of your text with the same length and structure:
Hello
My specifications are as follows:
Mobo: ASUS Maximus XI Extreme
CPU: i9-9900k
GPU: RTX 2070
PSU: Corsair AX1600i
Cooling: Custom Waterloop
RAM: Corsair Vengeance DDR4 4x8GB 4000Mhz 19-23-23-45 1.35v -> Gskill Trident Z DDR4 4x8Gb
4000MHz 15-16-16-36 1.5v
Initially, I’m just starting out with overclocking. AI Suite managed the CPU adjustments, but while experimenting with settings, I noticed my RAM (Corsair Vengeance DDR4 4x8GB 4000Mhz 19-23-23-45 1.35v) was clocked at 2133MHz. I studied RAM specs and timings, then dove into the XMP community. I realized my RAM wasn’t hitting 4000MHz because XMP wasn’t turned on in the BIOS. That seemed straightforward enough, but it didn’t work.
I checked the BIOS, activated XMP 2, saved the changes, and restarted. After rebooting, the motherboard displayed codes rapidly while the screen stayed black, then it rebooted itself. On the second restart, I reset both the BIOS and CMOS on the back of the motherboard. It worked again, but now everything returned to default with my RAM still at 2133MHz.
I tried tweaking voltage and timings manually in the BIOS, but nothing changed. Eventually, I ran a MemTest using Windows Memory Diagnostics. The test flagged a hardware issue, continued until about three-quarters of the way through before crashing. I ordered replacement RAM (Gskill Trident Z DDR4 4x8Gb 4000MHz 15-16-16-36 1.5v).
Today, the new RAM arrived. I installed it and booted up. I used Karhu RAM Test for about 30 minutes with 24448MB and 16 threads. It achieved around 500% coverage with no errors. Then I powered down the PC, entered BIOS, enabled XMP 2, and during the next boot it displayed a motherboard code of 55 before crashing. After resetting the BIOS to defaults, everything functioned properly.
Now I’m reaching out for expert advice since further self-study isn’t yielding results. Any suggestions would be greatly appreciated. Thanks!
DerrySellers,
On behalf of Tom's Moderator Team, welcome aboard!
Premium memory kits featuring rapid XMP profiles don't always deliver the promised speed, as performance hinges on factors like your processor's IMC quality and motherboard capabilities. Still, with dedication to stability testing, you can still optimize memory speed by determining "True Latency"—a straightforward approach to assess performance and adjust settings for better results.
For instance, we created a 9900K build for one of my relatives using a G.Skill...
The Intel website lists the maximum supported RAM speed for the I9 9900K as DDR4-2666. I assumed that exceeding this would be unnecessary since the CPU wouldn’t benefit from it. However, I recently watched a video by Linus using the I9 9900K and noticed something else. I believe a 9900K might struggle with 4000 Mhz RAM without significant modifications. It would likely require adjustments like tighter timings or sticking to XMP settings while lowering the clock speed. There doesn’t seem to be much advantage in going beyond 3600 Mhz, as shown in the comparison image from the thread I referenced.
DerrySellers,
On behalf of Tom's Moderator Team, we're pleased to have you here!
Premium memory kits featuring rapid XMP profiles aren't always as effective as claimed. This varies based on factors like the processor's IMC quality (silicon lottery), motherboard capabilities, and BIOS features. Still, if you're prepared to invest time in stability testing, you can still attain swift memory performance by determining "True Latency." This straightforward approach helps you assess memory performance and adjust your settings for optimal speed.
We recently worked with a 9900K built for one of my relatives using a G.Skill kit similar to yours—16 GB (two 8 GB units) Trident Z RGB DDR4 4000 at 1.4V. As you may know, this unit didn't boot at the advertised XMP 4000 settings. By evaluating various configurations and testing different timings, we managed to reach speeds nearly matching XMP, using the next fastest stable frequency with the lowest possible timings.
Let's clarify:
- Affordable kits often show a True Latency of 10.0–12.5 ns or more.
- Higher-quality kits approach 8.0 nS, while premium options hover between 8.0–7.5 nS or even lower.
True Latency = Time ÷ Frequency × Timing.
To calculate it (in nanoseconds), use these steps:
Time = 1 second
Frequency = GHz in single data rate (DDR4 at 3600 MHz is 1800 MHz)
Timing = CL (Column Latency, also known as CAS) from the module label.
For example, a DDR4 3600 kit with 16-16-16-36 at 1.867 GHz yields:
Time = 1 / 1.8 = 0.555…
× 16 = 8.88 nS (True Latency)
A DDR4 4000 kit with 16-15-15-35 gives:
Time = 1 / 2 = 0.5
× 16 = 8.00 nS (True Latency)
For the 9900K, we selected 3733 MHz (1/2 DDR @ 1.867 GHz) at 15-15-15-35:
Time = 1 / 1.867 ≈ 0.537
× 15 = 8.03 nS (True Latency)
Our chosen frequency and timing (with a slight voltage boost to 1.41V) proved stable during repeated MemTest runs, gaming sessions, and everyday use. You might find that 3800 MHz at the same timings works well for your kit:
Time = 1 / 1.9 ≈ 5.26
× 15 = 7.89 nS (True Latency)
Regarding voltage, Intel specifies DDR4 at 1.4175V ±0.5%. Raising the voltage affects not just the memory but also the CPU's IMC. Excessive voltage can cause electromigration and degrade the IMC, leading to memory instability, crashes, BSODs, or data loss. This applies equally to CPU core voltage (Vcore). Therefore, when overclocking both memory and CPU, proceed with caution and moderate voltage settings.
Always perform a full system backup before making adjustments, as changes can trigger software crashes.
When selecting a kit, remember that overclocking memory also impacts the CPU's IMC. It's often better to use two slots instead of four to ease the load on the IMC. For instance, a 32 GB 2 x 16 kit may perform better than a 32 GB 4 x 8 kit at the same timings. Also, ensure adequate cooling—overpacking all slots can restrict airflow between modules.
Welcome aboard!
Hello,
I share the same issue: I bought premium RAM (4166 MHz CL16) but can't turn on XMP because Windows won't boot. It starts at 3200 MHz. I managed to get it up to 3333 MHz and play PUBG, but it still doesn’t work.
RAM: G.Skill Trident Z Royal 2x16GB DDR4 4166 MHz CL16 - Gold
Motherboard: Asus Prime Z390-P
CPU: i7 9700k
Cooling: Corsair Hydro Series H150i PRO
I realize it’s normal that I’m not getting the full performance, but 3200 MHz feels a bit underwhelming. With the amount of memory and CPU I have, should I expect to run CL12? Would that be the best I could achieve?
Thanks,
Someone submitted their 4133 @ 18-19-19 to run at 3800 Mhz @ 18-19-19. It looks like a lucky draw—maybe due to luck or a good IMC. You might have spent a lot on premium RAM; consider returning it for a standard 3200 Mhz unit, or try working with what you have. Who knows if you can surpass his score?
Observe that enabling XMP for my Windows updates causes a significantly higher failure rate. The 3600mhz RAM sticks work fine, though they didn’t come with the high-speed models. I’ve seen about a 10% FPS improvement in certain games when using 2133mhz versus 3600mhz. Also, running RAM at that speed made my system feel unstable—I initially thought it was due to RAM issues, but now I suspect a faulty GPU I recently replaced. Not an expert on RAM, but it’s enjoyable to experiment with speeds.
I'll experiment with the timing settings. Before running the benchmark, what configuration would be optimal? Should I aim to reduce the total sum of the four primary timings, or focus on minimizing the first one, CL? Or perhaps a balance between them?
I can play PUBG using 14-19-19-39 timings for 4266 MHz. Using 12-18-18-39 allowed me to boot, though PUBG still crashed. I've tried various other combinations, like 13-15-15-33 on Windows, but didn't test PUBG.
Edit -
My DRAM voltage wasn't 1.5 V (the standard for my RAM), and with that setting I could reach 3733 MHz using 16-19-19-39.
Inc0wnage,
Hello, I am Tom's Moderator Team, and I hope you are welcome. Please note that your initial post violated our Forum Rules by not adhering to them. Specifically, it was about not hijacking another Member's Thread. I recommend reviewing the Forum Rules again, which can be found in the "Sticky" Threads at the top of each Forum. Instead of just responding to titles or the first post, take the time to read the entire Thread before posting, as someone else might have already shared a helpful answer.
If you're adding comments like the one above, which include your own questions without providing solutions for the original poster (OP), then please create a separate Thread.
CT