Should I keep my DDR5 ram on 6400 or 6000 ?
Should I keep my DDR5 ram on 6400 or 6000 ?
These are various standards that measure performance differently. In practice, most benchmarks reporting memory latency in nanoseconds aren't very accurate. Aida stands out for inconsistent results and doesn't match actual performance well. I haven't noticed much from Passmark when it comes to real-world usage. Probably because it defaults to a 2:1 clock relationship (also called Gear 2), which makes the memory controller run at half speed, causing latency spikes but allowing higher frequencies. However, unless you use DDR5 7600+, this 2:1 setting isn't beneficial since you won't exceed around 6800MT/s, and dual-rank memory rarely goes beyond that. So it's best to avoid 2:1 mode if possible. DDR5 6000 tends to default to 1:1 mode, offering much better latency, making it a stronger choice. You can manually adjust it to 1:1 for higher speeds, as most chips are designed for around 6200MT/s and some reach 6400MT/s in that setting. On ASUS boards, this option is hidden in the AMD overclocking settings—hard to find, but if you set it yourself, you can push frequencies safely without worrying about performance loss. Raising the voltage helps, up to 1.3V is safe, but if it fails, lower to 6200MT/s and move on. Custom sub-timings are possible but not practical for production systems. These boards use 24Gb DRAM, which performs worse than A or M DRAM, so aiming for 550MT/s is realistic, though ideally you'd target 600-700.
I’m realizing I missed the fact it’s a 2x48GB package, not 2x32GB. Even if I set it in the 600-700 range manually, it should still perform better than what the system automatically chooses.
I reviewed the BIOS settings and found no changes needed. The displayed values match the expected parameters. RAM speed adjustments aren't required unless you have specific performance goals. The DOCP/EXPO configurations differ mainly in optimization targets—some aim for stability, others for higher performance. The tweaked profile may offer marginal gains if your system supports it.
You should set everything to match exactly if needed. This approach is common in real-world use. TL;DR: It's about marketing. For a production setup, the stress tests would take too long to verify reliability, so lowering the speed to 6000 would be practical.
It seems you should refresh your profile info since the voltage isn't at 1.3v anymore. This is likely because of the Ryzen 7000's performance issues a few months back when the voltage exceeded the safe limit. The setting you mentioned is in the AI Tweaker menu under UCLK mode, where you adjust it to match the memclk frequency—currently around 3200 MHz. You're in the dram timing control section, but I couldn't find any screenshots showing those options. If the system hasn't been used recently, just update your bios first, set UCLK to memclk, and then fine-tune the other parameters.
The UCLK setting refers to the maximum clock frequency the system can handle, not the actual RAM speed. Your BIOS and 3DMark results might be showing a higher speed because they measure performance under different conditions. If you don’t adjust the UCLK from Auto, your processor won’t run at the intended speed, which could affect stability or performance. Changing the TRFC can help align your system’s capabilities with the settings you want. Your December 2023 BIOS update should resolve any compatibility issues.
ASUS monitors SOC voltage nearer to the VRM than silicon, with 1.32V equating to 1.3V on silicon. This introduces significant delay. All Ryzen 7000 processors should handle 6000MT/s using two DIMMs, which is reliable in single-chip configuration. Within Windows, you can verify the setting via HWInfo—look for UCLK readings and compare them to the MCLK (memory speed) to confirm a 1:1 or 2:1 ratio. For safety, set it permanently to 1:1. A 6400 MHz 2:1 configuration runs slower than 6000MT/s in 1:1 mode; if you avoid stress tests, stick with 6000 in 1:1. In production, at least 24 hours of testing is ideal. Practically, for a system with 32-48GB RAM, 4-6 hours suffices, but with double the memory and higher costs from crashes, aim for 48-72 hours of rigorous testing over the weekend. Don’t obsess over sub-timings—if you were in my shoes, I’d skip them. Avoid any references to tRFC, tREFI, tRAS or similar terms with a 't' at the start.
Hello everyone. I installed HWinfo and it displays the MCLK at 32.00 MHz and UCLK at 16.00 MHz. That suggests adjusting the UCLK in the BIOS? If I leave UCLK unchanged, how would latency impact performance in gaming, 3D animation, or rendering? Which software would you recommend for stress testing? The options referenced are mostly from 2010 and 2013 and haven’t been updated.
UCLK setting adjusted to match MEMCLK for better stability. Spent a lot of time testing performance, especially with animation and rendering—though I’m not as familiar with those details. The best results for spotting memory problems on DDR5 came from TestMem5 using the 1usmus_v3 preset (increase loop count to around 100 for about 24-48 hours) and Y-Cruncher with the VT3 stress test. Running both together on my system usually means 2-3 hours of VT3 and then TestMem5.