Not possible to boost memory frequency in the BIOS.
Not possible to boost memory frequency in the BIOS.
In my bios, memory settings higher than 6400 don’t apply and revert to that number. Earlier today it was 6500, so now it’s stuck at 6400. I just removed the battery and reinserted it. What’s happening? Short version: I can’t push memory speeds above 6400. After my first 14.7MHz died, I switched to stock mode. I still boosted my memory from 5600 to 6000 CL32 without affecting CPU voltage. I also stopped undervolting (though it’s safe for CPUs) and ran the board at 13th gen microcode to get around Intel’s B-series undervolt block, but I wasn’t sure if that was connected. Recently I got a blue screen—first time it happened, then more often until I couldn’t use my PC for minutes without a reboot. I told myself I’d keep running at whatever speed until it failed and then replace the RAM. I still have all power limits set correctly. But I changed the undervolt setting and tried 6600, but Testmem5 kept failing. I reset everything and removed the battery. Now, no matter what I do, I can’t go above 6400. I thought it might relate to CPU power or voltage, but increasing those didn’t help. I increased the RAM wattage and amperage, but nothing changed. I even tried reseating the RAM. I considered adjusting timings instead of frequency, but I’m stuck. The only other idea is a RAM power limit, but I’m not sure how much it would need. What should I do?
Question your Task Manager settings; it often misrepresents CPU speeds. I’ve seen it claim a 40GHz clock before—definitely not accurate. For memory frequency checks, try CPU-Z instead. The 6500 designation doesn’t apply to Intel CPUs, which is why it fails. Adjusting BCLK settings is necessary for the 6500 to function properly. Since you’re using a B760 board, this is likely only possible with a B760M PG Riptide. What model motherboard are you running? It probably isn’t Gigabyte—they claim support for the 6500 ratio even though it doesn’t work. Have you tried booting with the 6600 instead? Are you focusing solely on primary timings or also testing sub-timings? Those sub-timings often determine the biggest gains during memory overclocking. High-end single-rank DDR5 chips can draw up to 7W under load, typically around 5W at standard voltages. I’m not familiar with any motherboard that enforces RAM power limits, as RAM consumes very little power.
I'm ready to proceed. First, I'll save and apply the changes in BIOS then re-enter BIOS to verify if the frequency remains consistent. I've also experimented with many frequencies above 6500, but none have worked. I checked BIOS and Task Manager both confirmed the frequency was set to 6500 before I changed it. After resetting BIOS, I'm unsure what settings were altered. My motherboard is a Gigabyte B760 Aorus Elite Ax with DDR5, and I used 6600 MHz before swapping the battery. It booted fine then reported multiple memory locations as affected, but TestMemory5 showed numerous errors. I suspect the microcode was downgraded, allowing me to undervolt safely, which I shouldn't normally do. Currently, I'm using 40000 and 40 timings, aiming for stability without excessive heat risk. I've also tried other sub-timings without significant improvement. If you have any advice, I'd appreciate it. Honestly, the RAM is running smoothly now, but I'm still puzzled by this issue.
I discovered a setting named memory multiplier with choices Auto, 100, and 133. Earlier adjustments didn’t work at 133—1500, 1533, or 1600 didn’t click. I switched to 100 and it still didn’t lock in those numbers (confusing). However, when I returned to Auto mode, 1600 did work! I increased CL to 32 but it wasn’t enough, so I changed it to 34 and am now testing it. I’m unsure if adding four extra points is worth the 100MHz boost, but I’ll give it a shot. Regarding RAM voltage, I’m considering going above 1.5V since my Kingston Fury Beast 5600 CL36 (part KF556C36) runs fine at Hyixi M. I don’t really focus on this stuff much; I bought it cheaply expecting faster DDR5 speeds, which did happen. I’m still holding off on upgrading until we release single rank 64GB kits (if that ever happens). Also, searching BCL K didn’t yield much except voltage options, and the test failed. I might revert to 6400 and avoid another BIOS reset.
Yes and no. It got more limited after launch, but it never came with the built-in clock generator. For boards with external generators like XOC or B760M PG Riptide, you can adjust the BCLK, though not on older models. That logic makes sense—Gigabyte shows 6500 support even though it didn’t function properly. It seems to be a flaw that let it claim compatibility while actually running at 6400 most of the time. If cooling is active (such as a fan near RAM), it’s fine. The ideal voltage for Hynix 16Gb is around 1.43V, but it’s not crucial. Running it higher isn’t worth it unless you have good cooling and a capable board. If cooling isn’t present, stick to 1.4V as the safe limit, maybe 1.45V if the chips handle heat well.
For older Intel boards, some models had one strap that stayed stable longer (especially the 9th gen). This doesn’t really change performance much, just stability on certain chips. It’s mostly a minor detail.
The memory speed depends on BCLK × Strap × Memory Controller Multiplier × Gear mode. On B760, BCLK stays around 100MHz, and with DDR5 the gear mode is set to 2. So speed comes from Strap × MC × 200MHz, then doubled for data rate. The MC multiplier changes in halves, which is why 6500 isn’t a real frequency—it can’t be achieved with any combination. This setting usually doesn’t hurt performance much.
For older Intel chips, one strap could be more reliable than others. It rarely affects speed but might help stability on some boards. For reference, speed comes from BCLK × Strap × MC × gear mode. If you’re unsure about a frequency, leave it on auto and set tRFC to 300 for better results. I can share other timings that work across most chips if you’d like.
It seems you're having trouble with a lot of things at once. You appreciate the help so far and are glad you have the information. Just to clarify, maybe the 6500 reference is a bit unclear—otherwise it’s hard to pinpoint exactly what you’re asking about. Let me know how I can assist better!
gigabyte offers the option to adjust settings even when technical constraints exist. For an overclocking board, it seems necessary to return to gigabyte since the x58a udr3 has performed well in RAM overclocks, mainly because I can fine-tune parameters like <3:2 uncore memclk (5:4 11:8 1:1 and 1:1+1). Other boards often restrict these settings. This allows stable performance between 2800-2900 on Bloomfield with two channels, possibly reaching 2900+ on Gultown with three channels—up to around 2900 on the outermost channel if I secure a better RAM stick. On my current setup, the top sticks are rated at 1333, while the outermost is around 1600. The innermost barely holds about 200% stable over spec, making it a decent choice despite its limitations.
Some components are sensitive to board choices; for example, my 1gbit Hynix BFR struggles past 2300 even though it’s rated higher. Stock RAM speeds aren’t crucial if you’re using a JEDEC chip, as they’re essentially untouchable. The main bottleneck is likely the board itself.
The inner two sticks sit at 1333, and the outermost at 1600. My best performance comes from the middle stick, rated at 1333, which could hit around 2900 stable under optimal conditions. However, due to channel limitations, I’m capped near 2860C11 on the outermost channel. RAM ICs can be quite picky about board compatibility, which explains why my 1gbit model fails to boot beyond 2300 despite its potential.
In short, gigabyte remains a solid option for pushing these builds, especially if you’re targeting higher frequencies or need more stability.
It doesn't really fail technically—it just doesn't function properly. The team omitted the Gear 1 adjustments even though the DDR5 controller lacks support for them. These are the test timings from what seems to be one of the least ideal manufacturing runs (the primary voltages are close to their minimum at 1.43V). tRFC and tREFI remain quite flexible, especially since I didn't run a fan and avoided checking temperature effects. A 300 tRFC setting might have been sufficient.