Needs advice on overclocking for GA-X58A-UD3R+i7 920
Needs advice on overclocking for GA-X58A-UD3R+i7 920
Hello Experts,
This is my initial post. I’m new to overclocking and recently completed one with the specified setup. I’m seeking advice on stability.
Board: Gigabyte X58A UD3R revision 1 & 8 Gig DDR3 Memory
CPU i7 920 at 2.8 GHz increased to 4.01 GHz (21 x 191) with Vcore 1.4v, memory at 1.66 v.
I’m currently running a Prime 95 stress test in blend mode with 8 threads for the last 12 hours. Temperatures across all four CPU cores stayed between 58-68°C using HWMonitorSMC2.
According to various forums, everything seems fine so far. Could you please confirm if this overclocking is truly stable or if there’s something I’m overlooking? Any guidance would be greatly appreciated. Thank you.
Welcome to the forums, newcomer!
Going through these posts;
I’m trying to overclock an I7 920 to above 3.6ghz. My dad wants to push his Bloomfield chip further using a Hyper 212 EVO and I believe he needs the nzxt kraken X41 that I own for this purpose. I’ll update with his CPU specs when possible. I’m comfortable with overclocking newer chips. I have a 4790k that’s been overclocked to 4.7ghz, please...
https://www.overclock.net/threads/g...h3...CjcnBszQnl
I think you’re in the clear. For stability testing, you’ll need to run the tests overnight—longer is better. You can also try running daily tasks to check if your overclock is just generating heat or if your system performs better than it did before.
Also, which step are you focusing on? What cooler are you using? Are there any memory overclocks? What’s your motherboard BIOS version?
Thank you for the guide. I considered upgrading the i7 920 to 4 Gig, but it might be risky; if it doesn’t work stably, I’ll have to go down.
I’m also using H60 with a couple of G-skill 4 G stick @1600 and F6 BIOS version. After updating the BIOS from F1 to F6, I noticed some checksum errors persist. I tried a new CMOS battery, but the issues keep happening randomly.
Below are the details for the step up:
CPU ratio: 21x
BCLK freq.: 191
QPI ratio: x36
Uncore ratio: x16
XMP: Disabled
Memory multiplier: 8
Memory performance: Standard
DRAM timing: Auto
CPU vcore: 1.4 v
QPI/Vtt: 1.335 v (pink)
CPU PLL: 1.88 v
IOH & ICH: 1.2 v
DRAM: 1.66 (pink)
All other voltages should be manually adjusted to match the default or a slight increase.
Please let me know if you need any corrections for stability. Also, prime 95 is still running as I mentioned earlier. It’s been over 16 hours and the CPU core remains at around 68°C max.
Thanks.
I kept a guideline to maintain below 1.45v for the 45nm process, similar to Lynnfield, and around 1.425v for Bloomfield at 32nm. In fact, 1.400v is quite safe if you manage heat well with a good AIO or air cooler. VTT/QPI tends to be the tricky part; it can lead to rapid degradation above 1.35v on both 32nm and 45nm. I’d aim for under 1.325v (absolute max 1.35) on VTT, and consider increasing the VCore if necessary. Ensure voltage stability during idle and under load, adjust the Load line calibration accordingly.
I haven’t encountered any issues with a 1.65dram voltage, but avoid increasing it further for continuous operation. PLL voltages up to 1.90v are acceptable.
A lower QPI ratio improves stability, while higher frequencies boost performance—but too high causes instability and forces more aggressive VTT adjustments. After identifying the maximum stable CPU frequency, you can experiment with uncore ratios; I’m currently running 3.6ghz (uncore) on my Lynnfield X3470, though Bloomfield behaves differently and achieving those results is more challenging.
Thank you for the suggestion, I'll modify the voltage and continue testing. Yes, I'm using load line calibration Level1 beforehand. Currently, the 24-hour p95 test maintained temperatures under 70°C, which seems to be moving the OC part toward stability. However, I'm still facing challenges with the BIOS checksum issue. I restored the older version F1, but it keeps getting stuck in the checksum with OC settings. I'm not sure if the voltage adjustments are contributing to this problem.