Question Overclocking X5650 above 3.9Ghz
Question Overclocking X5650 above 3.9Ghz
Hello everyone!
I'm working hard to push my X5650 up to around 4.0 to 4.2Ghz using a GA-X58A-UD3R Rev 1.0 with BIOS F6.
Using the newest BIOS versions (beta) has caused problems accessing the BIOS menu, especially with USB or PS/2 keyboards, which is why I'm stuck on F6. It will boot but doesn't respond to repeated DEL keys.
My current settings are:
- CPU Multiplier: 20x
- BCLK: 8.0x
- Memory Multiplier: 195BCLK (1520Mhz) in Triple Channel, standard mode
- QPI and Uncore: Auto
- Vcore: 1.275
- RAM timings: manually adjusted, but not successful
- LLC Level 2: Vcore 1.275, lowered to stable voltage
I've also tried increasing QPI/Vt Voltage to 1.295 and adjusting the CPU PLL to 1.86, but the system still doesn't boot.
Some users managed to reach 4.4 to 4.6Ghz with similar chips, so maybe they had an advantage in hardware or settings.
I have 3x2GB DDR3 1333Mhz Corsair Dominator RAM in triple channel, installed new thermal paste, and the CPU is cooled by a Corsair H100i. Temperatures are under control.
Any suggestions on increasing BCLK beyond 195 to 200 or even 205/210? I'm curious about how to get past this barrier.
Appreciate any guidance, it would really help!
This component handles the boost multiplier. Some boards activate it only when one or two threads are using 100% of the CPU; otherwise, more than two threads trigger a reduction to 20x. This is probably the root of your problems.
Uncore equals 2x for DDR3 speed. At 195 BCLK and a minimum 6x multiplier for DDR3, you're around ~DDR3-1170. To reach ~2340MHz, you should aim for 12x. Going beyond this causes instability and potential crashes or failure. This is particularly problematic with high BCLK values (above 167 is considered "very high"). QPI/Uncore performance drops significantly, leading to crashes or non-starting.
You may need to explore further but the recommended LLC target should maintain consistent voltage during idle and load. For chips like X56xx, try overclocking only when necessary—otherwise, stick with default settings except for auto.
Check your OC limits; some boards support 20x at 167 (DDR3 8x Multi ~DDR3-1336). Boot into Windows, use CPU-Z to note the current VCore under low/idle, then stress-test with Prime95 or similar. Record the VCore again after stressing. Adjust accordingly.
Set VCore and QPI link speed to their rated values. Avoid altering them unless you're certain. Instability often stems from incorrect timings, QPI settings, or voltage mismatches.
VT-d/Virtualization support should be enabled if it's causing issues. Research shows that these factors can lead to boot failures unless RAM timings are correct or overclocking is properly managed.
QPI/VTT values should not exceed 1.35v for 32nm Westmere-EP chips unless you're using extreme cooling methods. Exceeding this can cause instability.
VCore specs suggest a max of ~1.375v, but real-world performance on these chips is usually lower (<1.35v). They tolerate up to 1.4v only briefly before overheating.
Uncore should never deviate from 2x DDR3 speed unless using exotic cooling solutions. Deviations here often result in crashes or failure.
QPI/VTT can become unstable with excessive voltage; lowering it may resolve boot problems.
DRAM voltage settings vary by board—some require higher voltages (e.g., i7-900 at 45nm), while others like the X58 need lower (1.58v). Always verify your chip's specifications.
QPI link speed must stay below 7,500MT/s unless you're using advanced cooling. Exceeding this can cause instability or hardware damage.
Maintain a balance between VCore, VCore, QPI, and QPI/VTT values. Keep them stable to avoid crashes. If issues persist, document BSOD codes for further analysis.
Wow, that was quite a lot to take in.
Thank you for the detailed explanations.
You definitely addressed some points I had already explored before starting an x58 project, such as the recommended voltages for QPI, VCore, etc.
Currently, my RAM runs at 1520Mhz and the uncore at around 3400Mhz. I can always adjust it down to better match the 2x ratio for a higher BCLK, so I’ll review your suggested settings to see how they perform and whether the system will boot.
When I mentioned the system wouldn’t POST, I should have been more precise about the fact it’s looping. The current setup is fully stable, and it took me some time to fine-tune the parameters. I was concerned that not seeing any beep codes beyond 196BCLK might be due to luck or issues with the CPU or motherboard.
I’ll try it tomorrow and let you know my results or findings.
I hope you’re right, and I should have been more thorough in my research on the x58 platform for overclocking above 4.0GHz, even though running at 3.9Ghz seems unnecessary compared to 4Ghz—it just feels a bit off.
Have a great day, and thank you again!
After many attempts and adjustments, it seems the system is still functioning intermittently, making a brief click before restarting after roughly 2 to 3 seconds. I followed all your suggested configurations and made minor changes, but nothing has resolved the issue. It’s frustrating. If you have any additional suggestions or insights, please share them. Thank you again.
VCore: 1.425v (Yes, I know this is really high but it's just a quick test)
Double-check that your QPI link speed is ~7,236MT/s (3,618MHz).
If this lets it boot then try 1.4v VCore.
Keep dropping the voltage by two steps until it wont boot.
When it wont boot, bump the voltage back
up
3
steps.
At that point, see if you can get to Windows.
If it still wont boot, even with 1.425v VCore, then try lowering the Uncore to one step below 2x DDR3 speed. Yes, even though you're warned this is a bad idea it can still work and sometimes gives you some more stability. Although, the cost is an obvious reduction in RAM performance and increase in RAM latency.
If it STILL wont boot then set the QPI link speed to "SLOW MODE". This mode is painfully slow (We're talking like 1/10 the speed here) but sometimes it will let the system boot. If it does then that means you need to greatly increase the QPI voltage. If this is the case then you can try QPI at 1.4v, whilst setting link speed back to 7,236MT/s, but keep in mind that you shouldn't stay at that voltage for very long. You want to stay at, or below, 1.35v on QPI. You can run above 1.35v, all the way up to 1.375v, but be warned that the IMC will degrade at that voltage and eventually make it harder and harder to stay stable (more voltage).
If it wont boot fully into Windows: Increase VCore 2 steps.
If it boots to Windows but you get a BSOD or straight up restart: Depends on the BSOD/restart but very highly likely you need to increase VCore (sometimes QPI, only increase one at a time... I have a cheat-sheet for the BSODs - See spoiler)
After that you're looking at doing the stress tests to see if it's stable.
My guess at this point is that you have poor chip quality. Kind of like my second X5670 that needs a ridiculous amount of VCore to even run at 4.2GHz (175x24 or 191x22 -> 1.4v VCore). It's just silicon lottery. The X5650's are the lowest bin of the series so it's not surprising.
Last thing:
There is a slight possibility that your overclock is tripping your motherboard's over-current protection (OCP) or you're tripping OCP on the PSU. What's the make and model of PSU you have? Are you using a 2x4-pin CPU EPS power cable or are you using a 2x2-pin?
Spoiler:
BSOD codes/names and their possible cause
0x101 = CLOCK_WATCHDOG_TIMEOUT = increase VCore
0x124 = WHEA_UNCORRECTABLE_ERROR = increase/decrease VCore (Usually increase) or QPI/VTT (Depends, usually increase)... You have to test to see which one it is.
0x0A = IRQL_LESS_OR_NOT_EQUAL = unstable RAM/IMC, increase QPI first, if that doesn't work increase VCore
0x1A = MEMORY_MANAGEMENT = It usually means a bad stick of Ram. Test with Memtest or whatever you prefer. Also, try raising your RAM voltage.
0x1E = KMODE_EXCEPTION_NOT_HANDLED = increase VCore
0x3B = SYSTEM_SERVICE_EXCEPTION = increase VCore
0x3D = INTERRUPT_EXCEPTION_NOT_HANDLED = increase VCore
0xD1 = DRIVER_IRQL_NOT_LESS_OR_EQUAL = QPI/VTT, increase/decrease as necessary
0x9C = MACHINE_CHECK_EXCEPTION = QPI/VTT most likely, but increasing VCore has helped in some instances
0x50 = PAGE_FAULT_IN_NONPAGED_AREA = RAM timings/Frequency or Uncore multi unstable, increase RAM voltage or adjust QPI/VTT, or lower uncore if you're higher than 2x
0x109 = CRITICAL_STRUCTURE_CORRUPTION = Not enough or too Much memory voltage
0x116 = VIDEO_TDR_ERROR = Low IOH (NB) voltage, GPU issue (most common when running multi-GPU/overclocking GPU)
0x7E = SYSTEM_THREAD_EXCEPTION_NOT_HANDLED = Corrupted OS file, possibly from overclocking. Run sfc /scannow and chkdsk /r
0x139 = KERNEL_SECURITY_CHECK_FAILURE = Corrupt/out of date driver, possible corrupt OS file (I've had a bad drive), possibly bad/badly seated RAM, maybe increase VCore
0xF7 = DRIVER_OVERRAN_STACK_BUFFER = Bad system file read/corrupt. Issue with the OS or the storage. -> "HDD bad", or increase VCore
0x7F = UNEXPECTED_KERNEL_MODE_TRAP = QPI/VTT, Vcore, or RAM voltage. Have to test which.
0xC0000255 = Strange boot up = Corrupt BIOS config/behavior probably by RAM overclock. -> load defaults or reset CMOS to fix.
My PSU might be the problem—it's an older Corsair CX750w with 4+4 for the CPU. I can attempt to replace it with a newer model that has more power and see if that improves things.
Yes, having a newer PSU would be beneficial. It doesn't have to exceed 750w, as I'm using a 650w one for "Gramps." I definitely need something above 550w based on my experience.
Power meter shows 570w at the wall (about 513w from the PSU assuming 90% efficiency). I've tried SLI OC'd GTX 970s with an OC'd CPU and it caused issues starting up on a 650w unit.
😛
Does the total amperage and wattage usage factor into this calculation, or is it solely about the reliability of the PSU? I'm checking if using a GT210 instead of the RX580 would still work, just in case. It would be helpful to know for future reference. Thanks.
It varies by power supply unit. Some are based solely on overall consumption, while others depend on specific components drawing more. Generally, most systems rely heavily on the +12v line. Power output equals voltage multiplied by amperage. When purchasing a new PSU, opt for at least 750w. I favor higher quality models from brands like Seasonic, Corsair, or certain EVGA units over cheaper alternatives with lower wattage.
If your PSU is likely causing problems, adding a low-power GT 210 can provide a solid buffer. The main concern is that many GPUs at idle already consume minimal power, so the difference may be negligible. Still, testing with a higher output unit could help. The old Corsair CX750 supports up to 62A on the +12v line, suggesting you're probably not overdrawing unless there are other hardware demands like multiple drives and fans.
In the end nothing changed, it didn't matter.
I still experience that strange clicking and no posting.