Problem Overclocking Threadripper 1950X

dpriest

Limp Gawd
Joined
Nov 18, 2014
Messages
131
With my original Windows 10 install, I overclocked my 1950X to 4GHz & 1.25V. The actual frequency says 3.9GHz This was stable with video encoding and I could also run other applications. After I reinstalled windows because I changed M.2 drives and changed the cooler from the Enermax AIO to the Noctua NH U14S TR4, my computer would occasionally freeze up after encoding for a little while (cursor was frozen, computer was unresponsive and I had to force shut down by holding power button down.) I tried to increase volts to 1.262V and still freezes up. I assume the freezing up is instability due to not enough volts? 1.275 volts seems stable but is running a little hot even with the Silver Arrow TR4 that I just installed. (Thanks Kyle for a great tutorial on applying the TIM). Again, the freezing up would not happen 100% of the time but would do this somewhat frequently.

What would cause the freezing up? Again, it didn't used to do this with the original install. I have the Asrock X399 Taichi and 4x 8GB Corsair RGB RAM overclocked to 3200 XMP Profile 2.0.

Looking forward to insight and a solution.

Thanks in advance.
 
What voltage you running the RAM at? Probably needs to be 1.35v to 1.45v to get that stable.
 
The RAM voltage is at 1.35V (XMP 2.0 Profile 1: DDR4-3200 16-18-18-36). For overclocking the CPU, the Overclock mode is set to manual, CPU Frequency and Voltage Change is set to Manual and SMT Mode is Auto. Is the reinstallation of Windows and changing the CPU cooler coincidental or not?
Run it at 1.45v on the RAM, see if that is the issue, then move to CPU. I would start pushing vCore up aby 0.05v at a time and see what happens.

Edit: 0.05v
 
Last edited:
Just dropped the cpu voltage to 1.25=Unstable (Froze Up) Going to 1.262 which is actually 1.256. I heard that changing the RAM SOC voltage could help with stability to 1.1 or 1.2V. Mine is currently at .0990V. I trust your judgment and so not touching SOC yet. Thoughts?
Unstable at 1.262V and RAM at 1.45V.:( Going with your next suggestion with is pushing vCore? How do I do this?
 
Also, are you sure the system is stable at stock clocks?
 
I took a screenshot of all the sensors and tried to upload but wasn't sure how to do that.
As it says in the edit box....just drag and drop your screen shot here. Or if you had it copied to memory, you can simply Ctrl-V to paste into the edit box.
 
vSOC is chipset voltage. Not sure what Asrock calls it. 1.1v is what I would set.
 
Still not stable at 100% cpu load:( Here is wwhat I did in the BIOS for overclocking:

Overclock Mode: Manual
APU/PCIE Frequency: 100
CPU Frequency and Voltage Change: Manual
CPU Frequency: 4000
CPU Voltage 1.262
XMP Profile: XMP 2.0 Profile 1
VDDCR_SOC Voltage: Fixed Mode (Fixed Voltage=1.1)
DRAM Voltage: 1.45
DRAM_CD Voltage: 1.45
VDDCR_SOV_S5: Changed from Auto To 1.1V

Thoughts?
 
Still not stable at 100% cpu load:( Here is wwhat I did in the BIOS for overclocking:

Overclock Mode: Manual
APU/PCIE Frequency: 100
CPU Frequency and Voltage Change: Manual
CPU Frequency: 4000
CPU Voltage 1.262
XMP Profile: XMP 2.0 Profile 1
VDDCR_SOC Voltage: Fixed Mode (Fixed Voltage=1.1)
DRAM Voltage: 1.45
DRAM_CD Voltage: 1.45
VDDCR_SOV_S5: Changed from Auto To 1.1V

Thoughts?
Up the vCore? You realize that is incredibly low for 4GHz on that CPU?
 
When referring to the vCore, you mean changing the value of the CPU Frequency in my BIOS screenshot above or another way? llc I asume is load line calibration. The CPU llc is set for 2. At 2.62V the CPU temp averages 63-65C and maxed at 68C which is a little too hot for my comfort zone. Based on my temps listed above, can I get any cooler temps from this heat sink? I can re-check the TIM but idle temps at this clock speed is 29C.
no the one labeled "cpu voltage" up it. which cooler do you have on, the aio or the hsf?
edit: from what I see most people are needing 1.35+ and llc 1-2 to get ~4GHz.
 
no the one labeled "cpu voltage" up it. which cooler do you have on, the aio or the hsf?
edit: from what I see most people are needing 1.35+ and llc 1-2 to get ~4GHz.
I've got the Thermalright Silver Arrow TR4 heat sink. I used the prolimatech PK-3 TIM and applied using Kyle's method. When increasing the CPU voltage, I enter 1.262, and the BIOS changes it to 1.25625 as in the screenshot above. Is there a way for me to get more manual control and increase incrementally? Any way to get temps lower? I am going to try to put another Thermalright TY-143 fan on the front of the heat sink. I have the Corsair RGB memory sticks which is tall so I may have to raise the fan a bit. Also, I ordered 2 140mm Noctua NF-A14 industrial fans to replace the standard Noctua A-14 fans on the front of my case as intake. Maybe this will help with temps.
 
When referring to the vCore, you mean changing the value of the CPU Frequency in my BIOS screenshot above or another way? llc I asume is load line calibration. The CPU llc is set for 2. At 2.62V the CPU temp averages 63-65C and maxed at 68C which is a little too hot for my comfort zone. Based on my temps listed above, can I get any cooler temps from this heat sink? I can re-check the TIM but idle temps at this clock speed is 29C.
vCore = CORE / CPU VOLTAGE.

Could also push up CPU LLC which is LoadLine Calibration. But if you start doing that, be very careful with what you are doing as it can get vCore out of hand quickly.

If you are 29C at idle on air, it is likely you have a very good TIM mate.
 
switch the overclock mode to manual and see if you can increase it.
I did switch it to manual. The CPU frequency can be changed from 100-200. Anybody familiar with the Asrock X399?

These are my current settings:

Overclock Mode: Manual
APU/PCIE Frequency: 100
CPU Frequency and Voltage Change: Manual
CPU Frequency: 4000
CPU Voltage 1.262
XMP Profile: XMP 2.0 Profile 1
VDDCR_SOC Voltage: Fixed Mode (Fixed Voltage=1.1)
DRAM Voltage: 1.45
DRAM_CD Voltage: 1.45
VDDCR_SOV_S5: Changed from Auto To 1.1V
 
your pics show overclock mode at auto. if you've changed it good. you need to watch the voltage when stress testing to see what llc is bumping it too. if its to high like 1.4+ then you need to back off. also most are doing ~4 on water so you might need to switch back if a second fan doesn't help.
 
your pics show overclock mode at auto. if you've changed it good. you need to watch the voltage when stress testing to see what llc is bumping it too. if its to high like 1.4+ then you need to back off. also most are doing ~4 on water so you might need to switch back if a second fan doesn't help.
What monitor sw will show llc? I'm a bit of a noobie to this so not sure what you mean by this. In terms of cooling, I started with the Enermax Liqtech TR4 360. Had problems with the first one. RMA'd the first and problems with the second. I've gotten better performance and much more reliable with air cooling.
 
What monitor sw will show llc? I'm a bit of a noobie to this so not sure what you mean by this. In terms of cooling, I started with the Enermax Liqtech TR4 360. Had problems with the first one. RMA'd the first and problems with the second. I've gotten better performance and much more reliable with air cooling.
your hwmon shows the cpu voltage, nothing shows the llc. llc takes the voltage you set and adds a bit more to compensate for vdroop under load and the settings are how aggressive it its. if you turn it off and then stress the cpu and see the cpu voltage drop thats vdroop, llc bring it back up. but as kyle said it can be overly aggressive and instead of bringing it back to what you set it at it can add too much.
at least that how I understand its operation...
 
your hwmon shows the cpu voltage, nothing shows the llc. llc takes the voltage you set and adds a bit more to compensate for vdroop under load and the settings are how aggressive it its. if you turn it off and then stress the cpu and see the cpu voltage drop thats vdroop, llc bring it back up. but as kyle said it can be overly aggressive and instead of bringing it back to what you set it at it can add too much.
at least that how I understand its operation...
Doesn't seem like there is much else to try. I can't seem to incrementally increase the voltage. The next bump up in CPU voltage is 1.275. Temps are a bit high at this voltage. Cannot push 1.3V because of high CPU temps.
 
Doesn't seem like there is much else to try. I can't seem to incrementally increase the voltage. The next bump up in CPU voltage is 1.275. Temps are a bit high at this voltage. Cannot push 1.3V because of high CPU temps.
back off then, wait for your second fan and try again. not sure that hsf is gonna cut it for a 4Ghz oc...
 
back off then, wait for your second fan and try again. not sure that hsf is gonna cut it for a 4Ghz oc...
OK. By the way, I stress tested the CPU by running 3 instances of Handbrake encoding video at the same time resulting in 100% CPU usage in all 32 threads. Normally when I encode video, I only use a single instance of Handbrake which results in all threads fluctuating percentages. If the 4GHz failed stability when stressed and it is stable mostly with only one Handbrake open, maybe I'll leave it at 4GHz. Is there any adverse effects to the longevity of the processor or system if it becomes unstable and freezes up? I know excess heat shortens the life of a processor. I'm wondering if freezing up does.
 
A couple of other thoughts. Do you think increasing the VDDCR_SOC and VDDCR_SOC_S5 voltage to 1.3V will help and is it safe to do so? What about adjusting LLC?
 
I dont think freezing up is gonna hurt it. try it and see but in small bumps.
 
I dont think freezing up is gonna hurt it. try it and see but in small bumps.
OK. Thoughts about this: Do you think increasing the VDDCR_SOC and VDDCR_SOC_S5 voltage to 1.3V will help and is it safe to do so? What about adjusting LLC?
 
OK. Thoughts about this: Do you think increasing the VDDCR_SOC and VDDCR_SOC_S5 voltage to 1.3V will help and is it safe to do so? What about adjusting LLC?
what is it at now? and leave llc at 1 or 2 for now.
 
OK. By the way, I stress tested the CPU by running 3 instances of Handbrake encoding video at the same time resulting in 100% CPU usage in all 32 threads. Normally when I encode video, I only use a single instance of Handbrake which results in all threads fluctuating percentages. If the 4GHz failed stability when stressed and it is stable mostly with only one Handbrake open, maybe I'll leave it at 4GHz. Is there any adverse effects to the longevity of the processor or system if it becomes unstable and freezes up? I know excess heat shortens the life of a processor. I'm wondering if freezing up does.

if it's working leave it, but check your encodes for any errors. if you start noticing weird things like hiccups in the encode, etc. then the overclocks probably right at the edge of being stable. as far as longevity sure heat can effect it but unless you plan on using that processor for the next 10 years it's not going to negatively effect it enough to matter. now if you were sitting at 95C 100% of the time 24/7 then yeah you might start seeing issues earlier. excessive voltage above what is recommended for the processor is what really shortens the lifespan of a processor and/or VRM's on the motherboard but you're no where close to that on temp or voltage so it shouldn't be an issue.

either way if it was me i'd definitely watch the voltages with the LLC being set to 1, usually setting it that high is reserved for extreme overclocking and if it's not stable at LLC 2 you might need to bump the cpu voltage up just a little more.. if what you said earlier in the thread is correct about your cpu sitting at 68C full load @ 4Ghz then you have plenty of thermal room to spare. i'd personally go up to 1.3V check temps(motherboard and cpu will protect it's self from exceeding the 95C limit on the processor). see if that ends up being stable with LLC2. if it is then work on lowering the voltage til you find the lowest voltage possible to keep it stable.. if it's not stable even at 1.3v with LLC2 then lower the overclock to 3.9Ghz and find the lowest voltage for that and call it a day.
 
if it's working leave it, but check your encodes for any errors. if you start noticing weird things like hiccups in the encode, etc. then the overclocks probably right at the edge of being stable. as far as longevity sure heat can effect it but unless you plan on using that processor for the next 10 years it's not going to negatively effect it enough to matter. now if you were sitting at 95C 100% of the time 24/7 then yeah you might start seeing issues earlier. excessive voltage above what is recommended for the processor is what really shortens the lifespan of a processor and/or VRM's on the motherboard but you're no where close to that on temp or voltage so it shouldn't be an issue.

either way if it was me i'd definitely watch the voltages with the LLC being set to 1, usually setting it that high is reserved for extreme overclocking and if it's not stable at LLC 2 you might need to bump the cpu voltage up just a little more.. if what you said earlier in the thread is correct about your cpu sitting at 68C full load @ 4Ghz then you have plenty of thermal room to spare. i'd personally go up to 1.3V check temps(motherboard and cpu will protect it's self from exceeding the 95C limit on the processor). see if that ends up being stable with LLC2. if it is then work on lowering the voltage til you find the lowest voltage possible to keep it stable.. if it's not stable even at 1.3v with LLC2 then lower the overclock to 3.9Ghz and find the lowest voltage for that and call it a day.
 
I think I found stability. The only thing new I changed was CPU LLC from 2 to 1. Ran 3 instances of Handbrake simultaneously with very high quality encoding settings, taxing the processor to 100% for 3 1/2 hours. My system did not freeze up and the max. CPU frequency hit 4.5Ghz. Does this mean that I can consider it stable? As I mentioned before, normally when I encode video, I only use a single instance of Handbrake which results in all threads fluctuating percentages but all threads do not sit at 100%. I would assume there would not be any encode errors with only one Handbrake running?

Not sure what I should be looking for with voltages. In HWMonitor, the CPU VID for all 16 cores were constantly fluctuating with min. values. I'm not sure what VID is. What are your thoughts based on the screenshot picture below?

Summarizing my OC settings:

Overclock Mode: Manual
APU/PCIE Frequency: 100
CPU Frequency and Voltage Change: Manual
CPU Frequency: 4000
CPU Voltage 1.262 (1.25625 Actual)
XMP Profile: XMP 2.0 Profile 1
CPU Load-Line Calibration: Level 1 (Level 2 Was Default)
VDDCR_SOC Voltage: Fixed Mode (Fixed Voltage=1.1)
DRAM Voltage: 1.45
DRAM_CD Voltage: 1.45
VDDCR_SOV_S5: Changed from Auto To 1.1

Capture.PNG
 
Back
Top