Aurora R8, Newer Bios Seems to Limit CPU Performance

You can always revert back to the bios version you were on before the upgrade to see if performance changes.

Thanks but I would not like to roll back to an older BIOS as the newer ones have implemented important security patches. I am just curious about what other factors can cause CPU throttling other than overheating or OS power management options. CPU-Z test results showed that the CPU voltage dropped from 1.217 to 1.194. All other things stayed the same. 

@cubicfang :

Unfortunately you're missing 3 critical pieces of information on the comparison snapshots you posted pre and post BIOS upgrade. They are the PL1 power limit, PL2 time limit or tau, and the power draw chart.

It is very possible that a BIOS update goes and reset the PL1 power limit or the tau to a more conservative default. With AWCC, you do not have the option to tweak that number. You will need to install XTU and tweak those numbers.

If I remember correctly, my R8 BIOS has a default of 95W PL1 power limit and tau of 28s. If you cannot change your power limit value due to the locked processor you have, see if you can extend the time limit from 28s to infinity.

FYI, my MSI Z390 MEG ACE which is a gaming motherboard made to support up to LN2 cooling for OC has infinity tau and 250W power limit right out of the box.

@Anonymous 

Yes, good catch! Water-cooling truly solved the overheating issue but somehow there is still a thermal-throttling-like symptom.

@GTS81 

Thanks! Unfortunately, XTU does not support a locked processor like my i7-9700. I don't think there is any way I can adjust those numbers. Maybe next time I should build a game PC from scratch. 

Build your own or buy a PC with an unlocked CPU.  Lot of Alienware PC's with unlocked CPU's.

@Anonymous 

Never thought VRM would be an issue. It definitely could be. But how can I monitor its temperature? 

Anyway, after the water-cooling upgrade, CPU-Z stress CPU test can last longer before the processor got throttled (15-20 seconds versus just 5 seconds). But Cinebench R20 test consistently hit the brake on core speed in about 6-7 seconds, no matter what CPU temperature is. 

@Anonymous :

Are you getting this issue when running OCCT? Because strangely, the orange julius is facing something similar. When OCCT running the Large Data Set benchmarks with my 9900K OC'ed to 5GHz, at 1.4V Vcore (yeah I know it's stupid high but I'm testing the orange juice), for the first 10 minutes, I'm holding 70C. Suddenly the VRMs jump from 58C to 85C, CPU spikes to 85C, and then I panic.

@cubicfang :

Dell did not apply thermal paste on my R8's VRM heatsink. They used thermal pad.

Aurora R8 VRMs and heatsink thermal pad

@Anonymous , @cubicfang :

I think both of you are getting limited by the BIOS as to how much juice (current) can flow through the VRMs.

For @Anonymous , it's probably by design the motherboard on the XPS 8930 doesn't have that beefy a VRM so spend less time with pseudo and more time with Cosmos.

For @cubicfang , it's probably the non-K CPU locking you in to Dell BIOS defaults. I plugged a 9600K into my R8 and I can change PL1 limits, tau, etc. in XTU.

@Anonymous 

Very interesting! What VRM heatsink did you get? I never checked VRM heatsink and do not even know if Dell applied any thermal paste on it.

yeah like really nobody puts thermal paste on vrm heatsinks, it's not designed for that type of gap filling application.  But you can check to make sure Dell removed the plastic liner from the thermal pads.   Remove Thermal Liner

@GTS81 

Thanks a lot for the info. I now think your early response that says "R8 BIOS has a default of 95W PL1 power limit and tau of 28s" seems to explain what I experienced. I just used the stopwatch app on my cell phone to accurately monitor the CPU-Z stress test. The processor power did stay at around 95W for about 28s before dropping to 65W. I sampled several times and got similar data varying between 25-30s. Though the time was always longer when the CPU is cooler. It does seem to be "Power Limit Throttling". Cinebench R20 results are also understandable since the processor power spiked to be above 125W at the beginning of the test. Then it would get lowered to 65W in a much shorter period of time (the accurate value is about 15 -16s). I don't know why Cinebench pushes processor power to such an extreme but that behavior makes more sense for benchmarking unlocked CPUs. 

@Anonymous 

Very impressive DIY stuff! Good luck with your experiment and please do share the outcome in this thread. 

You can buy mini mosfet heatsinks from amazon and stick it on with thermal stickey pads.  

Or I have some spare thermal plastic if you want an oem part to cover it with  

@r72019 :

But you can check to make sure Dell removed the plastic liner from the thermal pads

Can't say how glad I am that now your VRMs aren't choking.

@Anonymous :

Note how the CPU heatsink has a mounting arm for the fan that cuts into the VRM heatsink.

Does the mounting arm have a conductive piece that touches the exposed VRM(s) to cool them down?

@cubicfang :

I don't know why Cinebench pushes processor power to such an extreme but that behavior makes more sense for benchmarking unlocked CPUs. 

Cinebench R20 unleashes all of your 6 cores' AVX2 logic which is a boatload of power used to run. In the first few seconds all of your cores are firing away on AVX2 workload, it probably beat the hysteresis effect of the thermal sensors on the die and the time it takes for the power control unit to cycle through its microcode to hit the stop button on supplying > 65W to your CPU.

https://www.pcworld.com/article/3356179/meet-the-new-tougher-cinebench-r20-benchmark-we-test-it-on-xeon-and-threadripper.html