Rampage vi extreme i9 7980xe overclock avx too hot

[feedmeink]

Overclocking i9 7980xe. Looking for adaptive settings as to downclock and down volt when not under full load. Set turbo cpu voltage to 1.008, avx negative offset to 3 and 3. LLC is set to 4. Cache 27. Cpu input voltage 1.9 (don't confuse this with cpu core voltage. Core voltage set to 1.008). Power saving settings set to enable. Turbo on, speed step on ect. Setting each core ratio in specified per core. Depending on heat durring stress test and cinebench either downclocking or adjusting offset voltage negative increments. Trying to keep cores within range of 38-41multiples. Temperature for each core (monitored in core temp application) is steady mid to high 60c in aida64 cpu only stress test. Cinebench scores great. Almost too good to be true at 3,860. Temps in cinebench are reasonable. Fpu only stress test in aida 64 2 cores reached 92c before I shut it off. Other cores within 75-83c for fpu test. This is custom waterloop with full monoblock. 2 radiators. One 260mm, another 360mm. 6 fans blasting. One overclocked vega 56 with waterblock attached to the same loop. If anyone has any advice on how to pass this fpu test with reasonable temps given the settings above, it would be appreciated. What would you consider reasonable temp durring fpu stress test in aida64. Thanks in advance. This is my first time here and I'm very open to learning more about overclocking.

[feedmeink]

I figured out the issue. Set avx offsett to 8 each and it runs 65-70c durring fpu test. Score in cinebench went down a crap load. If someone can give me some insight on how to get a better handle on avx it would be much appreciated. If I control cache voltage will that have an effect on testing fpu?

[feedmeink]

Not a good first experience with these forums. But if someone is going through similar experience, I am going to adjust ac and dc to .01 when I get home from work to see if it will help prevent me from having to set avx offset to such a high number in adaptive mode. I doubt anyone is even going to lend a helping hand here, but I should not have to set avx offset to such a high level to maintain my heat output when running avx stress test. I don't need exact guidelines
Just looking for some input on what setting in bios control avx operation voltage besides avx negative frequency and llc.

[feedmeink]

There is no ac and dc option avaliable on x299 bios. Someone please help.....

[Silent_Scone]

feedmeink wrote:
There is no ac and dc option avaliable on x299 bios. Someone please help.....

Those options are not applicable to X299. They are designed to bring the voltage down closer to the user requested voltage as Intel's programmed VID will attempt to override what is being set, which is nessessary means on Intel's mainstream platforms. On X299 the voltage requested in adaptive mode will be close enough to what is being set by the user in the Additional Turbo Voltage field. That is unless the voltage being set is lower than the stock VID for a given ratio, which is not possible without use of the offset function.

It may be possible to reduce VCCIN to around 1.85v depending on the CPU to help reduce temps. However, if planning on running AVX workloads in tandem with overclocking the best answer in terms of tangable benefit would be to delid the CPU. Understandably if you're not comfortable invalidating your warranty then this isn't the solution for you.

Floating point tests such as AIDA's routine will push the CPU hard, you're asking a lot of the CPU. In fact I wouldn't recommend running this and stick to real world stability testing that mimics your workload more closely. I don't believe anyone in the tech media has provided any testing or oscilloscope shots with these CPU, but I'd imagine the current being pulled in those tests is quite eye opening.

[feedmeink]

Silent Scone wrote:
Those options are not applicable to X299. They are designed to bring the voltage down closer to the user requested voltage as Intel's programmed VID will attempt to override what is being set, which is nessessary means on Intel's mainstream platforms. On X299 the voltage requested in adaptive mode will be close enough to what is being set by the user in the Additional Turbo Voltage field. That is unless the voltage being set is lower than the stock VID for a given ratio, which is not possible without use of the offset function.

It may be possible to reduce VCCIN to around 1.85v depending on the CPU to help reduce temps. However, if planning on running AVX workloads in tandem with overclocking the best answer in terms of tangable benefit would be to delid the CPU. Understandably if you're not comfortable invalidating your warranty then this isn't the solution for you.

Floating point tests such as AIDA's routine will push the CPU hard, you're asking a lot of the CPU. In fact I wouldn't recommend running this and stick to real world stability testing that mimics your workload more closely. I don't believe anyone in the tech media has provided any testing or oscilloscope shots with these CPU, but I'd imagine the current being pulled in those tests is quite eye opening.

Thank you very much for this response. I am setting the voltage below stock voltage. Had to offset each core negative to keep voltage read out steady across all cores. I updated bios last night to the 1401. Previously it was 1301. I found it runs more consistant in testing when I lower the voltage more than stock. Right now max voltage is 1.02-1.021 on all cores durring basic cpu stress test in aida64 without avx functions. With fpu test the volts dropped to .870-.920 volt range depending on v-droop.

Vccin is set much lower than you recommended. Mine is 1.01. Is this an issue? This is my first time getting deeper into overclocking. I know originally I was using a higher value, but I got paranoid and started low

[Rob_W_]

I have my cpu input voltage set to 1.8volts, avx-3 avx512-5, on its own loop 420 rad, delided, most testing in the high 60’s to low 70’s at 4.6ghz, the biggest impression on temps was to delid cpu, such a hot running chip!