the process reduces speed to 4.5ghz from 5ghz at 70°c
the process reduces speed to 4.5ghz from 5ghz at 70°c
When the CPU isn't under any workload, all cores operate smoothly at 5GHz. However, when I start running prime95, they drop to 4.5GHz while the temperature rises between 65°C and 70°C. These readings feel unusually low, suggesting possible throttling. I suspect this might be related to BIOS configurations, but since I'm not very familiar with it, I need some guidance.
If you know why this occurs, please let me know! All relevant details should be available in the links provided below. My apologies for sending links instead of images directly here, but I’m struggling to view them properly. Even after following the instructions, the issue persists.
EDIT: I tested it while playing games and all cores remained at 5GHz even with a load around 35%. Yet, with prime95 running, they drop at 100% load.
Specs:
https://prnt.sc/rurp3s
prime95 = off
https://prnt.sc/rurppg
prime95 = on
https://prnt.sc/rurq6t
BIOS settings:
https://prnt.sc/rurqyr
https://prnt.sc/rurrat
https://prnt.sc/rurrk9
https://prnt.sc/rurrtr
https://prnt.sc/rurs4e
It's the AVX offset setting. When executing AVX instructions (Prime95 uses AVX), the system will slow down to that offset.
Your current screenshot sets it to auto, which often defaults to around 5. Running an AVX instruction with an offset of 5 will apply a multiplier of 50 and then subtract 5, resulting in a multiplier of 45.
You have the option to change this, but doing so increases heat generation and power consumption as the offset number decreases.
I observe that you are operating the 9900KS on a Z370 series motherboard. Although it works, the power consumption of the 9900K and particularly the 9900KS can impose significant demands on the VRM design built into the Z370 boards, which were not originally intended for the two most power-hungry CPUs in the 9th Gen lineup—your model being the most power-intensive among them. This may lead to the VRM overheating and causing the CPU to throttle. The CPU's 70°C temperature isn't an issue, but it could suggest that the VRM is also running excessively hot.
Keep in mind that Prime95, Small FTTs, especially when AVX is enabled, place considerable stress on both the CPU and the board’s power phase design, often exceeding normal everyday loads. If you're using a custom water cooling loop or an AI-O for CPU cooling, this could worsen the situation due to reduced airflow over the VRM—something typically absent with standard air coolers.
It's generally advised to opt for a Z390 series board with robust power architecture when dealing with the two i9 CPUs, given their high power requirements. While compatibility isn't an issue, power consumption often becomes the real concern in this setup.
Occasionally, people claim that Z370 boards may experience significant lifespan reduction under such conditions. However, without personal experience in this exact case, I can't confirm how critical this might be generally.
I reviewed your BIOS settings, but nothing immediately stood out as problematic. Still, a closer inspection would be wise to ensure everything is functioning correctly.
However, for me, using a Z370 board could pose a challenge that warrants further investigation, especially regarding the potential causes of throttling—most likely related to VRM temperature and load. There are numerous forum discussions describing similar issues.
I understand that power architecture varies across boards, and I've already noted this in my response. Even high-quality Z370 boards have reported struggles with this scenario—though it's more common on boards with average power designs.
Another potential issue could be power throttling instead of thermal throttling. Considering this, what PSU do you have? (Make and model + age)
Typically, there is a BIOS setting to allow higher power draw. It's been a while since I used an MSI board, so I'm not sure what that setting is called or if it exists.
To be honest, I'd start by checking the VRM temperature first. The combination of a Z370 board, the most power-hungry CPU compatible with it (but not originally built for), and Prime95 seems to create excessive stress on the system.
Thank you for your feedback. I reviewed the situation and found that VRM overheating or power throttling aren't commonly monitored through standard tools. Regarding your PSU, it's a new model—'corsair-600w sf600 sfx 80+ gold modular'—and I expect it will be replaced soon with a larger one. To verify if the system approaches the 600W limit when running Prime95, you could check the total power draw of your entire setup and compare it to the PSU's capacity.
It's the AVX offset setting. When executing AVX instructions (Prime95 uses AVX), the system will slow down to that offset.
Your current screenshot sets it to auto, which often defaults to around 5. Running an AVX instruction with an offset of 5 will apply a multiplier of 50 and then subtract 5, resulting in a multiplier of 45.
You have the option to change this, but doing so increases heat generation and power consumption as the offset number decreases.
Unfortunately it is not really a very common feature to have VRM temperature sensors. I believe some AMD x570 bords have taht feature.
So you would need a device which can read temperatures by placing sensors in various spots on the VRM, or something like an infrared thermometer like the ones that can measure body temperature in the ear.
I should have been more precise when I mentioned power throttling. Although the PSU is important to include when troubleshooting possible power related issues, what I actually had in mind, is that some motherboard has a defaullt BIOS setting, which limits CPU power draw to a certain number.
My own ASUS motherboard, actually throttles the stock i9 9900K quite a lot, because ASUS sticks very closely to Intels specified 95 TDP.
SO even with all requirements met, the 9900K throttles hard on my board. To remove t hat limit, there is a simple BIOS setting, and then everything runs perfectly fine.
Inmy own experience, this is not something I have run into on an MSI motherboard, I mostly mentioned it, because that could be something to look into, if the VRM temperature is not the issue.
If you complete system lacked power, I suspect a hard crash with BSOD would be an issue, or GPU related crashes, since it is very commonly seen in situations with poor power suplly.
Is the AVX Offset for the KS CPU really zero? That matches the standard documentation.
However, since it's configured as Auto in OPs post, there might be a problem.
Yes, I understand. Even though there are exceptions, based on my experience, Auto settings usually revert to CPU factory defaults, except for voltages, which often appear too high in many cases. However, I’m not trying to dismiss your explanation entirely—it seems like a very likely solution to the issue. The reason for asking is that I’m always open to learning new things.
I have not come across any information about a 300 series motherboard with a 0 default for the AVX offset. Where is the mention of the 0 default for the 9900ks? Intel's website only lists AVX2 support and does not specify a default offset.