Core 7 265 temperature spikes and current relation building Intel Linux kernel - continued

[Debugging]

Continued from https://community.intel.com/t5/Mobile-and-Desktop-Processors/Core-7-265-temperature-spikes-and-current-relation-building/m-p/1728629

I was  away for a while. Thank you for the clarification. The Mobo has 3 x 8 pin connector.. The PSU came with 2  8pin power cables. Intel docs. state that peak current for this CPU can be 34A for 265. (for a few milliseconds).  The Mobo support 336W per 8 pin power connector Per calculation 1x 12V connectors are not enough for 34A. So I added a 2nd and 3rd cable from the PSU.  The very short peaks are around 90C. Would this reduce chances on damage or  longlivity of the CPU. The active cooler installer was rated at 240W, and the CPU is rated at 65W. (but peak is 34A*12V=408W) but that cooler then still seems not enough to handle such peaks that seem to require 408 (34x12) W. I am not aware of LGA1851 coolers than 240W. Is there a way to bring these peaks down ?

[JohnM_Intel]

Hello Debugging,

Thank you for creating a new thread regarding your concern with the Intel® Core™ Ultra 7 265K. I understand how concerning overheating issues can be, and I appreciate you bringing this to our attention.

To help isolate the issue, I kindly ask that you follow the troubleshooting steps outlined in the article “Overheating Symptoms and Troubleshooting for Intel® Boxed Processors” These steps will help us better understand the root cause of the behavior you’re experiencing.

Once you’ve completed the steps, please let me know if the issue persists, and I’ll be more than happy to assist you further.

Regards,

John M.

Intel Customer Support Technician

[PC1997]

You’re on the right track thinking about power delivery, but the key thing that helps here is separating CPU core power from the 12 V input side. On the Core Ultra 7 265, the current numbers Intel publishes are for the CPU cores running at around 1 V, not for the 12 V coming in from the PSU.

The CPU never actually sees 12 V directly. The EPS connectors feed 12 V into the VRMs, and those VRMs step the voltage way down while increasing current. So when you see a current figure like 34 A, that’s happening at the core voltage, not at 12 V. That’s why multiplying that current by 12 V makes the power look much higher than it really is.

From the PSU’s perspective, even during boost, the amount of current being pulled from the 12 V rail is relatively low. A single 8-pin EPS connector is already more than enough for this CPU at stock settings, including short boost bursts. Adding extra EPS cables mainly helps with VRM headroom under sustained heavy loads rather than changing temperatures or transient behavior and is worth connecting those extra cables, better to have more headroom than not.

The short temperature spikes you’re seeing into the 80–90 °C range are normal for modern Intel CPUs. Those spikes happen much faster than heat can actually build up in the cooler, so a 240 W-rated cooler can handle them without issue. Cooling capacity is about sustained heat over time, not millisecond-long power spikes.

If you want to rein in the peaks a bit, things like lowering PL2, shortening Tau, or applying a small undervolt will make a real difference. Changing cabling or sizing cooling based on theoretical peak electrical power won’t really affect those brief events.

[Debugging]

 I referenced https://www.intel.com/content/dam/www/public/us/en/documents/design-guides/resellers-power-supply-design-guide-changes.pdf this document. without further details and the manual referencing 12V2 my assumption was this peak current was @ 12V. 34A . Thank you for the explanation, even the manufacturer did not hint at the 1V when I inquired about the power.

Regarding overheating:

• System boots up but shuts down automatically after a short period of time.
  • system does not have such issue
• Reported CPU operating frequency is less than expected.
  • system does not have such issue
• Evidence of CPU throttling.
  • system does not seem to have such issue, For example, PCI5.0 NVME Gen5 benchmark (Temp-ssd=34C) using  one of the PCIe 5.0 MCIOx8 port , the speed is @ 14.000+ MB/sec constant
• General slowness of system.
  • system appears not have such issue
• CPU/system fan noise is excessive.
  • I spend a lot of effort using high end fans with very low noise so that the unit is almost un-audible used in a silent room . For this, I modified the chassis and cooler fans. I got an IPMI document from  MB supplier to force fan to maximum speed, (up from 900 to around 2400rpm). I could use this to speed up the fan under high load even though the BMC apparently deemed it not necessary to increase the fan speed. Even then, the same temperature peaks occur but a just a few C lower

After cold power on idle running Linux the cores of the CPU are around 38C with some tricks changing the chassis I could get to 35C. I use the thermal paste from the supplier for the cooler. the fans are operating. When blocking the air flow, the CPU heats up, so the fans have impact . The CPU cools in my view a bit slow down after peak effort to idle as to get back to 35~38C takes quite some time (+10 minutes). I could improve the cool down time by holding a very noisy 8mm  fan beside the cooler.

All in all, there is probably nothing wrong but it was surprising to me to see these spikes, without or a poor a fan, they reach 105C. As far as I can recall this did not occur on older CPU's ,like and very old Sandy bridge I have using the same kernel build.

[PC1997]

Your idle temperatures are normal and can vary depending on what background processes are running.

The Intel Core Ultra 7 265K can briefly draw up to 250 W under Power Limit 2 (PL2) during short turbo workloads, typically for around 45 seconds. During these bursts, the CPU is designed to safely reach temperatures of up to 105 °C as part of Intel Turbo Boost Max Technology 3.0.

In simple terms, the processor will use as much of your cooler’s available thermal headroom as it needs. If your cooler is efficient, it may never hit those maximum temperatures. Run a CPU stress test to fully load all cores and verify that your thermal solution is properly mounted and sufficient.

Once turbo boost has commenced, base power (PL1) is set to 125 W. Refer to the motherboard BIOS to review or adjust the configuration.

[PC1997]

Correction: You are using the Intel Core Ultra 7 265 CPU, not the 265K. As a result, PL1 is 65 W, PL2 is 182 W, and Tau is ~20 seconds. I don’t have the exact turbo boost time window values for this CPU.

Ensure you have at least a 400 W power supply; 500 W is recommended.

[Debugging]

Perhaps an illustration helps for a typical build. Horizontal Time in seconds, vertical temperature. mprime results in  about the same "highest core" temperatures, but since mprime takes a long time to complete. IMHO a Linux build is a more close to practical use to evaluate temperature and performance. AR , LD, BTF are build stages/commands shown in the screen through the build . No t sure why these peaks occur exactly at the points. Perhaps it has to do with how these tools behave and use CPU cores.

[Debugging]

This is the same build on a PC from around 2012. No spikes in temperatures. 

[1] Why is this "normal" and does this happen on newer CPUs ?

[2] What is the mechanism that causes this behavior ?

[3] Is there any other example/data (from/by Intel?) running such test to confirm this behavior ?

[JohnM_Intel]

Hello Debugging,

You may notice spikes in CPU temperature whenever your processor experiences sudden increases in workload. This happens because higher CPU power consumption is converted into heat, quickly raising the core temperature.

Intel® processors are equipped with thermal and power management technologies, such as Intel® Turbo Boost and power limit controls (PL1 and PL2), which dynamically adjust frequency and power to manage these peaks. The effectiveness of your system’s cooling solution and airflow will also influence how high and how long these temperature spikes last.

During CPU-intensive tasks, like Linux builds, temperature spikes are expected especially during stages such as AR, LD, and BTF.

These tasks generate short bursts of high CPU utilization, unlike stress tests like mprime, which maintain constant load. Such bursts lead to sharp but brief temperature peaks that match the execution of specific build commands. This behavior is normal and reflects how real-world workloads use multiple CPU cores in parallel.

Please note that these temperature spikes are normal and generally not a cause for concern. Idle temperatures are lower, and higher temperatures during active workloads such as building software or gaming are expected. Intel processors continuously monitor temperature and adjust frequency and power to prevent overheating or hardware damage. The maximum operating temperature is 105 °C.

To maintain safe operating conditions, ensure that your system has an adequate cooling solution and proper airflow. If you notice persistent or unusually high temperatures, consider checking your cooling setup or contacting support for guidance.

Regards,

John M.

Intel Customer Support Technician

[JohnM_Intel]

Hello Debugging,

I hope you had the opportunity to review the information I posted. At your earliest convenience, please let me know so we can determine the best course of action to resolve this matter efficiently.

Regards,

John M.

Intel Customer Support Technician

[JohnM_Intel]

Hello Debugging,

Since I haven't received a response from you, I will be closing this inquiry. If you need further assistance, please submit a new question, as this thread will no longer be monitored.

Regards,

John M.

Intel Customer Support Technician

[Debugging]

Sorry, I was occupied on some other work. I will test one more PC. Though I would like to  understand the mechanism behind these spikes. for example, how to simulate such. for example I could create a tool to start and stop threads rapidly on a single or multiple cores ? 

[JohnM_Intel]

Hello Debugging,

When the processor suddenly has to do a lot of work at once, it uses much more power in a very short time. That extra power quickly turns into heat, which causes the CPU temperature to spike.

Some types of activity don’t keep the CPU busy evenly. Instead, they create short bursts of heavy work. These sudden bursts make the processor speed up quickly, use more power, and briefly get hotter than usual.

This is similar to what happens when a system performs demanding tasks in short phases, such as loading, processing, or finalizing data. The rapid changes between light work and heavy work are what cause the quick temperature jumps, not a constant high load.

Regards,

John M.

Intel Customer Support Technician

[Debugging]

Hello John, thanks for the reply. Here is the last PC I could test. An off the shelf laptop from 3 years ago. Indeed peaks are showing here as well. if this is related to cooling performance of the cooler design, a test script to log the fan, CPU cores speed would be very welcome  as it seems to be able to show major differences for the heat spikes by making minor changes to the cooler setup. For example I made  a small change to fix fan speeds through IPMI,changed the airflow direction and type of fan and different heatsinks. Some cheap  high RPM fan can help but those  are around 35-40dB and thus unacceptable noisy. I spent about a month to get the 265 in a 2U chassis almost noiseless  (~14dB) within an acceptable range of sustained temperature and lower the high peaks. it has now 4 NMVE Gen 5, 4 SATA III 2.5"  and 2 3.5" SATA III drives in swappable modules drives (these don't impact the temperature at all and all stay pretty cool)  it's a pity there isn't such a benchmark tool as I believe it would reply help manufacturers as well as hobbyists to evaluate their design or find issues . No problem to close this issue. I hope this topics has helped someone if they experience the same symptoms,

[JohnM_Intel]

Hello Debugging,

Thank you for sharing the detailed update and for taking the time to test this on another system. I appreciate the insights you provided regarding cooling performance, airflow direction, fan behavior, and noise considerations. It’s clear you’ve put significant effort into optimizing the setup, especially achieving low noise levels while maintaining acceptable temperatures.

Regarding benchmarking, we’re unable to officially acknowledge or validate third-party benchmarking tools. However, if you would like to further test the processor and check for any potential issues, you may use the Intel® Processor Diagnostic Tool (IPDT).

Thank you as well for your cooperation throughout this process. I will now proceed with closing this thread.

If you need assistance in the future, please feel free to reach out.

Regards,

John M.

Intel Customer Support Technician

[Debugging]

the AsusRock W880D4U BIOS manual does not list any setting for L1,L2 etc..

https://download.asrock.com/Manual/W880D4U.pdf

where to set these levels ?