X3470 thermal throttling issues
X3470 thermal throttling issues
Used MX4 thermal paste with a pea-sized amount in the center. Removed the heatsink, and the paste spread evenly. Reapplied to check for cooler temperatures, but results were unchanged.
The VRM transforms the 12V supply from the power source into a range of 1.5V to 1V for the processor. In the diagram below, some components are shown with highlighted chips. The process isn't perfectly efficient, causing energy loss mainly as heat. For a processor with a 95W power draw, the VRM might approach its maximum capacity, leading to significant heat generation. Typically, a standard CPU cooler directs air downward into the heatsink, through fins along the sides of the CPU socket, and over the chips in the highlighted areas. With a typical tower cooler, the fan sits a few centimeters above the motherboard and blows air only outward, directly toward the fan that expels it through the case. This results in limited downward airflow onto the chips in those orange sections, causing them to heat up more. These components are built to handle temperatures up to 125-150°C, but the BIOS keeps an eye on temperature, usually shutting down the system if it exceeds 100°C for safety. Long-term exposure to such high temps can damage the circuit board, especially the fiberglass solder, which may degrade over time. The BIOS might reduce CPU speed to lower power consumption, allowing the chips to cool slightly and prevent overheating. This is just a likely explanation, not absolute proof. It could be that the motherboard feels underpowered, or it's a cost-saving design with minimal cooling features (just two RAM sticks, PCI slots, IDE/floppy connectors, electrolytic capacitors, 4-pin CPU port, no heatsinks on the VRM, missing IO shield connectors). For testing, you could remove the fan from your case and place it outside, directing airflow onto the top corner of the board with orange squares. This extra cooling might lower the chip temperature and confirm if that's the cause.
I chose a P55 motherboard over the H55 because I believe the P series offers better performance. I’ll return the one I received thanks to eBay’s flexible return policy. I’m planning to test the new model to compare.
It might or might not be possible to find P55 motherboards that are just as powerful as others, especially with weak VRM or other issues. If you have multiple options, please list the models and I can help you decide which one looks more promising.
It felt odd that a stock CPU without overclocking would slow down. Performance was actually solid with an RX 570 when it wasn’t throttling. In Apex Legends I achieved 70-100 FPS on high settings. After roughly 20 minutes of play, it would occasionally drop to 1.2GHz and then the frame rate would fall to 25-35 FPS.
Installed my P55 motherboard and everything functions properly. Replaced the CPU cooler with a standard model but upgraded its performance significantly—quieter yet more robust. Thermal throttling is absent. Stress tests show the CPU reaches its peak temperature near 83°C.
I understand this thread is quite old. I recently acquired my own x3470 and built a board... P7H55-M LX plus The 4c8t lynnfield 95w 2.66-3.2 Base/Boost Xeon. The system runs on the latest version 406. It isn’t fully supported, but it’s functional. CPUZ indicates maxMulti as x27, whereas the motherboard lists my CPU’s maximum as x24. I had to manually adjust settings because it was overclocking or the sensors reported 3.9Ghz-4Ghz while a fixed setting of 3.58Ghz was applied. Under load, boosts occasionally reached 25.5x and 27x randomly. Because the board isn’t fully supported, I locked it at x24 and focused on raising the BCLK. The VDROOP is noticeable here. There’s no loadline. I set BCLK to 149; anything above 150 drops the Vcore setting instantly without a reboot—just jump to 150. That forced me to reach a full manual 1.25v max OC, which peaked at 3.58Mhz even when idle. Stress tests showed the board handled mid-40s temps well. With a 240mm 1100rpm AIO and 58-62°C clocks lasting four minutes, I experienced brief 1.2Ghz spikes and then reset. Airflow around the VRMs wasn’t optimal—I was outside a case and in a testbench. I mounted a 120mm 1300rpm fan near the VRM area. No throttling occurred. The board ran at 3580Mhz for hours, maintaining stable performance on idle. My minimum clock in AIDA64 and Cinebench looptests stayed solid. For these budget boards (especially with an AIO), proper VRM airflow is essential.