Turbo boost on Core i5-4590T migh cause lock up?

[ELige]

I recently upgraded my home server with a Core i5-4590T processor and started to experience lock ups during heavy CPU load. My system looks stable in normal conditions, but after a few minutes of H.264 video encoding, it freezes. No kernel panic, error message or reboot, just locks up. I ruled out the obvious root causes (overheating, faulty PSU, faulty RAM modules, outdated BIOS) and started to focusing on Turbo Boost.

According to the specification of this processor, base frequency is 2.00GHz which can go up-to 3.00GHz (for a single core) with Turbo Boost. I observed, that when lock up happens, all 4 cores are running on ~2.6GHz which seems to be a bit high. What is the maximum core frequency for this CPU when all cores are utilized? Is it possible that I need to manually limit the maximum frequency to stabilize my system? Could it be a BIOS or OS issue, or I just have a faulty CPU?

I heard that 6th generation of Core CPUs suffered from a bug which caused freeze during high load (Prime95) when using AVX instructions. Could it be a similar issue with 4th Gen i5 as well?

My mainboard is an ASRock Z97M-itx/ac with the latest BIOS and I even ensured that the most recent microcode update is loaded by the OS (Debian Linux with kernel 4.9.13).

I would appreciate any comments and ideas.

Thanks and regards,

Endre

[idata]

Hello Endre,

Thank you for contacting Intel Communities.In regards to your question "What is the maximum core frequency for this CPU when all cores are utilized?"You can click http://www.intel.com/content/www/us/en/support/processors/000005647.html here to check the maximum frequencies while the Intel® Turbo Boost Technology is enabled, please expand 4th Generation Intel® Core™ i5 Desktop Processors and look for the processor you have.You can disable the Intel® Turbo Boost Technology from the BIOS (if the option is available).In regards to the issue with Prime95 and the Skylake (6th gen) processors, it is not related to any other families of processors.For troubleshooting purposes, you can test the system in a Windows environment to run the https://downloadcenter.intel.com/download/19792/Intel-Processor-Diagnostic-Tool Intel® Processor Diagnostic Tool as the Linux* version is available for Fedora only and won't necessarily run in Debian, you can download the tool http://www.tcsscreening.com/files/users/IPDT_LiveUSB/index.html here.You can also test the processor in a compatible motherboard to see if the problem persists.If you tested the RAM modules using Memtest86 or any other memory testing tool I would recommend testing with different sticks of RAM.Please let me know how it goes.Regards,Juan Carlos

[idata]

Hi Endre,

I was reviewing this thread and I wanted to know if you were able to resolve the issue?Please don't hesitate in replying to this post if you need further assistance.Regards,Juan Carlos

[ELige]

Hi Juan,

First of all, thank you for giving some hints and advises. The issue is still under investigation, but there's some progress:

• Intel Processor Diagnostic Tool gives me "pass" on each run
• Core clocks looks normal, according to Turbo Boost Technology Matrix. If all 4 cores are fully utilized, each are running at 2.6 GHz.
• System is stable even under high load. The only workload which triggers the issue (so far) is running x264 video encoder. Based on my tests, it's unpredictable when it's causing freeze. Sometimes it takes only a few seconds, another time just after about 15 minutes.
• I did some tests on my desktop PC (Z97 chipset, Xeon E3-1240 v3, 2x8GB DDR3 modules), with the same workload (same OS, kernel, application, test video file) but unable to reproduce the issue.

Since I still suspect the CPU or the mainboard, the plan is to swap the CPUs and repeat the tests on both system.

Best regards,

Endre

[idata]

Hi Endre,

Thank you for the update.I will be awaiting your findings.Best Regards,Juan Carlos

[ELige]

Hi Juan,

After many hours of testing, here are the results.

Base systems

PC # 1: ASRock Z97M-ITX/AC, BIOS v1.80 (6/29/2016), 2x4GB DDR3-1600 (Kingston HyperX Blue), no add-on VGA

PC # 2: ASRock Z87 Extreme4, BIOS v3,40 (4/28/2016), 2x8GB DDR3-1600 (Kingston HyperX Blue), NVIDIA GTX1080

SystemCPUTurbo BoostOperating systemDurationWorkloadResultPC # 1Core i5-4590TEnabledDebian, Linux 4.9.131-15 minx264FreezingPC # 1Core i5-4590TEnabledDebian, Linux 4.9.135+ daysApache httpd, Tomcat, MySQL, Samba, QEMU-KVMStablePC # 1Core i5-4590TEnabledDebian, Linux 4.4.391-15 minx264FreezingPC # 1Core i5-4590TDisabledDebian, Linux 4.9.131-15 minx264FreezingPC # 1Core i5-4590TEnabledWindows 10 Pro x64120+ minx264, AIDA64 Stress test, Intel PDTStablePC # 1Pentium G3240TN/ADebian, Linux 4.9.1360+ minx264StablePC # 1Xeon E3-1240 v3EnabledDebian, Linux 4.9.1360+ minx264StablePC # 2Xeon E3-1240 v3EnabledWindows 10 Pro x64120+ minGeneric desktop usage (Office, gaming, etc.)StablePC # 2Xeon E3-1240 v3EnabledWindows 10 Pro x6460+ minx264, AIDA64 Stress test, Intel PDTStablePC # 2Core i5-4590TEnabledDebian, Linux 4.9.131-15 minx264FreezingPC # 2Core i5-4590TDisabledDebian, Linux 4.9.131-15 minx264Freezing

Observations:

• i5-4590T is freezing under Linux, regardless Turbo Boost state and mainboard
• PC # 1 is stable under Linux with Pentium G3240T and Xeon E3-1240 v3 under the workload where 4590T is freezing
• PC # 1 is stable under Windows 10 with i5-4590T

Notes:

• Windows port of x264 is slower under Windows, however it reports to use the same CPU instruction kits
• CPUID of Xeon E3-1240 v3 and i5-4590T is the same (306C3)
• BIOS of PC # 1 and # 2 contains the same microcode update revision which is 19
• Windows 10 Pro applies microcode update rev 1E (8/13/2015)
• Debian applies microcode update rev 20 (3/16/2016), however system behaves similarly without OS-based microcode update

What's next:

• Checking and comparing "Specification Update" documents for Xeon E3-1200 v3 and Desktop 4th Generation Core processor family
• Trying to apply microcode update rev 1E under Linux
• Re-ccompile x264 binary under Linux and repeat tests in PC # 1 with i5-4590T

Any comments and suggestions are welcome. I'm open to any idea what to check.

Best regards,

Endre

[idata]

Hi Endre,

Thank you for the information.

Seems that the recommendation about testing the system in a Windows environment worked:PC # 1 Core i5-4590T Enabled Windows 10 Pro x64 120+ min x264, AIDA64 Stress test, Intel PDT StableThis is because our drivers are fully supported while using Windows 7, 8 or 10*, on the other hand, while using a different OS, you may encounter different issues.Regards,Juan Carlos

[ELige]

Hi Juan,

What drivers do you mean? HD graphics driver for Linux-based systems, or something else? I know that the best support by most of the hardware vendors is given for Windows, but did you mean that Intel does not officially support 4th Generation Core i5 under Linux? Info of the kernel module says that the driver is created by Intel:

filename: /lib/modules/4.9.0-2-amd64/kernel/drivers/gpu/drm/i915/i915.ko

license: GPL and additional rights

description: Intel Graphics

author: Intel Corporation

author: Tungsten Graphics, Inc.

Anyway, back to the original topic. Since Xeon E3-1200 v3 series and 4th Gen Core i5 are built upon the same microarchitecture (even the CPUID signature and microcode update is the same), I would say they should behave similarly under the same workload. However the fact is that Xeon is working perfectly, but i5 is failing. Let's forget the obvious performance differences (core frequency, cache size) for a while, and let's focus on the differences based on the public info found on http://ark.intel.com/compare/75055,78928 ark.intel.com:

• Vertical segment (desktop vs server) -> should not be an issue, if Linux support is the same for both
• Hyperthreading -> Xeon is stable with disabled HT
• TDP -> should not be an issue
• ECC memory support -> no ECC memory is used in any of my PCs
• Integrated graphics -> PC # 2 has a PCIe graphics adapter and i5 is also failing in that configuration
• TSX-NI -> as I know, this feature is disabled on Xeon because of a faulty implementation

It looks like the major difference is the integrated graphics controller: E3-1240 v3 does not have it. Maybe HD Graphics controller is causing the trouble even if an external VGA is used? Errata section of http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/4th-gen-core-family-desktop-specification-update.pdf Specification Update released for Desktop 4th Generation Intel Core Professor Family mentions issues, like:

• HSD77 - Graphics Processor Ration And C-State Transitions May Cause a System Hang
• HSD139 - Concurrent Core and Graphics Operation at Turbo Rations May Lead to System Hang

How can I check if my CPU is affected or not?

Note:

I tried loading different microcode updates on i5-4590T, but no change.

Best regards,

Endre

[idata]

Hi Endre,

You are correct, I meant graphics drivers, for Linux users, the support is provided through the distros' community.Based on the tests you did, you got issues in both situations (with Turbo Boost enabled and disabled). Regarding the configuration in question (with the Intel® Core™ i5-4590T Processor), the system worked while using Windows® 10, then, at this point, we can't blame the Intel® Turbo Boost Technology.In regards to the Errata, I can do further research and then get back to you.Regards,Juan Carlos

[ELige]

Hi Juan,

I agree and convinced that the issue is not related to Turbo Boost Technology, but still curious what is the root cause of these freezes. I appreciate your efforst on checking Errata, maybe you'll find a clue why a CPU is affected and another from the same generation doesn't.

Thanks again in advance!

Best regards,

Endre

[idata]

Hi Endre,

I have sent you a private message.Best Regards,Juan Carlos

[idata]

Hello Endre,

I have sent you a private message.Regards,JC