What to do when machine fails a Prime95 stress test?

Question

In an effort to diagnose the cause of random resets and freezes I have been running a series of stress tests on my two year old computer. Thus far the machine has passed nine hours of the Prime95 Small FFT test, thirty minutes of the standard burn-in FurMark GPU stress test and thirteen hours (eleven passes) of MemTest86+. The random resets and freezing have been experienced under clean installs of both Windows 7 Ultimate and Windows 8 Pro. According to SpeedFan all case, GPU and CPU temperatures report within normal ranges.

Nevertheless, my computer fails both the Prime95 In-place large FFTs and the Blend stress tests. Sometimes the machine resets immediately. While other times it will run up to ten minutes. Regardless, it fails and either resets or freezes when running these tests.

These are my machine's specs:

• OS: Windows 8 Pro
• CPU: Intel Core i5-2500k Sandy Bridge 3.3GHz (3.7GHz Turbo Boost)
• Motherboard: GIGABYTE GA-P67A-UD3-B3 LGA 1155 Intel P67 SATA 6Gb/s
• GPU: XFX HD-577A-ZNFC Radeon HD 5770 1GB 128-bit GDDR5 PCI Express 2.1 x16
• RAM: G.SKILL Ripjaws X Series 8GB (2 x 4GB) 240-Pin DDR3 SDRAM DDR3 1600
• PSU: Antec EarthWatts EA750 750W Continuous Power ATX12V version 2.3
• Primary HD: OCZ Vertex 2 OCZSSD2-2VTXE120G 2.5" 115GB SATA II Solid State Drive (SSD)
• Data HD: Western Digital Black WD1002FAEX 1TB 7200 RPM SATA 6.0Gb/s Internal
• Case: Fractal Design Define R3 Black ATX Mid Tower Silent PC Computer Case

I have never overclocked the machine. The BIOS is set to the default configuration which results in Auto being set for most options.

This is the machine's BIOS configuration (screenshots):

BIOS

• BIOS Version: U1d (this is a UEFI BIOS though the issue was experienced under the original BIOS)

CPU

• CPU Clock Ratio: 33
• CPU Frequency: 3.30GHz
• Internal CPU PLL Overvoltage

Memory

• Extreme Memory Profiler: disabled
• System Memory Profiler: 16.00
• Memory Frequency: 1600MHz
• Performance Enhacne: Turbo
• Voltage: 1.5V
• Profile VTT Voltage: 1.05V
• Timing: 9-9-9-28 Newegg lists timing as 9-9-9-24 fails the same regardless of setting

Voltage Settings

• CPU Vcore: 1.85V
• PCH Core: 1.050V
• PCH Core: 1.050V
• CPU PLL: 1.800V
• System Agent Voltage: 0.920V
• DRAM Voltage: 1.500V
• DRAM Voltage Reference: 0.750V
• DRAM Termination: 0.750V
• Data Reference (CH A): 0.750V
• Address Reference (CH A): 0.750V
• Data Reference (CH B): 0.750V
• Address Reference (CH B): 0.750V

While the failure has given me hope that I have begun to isolate the issue I am still unclear as to what to test from here. Given recent reviews of the motherboard I am wondering if I just have a bad board and need to replace it. Could the processor actually be bad? Are there further tests I should run? What's my next step when my computer fails a Prime95 large FFT stress test?

Answer 1

Ultimately, the issue was bad memory. I finally determined that I was dealing with faulty memory by placing the DIMMs in a different motherboard and running the Prime95 In-place large FFTs and the Blend stress tests which resulted in the same system faults.

When I provided this information to G.SKILL they honored their warranty and sent me new modules.

Answer 2

I see no issues with BIOS settings or running your memory at rated specs. However, assuming your BIOS temp readings are accurate you likely have thermal issues. 61°C at idle in BIOS is extremely high for optimized defaults. Even your ambient of 48°C seems high. Likely ,your cooling in general needs upgrading. For sure , the heat sink needs upgrade or a remount at least and its possible that your case needs additional fans or cleaning those dust bunnies.

Large fft stress testing produces the most heat and will reveal thermal issues and has no doubt happened to you.

Answer 3

There is an answer on a related question BSODs and Prime95 failures where a user, ryantmer, experiencing the same issues increased the RAM voltage from the AUTO setting of 1.5V up to 1.65V and it solved the BSOD issues for them.

You don't have the same RAM but I think it's worth a shot:

I have the exact same RAM kit, and the problem is that it will automatically set itself at 1.5V, whereas it's made to run at 1.65V. Change this in your BIOS, and you should be golden! (Ha.)

Also try searching through the feedback on newegg for related voltage issues. Most people don't have any problems but you'll find some good suggestions in the reviews there:

Other Thoughts: I'm writing this for people who are experiencing random BSODs with this ram in dual channel mode. It seems this RAM, or maybe my motherboard or CPU (i5 2500K), really does not like even as little as ~.05 VCCSA/IO overvoltage. The motherboard defaults to 1.058v VCC on stock settings and this turned out to be too much! The only thing that made the BSODs stop was reducing the voltage down from default to 1.01v. That's contrary to most of the advice about memory errors, which suggests increasing the VCC voltage. Wish I had known this earlier! Hope this helps someone!

And even manufacturer responses:

Manufacturer Response: . Dear Customer

Thank you for your G.Skill purchase and review. Glad to hear the memory is working well in your computer. With an Intel i3 CPU, it does not fully support DDR3-1600 CL9, but you can still utilize the performance of the DRAM by lowering frequency and timings. This high performance memory is unlike standard plug and play memory; input DDR3-1333 8-8-8-24 1.50V in BIOS, and it will perform quicker. With a slightly higher DRAM voltage, it may be capable of even better. This is the tweaking ability of good G.Skill RAM. Should you have any further questions or concerns, please feel free to contact us directly.

Thank you GSKILL SUPPORT

Quality and customer service are our top priorities.

Tech Support Email: [e-mail] RMA Dept Email: [e-mail] G.Skill Forum: http://www.gskill.us/forum/