Is prime95 harmful to the Haswell-E processor?
Is prime95 harmful to the Haswell-E processor?
Some people think prime95 isn't the best tool for stress testing haswell-e. With a 5960x CPU and BIOS OC tweaker set to 4.4 GHz, I ran Cinebench and used Hwmonitor to check temperatures. The results showed what they expected. Voltages seem okay at that frequency. If you want to test processor stability or performance limits, there are other options to consider.
Yes, it works with any brand motherboard as per ASUS guidelines. I question if you might have caused any harm during a 17-hour session at 4 GHz. What voltage did you configure in your BIOS/UEFI? Do your CPU voltage levels fluctuate when you monitor them? I’m not well versed with Asrock boards, so I can’t confirm how the OC tweaker modifies settings. From what I understand, if your CPU speed shifts from idle to your set OC of 4.4 and your voltage adjusts accordingly, you’re likely using adaptive voltage control. If you’re using manual voltage, such as 1.3V, it will remain constant regardless of CPU speed. With adaptive at 1.3, it would stay around 0.75 during idle and rise to 1.3 under load. The "spikes" mentioned are actually intentional. When certain...
I started with ASUS Realbench. Initially, Cinebench performed well, but when I tried Realbench it triggered a BSOD and I had to reduce the overclock to pass the benchmark or stress test. I personally used a 5820k processor, as recommended by ASUS forum members for Haswell-E. With air cooling and a Noctua NH-U14S fan, I reached 91°C on the hottest core during the stresstest—this only lasted until I integrated it into my waterloop setup. The latest Prime model seems to draw more voltage (especially with adaptive voltage) and pushes high power through the VRMs when overclocked due to AVX instructions. This could potentially harm the CPU (according to ASUS). Here’s a link if you’re interested: http://rog.asus.com/365052014/overclocki...30k-5820k/ BTW, nice Cinebench score, got around 1303 at 4.4Ghz.
ASUS Realbench was effective with my 5820k processor. It helped me reach 91°C on the hottest core during a stress test, which lasted only briefly after integrating into my waterloop setup. The latest Prime version seems to draw more voltage (especially with adaptive voltage) and higher current through the VRMs when overclocked due to AVX instructions. According to ASUS, this could potentially harm the CPU. For more details, check this link: http://rog.asus.com/365052014/overclocki...30k-5820k/. Also, your Cinebench score was around 1303 at 4.4Ghz. My ASRock X99 Extreme 4 motherboard supports ASUS Realbench, but I’m unsure if adaptive voltage is active. The damage appeared gradually—after running Prime95 at 4Ghz for several hours, the power supply cut off unexpectedly.
Yes, it works with any brand motherboard according to ASUS. I question whether you caused any harm during a 17-hour session at 4 GHz. What voltage did you configure in your BIOS/UEFI? Does your CPU voltage fluctuate if you're tracking it? I'm not knowledgeable about Asrock boards, so I can't determine how the OC tweaker modifies settings. From what I understand, if your CPU speed shifts from idle to your OC target of 4.4 and your voltage adjusts accordingly, you're using adaptive voltage. If you're using manual voltage, say 1.3V, it will remain constant regardless of CPU speed. With adaptive at 1.3, it should stay around 0.75 during idle and rise to 1.3 under load. The "spikes" mentioned are intentional. When specific instructions are sent to the CPU (such as AVX), the fixed voltage can be bypassed. If you're running manual voltage at 1.3, even with AVX, it won't exceed that level but might affect your OC stability. On the other hand, adaptive voltage under similar conditions could briefly push it up to 1.4 for a brief moment. I verified this behavior myself while testing my 5820k. Realbench will exhibit the "spike" in certain games, including Metro LL and Battlefield 4, but not in Battlefield 4. You can monitor voltages and temperatures using HWMonitor Pro Trial edition.
Here are the important points rephrased while keeping the original meaning and tone: Two key aspects remain unaddressed: (1) What is your current ambient temperature? (2) Prime95 version 28.5 isn't recommended for 22 nanometer processors. Furthermore, the link mentioned earlier was helpful because it provided more details—Asus didn’t go into detail about which Prime95 versions are inappropriate for stress testing. They only noted excessive power consumption, which leads to unrealistically high core temperatures, but didn’t clarify the reason, which is related to AVX code. This issue is covered in the following Tom’s Sticky: Intel Temperature Guide - Remember that the guide references Prime95 version 26.6, which is the latest pre-AVX release. Small FFTs are used for a 10-minute thermal assessment only, not for stability checks. "... Section 12 - Thermal Testing @ 100% Workload Prime95 Small FFT's are commonly used for CPU thermal evaluation because they represent a stable 100% workload. This is the benchmark Real Temp employs to verify sensors. The link provided corresponds to version 26.6, ideal for all Core 2 and Core i models. CPU models from the 2nd, 3rd, and 4th generation support AVX instructions. Newer Prime95 versions execute AVX code on the FPU math unit, which can cause excessively high temperatures. AIDA64’s FPU test confirms these findings. Prime95 v26.6 generates higher temperatures on 3rd and 4th Gen processors compared to 2nd Gen ones, despite both supporting AVX. This is due to the soldered Integrated Heat Spreader and a larger die size in the latter. Prime95’s default benchmark, Blend, is designed for memory stability testing, while Large FFT combines CPU and memory assessments. Both cyclic workloads are inappropriate for CPU thermal evaluation. Other tests like Linpack and Intel Burn Test also involve cycles that fully utilize all registers, equating to an 110% workload—making them unsuitable for CPU thermal analysis. The software tool OCCT performs tasks such as Linpack and Prime95. Displayed from left to right: Small FFT's, Blend, Linpack, and Intel Burn Test. Observe the steady-state thermal patterns of Small FFT's, which enable precise Core temperature readings. Also displayed: Small FFT's, Intel Extreme Tuning Utility CPU Test, and AIDA64 CPU Test. The charts in SpeedFan illustrate 13 minutes of data, highlighting how each test produces distinct thermal profiles. Intel Extreme Tuning Utility is also a cyclic workload. Although AIDA64’s CPU test is steady-state, the duration isn’t sufficient..."
CompuTronix: iopihop, Two key points remain unaddressed: (1) What is your ambient temperature? (2) Prime95 version 28.5 isn't recommended for 22 nanometer processors. Furthermore, the link provided by 1LiquidPC was insightful; Asus didn’t go into detail about which Prime95 versions are inappropriate for stress testing. They mentioned excessive power consumption, which leads to unrealistically high core temperatures, but didn’t clarify the root cause—AVX code. This issue is covered in Tom's Sticky: Intel Temperature Guide - Be aware that the guide references Prime95 version 26.6, the latest pre-AVX release. Small FFTs are intended for a 10-minute thermal assessment only, not for stability evaluation. "... Section 12 - Thermal Testing @ 100% Workload Prime95 Small FFT's are commonly used for CPU thermal analysis since they represent a constant 100% workload. This is the benchmark Real Temp employs to evaluate sensors. The link provided corresponds to version 26.6, which is compatible with all Core 2 and Core i models. CPU models from the 2nd, 3rd, and 4th generation include AVX instructions. Newer Prime95 versions execute AVX code on the FPU math coprocessor, resulting in excessively high temperatures. AIDA64’s FPU test confirms these findings. Prime95 v26.6 generates higher temperatures on 3rd and 4th Gen chips compared to 2nd Gen models, which also feature AVX but are less prone to thermal spikes thanks to a soldered Integrated Heat Spreader and a larger die size. Prime95’s default benchmark, Blend, applies a cyclic workload for memory stability checks, while Large FFT combines CPU and memory assessments. Both cyclic tests aren’t ideal for CPU thermal evaluation. Other stability tests like Linpack and Intel Burn Test include cycles that fully utilize all registers—effectively simulating 110% load—and are therefore unsuitable for CPU temperature analysis. The utility OCCT also runs Linpack and Prime95. Displayed from left to right: Small FFT's, Blend, Linpack, and Intel Burn Test. Observe the steady-state thermal patterns of Small FFT's, which enable precise Core temperature readings. Also displayed: Small FFT's, Intel Extreme Tuning Utility CPU Test, and AIDA64 CPU Test. The charts in SpeedFan illustrate 13 minutes of data, highlighting how each test produces distinct thermal profiles. Intel Extreme Tuning Utility itself is cyclic. Although AIDA64’s CPU test is steady-state, the duration isn’t adequate... " CT do not possess a device for measuring ambient temperature... I executed Prime95 version 26.6 as you advised; small FFT results showed these temperatures, which seem unusually high. Did I inadvertently reduce my processor lifespan? 🙁 http://i.imgur.com/i1VPenE.png After about 10 minutes it might be closer to 20 minutes; I’m hesitating and should stop now—scared. Additionally, the Blend test I ran recently at 4.0 GHz overclock used Prime95 version 28.40. http://www.guru3d.com/files-details/prim...nload.html
I don’t have the necessary equipment for measuring ambient temperature... What’s the approximate temperature in your room? Wall thermostat in the hallway?
Perhaps this will help:
Here are the temperature conversions and a brief scale:
Cx9/5+32=F ... or ... F-32/9x5=C ... or more simply ... an increase of 1C = an increase of 1.8F
30.0C = 86.0F
Hot
29.0C = 84.2F
28.0C = 82.4F
27.0C = 80.6F
26.0C = 78.8F
Warm
25.0C = 77.0F
24.0C = 75.2F
23.0C = 73.4F
22.0C = 71.6F
Standard
... or ... 22.2C = 72.0F
21.0C = 69.8F
20.0C = 68.0F
Cool
I ran Cinebench and used Hwmonitor to check the temperatures. Here’s what they showed:
http://i.imgur.com/54N7KeH.jpg
Sorry, but this image displays your processor at idle, meaning we can’t compare it directly to the next screenshot, which is Prime95 v26.6 under full load.
I also ran Prime95 26.6 as you suggested; these results seem unusually high. Did I inadvertently reduce my processor’s performance?
🙁
http://i.imgur.com/i1VPenE.png
Not to mention, 26.6 is pre-AVX and doesn’t support AVX instructions on your chip, so it’s safe. Also, your device will automatically slow down at 100°C to prevent thermal damage, which means you’re still within acceptable limits.
Hi CT,
previously you edited an excellent subject for calibrating CPU sensirs, which I successfully applied to my Core 2 Quad Q9650.
Now, I purchased a new machine with Core i7-5930K and I was wondering if you could adjust this calibration method for the latest CPU generation?
Thank you.
Running an overclock can push your CPU beyond its safe operating range, potentially causing instability. Use the instruction set Advanced Vector Extension 2 and run P95 v26.6 or earlier to prevent this. For better results, try version 27.9 to support a lighter AVX workload. Many users with unstable overclocks report damage at higher versions like 28.3 or 28.5, but the system can still reach its limits with AVX2.