• Happy holidays, folks! Thanks to each and every one of you for being part of the Tom's Hardware community!

[SOLVED] i7 7700K OC, 4.8 GHz @ 1.350 Volts ISSUE

Aug 6, 2019
27
0
30
So, I just overclocked my 7700K to the amounts specified in the title and when I run Prime95 I get a FATAL ERROR: Rounding was 0.5, expected less than 0.4. The next message is Hardware failure detected, check stress.txt. Please help.
 
Solution
... prime95 v29.8, build 3 ... "Blend" test on all 8 threads. I gave up on trying Prime95 ... 4.6GHz at 1.265 Volts. Above that, the CPU hits 90 ... Fractal Design Kelvin T12 AiO.

It's highly likely that your unfavorable experience with Prime95 was due to the "AVX" issue, which trips up many users who are caught unaware, so don't feel bad. Although many users that mention Prime95 on various forums are aware of the AVX issue, most assume everyone else is also aware. Consequently, they fail to be specific about which version, which test, and of course, the all-important detail concerning AVX.

The short answer is don't use AVX, but that doesn't say why, so here's a clear and complete explanation ...

“Stress” tests vary...
Have you researched how to OC, how to stress test and how to dial in the OC?
Indeed I have. I read some independent articles, I watched some YT videos and read an overclocking guide for my motherboard (Gigabyte Z270X Gaming K5) that was made by Gigabyte. I adjusted all the settings and Prime95 still shows the error. I tried it with a few other CPU burning tools and every other worked fine. I currently toned down the OC to 4.5GHz @1.265 Volts and one of the workers (Worker #8) still errors out in Prime95.
 
I am using prime95 v29.8, build 3 and am running the "Blend" test on all 8 threads. The ambient temperature is about 23 Degrees Celsius. I gave up on trying Prime95 but with regular testing programs like Aida64 and FurMark CPU Burner, the best OC I can get is 4.6GHz at 1.265 Volts. Above that, the CPU hits 90 degrees. I am using a Fractal Design Kelvin T12 AiO.
 
... prime95 v29.8, build 3 ... "Blend" test on all 8 threads. I gave up on trying Prime95 ... 4.6GHz at 1.265 Volts. Above that, the CPU hits 90 ... Fractal Design Kelvin T12 AiO.

It's highly likely that your unfavorable experience with Prime95 was due to the "AVX" issue, which trips up many users who are caught unaware, so don't feel bad. Although many users that mention Prime95 on various forums are aware of the AVX issue, most assume everyone else is also aware. Consequently, they fail to be specific about which version, which test, and of course, the all-important detail concerning AVX.

The short answer is don't use AVX, but that doesn't say why, so here's a clear and complete explanation ...

“Stress” tests vary widely and can be characterized into two categories; stability tests which are fluctuating workloads, and thermal tests which are steady workloads. Prime95 v29.8 Small FFT's (all AVX selections disabled) is ideally suited for testing thermal performance, because it conforms to Intel's Datasheets as a steady 100% workload with steady Core temperatures. No other utility can so closely replicate Intel's thermal test workload.



Utilities that don't overload or underload your processor will give you a valid thermal baseline. Here’s a comparison of utilities grouped as thermal and stability tests according to % of TDP, averaged across six processor Generations at stock settings rounded to the nearest 5%:



Although these tests range from 70% to 130% TDP workload, Windows Task Manager interprets every test as 100% CPU Utilization, which is processor resource activity, not actual workload. Core temperatures respond directly to Power consumption (Watts), which is driven by workload. Prime95 v29.8 Small FFT’s (all AVX selections disabled) provides a steady 100% workload, even when TDP is exceeded by overclocking. If Core temperatures don't exceed 85°C, your CPU should run the most demanding real-world workloads without overheating.

AVX - Advanced Vector Extension (AVX) Instruction Sets were introduced with Core i 2nd Generation CPU’s, then AVX2 with 4th Generation and AVX512 with later Generations of High End Desktop (HEDT) CPU’s as in certain X-Series, Extreme, i9’s and i7’s. Running versions of Prime95 with AVX enabled imposes an unrealistic 130% workload which can adversely affect stability and severely overload your CPU. 2nd and 3rd Generations are less affected, but Core temperatures on 4th through 9th Generations may be over 20°C higher.

Many 6th through 9th Generation motherboards address the AVX problem by providing “offset” adjustments (downclock) in BIOS. -3 (300 MHz) or more may be needed to limit Core temperatures to 85°C. Since 4th and 5th Generations don’t have AVX offsets, you can create individual BIOS Profiles for AVX and non-AVX software. Except for a few utilities and specialized computational apps, real-world apps (such as rendering / transcoding) with AVX and recent games with AVX shouldn’t exceed Prime95's test workload without AVX.

As per Intel’s Datasheets, TDP and Thermal Specifications are validated “without AVX. In Prime95 versions from 27.7 through 29.4, AVX can be disabled by inserting CpuSupportsAVX=0 into the local.txt file, which appears in Prime95's folder after the first run. However, since Core temperatures will be the same as 29.8 without AVX, it's easier to just use 29.8. You can also use 26.6 which doesn't have AVX.

Shown below from left to right: Small FFT's and Blend (no AVX), as well as Linpack and IntelBurn Test on a 7700K.



Note the steady thermal signature of Small FFT's, which allows accurate measurements of Core temperatures. A steady 100% workload is key for thermal testing so the CPU, cooler, socket, motherboard and voltage regulators can thermally stabilize.

You might want to give this a read, with emphasis on Section 11 - Thermal Test Basics: Intel Temperature Guide - https://forums.tomshardware.com/threads/intel-temperature-guide.1488337/

Still Sceptical? Give Prime95 another try, but this time use Small FFT's with all AVX test selections disabled. I'm quite confident that your results will be much more palatable.

CT :sol:
 
Solution
It's highly likely that your unfavorable experience with Prime95 was due to the "AVX" issue, which trips up many users who are caught unaware, so don't feel bad. Although many users that mention Prime95 on various forums are aware of the AVX issue, most assume everyone else is also aware. Consequently, they fail to be specific about which version, which test, and of course, the all-important detail concerning AVX.

The short answer is don't use AVX, but that doesn't say why, so here's a clear and complete explanation ...

“Stress” tests vary widely and can be characterized into two categories; stability tests which are fluctuating workloads, and thermal tests which are steady workloads. Prime95 v29.8 Small FFT's (all AVX selections disabled) is ideally suited for testing thermal performance, because it conforms to Intel's Datasheets as a steady 100% workload with steady Core temperatures. No other utility can so closely replicate Intel's thermal test workload.



Utilities that don't overload or underload your processor will give you a valid thermal baseline. Here’s a comparison of utilities grouped as thermal and stability tests according to % of TDP, averaged across six processor Generations at stock settings rounded to the nearest 5%:



Although these tests range from 70% to 130% TDP workload, Windows Task Manager interprets every test as 100% CPU Utilization, which is processor resource activity, not actual workload. Core temperatures respond directly to Power consumption (Watts), which is driven by workload. Prime95 v29.8 Small FFT’s (all AVX selections disabled) provides a steady 100% workload, even when TDP is exceeded by overclocking. If Core temperatures don't exceed 85°C, your CPU should run the most demanding real-world workloads without overheating.

AVX - Advanced Vector Extension (AVX) Instruction Sets were introduced with Core i 2nd Generation CPU’s, then AVX2 with 4th Generation and AVX512 with later Generations of High End Desktop (HEDT) CPU’s as in certain X-Series, Extreme, i9’s and i7’s. Running versions of Prime95 with AVX enabled imposes an unrealistic 130% workload which can adversely affect stability and severely overload your CPU. 2nd and 3rd Generations are less affected, but Core temperatures on 4th through 9th Generations may be over 20°C higher.

Many 6th through 9th Generation motherboards address the AVX problem by providing “offset” adjustments (downclock) in BIOS. -3 (300 MHz) or more may be needed to limit Core temperatures to 85°C. Since 4th and 5th Generations don’t have AVX offsets, you can create individual BIOS Profiles for AVX and non-AVX software. Except for a few utilities and specialized computational apps, real-world apps (such as rendering / transcoding) with AVX and recent games with AVX shouldn’t exceed Prime95's test workload without AVX.

As per Intel’s Datasheets, TDP and Thermal Specifications are validated “without AVX. In Prime95 versions from 27.7 through 29.4, AVX can be disabled by inserting CpuSupportsAVX=0 into the local.txt file, which appears in Prime95's folder after the first run. However, since Core temperatures will be the same as 29.8 without AVX, it's easier to just use 29.8. You can also use 26.6 which doesn't have AVX.

Shown below from left to right: Small FFT's and Blend (no AVX), as well as Linpack and IntelBurn Test on a 7700K.



Note the steady thermal signature of Small FFT's, which allows accurate measurements of Core temperatures. A steady 100% workload is key for thermal testing so the CPU, cooler, socket, motherboard and voltage regulators can thermally stabilize.

You might want to give this a read, with emphasis on Section 11 - Thermal Test Basics: Intel Temperature Guide - https://forums.tomshardware.com/threads/intel-temperature-guide.1488337/

Still Sceptical? Give Prime95 another try, but this time use Small FFT's with all AVX test selections disabled. I'm quite confident that your results will be much more palatable.

CT :sol:
Wow man thanks for explaining everything with so much detail. I really appreciate it! Now the only thing I need is a better cooler and then it's time for 4.8 GHz. I literally can't thank you enough for this!
 
There are 3 choices:

(1) Upgrade to big air or at least a 240mm AIO.
(2) Delid the CPU yourself, or send it to Silicon Lottery for worry-free delidding.
... OR ...
(3) If you want every bit of performance your CPU is capable of delivering, then do BOTH.

Your biggest bang-for-the-buck is delidding, which including shipping is about $50.

Silicon Lottery is a reputable company that does a professional job with quick turn-around time and a warranty. At 100% workload, delidding will drop your Core temperatures by about 18°C, which means at your current overclock, it would drop to less than 70°C, and even lower with better cooling. This would give you the overclocking headroom I think you're looking for.

However ... to put it all into perspective, you need to ask yourself, is 200MHz worth about $150 for shipping, delidding and a better cooler? The difference between 4.6 and 4.8 is less than 4.4%, which is a meager increase in the big scheme of overall system performance. In terms of an increase in gaming performance at 1080p, aside from benchmarks, it would most likely be difficult to notice. Give it some thought.

CT :sol:
 
There are 3 choices:

(1) Upgrade to big air or at least a 240mm AIO.
(2) Delid the CPU yourself, or send it to Silicon Lottery for worry-free delidding.
... OR ...
(3) If you want every bit of performance your CPU is capable of delivering, then do BOTH.

Your biggest bang-for-the-buck is delidding, which including shipping is about $50.

Silicon Lottery is a reputable company that does a professional job with quick turn-around time and a warranty. At 100% workload, delidding will drop your Core temperatures by about 18°C, which means at your current overclock, it would drop to less than 70°C, and even lower with better cooling. This would give you the overclocking headroom I think you're looking for.

However ... to put it all into perspective, you need to ask yourself, is 200MHz worth about $150 for shipping, delidding and a better cooler? The difference between 4.6 and 4.8 is less than 4.4%, which is a meager increase in the big scheme of overall system performance. In terms of an increase in gaming performance at 1080p, aside from benchmarks, it would most likely be difficult to notice. Give it some thought.

CT :sol:
Yeah, about delidding, I am from Southeastern Europe so shipping would kill me, and also, i don't wanna risk my $400 chip. And TBH something is wrong with my cooler so I will replace the tim as soon as possible. For now, the next upgrade i am looking at is a new GPU.