Question first, overclock to 5ghz. Can you fail OCCT LINKPACK and still be "stable" OCCT?
Question first, overclock to 5ghz. Can you fail OCCT LINKPACK and still be "stable" OCCT?
hey guys, i just completed my first overclock using the instructions from a der8auer video for my i7700k and asus apex ix. motherboard (i also have a kraken x62 aio cooler). so far it’s been stable, fluctuating between 4.7-5ghz (mostly around 4.7 unless it gets stressed), at normal idle speeds. when running OCCT, the max temperature across all cores stays at 67c except the first core which briefly reached 81c. no errors after 15 minutes. same results with firestrike extreme, userbenchmark, and intel diagnostic tool. gaming in rage 2 keeps it closer to 4.7 with occasional jumps to 5ghz, but still around 65c. if i run the OCCT linkpack test, it instantly spikes to 90c and stops. i haven’t tried prime95 yet because i thought it would be even more demanding. is the linkpack just for extreme overclocking or something else? i’m nervous about stability but new to this. i’m worried it might hit 90c, shut down, or worse fail.
Gamers are extremely upset due to sudden temperature increases in Intel Core i7-7700 chips. These powerful processors often exceed their thermal limits unexpectedly. Check the forum discussion for details.
True. Prime95 was once the go-to tool for stress testing, when CPUs were simpler and less powerful than they are now, making P95 less capable in that era.
Nevertheless, Prime95 small fft remains the top choice for temperature checks, since it applies a full load using identical instruction sets used by games, unlike Aida 64, IBT, or OCCT, which incorporate parts of linpack to cause the CPU to fluctuate between 80% and 130% loads without maintaining consistency.
This approach suits stress testing better than gaming scenarios or heavy tasks like compiling or rendering.
Linpack performs well in stress tests due to its intense workload, but temperatures should be disregarded because of its extreme nature.
Performance will generally exceed any gaming situation or demanding operations, yet it won’t reach the 130% load levels typical of Prime95 small fft runs.
Between these two, stability and temperature now differ significantly—running the CPU at around 100% load during stress tests is uncommon, and Prime95 small fft sessions are unlikely to hit that level.
The only concern with Prime95 is its handling of AVX technologies; games rarely use them, and there’s no evidence of AVX2 or AVX-512 support, so disabling these features provides a more accurate baseline.
The default setting of the OCCT link pack shuts down when it reaches 90c. I'm worried about disabling it, so I'm unsure how close to 100c I get. When you're trying the avx offset, do you mean reducing it? I've set it to 3, but this is my first overclock and I'm really new to it. I followed every step in the guide except for RAM and RAID.
Hey, thanks for the reply. That’s why I considered dropping it since it doesn’t look like anything I can reproduce with the link pack test. Everything else, including OCCT regular, tops out at 81c and that’s a big jump. Just like I told the other person, I’d be lying if I said I fully grasped the avx offset. I just uploaded the video and put it to 3. Maybe I should change it?
Prime95 was leading to Intel CPUs overheating and Haswell chips being believed to lose performance. That's why some users switched to Prime95 versions without avx, and the avx variant also caused overheating problems. Instability from overheating means a stable system should handle any workload.
Running Prime95 for 10 hours of 8k FFT with a maximum of 10 hours, if no errors occur, usually indicates stability. However, you might still face reboots or temperature issues. Testing PCIe and other components is important. You can pass Prime95 and AIDA64 but may still crash in games.
On my old 4.5GHz 4930k (x79), the vcore needed increasing with each new version. I’d pass most programs except Prime95, as skipping it could cause problems later. See the video encoding thread for more details.
Some Intel CPUs failed Prime95 at stock due to issues with certain versions.
You can find more info here: https://www.overclockers.com/forums/show...-Stability
Not using Prime95 is tied to a warning from an ASUS engineer about stability in stress tests. According to the wiki, it’s crucial to select testing software wisely to avoid problems: Prime95 can make systems stable in other tests but may trigger errors with HEVC. HEVC can be as hot as Prime95 with small FFTs and uses AVX instructions. AIDA64 would work but would fail with Prime95. AIDA64 also supports AVX and runs cooler, though it lacks the full features of Prime95. If you could rely only on AIDA64, I’d assume a 4.5GHz all-core 3800x setup is possible.
The way you stress test matters, but aim for no issues with any program—especially temperature-related ones.
Intel understands how intense AVX instruction use can become on a CPU. Most motherboard makers share this concern too. Therefore, newer BIOS versions added an AVX offset to address the issue. This adjustment reduces your clock speeds by a fixed amount to counteract the extra power consumption and heat generated. With an offset of 3, it lowers clocks by 300MHz and adjusts voltages each time AVX is used. Some users find this helpful, while others believe it disrupts their overclocking stability and choose to disable it, only managing short-term temperature spikes due to limited usage. My main recommendation is to review your BIOS settings, understand what they do, study OC techniques, and read discussions about unclear options.
for
<Mod Edit>
and giggles I improved stock speeds and executed OCCT LINKPACK tests on stock, achieving a max temp of 79c. to me that's quite achievable to reach upper 90s at 5ghz, considering my aio cooler and case fans? possibly I just really missed the silicon lottery?
Getting high temps is part of overclocking and having very high temps in some stress tests is normal. So long as the temps stay within safe margins then you don't worry. If you were hitting 100c and thermal throttle then dial it back, sometimes avx offsets can help or if you can reduce vcore. If the vrms overheat and the board shuts down then you really have to dial it back.