Is OCCT Linpack stress test still reliable?

[sbakic]

Hi folks.

I was trying to OC my CPU and I was wondering something.

CPU: Ryzen R7 1700X
CPU Cooler: Kraken X62

 

Tests:

1. CinebenchR15 - Benchmark
2. PassMark - Benchmark
3. RealBench - Benchmark
4. RealBench - Stress test (min ~1h, max ~24h)
5. AIDA64 - Stress test (min ~1h, max ~24h)
6. prime95 - (Blend test) (min ~1h, max ~24h)
7. OCCT CPU - Stress test  (min ~1h, max ~24h)
8. OCCT Linpack- Stress test  (min ~1h, max ~24h)
9. Memtest86 - Stress test (~8h: 4 tests)
10. mprime (Linux) - Stress test (min ~1h, max ~24h)

Short story: I consider stable OC only if all benchmarks and stress tests from above are passed. On default setting all test are passed (for example OCCT Linpack was running for ~6h, and prime95 for ~24h).

Results: 40x multiplier, Vcore offset +0.01250, LLC3. -> idle Vcore 1.369V, load Vcore 1.337V, min temp 22-25, max temp 74C. All test are passed except OCCT Linpack. With Linpack I run 5 tests for 20 mins (success rate 3/5), and 4 tests for 2h (success rate 1/4, one failed at 1h:57min). The problem with this test is that at full load Vcore drops to 1.331V and temps rise to 74C, in any other test at full load Vcore is at min 1.337V with temps 72C. Even if I go higher offset +0.01785, idle Vcore is 1.375V, load Vcore 1.344V, Linpack lower the Vcore at load by 0.006V to 1.337V.

I consider this OC stable at 99,9% because of succes rate of 50% for OCCT Linpack. Is this stress test even valid or it's outdated? Remember this is the only stress test that lower Vcore a little more than any other stress test.