Ciao, confermo :
[*]5900X:
[*]PPT: 142 W[*]TDC: 95 A[*]EDC: 140 A
Quanto alla guida, direi ottima!
Qualche appunto :
- per terminare correttamente core cycler basta premere ctrl+c (così che dia un recap del risultato complessivo). Generalmente riesce anche a killare lo stress test.
- quanto al test "finale" - detto che la sentenza finale sulla bontà dell'oc è data dall'utilizzo giornaliero, protratto per qualche settimana- l'autore di Core Cycler si è espresso così :
Quote:
Hm. I developed the tool while testing my own setup, so basically whenever I implemented a new feature I switched to this to check if it's working. So I couldn't really establish a specific order of sequence.
However if I were to run this on a new computer,
1) I would probably first do a couple of runs with the default settings (SSE, huge FFT, 6 minutes) to find a seemingly stable setup that doesn't throw any immediate errors.
2) If this has run successfully for a couple of hours / over night without an error, I would then switch to 2 threads and repeat the test.
3) When this seems stable, it's time for the real stability test, so I would switch to 1 thread, SSE, All FFT and an automatic runtime and let it run at least one full iteration.
4) Then switch to AVX and repeat.
5) And then switch to AVX2 and repeat.
6) And then repeat this with 2 threads.
I found most of my errors with SSE and 1 thread, but AVX/AVX2 and 2 threads did find some later in the process for me as well, so for a thorough stability test I think it's necessary to include them.
Another alternative for quick stability testing would be Small FFT and 2 minutes per core. I actually found many of my errors with this setting, before I added the more advanced features in CoreCycler.
y-Cruncher and Aida64 were added relatively late in the development process, so I don't have too much experience with these. But y-Cruncher either completely crashed my PC or ran through fine during my somewhat limited testing runs (i.e. when artificially setting the Curve Optimizer to unstable values).
If a game like Doom Eternal brings quick(er) results for you, I would stick to it as much possible. Playing a game for stability testing is undeniably more fun than letting the CoreCycler run, but of course you have to approach it with the mindset of actually wanting to test the stability, and not with the expectiaton of the game running flawlessly.
|
Il vantaggio di core cycler è anche il suo limite, giacché testando per singolo core i tempi diventano biblici ...
Il pacchetto completo teorico prevederebbe su un 5900x 144h ore di test (12h di p95 per ciascun core). Abbastanza folle...
Credo che stiamo assistendo al tramonto dell'overclock toutcourt... il rapporto tempo speso / beneficio è già ora svantaggioso a mio avviso.