maybe u should use some like TOSHIBA with LifeTime warranty...KingSton which has a few y warranty?

I'm sorry for intervening but on my PC I see continuous writing.

Through Windows task manager in Win 8.1 while the notebook (with an HDD) is crunching two CEP WUs at a time I see two instances of wcgrid_cep2_qchem_prod_win32.exe.7.00 writing to disk almost continuously at a total rate (BOINC is the only program rurring while I observe this) varying precisely every second in a range of 10 KB/s to 10 MB/s. It increases, reaches a maximum and than decreases every two seconds when the maximum is up to 2 MB/s and every six seconds when the maximum is up to 10 MB/s. My on-site settings is Write to disk at most every: 60 seconds.

Sorry, I've made a mistake. Taking a closer look I've noticed there's a few other executables (mousepad, System.exe and similar ones) that contribute in the behaviour I've seen and Q-Chem executables don't write to disk almost continuously as I first thought.
But is still true that between a write and another done by the same Q-Chem executable passes less than a minute. Since their write rate is lower than what I previously stated (and It doesn't causes me any problem because I don't use the PC while it crunches) I think there's no need to bother you with this matter. Sorry for being impetuous.

I'm aware that default setting is strongly reccomended but fortunately crunching two WUs simultaneously, even if I see some 0x3 or 0xc0000005 RC (however WUs always get validated), my fail rate is extremely low * and doing that way doesn't affect my work since I crunch when I don't use the PC. So I think it isn't a bad idea to do so; I've seen three at a time is too much for me but two fits in.
* I remeber less than five failed WUs among over two hundred completed and they all failed on all the wingmen so it wasn't due to my setting

Should I ever encounter any problem due to this, I'll revert to one at a time.

The latency collapse, as seen in the second picture, is ended an extreme point, but try it with 12 CEP2 threads and an SSD's garbage collector could very well run into such problems--especially if the BOINC Data file is on the same disk as the OS.

In the beginning, when most of the blocks are empty, the garbage collection processes has more time to do its thing, but what happens to the latency when there are no longer any free blocks and every block written to first has to be erased? This is why old SSDs become noticeably slower. Running just one CEP2 thread, like some suggest we should do, and which I disagree with since clean energy is so important, doesn't cause any latency problems.


I do have some CEP2 errors from time to time--less than 2% of all CEP2 wus--but that's a price I'm willing to pay. My processor's utilization is around 99%, which is OK.