you guys supposedly test your code on linux machines i read somewhere. it seems that the tasks check out ok on linux. but then when they come to the windows machines they crash with this -1073741819 (0xc0000005) error which i have had 9 tasks crash in the last 24hrs in this manor.

why doesn't this show up on your linux boxes?
perhaps that person that was talking about the team getting more windows boxes should persuade the right people to start purchasing them.

this is the highest error rate i have had on my new system as well as my old one that i gave away. it is getting annoying to see my RAC drop and waste cpu time on stuff that errors out. if you think that you are getting good information out of the completed models that good. but i don't get any credit for crashed tasks. i would love to see credit granted for tasks that run at or near 50% of run time to get the claimed credit granted. then i wouldn't be so upset and feel cheated.

Huh ?

Ubuntu - thought that was Linux !

Prime95 and Memtest86+ v2.10 clean several hours run so it's NOT the machine

Stop whining about Windows, it's more than just you being affected !

---

you guys supposedly test your code on linux machines i read somewhere. it seems that the tasks check out ok on linux. but then when they come to the windows machines they crash with this -1073741819 (0xc0000005) error which i have had 9 tasks crash in the last 24hrs in this manor.

why doesn't this show up on your linux boxes?
perhaps that person that was talking about the team getting more windows boxes should persuade the right people to start purchasing them.

this is the highest error rate i have had on my new system as well as my old one that i gave away. it is getting annoying to see my RAC drop and waste cpu time on stuff that errors out. if you think that you are getting good information out of the completed models that good. but i don't get any credit for crashed tasks. i would love to see credit granted for tasks that run at or near 50% of run time to get the claimed credit granted. then i wouldn't be so upset and feel cheated.

I would guess that most testing is done on fairly bone stock machines. While you are running your machines oc'd. Oc'ing is always a quirky thing, two identical machines will not oc exactly the same, despite having exactly the same components in them. In the other thread you stated you are backing off your oc'ing just a bit and I said I thought it would fix the problem, I still do.

---

you guys supposedly test your code on linux machines i read somewhere. it seems that the tasks check out ok on linux. but then when they come to the windows machines they crash with this -1073741819 (0xc0000005) error which i have had 9 tasks crash in the last 24hrs in this manor.

why doesn't this show up on your linux boxes?
perhaps that person that was talking about the team getting more windows boxes should persuade the right people to start purchasing them.

this is the highest error rate i have had on my new system as well as my old one that i gave away. it is getting annoying to see my RAC drop and waste cpu time on stuff that errors out. if you think that you are getting good information out of the completed models that good. but i don't get any credit for crashed tasks. i would love to see credit granted for tasks that run at or near 50% of run time to get the claimed credit granted. then i wouldn't be so upset and feel cheated.

I would guess that most testing is done on fairly bone stock machines. While you are running your machines oc'd. Oc'ing is always a quirky thing, two identical machines will not oc exactly the same, despite having exactly the same components in them. In the other thread you stated you are backing off your oc'ing just a bit and I said I thought it would fix the problem, I still do.

thanks for your input.
i will let things run at the lower speed for a day and see if that stabilizes things. i was beginning to suspect it might be OC speed. just wasn't sure.