- Jun 30, 2004
- 16,652
- 2,033
- 126
Last night, I popped my Tracer-800's back in, and took out the RMA-replacement Ballistx-1000's.
So we're back to viable setting of CMD = 1T.
This time, I'm trying to tweak the voltages so that they're just adequate, trying to take the heat off the VDIMM setting. I don't want VDIMM to go higher than 2.125V.
So this thing runs PRIME95 Blend-Test for 9 hours, 25 minutes. I wake up at 5AM, to visit the LCD monitor and see what gives.
Three of four threads post "rounding error," stopping at 30 minutes before I woke up. Something like 8 hours, 55 minutes on the three dead threads.
Never saw three threads stop all at once like that. My intuition just said "Tweak the NBCore voltage up +0.05V. Now, it's headed toward a 10 hour running-time.
Why does my intuition say "NBCore?"
And since PRIME95 runs a discrete selection of Lucas-Leymer iterations, there is a non-random component to the stress-level. I'd say that's why, if the errors posted after an hour and we tweak a voltage such that the next failure takes three or four hours, we pat ourselves on the back and say "Good job" before we restart and tweak again.
But I'm wondering if there isn't a random component, which could mean that the tweak we thought had fixed things . . . . didn't really fix things. Still, three threads, all halting exactly at the same time -- that ain't random. No, sir. There's an assignable cause to its occurence, even if the time it occurred has some random distribution.
So why does my intuition say "NBCore voltage?"
So we're back to viable setting of CMD = 1T.
This time, I'm trying to tweak the voltages so that they're just adequate, trying to take the heat off the VDIMM setting. I don't want VDIMM to go higher than 2.125V.
So this thing runs PRIME95 Blend-Test for 9 hours, 25 minutes. I wake up at 5AM, to visit the LCD monitor and see what gives.
Three of four threads post "rounding error," stopping at 30 minutes before I woke up. Something like 8 hours, 55 minutes on the three dead threads.
Never saw three threads stop all at once like that. My intuition just said "Tweak the NBCore voltage up +0.05V. Now, it's headed toward a 10 hour running-time.
Why does my intuition say "NBCore?"
And since PRIME95 runs a discrete selection of Lucas-Leymer iterations, there is a non-random component to the stress-level. I'd say that's why, if the errors posted after an hour and we tweak a voltage such that the next failure takes three or four hours, we pat ourselves on the back and say "Good job" before we restart and tweak again.
But I'm wondering if there isn't a random component, which could mean that the tweak we thought had fixed things . . . . didn't really fix things. Still, three threads, all halting exactly at the same time -- that ain't random. No, sir. There's an assignable cause to its occurence, even if the time it occurred has some random distribution.
So why does my intuition say "NBCore voltage?"
Facebook
Twitter