- Joined
- Apr 26, 2009
- Location
- Athens, Georgia
I have recently taken up the ddr3 and am3 route and purchased a new mb to go with my 720 while all the while eyeing a quad core whenever they substantially come down in price or a new revision is released that can doo really high memory clocks.
I decided to go my standard prime 95 blend route and ran it the past weekend about 10 consecutive times until I ended up getting a 3610mhz at 1.45v stable for 16 hours. Memory and NB-CPU were pretty maxxed out also (see sig.). I also ran about 50 passes linx max memory for a second opinion. Remembering all the times I had random freezes in my pass thru of fallout 3 a few months back, I figured I might as well get a 3 opinion before I called it a day and started gaming again. I ended up remembering a website that gave a overclock synopsis and the reviewer said he ran system for 12+ hours to stress the cpu and memory in prime (I assume blend was what he was using because he said both memory and cpu), plus ran a 2 hour torture test of prime + furmark.
Considering that to be the ultimate system test that would bring my system to knees and show any instability I gave it a try. 5 times it failed at 896k in prime after about 5-25 min, each time as I loosened timings it ran a little longer. I finally played around and upped chipset voltage a bit and It ran for about an hour and a half and errored out at 768k. I then tweaked timings a little bit more (each time I say tweaked timings It was making small steps to subtimings) and It finally ran for 2 hours and 2 min.
This definitely passed the 2 hour mark, but not by much. I have decided to go to pure multi overclocking raised cpu-nb to 2600 and reduced memory down to a standard 1333mhz and tightened back timings a bit (yielding end result of about 50 less on everest memory bandwidth score and a latency of about .4 ns more (less than about 1 percent loss). Question is....On something as stressful as this, how long is enough. should I have called it a day at the settings that errored out at 2 hours 2 min and just left it there, or should any error no matter the amount of time indication of a problem and best tweaked or overclocked reduced till no errors. I feel that maybe inherently running prime + furmark together is highly prone to fail no matter what, but I would like second opinion as to thoughts about this test method and when should I call it quits If I want a system that will never be problematic in gaming but am not going to run any FOH or SETI.
Note: Considering the settings that errored out the quickest in prime with furmark running were the ones that ran 16+ hours blend by itself, and the only thing I have changed is upping chipset voltage by .02 (to 1.26v) and loosened timings from 6-6-6-18-30 with trfc at 110ns to 6-6-6-20-33 trfc 180ns, this too should pass a long time run of prime by itself because it is reducing memory oc by loosening timings. I know this is not a memory thread but everyone who replied to a similar post of mine in the memory section seemed to be on intel platform and I want opinions of people using similar equipment.
I decided to go my standard prime 95 blend route and ran it the past weekend about 10 consecutive times until I ended up getting a 3610mhz at 1.45v stable for 16 hours. Memory and NB-CPU were pretty maxxed out also (see sig.). I also ran about 50 passes linx max memory for a second opinion. Remembering all the times I had random freezes in my pass thru of fallout 3 a few months back, I figured I might as well get a 3 opinion before I called it a day and started gaming again. I ended up remembering a website that gave a overclock synopsis and the reviewer said he ran system for 12+ hours to stress the cpu and memory in prime (I assume blend was what he was using because he said both memory and cpu), plus ran a 2 hour torture test of prime + furmark.
Considering that to be the ultimate system test that would bring my system to knees and show any instability I gave it a try. 5 times it failed at 896k in prime after about 5-25 min, each time as I loosened timings it ran a little longer. I finally played around and upped chipset voltage a bit and It ran for about an hour and a half and errored out at 768k. I then tweaked timings a little bit more (each time I say tweaked timings It was making small steps to subtimings) and It finally ran for 2 hours and 2 min.
This definitely passed the 2 hour mark, but not by much. I have decided to go to pure multi overclocking raised cpu-nb to 2600 and reduced memory down to a standard 1333mhz and tightened back timings a bit (yielding end result of about 50 less on everest memory bandwidth score and a latency of about .4 ns more (less than about 1 percent loss). Question is....On something as stressful as this, how long is enough. should I have called it a day at the settings that errored out at 2 hours 2 min and just left it there, or should any error no matter the amount of time indication of a problem and best tweaked or overclocked reduced till no errors. I feel that maybe inherently running prime + furmark together is highly prone to fail no matter what, but I would like second opinion as to thoughts about this test method and when should I call it quits If I want a system that will never be problematic in gaming but am not going to run any FOH or SETI.
Note: Considering the settings that errored out the quickest in prime with furmark running were the ones that ran 16+ hours blend by itself, and the only thing I have changed is upping chipset voltage by .02 (to 1.26v) and loosened timings from 6-6-6-18-30 with trfc at 110ns to 6-6-6-20-33 trfc 180ns, this too should pass a long time run of prime by itself because it is reducing memory oc by loosening timings. I know this is not a memory thread but everyone who replied to a similar post of mine in the memory section seemed to be on intel platform and I want opinions of people using similar equipment.
Last edited:
