I am trying to achieve and stabilize my final overclock for my Athlon 64 X2 3800. I have tried a few different things to get it to be completely stable. It seems I am getting closer and closer, but just may not be quite there yet.
First of all, I have the dual core hotfix applied and the registry tweak set to 1 to enable it. I also have the latest NVIDIA 6.70 chipset drivers installed and the AMD X2 driver installed.
I have the latest 1009 BIOS for the Asus A8N-SLI Premium motherboard.
I have good airflow in my Antec P180 case. I have a Thermalright XP-120 heatsink on my CPU and cooled with a 57CFM airflow 120mm fan. I have a SeaSonic S12 600WATT PSU. I have four 512MB DIMMS of PC4400 (550MHz rated) Corsair XMS memory. DIMM voltage is set at 2.80V.
Now to overclock, I set the FSB to 250 in the BIOS with the default maximum CPU multiplier of 10. I leave the RAM index value set at 400MHz which means that the RAM will overclock at the same rate as my FSB. That is ok though because my RAM can handle speeds as fast as 275MHz (550MHz efffective DDR), and my RAM will only be running at 250MHz (500MHz effective DDR). I set the RAM timings to 2.5-3-3-6. The 1T/2T timing is set at 2T. I think my RAM should be able to run at these timings because it is rated at only slightly lower timings at 25MHz (50MHz effective DDR) faster. And since my RAM is running 25MHz slower, I would think it should be able to easily handle 2.5-3-3-6 instead of 2.5-4-4-8. The RAM is rated at 2.5-4-4-8 for running at 275MHz (550MHz DDR).
I began by upping the VCORE to my CPU little by little until dual instances Prime95 is completely stable with my CPU speed overclocked to 2.5GHz (250*10). First, I set the VCORE at 1.425, and ran dual instances of Prime95. After about three hours, one instance failed. I tried upping the VCORE to 1.4375, and after a couple of hours, one instance of Prime95 failed again. The other instance kept going strong without any failures. I upped the VCORE again to 1.450 and the same thing. Finally, I upped the VCORE to 1.50 anf ran dual instances of Prime95. Both instances ran for longer than before without any errors. But after 10 hours and 20 minutes, one instance halted because of an error. The other instance kept going. I restarted the instance that failed, and it has been successfully going for an hour since while the other instance has continued to go without failure this whole time.
Now, both instances passed the large FFTs all the way up to 4096K. After it started over its round over again on the 1024K FFT length after 10 hours and 20 minutes, that is when one instance failed. So, it seems that both instances of Prime95 passed all FFTs from 8K to 4096K. But one core couldn't keep going once it started its cycle over again back at 1024K FFT. But about 7 minutes after one instance failed and the other was still going, I decided to start the instance that failed again with the same settings. And it started right at 1024K FFT test and it passed and continues to keep going without any errors.
Just so you know, the torture test I have always selected for eavh instance was Custom. I set 890MB of total system RAM for each instance assigned to each separate CPU core and left the minimum and maximum FFT lemgth at the default 8K and 4096K respectively. So each CPU core was set to use 890MB of system RAM. The amount of RAM used upon system startup is 110MB, so I set the RAM usage for each instance of Prime95 low enough so it shouldn't hit the page file.
Now, it seems I am so close to a perfectly stable overclock but maybe just missing something? Why would one instance fail after it starts back at 1024K FFT? I have my CPU VCORE set at 1.50V. This motherboard BIOS doesn't allow you to set the VCORE any higher? Or is there some secret trick to set the CPU VCORE to higher than 1.5 on this Asus motherboard? Am I all set and ready to go, or is it that important that Prime95 is 100% stable on both cores utilizing tons of physical RAM forever? Could there be a bug in Prime95 where one instance of dual instances fails after it starts over again at 1024K FFT? Or is it just that my system is almost 100% rock solid stable, but just barely not quite there?
My CPU temperature never reads any higher than 49C when both cores are under full 100% Prime95 stress. When one CPU core is under full Prime95 stress, the temperature reads 40C. When at idle, the CPU temperature reads 31C. These readings are all from NVIDIA Monitor.
Help greatly appreciated.
First of all, I have the dual core hotfix applied and the registry tweak set to 1 to enable it. I also have the latest NVIDIA 6.70 chipset drivers installed and the AMD X2 driver installed.
I have the latest 1009 BIOS for the Asus A8N-SLI Premium motherboard.
I have good airflow in my Antec P180 case. I have a Thermalright XP-120 heatsink on my CPU and cooled with a 57CFM airflow 120mm fan. I have a SeaSonic S12 600WATT PSU. I have four 512MB DIMMS of PC4400 (550MHz rated) Corsair XMS memory. DIMM voltage is set at 2.80V.
Now to overclock, I set the FSB to 250 in the BIOS with the default maximum CPU multiplier of 10. I leave the RAM index value set at 400MHz which means that the RAM will overclock at the same rate as my FSB. That is ok though because my RAM can handle speeds as fast as 275MHz (550MHz efffective DDR), and my RAM will only be running at 250MHz (500MHz effective DDR). I set the RAM timings to 2.5-3-3-6. The 1T/2T timing is set at 2T. I think my RAM should be able to run at these timings because it is rated at only slightly lower timings at 25MHz (50MHz effective DDR) faster. And since my RAM is running 25MHz slower, I would think it should be able to easily handle 2.5-3-3-6 instead of 2.5-4-4-8. The RAM is rated at 2.5-4-4-8 for running at 275MHz (550MHz DDR).
I began by upping the VCORE to my CPU little by little until dual instances Prime95 is completely stable with my CPU speed overclocked to 2.5GHz (250*10). First, I set the VCORE at 1.425, and ran dual instances of Prime95. After about three hours, one instance failed. I tried upping the VCORE to 1.4375, and after a couple of hours, one instance of Prime95 failed again. The other instance kept going strong without any failures. I upped the VCORE again to 1.450 and the same thing. Finally, I upped the VCORE to 1.50 anf ran dual instances of Prime95. Both instances ran for longer than before without any errors. But after 10 hours and 20 minutes, one instance halted because of an error. The other instance kept going. I restarted the instance that failed, and it has been successfully going for an hour since while the other instance has continued to go without failure this whole time.
Now, both instances passed the large FFTs all the way up to 4096K. After it started over its round over again on the 1024K FFT length after 10 hours and 20 minutes, that is when one instance failed. So, it seems that both instances of Prime95 passed all FFTs from 8K to 4096K. But one core couldn't keep going once it started its cycle over again back at 1024K FFT. But about 7 minutes after one instance failed and the other was still going, I decided to start the instance that failed again with the same settings. And it started right at 1024K FFT test and it passed and continues to keep going without any errors.
Just so you know, the torture test I have always selected for eavh instance was Custom. I set 890MB of total system RAM for each instance assigned to each separate CPU core and left the minimum and maximum FFT lemgth at the default 8K and 4096K respectively. So each CPU core was set to use 890MB of system RAM. The amount of RAM used upon system startup is 110MB, so I set the RAM usage for each instance of Prime95 low enough so it shouldn't hit the page file.
Now, it seems I am so close to a perfectly stable overclock but maybe just missing something? Why would one instance fail after it starts back at 1024K FFT? I have my CPU VCORE set at 1.50V. This motherboard BIOS doesn't allow you to set the VCORE any higher? Or is there some secret trick to set the CPU VCORE to higher than 1.5 on this Asus motherboard? Am I all set and ready to go, or is it that important that Prime95 is 100% stable on both cores utilizing tons of physical RAM forever? Could there be a bug in Prime95 where one instance of dual instances fails after it starts over again at 1024K FFT? Or is it just that my system is almost 100% rock solid stable, but just barely not quite there?
My CPU temperature never reads any higher than 49C when both cores are under full 100% Prime95 stress. When one CPU core is under full Prime95 stress, the temperature reads 40C. When at idle, the CPU temperature reads 31C. These readings are all from NVIDIA Monitor.
Help greatly appreciated.
Facebook
Twitter