- Joined
- Mar 14, 2007
- Location
- Katowice, Poland
An easy calculation says that a memory with a 1.4 ms access should be rated at it's maximum frequency of 1000 / 1.4 = ~714 MHz. Things are not that easy though (in a positive, but a bit confusing manner).
My PX7900GS TDH was shipped with a Hynix FP-14 GDDR3 chips and I am currently running it @ 756 MHz. Since I don't have a multimeter, I wasn't able to measure memory voltage on my vidcard (on the other hand - I've not voltmodded it for the very same reason).
Now, before applying current OC settings, I was playing around with ATITool, scanning for artifacts and trying to find maximum memory clock available... It lasted about two hours (I was advancing slowly with small, manual steps, leaving each frequency active for about 3 minutes), my mem hitted nearly 900 MHz with no artifacts visible and that was the maximum -> I <- could stand. I was just afraid of pushing it further...
The same tactics for core clock resulted in maximum frequency found at 549 (+18 for geometric unit). Artifacts started appearing on the cube, so I cancelled further testing, backed up a bit and set 522 (540 gu) for a day-to-day operation.
Back to the memory settings... Was my approach incorrect in some way? Did I miss something? My current settings (756) work great, but I have not discovered the maximum - and that means I don't know how much I'm allowed to set to be safe (in other words - how much should I back it up from unknown maximum)...
And what's the real relationship between access time and a real frequency? Almost 900MHz hitted looks more realistic for a 1.2ms memory than my 1.4...
Could someone be so kind and explain it to me like to a newborn baby?
My PX7900GS TDH was shipped with a Hynix FP-14 GDDR3 chips and I am currently running it @ 756 MHz. Since I don't have a multimeter, I wasn't able to measure memory voltage on my vidcard (on the other hand - I've not voltmodded it for the very same reason).
Now, before applying current OC settings, I was playing around with ATITool, scanning for artifacts and trying to find maximum memory clock available... It lasted about two hours (I was advancing slowly with small, manual steps, leaving each frequency active for about 3 minutes), my mem hitted nearly 900 MHz with no artifacts visible and that was the maximum -> I <- could stand. I was just afraid of pushing it further...
The same tactics for core clock resulted in maximum frequency found at 549 (+18 for geometric unit). Artifacts started appearing on the cube, so I cancelled further testing, backed up a bit and set 522 (540 gu) for a day-to-day operation.
Back to the memory settings... Was my approach incorrect in some way? Did I miss something? My current settings (756) work great, but I have not discovered the maximum - and that means I don't know how much I'm allowed to set to be safe (in other words - how much should I back it up from unknown maximum)...
And what's the real relationship between access time and a real frequency? Almost 900MHz hitted looks more realistic for a 1.2ms memory than my 1.4...
Could someone be so kind and explain it to me like to a newborn baby?
