This afternoon, I did another series of ram tests and the results were quite interesting. It shows that always aiming for the lowest possible timing is not the best option in AMD platform, sometimes its just wasted effort . A balanced setting can be as effective, and yet much less stress on ram + less voltage !!
If you can get a very high NB freq, it actually reduces the significance of low timings a great deal !
I spent hours today to get my ram stable at 1784 6-8-6-21-24 @ 1.73, and I did. But my WOW just disappeared after an hour ^^ . you will see why !
Generic 1664 9-9-9-24-1T - NB 3100+ Mhz VS 1600 6-6-6-18-1T-NB 3009 Mhz
1664 CL9

1600 CL6

The Generic 1666 CL9 only lost in latency and its overall score beats 1600 cL6.
that lower latency is noticeable in desktop application and...... but that generic 1664 CL9 performs really good !
--------------------------------------------------------
1800 7-8-7-21-24-1T @ 1.63v --> NB 3100+ Mhz VS 1784 6-8-6-21-24-1T @ 1.73v --> NB 3100+ Mhz
1800 CL7 @ 1.63v

1784 CL6 @ 1.73v

I had to add 0.1 extra voltage too get that 1784 CL6, But the gain is almost zero, and overall scores only differ by 0.01 !!! The latency difference is minimal too !
-----------------------------
My conclusion
-It is common knowledge that very low timing rams always fare better than rams with higher latencies, though What I see form these bechies is the opposite . Take the NB frequency factor into account and a generic ram can perform really well.
- NB frequency is very very crucial in ram overclocking. We see that a High NB can make up for lower latencies splendidly. In the second test I also added voltages. 1.73v for CL6, while there is barely any difference from 1.63v CL7. If it is me I go with CL7 1.63v. lower voltage is better anyway, and the latency difference wouldn't be noticeable either.
------------------
I'm open to all idea and opinions, so feel free to speak your mind about tests and results.
I know its not a 100% fair or anything but my point is just to show that NB frequency should not be forgotten if you're adjusting your timings, unless you want to end up like this
1800 CL7 - NB 2025 Mhz

+30% bottleneck
Edited by luches - 2/5/11 at 1:26pm
If you can get a very high NB freq, it actually reduces the significance of low timings a great deal !
I spent hours today to get my ram stable at 1784 6-8-6-21-24 @ 1.73, and I did. But my WOW just disappeared after an hour ^^ . you will see why !
Generic 1664 9-9-9-24-1T - NB 3100+ Mhz VS 1600 6-6-6-18-1T-NB 3009 Mhz
1664 CL9

1600 CL6

The Generic 1666 CL9 only lost in latency and its overall score beats 1600 cL6.
that lower latency is noticeable in desktop application and...... but that generic 1664 CL9 performs really good !
--------------------------------------------------------
1800 7-8-7-21-24-1T @ 1.63v --> NB 3100+ Mhz VS 1784 6-8-6-21-24-1T @ 1.73v --> NB 3100+ Mhz
1800 CL7 @ 1.63v

1784 CL6 @ 1.73v

I had to add 0.1 extra voltage too get that 1784 CL6, But the gain is almost zero, and overall scores only differ by 0.01 !!! The latency difference is minimal too !
-----------------------------
My conclusion
-It is common knowledge that very low timing rams always fare better than rams with higher latencies, though What I see form these bechies is the opposite . Take the NB frequency factor into account and a generic ram can perform really well.
- NB frequency is very very crucial in ram overclocking. We see that a High NB can make up for lower latencies splendidly. In the second test I also added voltages. 1.73v for CL6, while there is barely any difference from 1.63v CL7. If it is me I go with CL7 1.63v. lower voltage is better anyway, and the latency difference wouldn't be noticeable either.
------------------
I'm open to all idea and opinions, so feel free to speak your mind about tests and results.
I know its not a 100% fair or anything but my point is just to show that NB frequency should not be forgotten if you're adjusting your timings, unless you want to end up like this
1800 CL7 - NB 2025 Mhz

+30% bottleneck
Edited by luches - 2/5/11 at 1:26pm












