[Update] different memory ratio and the effect of jumper on NB and RAM
Yes, they are Crucial 8500.
Update
My previous post DDR2-1500 @2.4v has shown this board is very good at 5:8 ratio. What about the others? I chose 5:6, 2:3 and 4:5 and did a little research. Let's look at these three screenshots first. Settings were all the same except FSB to run memory at 600MHz, vDIMM=2.1v, NB=1.45v. CPU is E8400 and memory is Team 800 cas4 D9GMH. For comparison, This pair of RAM does 600MHz ORTHOS Blend on P5K premium with 2.1v in BIOS.
PIC 1 5:6
http://img337.imageshack.us/img337/4...thos10hrl8.png
PIC 2 2:3
http://img179.imageshack.us/img179/7...rthos2hhb0.png
PIC 3 4:5
http://img179.imageshack.us/img179/3...hos1h33bb9.png
Seems like no difference, they were all stable @ 600MHz. Actually no, there IS difference. Here's how I did the test exactly.
1, I set vDIMM =2.1v, NB=1.40v. 5:6 ran over 1 hour before an error was detected; 2:3 detected error in 5 seconds; 4:5 was the worst, blue screen right after I clicked Start button. So mem raito difficulty is 5:6 < 2:3 < 4:5
2, I raised NB to 1.45v, 5:6 ran over 10 hours and no error; 2:3 ran 2 hours and no error; 4:5 was still the worst, error detected in 5 minutes and then system automatically restarted.
3, I thought NB still need more voltage to stablize @ 4:5, so I set NB to 1.50v. No, still error in 3 minutes, but no auto restart this time.
4, The prebious 3 setps were done with FSB jumpers taken off, which means 400 FSB. In this step, I set those jumpers to 333 to see if there's any change.
And there IS. With NB=1.45v, 4:5 could run over 1 hour and no error, like PIC 3 shows. Apparently, 333FSB jumper made it easier for the northbridge to run high memory clock.
5, To confirm the conclusion from 4, I used 333FSB jumper to go back and test 2:3. NB set to only 1.40v, 5 minutes no error. Compared with 1st step, which detected error in 5 seconds, 333FSB jumper also made NB more stable.
6, I used another pair of DDR2, Bufflo 800 D9GMH. It's not as good as those TEAM. I set vDIMM=2.2v, NB=1.45v. With 400FSB jumper, it couldn't even successfully log on Windows, just like on P5K premium. While with 333 jumper, it was able to run 16M superPi.
http://img412.imageshack.us/img412/8...016m333mj1.png
7, From 400 to 333, the performance did not fall. Everest read bandwidth was still 102xx MB/s.
Now we can see these FSB jumpers are not just for "easy overclocking", they do have some hidden effect on the memory controller. In my test, 333FSB is easier than 400 and gives the same performance. But with 333 jumper, the board has "double-boot" issue, which does not happen in most cases when set to 400.
I also did a small comparison between 333 and 266 jumper. It did not seem to have any improvement to me. But you may get different results if you are running a 1066FSB or 800FSB CPU.
This board is ecellent in dual core CPU and memory overclocking. It beats ASUS P5K-Premium (also very good IMO, except mem clocking) and blows Maximus Formula away, not even to mention Gigabyte. I am really suprised that a sub $100 board could be so much fun. It has some issues, but overall it's the best mobo I've owned so far. Good job MSI:up: