Well I found this morning that I had set the slow boot mode to enabled from when I had my previous cpu.
It wasn't the cause of the aida64 bandwith prob, but disabling it helped make the board a bit snappier.
I found it's a setting called qpi link speed, it's just a manual switch for it.
And this new cpu doesn't seem to care either way, it ain't helping it boot high bclk like my previous cpu (previous didn't clock as high, bclk and cpu speed).
Before I went to bed I had ran alot of y-cruncher.
I got vccsa down to 0.050v and got 10hrs+ of that tested, 114mhz bclk and just past 2400mhz ram at cas9-11-11 cmd1.
1.65v ram too.
Now I'm testing 0.025v for vccsa.
Over a half hour into y-cruncher now with that.
Bandwith is over 65k across the board in aida64.
Ok here's the thing that gets me, over 10hrs y-c stable at 114mhz bclk, but cannot post 115mhz.
It doesn't even try, it goes straight to post code 00, nothing before that either.
Anyone got any ideas how to boot a bclk that does a post code 00?
To me, the way to go is, higher cpu speed 1st, try for highest speed at stock voltage.
This will get rid of any major slow downs, like the mouse stuttering that happens on stock (why I don't know, but it's horrid).
Higher bclk will get you alot more bandwith for the mem speed, and will cut your dpc latency in half or even by 2/3rd's.
Y-cruncher seems to still be the king for testing the mem controller and perhaps even the mem too.
I did hours of gaming at 110mhz bclk and 2600mhz+ mem, but within mins y-c said it wasn't stable without alot more vmem.
I did mess with 114mhz bclk and over 2700mhz mem.
But it want's more vccsa and alot more vmem, couldn't get it stable enough to try for a 24.7 setup yet.
These trident-x heatsinks are nice, tested over 1.85v and they just got slightly warm and that was all, no active cooling at all.
Some of my notes from this morning:
Bandwith bug also happens on unstable shutdown when pressing reset button
114mhz bclk 2736mhz mem needs +0.075v vccsa to complete aida64 bench