Much better
try running Prime95 on BLEND, to make sure the entire memory subsystem is "bench stable" 15-30 minutes is all I ever use. Usually less. (only has to stable long enough for a wprime 1024, or spi32M
If you are looking for total system stability thn run prime95 for 8-12 hours. (keep an eye on temps of course, CPU, VRM and chipset
)
You are approaching 200 BCLK, so if you are still unstable you have 2 options, give it a little more QPI PLL voltage or drop your BCLK and try a higher MEM multi
(remember though that uncore is going to raise its multi as well, so a little extra VTT might be needed if you find yourself getting unstable).
Take note of the voltages and speed you have set (specifically VTT, Vdimm and multis.) this way when you start concentrating on just BCLK or just CPU speed (pretty closely tied on a 920 with good cooling) you will already know the limits of your memory.
Make sure to run mem tests at 8-8-8 and 9-9-9 timings too, so you can find your max mem speeds for other component clocking you will do later....
Like Max uncore
Use your multipliers to clock this up as much as you can, then fine tune with BCLK
Next work on BCLK and then work on CPU.
Once you know what everything can do on its own, you can figure out how to put it all together
Just don't get frustrated and sell off something until you are sure it is the limit. If you have clocked AMD at all over the last few years, i7 is just like it. Getting memory and IMC stable is paramount if you do not have a high CPU multi to take advantage of
(Personally I would rather have a lower multi, cheaper chip than a higher multi 2-4 times as expensive chip to start with anyday. It is much more satisfying to get a 75% OC on a $200 dollar chip than it is to get higher speeds, but a smaller percentage on a chip that costs 2-6 times more
You also get a better understanding of the architecture knowing that every component is going to be pushed to its limits
) JMHO