*HOT* | Corsair DOMINATOR GT | 6GB (3x2GB) | DDR3-2000 7-8-7-20 @ 1.65V | *HOT*

[cstkl1]

check the BR review on xcpus.com, they compared it to the classified and saw the same weird perf drops, it only seems to affect some memory dividers... it could be that you have to manually adjust the mch straps for the diferent memory dividers... i dont know why clock for clock the classified is slower, but i wouldnt be surprised if its caused by some relaxed timings they use to get higher bclock/qpi speeds? i dont know if you can actually adjust this and get back to tight timings and good clock per clock performance for people who really dont need high bclocks...

yeah i thought so too, but i dont think its turbo since it affects diferent mem speeds... some speeds are fine, some are notably slow

thats odd about dfi btw, are you saying they decide when to chose what strap and dont allow users to adjust it manually in bios? thats kinda weird for dfi...

but with lower memory multiplier and same uncore multiplier you can get higher? try more vtt and... i dont know how they are called on the evga board, but increase the x58 voltages, theres a 1.1 and a 1.5v voltage, increase both and see if it helps... can you get higher with cas8? i was quite surprised when raja showed how much memory can hold you back bclock wise... he switched from some good samsung to some good elpida sticks and it gave him 10mhz higher bclocks or even more iirc... and he couldnt get those even with relaxed timings before...

its weird, and i dont really get why... but several people had the same experience, they couldnt get high bclocks even with high vdimm, vtt and loose timings, then switched to another set of mem and could get noticably higher bclocks then...

whats the highest bclock you can boot at SAE?
your on phase change right?

dfi mem skews are which are evga mem strap equivalent are set by bootup clocks.
so far tested that up to 204 bclk dram 2040 is perfect.
bclk 205 dram multi 10 above theres a issue. it will pass linx 100 runs
but will fail prime one thread error.
jimmy send me one relax mem skew bios 06/19. so far no diff. the rtls were relaxed as it is related to mem skew strength also.
so the problem could be this.
at bclk 205 onwards .. bios set 205, bios bootup at 204 and software read is 205. so i asked him whether is it because of this -1 that is causing the issue wince bootup clock would be 190 for 205 and would be exact.
tested with bios bclk 204, bootup 190 , bios bootup read 204, software read 204 and then used setfsb to up the bclk to 204.8-205.2. still the one thread error on prime 95 large fft custom 5.4gb

so no idea whats wrong atm.

also for the evga classified. the performance drop could be related to the vtt pwm frequency. hear me out on this. On the dfi lower pwm frequency gives better results on sp1m. infact its the key for stability and lowering vtt and vcore. so far i find for my board and proc setting -15 on vtt and nominal on dram is the best. heck vista boot up gain is by one bar by just doing this.

dfi mem sub timing are interesting say for C7
mem multi 6,8,10 , for TRFC,TWR,TWTR,TRRD,TRC,TRTP,TFAW on the bios would be
88,6,4,4,15,3,15
88,10,5,5,25,5,25
88,12,6,6,30,6,30

what i found interesting on the evga and dfi is say for bclk 205 i choose dram multi 8 but i set timings according to dram multi 10. it is more relax right.. but this will get prime95 custom 5.4gb thread errors. evga has no problem with this.

found that interesting.
and the evga the twr according the cpu tweaker i can change it but the dfi i cant atm on software read