New Posts  All Forums:

Posts by Sam OCX

Specs of 3466C16 and 3600C17 models actually look much better than of 3800+ rated stuff.
Testing the 32GB version of HyperX Fury 2666C15. Memtest (best scores): SuperPI 32M:
The optimal Hynix DDR4 temperature varies on the benchmark. For 32M it's not likely that you will be able to run below -75c with some sticks opting out around -50c already. For suicide shots / Maxxmem you can go near CB which usually is around -150.
DDR3 SODIMM on Impact
Getting 1666MHz 13-17-16 to do 32M is quite easy (even CL12 is likely to work on well-binned kits). Anything above 1666 depends more on the IMC than the mems now.
Old batches of 3200C16 were Hynix-based but yours are Samsung (also confirmed by xxxxx5xxxxxxxxx serial numbers).
I use/used Channel B for individual module binning as well. But when I tried 1600c11 in B+D a.k.a. "the hardest slots", I couldn't pass 32M no matter what. But it worked well in A+C. On MSI, A and C have looser RTLs (47/8 at 1600c11) compared to B and D (45/8 at 1600c11), this is what has most likely made things easier for mems and/or IMC.
Some boards like ASUS and MSI also set CL wrong in the BIOS with BIOS versions newer than ~Feb2015. That is, you boot CL10 but your minimal stable voltage and RTLs are exactly the same as with CL11 with performance getting only a marginal boost (0.1-0.2s in 32M, at best). If you use an older BIOS version then CL10 is "real".
9/32 is actually not a bad ratio for 1600Mhz CL11. I gave my ES sticks a try the other day (which for 1500MHz CL11 need 0.1V less than the best retail kit that I ever had) and they needed 1.83V.The real 1500MHz CL10 should be faster but I haven't yet seen anyone pull it off on air. Given that 1500MHz CL11 needs more voltage than 1600MHz CL12, 1500MHz CL10 should need more than 1600MHz CL11 which ends up in 1.9V+ even on super handpicked sets.
You have absolutely no chip maker guarantees on a spec like 2666C15, it can be anything.
New Posts  All Forums: