Originally Posted by CaptainCab21
Nice score! I might have to get some bdie that score is good, only getting 3940 on my 1700x at 4ghz
Samsung B-die definitely makes a difference. I had a Hynix A die that was actually good for Hynix as I was able to overclock the 3200Mhz kit to 3600Mhz, however I couldn't get tight timings with it. After upgrading to Samsung B-die I simply used the Ryzen Ram calculator and input the "Fast" timings for 3200Mhz and then just increased the speed to 3400Mhz. Basically I ended up with 3400Mhz running really tight timings. My Cinebench R15 scores instantly got about a 30 point boost with the Samsung B-die. I've been doing more testing and tweaking of my settings, I was able to hit just over 4500 @ 4.3Ghz (really I was running at 4.295Ghz across all cores) but continued testing and didn't get a screen shot. Haven't been able to hit that score again as for some reason I started to have BSOD issues that require a hard reboot.
I find Cinebench R20 to be very interesting in that it seems to push harder than stability tests. I can run my system at 4.35Ghz across all cores and hitting a top temp of 75C, and I can run normal rendering projects, video and photo editing, gaming, all the normal suite of benchmarks, and will even pass AIDA64 and Prime 95 stability tests, but I can't run Cinebench R20 past 4.295Ghz, and if I push it to 4.30Ghz I'll get a black screen of death. Not sure if I have slight instability past 4.295Ghz or if the program itself still has some bugs they are working out of it. I have no problems with any other program, its just Cinebench R20.
I guess in reality it really doesn't matter as I can adjust my settings and hit 4.35Ghz across all cores and haven't had any issues with any other programs, however my normal settings are 4.275Ghz across all cores and this rig will crush anything I throw at it and won't exceed 65C. I think for 24/7 use I'm more comfortable running at 65C than 75C. At 4.275Ghz I get ~1985 with Cinebench R15 and at 4.35Ghz I get ~2020, but in the real world when I'm really rendering the time difference is negligible.