Swapped CPUs on the bench - now 3900x
Pulled the 3800X and put the 3900X
Not as happy with the 3900x as with the 3800x - chip for chip the 3800 pulls more weight minus its 4-cores-less limitation. It boosts much better, reaching its rated max at some point on all cores, and under load the all core boost was 4.250 (higher on some individual cores). The UMC is strong easily 1:1:1 (DDR4 3800/MTs cl16 aka; MCLK 1900: UCLK 1900: FCLK 1900) I've posted some screens of that system here before. Also its single thread scores are better - multi-threaded it obviously has a disadvantage.
The 3900X is also good in the UMC - Using the same mem kit and settings as the 3800X and its stable (3000% coverage run using Karhu mem test), R/W/Copy stats all similar with latency values only 1ns less, and that difference is well... chalk it up to alignments of the moon and stars.
The boosting behavior... That is different matter. I have yet to see any core hit rated (I suppose I am hitting the "new" 50MHz less rating) even under light threaded loads. I have a couple cores hitting in the 4.550's-4.575's - but from what I've captured only 3 of 12 cores. All core under load is @ 4.150 (some cores higher at times). Temps are OK I think, but I may be outgrowing my 280mm AIO with this CPU - I'll swap in a 360 later but so far I am not seeing anything outrageous for temps and I don't think close enough to the 80c threshold to limit me to only 4400 MHz but... (did get up to 78c in prime95 w/no avx... 22c ambient)
I ran some Cinebench both r15 and r20 - the scores aren't going to win prizes but also don't suck. Same with Geekbench not a champion but nothing to cry about either. With that said however, I am not really too concerned with cores and boosts etc., as long as the chip delivers some performance. Which it clearly does and I believe still worth what I paid for. It's just not as exciting as the 3800x was for me. I can't wait to get a 3950x and see what it's all about...
The basic setup followed is as in the beginning of this thread and the link from Level1 Techs - Thank Heuchler for all the work he put into gathering info into one spot.
Not doing any all core clocking - not doing anything with PBO etc
Changed Vcore to normal
Did not set XMP !
All Memory setting were input manually per DRAM Calculator and following the method detailed at the beginning of thread.
Lian Li T60 Bench:
Gigabyte Aorus Master x570 F5
Ryzen 7 3900x
G.Skill F4-4000C19D-32GTZKK (Bdie) 32GB (2x16) DDR4 [email protected]
Sabrent Rocket 4.0 M.2 1TB PCIe 4.0 x4 NVMe
Samsung EVO 850 1TB sata SSD
Corsair H115i Pro
Radeon 5700 XT
Sorry I am not smart enough to figure out the spoiler tag to hide my images.
A look at the System and some baseline runs.
Aorus Master x570, 3950X, 32GB GSkill c16 @3800, Sabrent PCIe4 1TB, Sabrent 1TB, 860evo 1TB, 2080Super, HX1200i, Celsius S36, Define S2, pg348q
Aorus Master x570, 3900X, 32GB GSkill c16 @3800, Sabrent PCIe4 1TB, Sabrent 1TB, 860evo 1TB, 5700XT, HX1000i, H150i Pro, Phanteks p600s, lg32gk650f-b
MEG x570 Unify, 3800X, 32GB GSkill c16 @3800, Sabrent PCIe4 1TB, 860Evo 1TB, 1080Ti, RM850x, Celsius S36, Define R6, pg279q
C7H, 2700X, 32GB c14, 960pro, 860evo, 1080Ti, HX850, S36, DefineR6, u3415w