Why settle? That last 25MHz might increase your fps by .6 or even .9 depending on the game. Might even get another couple fps if your a 720P gamer! /s
Yea, I love the fact that some posters here thing that AMD should have said "boosts up to 4575 mhz" instead of "boosting up to 4.6 ghz"Having said I dont tweak anymore I did go into the bios and turn on PBO. None of my cores would boost beyond 4200Mhz, I turned it off and Im back to boosting to 4575.
For whatever reason I get better performance with 3600/CL16 than with 3600/CL14.
I would love it better if these chips (3900X) would actually boost up to the 4.65GHz they are fused at...
View attachment 8851
... And even better if it was for more than a few milliseconds. This is with PBO disabled. Though admittedly I'm using 3733 1:1 IF so I have no idea if my OC impacts the duration of the boost or if it's my old AGESA (1.0.0.2). I also have no idea if they'll ever get PBO working better than stock or not. At least with the 3900X it seems pretty much tapped out at stock.
I won the silicon lottery on 1 chiplet and got "meh" on the other. I suspect people who wait to adopt Ryzen 3000 anywhere from 6-12 months later will likely benefit from a maturing process which would make the better chiplets more common. As I suspect AMD intentionally binned 1 "good" chiplet and 1 "meh" chiplet on the 3900X. They're both salvaged from 8c chiplets so I bet the better 8c chiplets are going to Epyc and Threadripper.
I would be surprised if the 3950X was substantially better, but who knows.
One other theory I have heard is that there is a sweet spot with these chips when it comes to memory timings, and going tighter will put more strain on the memory controller, which in turn will cause more power consumption. And these chips behave the way we suspect them to do: lower/raise the boost frequency when there is a shift of power balance within PPT (package power tracking). Since the memory controller is part of the package, its power budget is shared with the cores, thus extra power consumption by the memory controller lowers the power budget of the cores.
I have a theory- you know why the highest voltage is single core idle? to reach with 0 weight of load advertised frequency
so the regulation has its "singularity" 1 core simplest load
https://www.msi.com/Motherboard/support/B450-TOMAHAWKMaybe, I dunno, lol
Well, you can have high volts and low amps, high volts high amps is what would kill the chip.
How is the 3900X and B450 Tomahawk?
What AGESA version are you using?
https://www.msi.com/Motherboard/support/B450-TOMAHAWK
the latest BIOS from 10.7.2019, 7C02v19 version
however, I am not touching anything until all of the crap is solved
we all have experience when you change BIOS settings that work to new that dont and back, it suddenly doesnt work
I am observing the same as others, so nothing special
Depends on what you are doing. Lapping IHS and HSF in a conventional setup with "normal" paste is good for maybe 1-2C, depending on which paste you use. In the "old days" when people were still using Shin Etsu products, stuff like G751 did not respond well to lapped surfaces, while x23 variants (notably x23-7783D) did. Sometimes you could get very narrow bondline thickness without air bubbles or gaps.
If you use liquid metal (CLU, Conductonaut), bare copper surfaces are essentially resurfaced, so the lapping doesn't much matter. But Conductonaut in particular likes to pump out when applied to nickel (it has no affinity for nickel). My goal was to expose copper on the IHS so that I could use Conductonaut between IHS and copper waterblock to avoid pump-out. Using LM products between copper surfaces tends to "soft weld" the surfaces together.
with the board I am satisfiedOk, are you stable and all that, happy with the board, or do you regret getting it?
Just curious, as my used B350 / PCI slot board purchases (ComboPI 1.0.0.2 and 1.0.0.3) have been surprisingly good with Zen2.
I understand liquid metal is awesome and all, does it make a big difference when you are cooling like this versus using a diamond compound?
I still cannot believe you killed your 3900X after you revived my 3700X!Depends on the application. In my case, custom water (close to CLC):
https://www.tomshardware.com/reviews/thermal-paste-comparison,5108-8.html
Only two pastes I have are conductonaut and Kryonaut. I'm hesitant to use Kryonaut since it can degrade above 80C, and I suspect I might hit hotspots higher than that. Maybe. Kryonaut should be safer on my Radeon VII (ironically).
I may still be able to rescue the dead 3900x . . . but I doubt it.
Very interesting. That's the exact ram speeds I'm running. I'll have to go do some boost testing of my own now.