PCTuning fórum

havli píše:Nejen teploty, ale i vykon...

(...)

Tohle je zrovna pripad, kdyz je SW hodne dobre optimalizovany, ale rozdily jsou treba i v x265 a to uz je real-life aplikace. Takze AVX opravdu neni v CPU pritomno jen kvuli vetsimu vytopu.

R7 2700X = 15,3 fps http://hwbot.org/submission/3837723_too ... 15.313_fps
i7 7820X = 23,3 fps http://hwbot.org/submission/3665257_spl ... _23.31_fps

The Stilt, post: 39149633, member: 366210 píše: RZN vs. CFL vs. SKL-X in X265 2.5+31:

RZN: /w AVX2 = 100.00%, /wo AVX2 = 105.21%
CFL: /w AVX2 = 130.61%, /wo AVX2 = 101.13%
SKL-X: w/ AVX2 = 135.47%, /wo AVX2 = 105.21%

Ryzen's performance without AVX2 is impressive, but it is sad to see that there is still a penalty (like on Excavator) when running 256-bit code.

havli píše:To je zajimava informace. Pocitam, ze casem to vyvojari x265 zaregistruji a pridaji vyjimku. Takhle to funguje treba na Pentiu M, kde je pomala implementace SSE2 a proto se v x265 ani x264 nepouziva.

Prozatim ale holt smula - nemyslim si, ze v beznych GUI aplikacich by se dalo ovlivnit vyuziti instrukci CPU, aspon ne nejakou uzivatelsky prijatelnou formou. V x265 benchmarku to sice jde, ale i tak jsem jeste nevidel (mimo me ) nikoho, kdo by si nastavil instrukce rucne.

Na Excavatoru jsem vliv AVX2 na vykon zkousel a rozdil byl v ramci chyby mereni, aspon pri nastaveni encoderu, co se pouziva v benchmarku. S jinym nastavenim to dost mozna muze jet jinak, moznosti je spousta.

flanker píše:netuším, ale ten LEVEL1+LEVEL2 je něco jako profil AMD ve smyslu Intelovských edhanced multicore. Takže ty první dva by teoreticky mohli mít i jiní výrobci. To by zas mohl poradit CageJ jako zástupce MSI

I ran into some failures using PB2 and I was able to track these down to NOT being PB2's issue, but rather a Windows Bug. We are using the latest version of Windows 10 64-bit and all its updates that were available on April 17th. We knew we were going to be doing a lot of testing, so we froze our OS updates at that point. What I was finding is that I would get these random power-downs using Cinebench, HWinfo64, and CPUz at the same time. I could not replicate the error without these three programs running simultaneously. At time I just assumed that I was beating on the CPU hard enough to make it fail, until it went into a hard power-down while sitting idle at the desktop, and I could replicate this issue at idle. Talking with AMD and ASUS about this, they asked me to work through the other power profiles we were not using. We use "High Power" for all our testing here. I moved to the Balanced profile, and it still happened. I then moved to the Ryzen Balanced profile, and it was still happening. Once I moved back to the High Performance profile again, I could not repeat the error. I could not replicate the error in Balanced or Ryzen Balanced either. ASUS let me know that there has been a Windows bug identified with this issue. The current solution to the issues seems to be to switch power profiles one or two times and it will correct itself. So if you are having some odd shutdowns, do not assume it is anything hardware or heat related.

flanker píše:netuším, ale ten LEVEL1+LEVEL2 je něco jako profil AMD ve smyslu Intelovských edhanced multicore. Takže ty první dva by teoreticky mohli mít i jiní výrobci. To by zas mohl poradit CageJ jako zástupce MSI

There are two settings in our Crosshairs VII EFI (BIOS) that are specific to Precision Boost2 and XFR2, and currently both are not clearly labeled, and we have asked ASUS to be a bit more transparent on these labels. Currently you will see "Core Performance Boost" and "Performance Enhancer."

Core Performance Boost = XFR2

Performance Enhancer = Precision Boost 2

CPB/XFR2 has three settings, Auto, Enabled, and Disabled. From what I can tell there is no difference between Auto and Enabled. Using these turns on XFR2. We have used Enabled and Disabled for our results here today.

Performance Enhancer is a bit more convoluted. There are settings for Auto, Default, Level 1, Level 2, Level 3 (OC), and Level 4 (OC). Levels 1 and 2 are using rules sets specified by AMD as to how Precision Boost 2 is supposed to work. Levels 3 and 4 are actually overclocked settings as noted. For our testing here, we have used Level 2 which is part of AMD's Precision Boost 2 specification. Quite frankly, we did not touch Level 1 as it is supposed to be a "lesser" BP2 setting in that Level 2 allows a slightly higher TDP value. Levels 3 and 4 are technically overclocking in that these profiles go above and beyond TDP and clock settings. Level 3 as you will see, requires even better cooling than what we have used here with our XSPC custom cooling loop.