Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion. Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules. |
|
|
Thread Tools | Search this Thread | Display Modes |
25th June 2019, 11:35 | #41 | Link |
RipBot264 author
Join Date: May 2006
Location: Poland
Posts: 7,816
|
I use default settings. You must be doing something wrong. BTW. Using single stick 16GiB RAM on CPU with Quad Channel was extremely weird! I hope that you used atleast fast memory sticks because Threadrippers really need high bandwidth.
__________________
Windows 7 Image Updater - SkyLake\KabyLake\CoffeLake\Ryzen Threadripper Last edited by Atak_Snajpera; 25th June 2019 at 11:39. |
25th June 2019, 12:33 | #43 | Link |
RipBot264 author
Join Date: May 2006
Location: Poland
Posts: 7,816
|
Was that the same company which installed only single stick and thus forced 1950x to work in single channel mode?
__________________
Windows 7 Image Updater - SkyLake\KabyLake\CoffeLake\Ryzen Threadripper |
25th June 2019, 12:37 | #44 | Link | |
Registered User
Join Date: Nov 2013
Posts: 577
|
Quote:
I told them to consider 3200Mhz as well but they answered that I could be sure that the rig would not be able to benefit from that. Last edited by mparade; 25th June 2019 at 12:52. |
|
25th June 2019, 13:38 | #45 | Link |
RipBot264 author
Join Date: May 2006
Location: Poland
Posts: 7,816
|
LOL! What a pros Everybody knows that Infinity Fabric in Zen is tied with memory speed.
__________________
Windows 7 Image Updater - SkyLake\KabyLake\CoffeLake\Ryzen Threadripper |
26th June 2019, 12:01 | #46 | Link | |
Registered User
Join Date: Aug 2010
Location: Athens, Greece
Posts: 2,901
|
Quote:
They must be paid by Intel even when they are selling Threadripper CPUs that way, I think.
__________________
Win 10 x64 (19042.572) - Core i5-2400 - Radeon RX 470 (20.10.1) HEVC decoding benchmarks H.264 DXVA Benchmarks for all |
|
26th June 2019, 20:25 | #47 | Link | |
Registered User
Join Date: Nov 2013
Posts: 577
|
Quote:
|
|
29th June 2019, 12:23 | #49 | Link | |
RipBot264 author
Join Date: May 2006
Location: Poland
Posts: 7,816
|
Quote:
__________________
Windows 7 Image Updater - SkyLake\KabyLake\CoffeLake\Ryzen Threadripper |
|
30th June 2019, 05:05 | #51 | Link |
Registered User
Join Date: May 2009
Posts: 331
|
Dude, where do you keep getting these totally awesome informational graphs and pictures? Do you build them yourself with all the benchmarking software you have? This one is glorious, and I wish we had more of these curves for more processors haha. It would allow people to know about getting that lower end CPU and overclock it and see if it's worth just paying for that next one up.
|
30th June 2019, 11:43 | #52 | Link |
Registered User
Join Date: Dec 2008
Posts: 589
|
you can make one yourself.
All it costs is a power meter and a lot of time. Here's example of a power meter : https://www.amazon.com/P3-P4400-Elec...dp/B00009MDBU/ You plug the computer into it and it will tell you how much power the system consumes. From there, it's a repeat process. 1. Find the lowest voltage the CPU can be configured at where it will be stable at a particular frequency (ex 3000 Mhz) 2. Run benchmark for a reasonable amount of time, let's say 1h. 3. Look at the power meter's peak Watts value or if the meter has such feature, divide the amount of power (kWh) by minutes to get average energy consumption per minute) 4. Increase cpu frequency by 50/100Mhz and increase voltage on CPU until it's stable again with this new higher frequency. (see step 1) The results will not be perfect, because the power meter measures the power consumption of the whole system, which includes the inefficiency of the power supply (the power supply doesn't have a fixed efficiency, for example it may be 85% efficient at 200w and 87% efficient at 400w) You could use a better clamp meter or multimeter to measure the energy taken by CPU VRM through the 8pin EPS connector - that will give you even better numbers. Though, these will still include some very small errors due to variation in CPU VRM's efficiency (vrm = dc-dc converter which converts 12v from psu to 1v..1.4v for the cpu) but this variation in efficiency should be so small that would not affect your graphs in any significant way. |
30th June 2019, 11:44 | #53 | Link | |
RipBot264 author
Join Date: May 2006
Location: Poland
Posts: 7,816
|
Quote:
__________________
Windows 7 Image Updater - SkyLake\KabyLake\CoffeLake\Ryzen Threadripper |
|
9th July 2019, 17:17 | #54 | Link | |
Registered User
Join Date: Jan 2015
Posts: 118
|
Quote:
The encoders sets frame-threads to "5" in my setup when it is not otherwise specified, so automode is "5". If I manually reduce that to a max of "3" it certainly won't degrade quality compared to the defaults. Primary question for me is if the option "numa-pools=48" does reduce quality as I manually raised it from the default of 32 to over 50 which gave me an incredible performance boost. Last edited by blublub; 9th July 2019 at 17:39. |
|
9th July 2019, 18:48 | #55 | Link |
RipBot264 author
Join Date: May 2006
Location: Poland
Posts: 7,816
|
Did you setup your Threadripper in NUMA mode instead of UMA?
__________________
Windows 7 Image Updater - SkyLake\KabyLake\CoffeLake\Ryzen Threadripper |
10th July 2019, 18:04 | #56 | Link | |
Moderator
Join Date: Jan 2006
Location: Portland, OR
Posts: 4,771
|
Quote:
Lookahead threading is reduced in slower presets, but I'm fuzzy on if and how it actually impacts quality much. Other than that, I don't think that thread count impacts quality so much. For Intel dual-socket systems at least, I've often found that pinning a job to just one socket ala --pools "-,+" has minimal impact on performance, although that probably varies with encoded frame size and number of cores per socket. |
|
12th July 2019, 18:27 | #57 | Link | |
Registered User
Join Date: Dec 2002
Location: Region 0
Posts: 1,436
|
Quote:
Do you also limit the total number of threads per encode? Like if you're going to run 4 -F 1 concurrent encodes you set thread pool for each encode to logical cores / 4? |
|
13th July 2019, 12:41 | #59 | Link |
RipBot264 author
Join Date: May 2006
Location: Poland
Posts: 7,816
|
Always use UMA. Video encoders unlike games do not need super low latency when accessing memory.
__________________
Windows 7 Image Updater - SkyLake\KabyLake\CoffeLake\Ryzen Threadripper |
|
|