Case Study: Parallel Processing

by on Oct.06, 2020, under How to, Linux, Programming, Server, Themes, Ultimate Edition

Case Study: Parallel Processing

Our computers did the leg work so you do not have to. I do have other computers if I need further data. Feel free to post your results as well, I did spend a chunk of my life writing the software and testing. I will not rob you of system info (minus mac addresses). Both are nasty computers.

I wrote an application uetheme a sub-application of tmosb (TheeMahn’s Operating System Builder) to take advantage of multi-core / multi-threaded Central Processing Unit(s) (CPU), you will also need Repostorm installed if you intend to do your own testing. It will build each of the 150 deb(s) perfectly. I will point that out later. Results are here, software is still advancing.

I intended and succeeded to demonstrate what a massive difference in speed and time such a technology makes, when this approach is mastered. You do not have to go through what I did to see the difference:

uetheme --benchmark # Launches 1 CPU thread, 2, 3.. until you run out of CPU threads & records results for each thread. (Will take a long time, read on) How 99%+ software is written.

I would not suggest doing the above command. The application will download ~1/2GB of theme source code data, extract it, start a timer and begin building. It took a AMD Ryzen 9 3950X 16 Core 32 thread CPU 4 hours, 40 minutes and 53 seconds to complete (32 times built). Critical data was gathered during the process to make the chart above, the same was also launched on a AMD Ryzen 9 3900X 12 Core / 24 thread CPU also gathering data 3 hours 40 minutes and 36 seconds (24 times, not 32 times).

For Example: an AMD Epyc 64 core / 128 Thread would do it 128 times (perhaps take longer), creating a point of data & would totally demolish the data I provided above.

I am also providing a spreadsheet of the results in ods format (Libre Office) & pdf format (Adobe Acrobat).

If you want to test it yourself. I would suggest manually (open a terminal):
uetheme --benchmark # Let one thread finish will report:

Timer: 0h54m9s (3249 seconds) for example. Painful was it not?

(Press CTRL+C) to cancel it will return you to the folder you executed it in.
cd themes/
uetheme --clean # erases build files
uetheme --all
# Automatically will detect the number of CPU core(s) / Thread(s) you have & use them all. (will build 150 deb packages – 1.3 GB in size total)

Timer: 0h4m00s (240 seconds) 13.5375 times as fast. Results will vary depending on the CPU used.

Observations

While the Ryzen 9 3900X 12 Core / 24 thread CPU initially laid the smack down on the Ryzen 3950X 16 core / 32 thread, the AMD 3950X did eventually beat the 12 core & the 12 core had an extreme advantage. I swear I did tell you about that, 12 days uptime did not bother you or the Terabytes of space? The 12 Core is running Ultimate Edition 6.6 Server (No GUI) – minimal applications running in the background. The 16 Core was running Ultimate Edition 6.7 Developer and many applications running in the foreground, so a margin for error does exist, it is very minimal & weighs heavier to the 3950X opening that gap for the positive (better results for the 3950X).

Reasons for the observation?
Different load as explained above & different base clocks both are 7nm technology.

Disk I/O activity is not considered both are running off Non-Volatile Memory Expansion (NVME) drives & both have identical memory, both have 20 Gigabit Ethernet (timer does not start until after the download finishes) therefore network speed is not a part of the calculation / equation.

Results in my case are not scalar.
The AMD Ryzen 9 3950X 16 core / 32 thread CPU is NOT 32 times as fast as a single core / thread: 13.5375 times as fast.
AMD Ryzen 9 3900X 12 Core / 24 Thread: 11.6774193548387 times as fast.

Feel free to test my thesis / case study & report your own case study. Enjoy faster software inbound. That part is a fact. I only have a quad core TheeMahn? Did you see the difference between 1 and 2 threads? 99% of software is written to take care of a single core. I am going to change that. Move yourself up that list to 4 cores (oh, I have 4 cores and 8 threads) do you see where I am going?

I almost forgot to tell you: I do not care if you have an Intel CPU. Results will be the similar. Ghz did the 12 core justice in single core, what Intel does best currently. That gap will broaden, Intel will go down. I am sorry Intel fans. Intel does not make a 64 core / 128 thread CPU. Glad to be a cutting edge programmer. Did you see the chart I provided above? The application is indiscriminate.

Multi-core processing is now in:
Repostorm
Roku-Ripper (Limited to 16 cores – complain to ffmpeg devs)

Blow off steam TheeMahn

You are welcome for faster software (Intel or AMD should not matter), I spent a long time writing the software, and recording the results. Am I missing Intel from the results? If I insert them it would look all bad. I own an Intel / NVidia ASUS ROG (Republic Of Gamers) Laptop. An 4 core / 8 thread (7700HQ) CPU. Please tell me it would get it done in 4 minutes? I doubt even an hour & also has 16GB of ram and NVME technology. Yes, a Laptop. Send me a 9900 (K)eep (S)pending chip I will dust it, Intel does have faster today, nothing to even think about touching me.

No pun intended, Intel will struggle to touch AMD 3950X, the Server world they are screwed & that is a fact. That is a desktop CPU v/s their top server CPU. AMD has them and by the short and curlies & faster is coming. I won’t buy one it has a lower “base clock” then the 3950X. Right now, software is not geared to take advantage of multiple cores / threads. I intend to change that once again.

If I get the opportunity I will test it, the wife has both of my laptops. Both have NVME technology. Both have Intel technology. The Yoga is also no joke has a 4K touchscreen.

My rigs will spin rings around both. Do you want to see where Intel falls in? You are screwed, that is the last thing they wanted to provide their users, more cores unheard of. I want to keep raping our users we have no competition, that is long gone.

Nvidia likes to think they are in “The Cool Zone”, congratulations for now. AMD will hand you your ass too. When NVidia seen AMD hand Intel it’s ass. NVidia Stepped up it’s game, the first time in many a year. In the end NVidia will lose. I am not just a programmer.

This is a pill you will be unable to swallow (if you are an NVidia Fella):
in the GPU arena (NVidia / AMD) That is Billions of dollars, AMD is going to take NVidia down, not with your 3090x, but raw power. What does your next Xbox have in it? How about your Playstation 5? I am not letting you off the hook how about your next cellphone Samsung is licensing from AMD?

NVidia please show me your sales next quarter, real competition is coming, did you blow off Intel’s loss as fruitless? I have been waiting to fill my pockets. Blow everything I said off, that will soon be the standard.

In the server market you are truly screwed. HBM2e will ensure that, who developed that technology as you tried DDR6? When you play with servers we do not care about money right?

Try once to Innovate, just once NVidia. I think I did a pretty good job, feel free to correct me. Nvidia attempts to buy ARM to be competitive to AMD, AMD rolls to be the worst thing you have seen $50 Billion purchase. We don’t have to worry about buying that technology. It is not like AMD is not doing the same, a real thorn to NVidia if it goes through & Intel for that matter.

TheeMahn

P.S. Did you know AMD patented project Quantum computing second time, w/o a doubt they mean business. AMD has the resources now. You are both screwed. Funny, I did not get in it for the money. I will absorb it. $6.97 a share in & hanging tight, what is it now? $85.35 +0.99 (1.17%), that is just today. I could just bail out & pay off my house.

It is not done, I am not going anywhere. AMD Stock is going no-where but up. Don’t blame my software.



Leave a Reply

You must be logged in to post a comment.

Back to top ↑