Gather around children!
Let me spill some tea about CPU performance myth of magical single core performance metric.
(I won't touch the topic of overclocking because majority of users don't overclock, we are talking stock performance)
Let's establish the basics:
The single core performance is the performance a CPU would have if you disabled all except one of its cores.
Single-core performance is a test for frequency scaling technologies:
like AMD-s
AMD Ryzen™ Technology: Precision Boost 2 Performance Enhancement
and Intel-s
Intel® Turbo Boost Max Technology 3.0
Use case of single-core performance metric is beneficial in mobile space mostly, representing "snappiness" of the laptop and speed of lunching apps.
Today there are just few applications that prefer single core performance, most notably older games (StarkCraft, Age of Empires and similar) and some in Adobe Creative Cloud like Photoshop and Acrobat.
FL Studio is not one of those app-s.
FL project are always multi threaded.
In FL bottleneck starts when your multi-core project maxed out one core in a multi threaded workload.
Multi threaded workload is not represented by single-core performance result multiplied by number of cores.
More accurate representation would be multi threaded load score divided by number of cores (IMHO physical not logical (cores not threads)).
IL recommendation for FL users from IL knowledgebase:
"Choose your own - Search for a CPU with the fastest single-core performance you can afford, in a package with 8 to 12 physical cores. Why? Because if any one core becomes overloaded, it does not matter how many other cores you have, there will be audio glitches in real-time playback. Here's how we grade multi-core scores for CPUs with 12 or less cores - Weak: Less than 4,999. Medium: 5000 to 9,999. Strong: 10,000 to 19,999. Very strong more than 20,000. For example: An 8 core CPU (14,400) with a single core score of 1800 is probably less well suited to music production than a 6 core CPU (12,000) with a single core score of 2600, since much of what happens with audio-processing can't be computed in parallel. Ideally, you need a CPU in the Strong or Very Strong category."
That statement refers to single core performance score obtained in a vacuum,
in the scenario where just one core is active, which never happens in FL projects.
PassMark Single Thread Performance
is obtained in short ~20 seconds Single Thread test.
That score is not scalable and in the context of FL load it is irrelevant.
PassMark table for generating scores is as fallows:
CPU Mark = 1 / AVERAGE ( test values ) * 1.65
Overall CPU mark is scaled up to be closer to PT9 result
1 / ( 0.149001983 * CPU Integer Math)
1 / ( 0.271311984 * CPU - Floating Point Math)
1 / ( 296.1871085 * CPU - Prime Numbers)
1 / (0.758484696 * CPU - Extended Instructions (SSE))
1 / (0.034970332 * CPU - Compression)
1 / (0.8766779 * CPU - Encryption)
1 / (17.81818182 * CPU - Physics)
1 / (0.392310713 * CPU - Sorting)
1 / (4.908772082 * CPU - Single Threaded)
As we can see Single Threaded result is added to overall score of a CPU.
Honestly this single core metric (way the score is measured by loading just one core) is not relevant in DA workstation environment.
I will take two flagship CPU from both x86 manufactures as a example:
Intel® Core™ i9-12900K boasts
Max Turbo Frequency 5.20 GHz (this is the single core load clock used for calculating Single Threaded score metric)
8 Performance-core Max Turbo Frequency 5.10 GHz
8 Efficient-core Max Turbo Frequency 3.90 GHz
AMD Ryzen™ 9 5950X boasts
Max. Boost Clock Up to 4.9GHz (this is the single core load clock used for calculating Single Threaded score metric)
All core Frequency ~3.95GHz (independent source (5950X all core frequency is not disclosed by the company))
The moment more than one thread is utilized single thread performance nosedives, transforming it in to, by my estimate, multi-core score divided by number of physical cores.
Not the "PassMark - CPU Mark" but rather something like 7zip or Cinabanche r23.
Here are results from Cinabanche r23 multi-core score divided with number of physical cores for top CPU-s:
https://www.cpu-monkey.com/en/cpu_benchmark-cinebench_r23_multi_core-16
AMD Ryzen 9 5950X Multi - 28577/16 - 1786
AMD Ryzen 9 5900X Multi - 21878/12 - 1823
AMD Ryzen 9 5800X Multi - 15228/8 - 1903
AMD Ryzen 9 5600X Multi - 10988/6 - 1831
Intel Core i9-12900K Multi - 27472/16 - 1717
Intel Core i9-12700K Multi - 22812/12 - 1901
Intel Core i9-12600K Multi - 17660/10 - 1766
Intel Core i9-12400F Multi - 12454/6 - 2075
Laptop CPU:
Intel Core i9-12900HK Multi - 18165/14 - 1297
AMD Ryzen 9 6900HS Multi - 13977/8 - 1747
and logical cores
AMD Ryzen 9 5950X Multi - 28577/32 - 893
AMD Ryzen 9 5900X Multi - 21878/24 - 911
AMD Ryzen 9 5800X Multi - 15228/16 - 951
AMD Ryzen 9 5600X Multi - 10988/12 - 916
Intel Core i9-12900K Multi - 27472/24 - 1145
Intel Core i9-12700K Multi - 22812/20 - 1140
Intel Core i9-12600K Multi - 17660/16 - 1104
Intel Core i9-12400F Multi - 12454/12 - 1037
Laptop CPU:
Intel Core i9-12900HK Multi - 18165/20 - 908
AMD Ryzen 9 6900HS Multi - 13977/16 - 873
I am not sure which result is more reliable (I am personally leaning towards physical cores result, you can share your opinion below) in the one hand Intel has a lead with performance cores but also disadvantages with efficiency core so it is up for debate.
I added two scores up and divided them by two just for lol-s:
Intel Core i9-12900K - 1431
Intel Core i9-12700K - 1520
Intel Core i9-12600K - 1435
Intel Core i9-12400F - 1556
AMD Ryzen 9 5950X - 1340
AMD Ryzen 9 5900X - 1367
AMD Ryzen 9 5800X - 1427
AMD Ryzen 9 5600X - 1373
Laptop CPU
Intel Core i9-12900HK - 1102
AMD Ryzen 9 6900HS - 1310
Comparison between 14 cores 20 thread Intel CPU-s one laptop and one desktop on PassMark:
PassMark - CPU Mark for:
Intel Core i9-12700K - 34188
Intel Core i9-12900HK - 29867
PassMark Single Thread Performance
Intel Core i9-12700K - 4050
Intel Core i9-12900HK - 3917
PassMark Single Thread Performance puts this Intel Core i9-12900HK at the top of the chart of laptop CPU-s and almost at the top of desktop ones
(one would think this CPU is magical at least chart said it is).
Comparison between 8 cores 16 thread AMD CPU-s one laptop and one desktop on PassMark:
PassMark - CPU Mark for:
AMD Ryzen 9 5800X - 28346
AMD Ryzen 9 6900HS - 24390
PassMark Single Thread Performance:
AMD Ryzen 9 5800X - 3484
AMD Ryzen 9 6900HS - 3354
Point being PassMark single thread performance test and other similar single thread tests favors heavily laptop CPU-s and better implementation of dynamic voltage and frequency scaling technologies and gives them performance that they don't have in the "real world".
Emphasis on PassMark Single Thread Performance metric, as most important as stated in the knowledgebase, is just wrong in my opinion and should be avoided. As knowledgebase states "Because if any one core becomes overloaded, it does not matter how many other cores you have" it is a valid point but that load will never be single core test related rather "multi-core score divided by number of cores" related. Method for obtaining single thread score (testing it in a vacuum) makes it inaccurate and unreliable.
When you are in the market for new CPU, ask your self how big are your projects going to be and choose CPU with adequate amount of cores relative to your project size estimation, difference between this modern desktop cores in FL project workloads is negligible. Don't forget that laptop CPU-s are deceptively fast, single threaded performance seems comparable with desktop but multi-threaded one is nowhere near.
Meaning performance is nowhere near.
You may think I am going too far but I think we are not going far enough.
I am not gonna promote my request, that was not the point of this post, but it is food for thought.
As some guy named TED once said "Ideas Worth Spreading".
Thank you for your time FLooper Reader
...and FL makers of course
Single Thread Performance Myth
Return to “FL Studio Users Forum (Looptalk)”
Forum rules
Please read them here.
Please read them here.
[You can only see part of this thread as you are not logged in to the forums]
Re: Single Thread Performance Myth
You can't simply divide it by the amount of cor...
Re: Single Thread Performance Myth
single core performance is not that important a...
Re: Single Thread Performance Myth
sql_error_uname wrote: ↑Thu May 12, 2022 5:11 p...
Re: Single Thread Performance Myth
Getbuzzed wrote: ↑Thu May 12, 2022 5:49 pm
sql_...
Re: Single Thread Performance Myth
Getbuzzed wrote: ↑Thu May 12, 2022 5:49 pm
They...
Re: Single Thread Performance Myth
sql_error_uname wrote: ↑Thu May 12, 2022 5:53 p...
Re: Single Thread Performance Myth
Well while multi thread performance is obviousl...
Re: Single Thread Performance Myth
You can talk measurements all you want, but if ...
Re: Single Thread Performance Myth
visiblenoise wrote: ↑Thu May 12, 2022 11:34 pm
...
Re: Single Thread Performance Myth
Getbuzzed wrote: ↑Thu May 12, 2022 6:56 pm
I am...
Re: Single Thread Performance Myth
sql_error_uname wrote: ↑Fri May 13, 2022 3:38 a...
-
- Content Creator
- Fri May 13, 2022 9:04 am
Re: Single Thread Performance Myth
Single thread performance is far from a myth. I...
Re: Single Thread Performance Myth
shroomhead_one wrote: ↑Fri May 13, 2022 9:04 am...
You do not have the required permissions to view the files attached to this post.
Re: Single Thread Performance Myth
This is re-post from the other topic, just to k...
You do not have the required permissions to view the files attached to this post.
Re: Single Thread Performance Myth
This is re-post from the other topic, just to k...