DAW CPU Performance Database - We Need It!

Envato Authors AudioJungle
1
  • OUR PROBLEM:

Hi everyone! This is AudioTrend. Please let me start :blush:

We all know that CPU is the most IMPORTANT thing in our computers when it comes to DAW performance. Synths, sample libraries, and effects hugely rely on it. Choosing RAM, SSD or HDD for our purposes is generally not a big problem. However, things are getting harder when you start choosing a CPU. I found that there are almost no fresh comparisons of different CPUs in terms of audio processing performance. For gamers and video creators there are tons of benchmarks, but generally no useful tests for us, audio engineers and music producers.
Therefore, it is HARD to make a right decision choosing CPU for your audio workstation PC.

  • SOLUTION:

So I came up with the idea to create some sort of HUGE database where EVERYONE can test his CPU and submit the result for everyone to see it. This way we will HELP others and ourselves choosing the most appropriate CPU for audio needs.

  • DO I NEED IT?

Just think about it. You may not need that information right now because you are pretty OK with your current system. But imagine how glad you will be in future when you will think about upgrading your outdated CPU and knowing that there is SUCH database like this. It will save you TONS of time and effort.

  • EXPERIMENT:

I was thinking about making this test fast and easy for everyone. I know most of us don’t have a lot of spare time. After some research, I finally found a FREE synth and a FREE effect. Both are CPU-hungry and easy-to-setup. Actually, it was pretty difficult to find such plugins available for Win32, Win64 and Mac at the same time. We got “SGA1566 Tube Preamp” EFFECT and “Tyrell N6” SYNTH.

I downloaded versions for 3 systems to my Google Disk (Win32, Win64, and macOS). I did it because we all want everyone to run a test with EXACT SAME copies of these VSTs (not modified, updated, etc). So please don’t download these plugins from anywhere else. You may not worry about the viruses - Google automatically searching for viruses once you click download. You can additionally upload them to Virustotal if you wish. Also, I have to add that it is highly important that everyone does exactly SAME things during this experiment if we want to get relatively objective and useful results. Here’s how I suggest running the test.

  • HOW TO:

I spent about 2 days designing the experiment and building a nice Google form. So you should go to GOOGLE FORMS and complete ALL the steps RIGHT THERE (the link is below). One step is done? Right away reply to the question in the Google Form. I will post all the steps here just to show what this test looks like. You can ALSO duplicate your results here, but it is ESSENTIAL to INPUT your data into Google FORM. That way we will get TONS of useful data and then ANALYZE it using Google Sheets.

  • RESULTS

When we will get some decent number of completed tests - I will post the raw data in one of the formats: XLSX (Excel), CSV, ODS or PDF. Then you can open it and easily sort all the results by CPU name and GET that DESIRED information! You can also send me an e-mail and I will make sure to send you the data in a first place!
The more people participate - the sooner I will post that full data!

  • EASIER THAN IT SEEMS:

You may feel that there are a lot of letters in the post but actually, the test is pretty SIMPLE and QUICK. You will see it for yourself.

  • THE MAIN PAGE:

Link to the experiment itself - http://bit.ly/DAW-CPU-DATABASE
OR CLICK THE BUTTON
rounded_corners

The most important thing is to CONDUCT the EXPERIMENT in GOOGLE FORMS, not here or somewhere else.

  • SHORT PLAN i.e. WHAT TO EXPECT:

  1. The first page contains several easy questions about your software and hardware.

  2. On the second page you will find 2 plugins and some quick recommendations to prepare (really quick!)

  3. The third page is a SYNTH test (takes about 1 minute to complete)

  4. The next step is a combined SYNTH + EFFECT test (takes even less time - just about 30 seconds)

  5. And the final test is called MAX LOAD (may take from 20 seconds to 2 minutes depending on your CPU)

Last 3 steps contain short video walk-throughs in case you not sure how to do it right.

  • THAT’S IT!

As you can see this test should take no more than 5 minutes of your time.

  • CONCLUSIONS - PROFIT FOR EVERYONE

If EACH of us will participate in that experiment we will get tons of useful information.
You may ask “What will I get by doing the test?”

You will get an amazing database allowing you to compare how good or bad different CPUs for music production and sound engineering purposes.
One day you will need to get a newer CPU and you will be happy to have such highly useful database.
As I said before, there is almost no such info on the web. So we are building a pretty exclusive thing here together.
And also, this is wonderful that by doing this test you are making a decent contribution to the worldwide sound producer’s community.

P.S.

  • GREAT! WHAT NEXT?

You can conduct the experiment right away, you got the link :blush:
One more thing: it is HIGHLY important to keep this thread alive in order to get more and more NEW CONTRIBUTIONS!

You may post here a lot of useful things:

  • your thoughts on the whole idea
  • your ideas
  • your results
  • criticism, ideas to change/fix something
  • anything related to the topic

THANK YOU!!

1 Like
2

Bummer…I went all steps and in the end realized that Bitwig Studio doesn’t show CPU load in % :joy:

On the other hand i don’t think this is an accurate test as many DAWs has some additional performance settings rather than buffer size. No guarantees it will match.

1 Like
3

Sorry to hear that :frowning: Well, you could at least complete the latest test, it doesn’t need % meter :blush:

Sure, this test is not ideal. But I’m assuming that most users have set up everything correctly inside their DAW to achieve the best performance possible. I looked at settings of several DAWs and there are not many CPU-related settings that people would use incorrectly. I agree that there is a lot of possible differences in results. But the main point is to get so many results so we can estimate some average performance of a certain CPU. Great accuracy is not the main goal here. Usefulness is :blush:

Anyways, we will be able to sort the results by CPU name, DAW name, OS type. So we should get a nice estimated picture in the end :floppy_disk:

4

Thank you for taking the time to do this, what do you want the sample rate and buffer size at? Eg. 44.1 Khz & 32 Samples… because that plays a role on your CPU performance too, not just that your sound card I/O driver and daw all have variables to add to the equation. check this video out too, it just puts more spanners into the works…

5

I know this is a long shot, but if someone could develop a program similar to Cinebench as a standalone application that utilizes CPU intensive Audio tasks, all these variables might fall away, Code Canyon? :smiley:

1 Like
6

The test contains detailed steps and instructions (yet it’s quick), including a sample rate and a buffer size parameters. In that test they are 128 samples and 44.100 Hz. I think these settings are pretty realistic and commonly used.

I don’t think that there are a lot of things that matter here. This test relies 99% on CPU. These VST synth and effect essentially don’t use RAM, SSD or HDD resources. Sure, the test doesn’t try to run 200 instances of some small VST (like DawBench). And I don’t think that this is needed because (I believe) that the main problem of sound producers is a pretty little number of CPU-hungry VSTs (HQ reverbs, oversampling limiters, saturation, etc) and not thousands of small EQs or panning effects. Or 300 sine wave synth instances. Who does that at all? Besides, even if I would create a test like that, who would participate? It would take about 1 hour to complete such experiment.

So yes, this test doesn’t take into consideration all possible variables. Only main variables. The main point is to get tons of data so we get look at at average performance of a certain CPU.

On of the advantages of this test (I believe) is close-to-life conditions and ease of participation. It takes max 5 minutes to complete :blush:

I support your idea to create “Cinebench” for audio. However, we need someone who can do that and promote that software so we could get tons of data.

1 Like
7

I took the test & I look forward to the results!

1 Like
8

Thanks, mate :blush: Appreciate your participation. I will publish all the results as soon as we collect a nice number of completed experiments.

1 Like
9

It is to praise your effort and your good intentions trying to help us choosing the best CPU for audio work. But there are so many nuances that must be taken into account that’s almost impossible to get accurate results and come to trustful conclusions on this matter running the tests the way you’re proposing. Unless the tests will be ran in the same “virgin” machine, I mean: the same computer with a fresh installed OS with a (well tuned) DAW (or different DAWs), the same sound card, the necessary plugins, the same MIDI / audio files to run the tests, and nothing else running under the wood. Now, lets try to get a bunch of different CPUs (Intel or AMD… the latest implies having another compatible MB)… and experiment them. Only now we can get some accurate results and can take trustful conclusions.
Well, as a rule of thumb to choose a CPU for MIDI / audio work, the highest clock rate, number of cores an number of threads per core your money can buy, the better!
But, as I am a really curious guy and it will take only some minutes of my time, I will run your tests anyway! At least I’m also giving my tiny contribution to a really nice and friendly community here! :wink:
Thanks a lot for your effort and for helping us all here @AudioTrend! :slight_smile:

1 Like
10

Thank you for the participation SonicFox! This very nice you decided to do it despite the fact you do not agree with the methodology used.

And thanks for the criticism. You are right that this experiment is not highly accurate. A lot of variables. However, I believe that data is still can be useful if a number of completed tests will be huge. We will be able to compare an average performance of each CPU.

11

What would the sequencer have worked clearly as a clock! We need a stationary computer and an Intel processor!

12

Yeah its a nice idea and I commend you for putting the effort into making this but, there really are so many variables that will affect performance here. The most important thing isn’t cpu choice, so much as how well integrated all the components are on a motherboard that plays well with them.

Not to mention audio interface can have quite an impact on ASIO performance (Focusrite Scarlet 2i2 for example being particularly bad choice vs say a high end RME). Then there’s all the particular drivers installed on the machine for all the components and whether they’re causing DPC latency interrupts that are also affecting performance on that machine. And finally DAW choice also makes a big impact, Reaper for example being especially good at cpu efficiency.

In short, two PC’s with the same cpu could give wildly different results because of all the other component choices , DAW used, audio interface and drivers installed.

1 Like
13

That probably explains some of the pain I’ve been suffering recently when I’ve been working on some more complex tracks! Thanks for the info as I hadn’t realised that the choice of audio interface could have such an impact. :slight_smile:

14

Audio interface can have a huge impact, somewhat so that you can have a beast machine capable of NASA type calculations, but if the audio interface is not up to scratch, it causes bottle necks and does not use the beast machine to its full potential.

15

For better CPU performance use mac platform. Just my small advice.

16

Page 2:
"- no applications are running in the background (e.g. browser, antivirus, player, etc)
NOTE: leave only this tab opened"
I’m not sure that all “testers” can leave this tab open - crackle and overloading CPU with any browser opened in my case (Reaper+XR18 WASAPI exclusive, Windows 10 x64).
“WASAPI shared” also isn’t the solution for it’s impacts on CPU and mixing quality.