System upgrad with faster cpu won't improve performance :-(

Hi people,

I’ve been running Cubase 8 on a 2011 imac with an i7 3,4ghz processor. Recently I have upgraded to a system with an i7-4790k 4ghz processor. In cpu benchmarks I get about 40-50 percent improvement over the old system so I had high expectations for Cubase.

Sadly after testing several projects I’ve come to the conclusion of an average improvement of only 7-10 percent. I’m checking the average performance in the performance meter in Cubase to make these conclusions. Very sad and poor result indeed. Should performance really scale this poorly over faster cpus on Cubase? I’m running osx like I said and I haven’t ever tried windows Cubase as osx is my environment.

This was a very disappointing update :frowning: tried finding good benchmarks to run for compassion but they all need access to a variety of sample libraries and paid plugin I don’t have.

EDIT:

So I read up on the whole topic of ASIO meter and overall performance with threads and what not. Measuring cpu use instead to compare the two computers I get a puzzling result.

First off, 3 song projects:

Cpu usage playing the song at the same latency gets the new system an embarrassing 0-2% edge over the older system. EVEN WORSE than I first calculated when solely looking at the ASIO meter.

These projects do not use more cpu power than 10-15% on either machines, so I decided to make something heavier. I load up a channel with a kontakt 5 player, a lead-sample and add a multiband compressor to the channel. Then I duplicate this channel so there is 20 in total. Here the machines are starting to give different results:

20ch: new system claims 14% lower cpu use compared to the older system.
40ch: new system claims 20% lower cpu use.
60ch: new system claims 38% lower cpu use. Here the older system is just barely making it and when the mail.app puts a notification on the screen it gives up and won’t recover until tracks are deleted.
80ch: can’t run at all on old system, but works on new system with 86% cpu, further suggesting the new system is capable of somewhere around 35-40% more.

WHY HOWEVER does this only appear when under heavy load by an artificial project? I’m not even sure this would translate well to a real project with a much bigger variety in the tracks. Any suggestions here on how to interpret this?

EDIT 2:

Now tried finding this scaling in a normal project by duplicating tracks with a mixture of midi+audio tracks and I see absolutely NOTHING of it. The new system sticks to being about 2% better than the old. I hope I’m missing something or this is a total upgrade fiasko!

You are mixing all kinds of things.
Your findings on the last post were logical though.

  1. base clockfrequency does make a difference overall. So a lower base frequency gives you in day to day usage a littlebit more resources under the same testing conditions. The difference is not liinear, so f.e. a I7 3,4 Ghz is not 25% slower as a i7 at 4Ghz. It’s maybe noticeable, but hardly.
  2. I7 is only raising the clockspeed when needed (turbo function) When the processor notices data queing up in the cache it will start raising the clockspeed so things get calculated more quickly. So yes, when you start multiplying your test project you start to see the difference between the two. But 40% is huge imho. It depends on what you are doing and what is being played back.

A bigger difference between the two i guess is the fact that processor 1 is a mobile one and the other is a desktop processor. Now that is a big difference. A mobile one at the same base clockspeed does not have the same performance possibilities as the counterpart from the desktop.

Look here for some info on it: http://superuser.com/questions/808776/whats-the-difference-between-mobile-and-desktop-processors

So replacing an i7 for another i7 is very often not worth it, but it could be worth to change because it’s clock speed is much higher, and it has more cores or it has much more cache. But overall, you should look to what you really need. There is not a single reason to go for a 4790k, which is one of the more powerfull processors in term of possible resources, if you are not going to use it.

An example where it is usefull is when you need f.e. more then a hundred active vsti’s in real time. But that is just an example.

So most of the people have a way to powerfull processor for what they need. I don’t care, but hey, it’s a bit snobbery after all. And intel is very happy with this. :slight_smile:

It’s a bit the same as with cars. :slight_smile: Why buy a ferrari if the speed limit in the country is 100 mph. Because it’s cool to have one.

You are the proud owner of two very vast i7’s! Now it’s time to find a way to use them and be happy. :slight_smile:

kind regards,
R.

Thanks for a good answer. It may be that I have had unrealistic expectations from the start, but I’m still quite puzzled.

I’m not comparing the two cpus, and systems, by pure clockspeed. That wouldn’t be true. I’m comparing my results in cpu benchmarking apps which are telling me that my newer system is about 40% faster than the old cpu-wise. A DAW in my mind, is all about calculations and therefore I was expecting a significant performance increase. Maybe not 40% but well, at least 20-30? I’m not sure whether the cpu in the old iMac is a mobile or desktop version.

Regarding the clockspeed. I’ve got this checked as well. I’m using Intel Power gadget to verify what frequency the cpu is working at during the tests. It’s always sitting at 4,4ghz which is maximum turbo. It’s hanging around at 4,4ghz even at the projects that are only using 10% of the cpu, so this is probably not the reason I’m seeing little performance gains.

Also, as you see from my testing, I DID manage to squeeze out some extra performance with that test-project playing a lot of kontakt 5’s with a multiband compressor. Here I’m perhaps getting the expected performance but this did however not translate over to a real project where I duplicate a mixture of channels featuring both audio and instruments. There, increase in performance stayed at the pathetic 2% mark. This would mean, I assume, that if I get close to the max my old rig could do I wouldn’t be any happier with my brand new machine and that’s just sad when there’s 5 years between the two computers.

I’ve rarely gotten close to the limit but things evolve and what cuts it now might not with the evolution of tomorrows plugins. I’m already noticing an ever increase in average cpu use in my projects due to bigger samples and more advanced plugins. I upgraded not only to get a better headroom but some of the reason for upgrade was that, to futuresafe my workstation and that as it seems, is just not going to happen.

Hey, don’t worry. With the 4790 you will be good for a very long time. It is one of the faster ones and it will cover your needs for a very long time.
As said: do not look at a processor as something that is the limiting factor.
When doing a normal project you will probably never need so much power, and it doesn’t change when you do the same project on a slower processor. They will perform just the same.

All those benchmarks are theoretical approaches of what a processor in theory can handle, not what someone is really doing with it. So it delivers more resources then you need for now, and that’s ok.

The more you learn of cubase, the more you will benefit from the fact that your system is a capable system. The bigger your projects will get, the more fun it will be.

The system is not there for it’s own existence, but for what YOU want to do.

kind regards,
R.

°double post°

Puzzled by these results I decided to turn to Logic Pro x for a comparison between the computers there. I’ve read so much about how lousy Cubase is performance-wise on osx and we all know that Logic is the king in the performance domain on mac. Now I’m not really interested in what daw is most effecient, but I’m trying to figure out if it might be Cubase osx that just won’t scale to better cpus or if it’s something else going on.

Loaded up the helena beat demo project on both old and new computer. A pretty light project with audio and some synths. Results? new computer is performing 22% better than the old. I raised an eyebrow. Intrigued I tossed my biggest logic-project into both computers. It should be noted that I moved from Logic to Cubase some time ago so thankfully I had some typical projects left to test with. This is a larger project with quite some effects mixed with instruments and vocals, choirs etc. The result? 37% better performance on the new machine. I raised two eyebrows and stare at Cubase in the dock, what the hell is going on here?

Finally I downloaded a synthetic performance test for logic as well of the same type as I did on Cubase which was the only area there where the newer computer showed any gains. Logic side of things? average of 40,5% improved performance on the new computer which aligns close to perfect with the cpu benchmarks between the systems: 877 vs 600 cb in Cinebench cpu test (46%).

Now tell me: what the heck is this? Unless I’m missing something in my testing, Cubase on osx seems severely flawed in handling the latest more powerful cpus for anything but the most synthetic of tests. Unfortunately I do not have the tools at hand right now to test Cubase in a windows installation on the same computers because it sure would be interesting doing so.

Would love some comments on this and let’s not make it a lpx vs cubase on mac. I love Cubase. I want the value of my new hardware investment though.

Roel: I understand your point and it’s a perfectly valid one. This machine is perfectly good enough in Cubase for my needs. Still, it angers me somewhat to not see any performance gains for future use from my quite expensive upgrade while as it appears: another daws is showing performance boosts as expected.