Latency on different CPUs

Hello, a little silly question: I use Cubase 7.5 on a Windows 7 PC (Asus motherboard with H55 chipset and Intel i3 550 3.4 GHz CPU - a little dated, I know …).
I want to update the CPU (for various reasons I don’t want to change the motherboard), and the more powerful CPU compatible with my socket LGA1156 would be the Intel i7 880 3.07 GHz.

I am not sure if this CPU will also improve the ASIO latency (the CPU frequency is slightly lower than that of my i3). From what I understand, the i7 manages simultaneous processes faster, but will I get an improvement (having almost equal frequency) even on individual processes (given that the ASIO latency depends on a single process, I’m not sure …)?

Can anyone help me? Thanks!


Yes, the two additional cores will help considerably.

Technically, what you would need to set the sample buffers to for glitch-free DAW operation would depend upon what is likely to be the worst instantaneous performance of any of the cores. Not easy to determine, which is why the ASIO Performance meter exists.

The more cores you have, the less likely any one core will be saturated, so for a given CPU clock rate (and instructions per clock cycle), more cores eases the burden.

However, I have noticed that Cubase does slightly favour one thread (which may be due to the GUI handling) and hence that may result in a tendency to push the ASIO meter more than if the CPU usage was completely even-handed.

Thanks Patanjali, may I “translate” your post?: in theory, the answer to my question would be “yes i7 would be better”, but based on how Cubase manages the different threads, it could be “not much difference”. Did I get it right?


I do not agree with the conclusion. Additional cores will always be used by Cubase if available and loads will be fairly equally distributed under most circumstances with the exception of Core 1 thread 1 which appears to bear the brunt of the UI.

Basically, yes.

Unless you really up the ante, it is a bit of a crap shoot, as there are so many things in the real-world operation of a computer that can mitigate against reaping the benefits of upgrades.

Which is as I said:

and Paolo concluded:

But one doesn’t always get the benefits one expects. :cry:

Yes, there are many variables which must work in harmony to derive superior results, no question about it.

Yes. Just by looking at SSD performance figures, one might expect that installing them should turn any PC into an outright speed demon. Instead, except for a few particular times entirely dependent upon disk speed, most improvements are nowhere near the order of magnitude one might expect, mainly because the OS and programs are interleaving the disk accesses with their own processing, so the improvements are only occurring in the small time slices between that processing.

When comparing generations of CPUs, one cannot just look at the clock frequency, but performance will depend upon how many instructions are executed per clock cycle. In earlier days, there were many clock cycles per instruction.

When the i386s were around, they actually free-wheeled between cycles (that is, executed as fast as the internal logic could flip), which meant that the actual performance could vary wildly from chip to chip, with the Norton SI speed utility registering variations from 18 to 22. The computer manufacturer I worked for at the time used to hand select chips to get the best reviews. What a crap shoot! Imagine trying to build a better DAW with that sort of variation?

Yes, that any of this works at all is a miracle to me. In 1995, an infinite amount of money could not have bought what can be done today inside a 2k PC. Amazing and exciting. One of the few bright spots in today’s world.

Given the number of possible points of failure, I am always amazed that they work at all.