Confusion with hardware specs and their performance.

[Luxuria]

I built a nice new gaming/music production desktop around November lats year. Before that, I had a macbook pro retina (2012) that ran Reason really well, still does actually.

I knew something was up when I first started using Reason(I'm on 8.3) on the desktop because I'd see the DSP meter flicker and spike more often than I ever saw it on the macbook.

Same settings on both systems. Same interface, CPU uncapped max out in settings. 44.1/1024 buffer. Ran the same project sessions through them both.

I float around the forums enough to know Pheads aren't the greatest programmers, I'm not trying to beat a dead horse deader.

I just want your guys' opinion on what might be the cause of the performance discrepancy?

Desktop CPU: Intel i3-8350k @ 4.00Ghz
Macbook Pro Retina CPU: i7-3820QM @ 2.70Ghz

Both are running Windows 10. Both have 16 gigs RAM, desktop is much newer DDR4. GPU doesn't matter much here but the desktop has a GTX 1060SC and the mac is a GT650M.

You'd think I should be getting better performance with the desktop right?

Here's the userbenchmarks on the CPUS http://cpu.userbenchmark.com/Compare/In ... 4339vs3935

[aeox]

It looks like the i7-3820QM has 4 cores/8 threads and the i3-8350k has 4 cores with no hyperthreading if i'm reading the specs right. The i7-3820QM also can turbo to 3.7ghz.

1.png (19.78 KiB) Viewed 2460 times

Although the i3-8350k seems to have better benchmark score, maybe Reason just works better with that CPU with 8 threads?

[antic604]

You'd think I should be getting better performance with the desktop right?

Yes. The question is are all the other Reason settings the same accross the MacBook and Desktop PC: multi-core and hyper-threading, bit depth, audio rate and audio buffer. Also, is the PC in high-performance mode?

[Luxuria]

Although the i3-8350k seems to have better benchmark score, maybe Reason just works better with that CPU with 8 threads?

I think that would be the only logical answer. The i7 runs projects that are about twice the track count and twice the amount of insert fx.

My first thought was that it was the interface causing me trouble. The Focusrite 2i2 (gen 2) I have seems to stutter at times if it isn't set to 96k/1024. I've been eyeing the RME Babyface pro as the upgrade to the 2i2. RME gets a good rep for stable drivers and really low latency speeds.

Maybe I should hold off on the interface and upgrade to one of the new i5's or i7's.

My only concern is my lack of knowledge in how much CPU architecture plays a factor in DAW performance? The new Coffee Lakes, which my i3 is, should be giving me closer results to the i7 on Ivy Bridge architecture. (edit)
The i3's performance is near equivalence of an i5-7600k.

Anyone with a Coffee Lake chip getting worse performance than expected?

[antic604]

It looks like the i7-3820QM has 4 cores/8 threads and the i3-8350k has 4 cores with no hyperthreading if i'm reading the specs right. The i7-3820QM also can turbo to 3.7ghz. Although the i3-8350k seems to have better benchmark score, maybe Reason just works better with that CPU with 8 threads?

Hyper-threading doesn't "magically" add extra performance - it just effectively "splits" the physical cores in half (in terms of available processing per second). This might be useful when running multiple applications as some of them have the tendency to take over the core even if they don't use it in full or when there's significant performance penalty when switching the core's context (as in which application is using it) - then it's better to have more cores.

In DAW, where ideally you're not running anything else along with it, hyperthreading doesn't really help and might actually cause problems if single device chain (instrument+effects) is so demanding that a split core isn't able to process it in time, whereas a "full" core would manage this.

Lastly, there's this - at least when I enable hyperthreading on i7-6650U (Surface Pro 4) - pay attention to the CPU core utilisation graph in the middle of the status bar to the right - Reason only uses 3, while Bitwig and Live use all 4:

[Luxuria]

You'd think I should be getting better performance with the desktop right?

Yes. The question is are all the other Reason settings the same accross the MacBook and Desktop PC: multi-core and hyper-threading, bit depth, audio rate and audio buffer. Also, is the PC in high-performance mode?

Yup! All settings are the same and the PC is in high-performance mode.
I use 8.3 so neither system has the option to enable hyper-threading, unless Reason uses the HT on the i7 without needing an option in the preference window.
Both have multi-core rendering enabled.

Could be just the way the program runs and it getting more power out of the i7 I guess.
Just would be nice, not having to print to new track after 5 tracks with effects applied to them.

[antic604]

The i7 runs projects that are about twice the track count and twice the amount of insert fx.

This really shouldn't be happening, because - on paper at least - the i3 beats the i7 in all loads: single core, quad-core and multi-core by 13 to 64%.

So maybe it does have something to do with drivers for the audio interface? Have you tried - just for testing purposes - using the generic ASIO driver (http://www.asio4all.org/) or even not using the audio interface at all (I'm assuming the PC has some on-board audio card)?

Also, can you post a screenshot of your Task Manager when running a song that results in high DSP on the PC? Like so:

[EnochLight]

I use 8.3 so neither system has the option to enable hyper-threading, unless Reason uses the HT on the i7 without needing an option in the preference window.
Both have multi-core rendering enabled.

Nope - Reason didn't begin supporting Hyperthreading - as an option - until 9.0/9.5. You can enable it manually, but it's a messy process that deals with manually modifying a file. It's unlikely that's your issue. Antic604 has some great points, though.

[Luxuria]

The i7 runs projects that are about twice the track count and twice the amount of insert fx.

Also, can you post a screenshot of your Task Manager when running a song that results in high DSP on the PC? Like so:

Here's a high dsp part captured

I have tried the basic ASIO4ALL drivers and resulted with disappointment.

Let me try updating/resetting the drivers for the 2i2. I don't remember checking if the drivers had any updates.

[antic604]

Thanks, that looks pretty normal with one exception - just like in my case, the 4th core seems to be significantly underutilised. While in my case this is a 'virtual' core with effectively a 1.5GHz clock, in yours it's a full physical core with 4GHz.

Intriguing, but difficult to make any binding conclusion from that...

[aeox]

It looks like the i7-3820QM has 4 cores/8 threads and the i3-8350k has 4 cores with no hyperthreading if i'm reading the specs right. The i7-3820QM also can turbo to 3.7ghz. Although the i3-8350k seems to have better benchmark score, maybe Reason just works better with that CPU with 8 threads?

Hyper-threading doesn't "magically" add extra performance - it just effectively "splits" the physical cores in half (in terms of available processing per second). This might be useful when running multiple applications as some of them have the tendency to take over the core even if they don't use it in full or when there's significant performance penalty when switching the core's context (as in which application is using it) - then it's better to have more cores.

In DAW, where ideally you're not running anything else along with it, hyperthreading doesn't really help and might actually cause problems if single device chain (instrument+effects) is so demanding that a split core isn't able to process it in time, whereas a "full" core would manage this.

Lastly, there's this - at least when I enable hyperthreading on i7-6650U (Surface Pro 4) - pay attention to the CPU core utilisation graph in the middle of the status bar to the right - Reason only uses 3, while Bitwig and Live use all 4:

I never said hyper threading magically adds performance but whatever project file he is testing on is obviously benefiting from having 4 more threads at it's disposal, right? If all else is equal, how can it not be true?

[antic604]

I never said hyper threading magically adds performance but whatever project file he is testing on is obviously benefiting from having 4 more threads at it's disposal, right? If all else is equal, how can it not be true?

But that's physically not possible (also, as EnochLight said Reason wasn't supporting hyperthreading until v9 and OP is on v8.3). I'm betting on audio interface drivers.

BTW OP, what's your screen resolution on the PC? Reason doesn't use GPU to render its GUI, so maybe that's where some of the theoretical performance is vanishing if it's a 4K screen(s).

[Luxuria]

Update: So I was pleasantly surprised that the desktop did not have the most up to date focusrite drivers!

That helped only partially though. I tested a different song with all midi info I had on both machines.Same project file in both machines with same interface, same audio settings and came up with weird data.

The desktop now runs at 3 dsp bars and around 48% CPU utilization.

The macbook pro runs at 4-max dsp bars resulting in glitches while displaying only 45% CPU utilization.

I'm happy that my performances now reflect my gear a little bit better, but still am puzzled as to how the macbook runs 3% less extensively yet struggles keeping the dsp bars down within Reason?

[Luxuria]

BTW OP, what's your screen resolution on the PC? Reason doesn't use GPU to render its GUI, so maybe that's where some of the theoretical performance is vanishing if it's a 4K screen(s).

PC runs at 1080. The macbook does have retina display which is running at 2800x1200.

[antic604]

I'm happy that my performances now reflect my gear a little bit better, but still am puzzled as to how the macbook runs 3% less extensively yet struggles keeping the dsp bars down within Reason?

DSP is not only about CPU, but also HDD speed, communication between various computer components: busses, memory, etc. or the strain that's being put on the computer by background processes.

Also, comparing CPU utilisation between OSes might be misleading, because you can't be certain they measure it in exactly the same way.

Anyway, happy to see that your experience has improved!

[aeox]

I never said hyper threading magically adds performance but whatever project file he is testing on is obviously benefiting from having 4 more threads at it's disposal, right? If all else is equal, how can it not be true?

But that's physically not possible (also, as EnochLight said Reason wasn't supporting hyperthreading until v9 and OP is on v8.3). I'm betting on audio interface drivers.

BTW OP, what's your screen resolution on the PC? Reason doesn't use GPU to render its GUI, so maybe that's where some of the theoretical performance is vanishing if it's a 4K screen(s).

Oops I missed the part about v8.3

[Luxuria]

I'm happy that my performances now reflect my gear a little bit better, but still am puzzled as to how the macbook runs 3% less extensively yet struggles keeping the dsp bars down within Reason?

DSP is not only about CPU, but also HDD speed, communication between various computer components: busses, memory, etc. or the strain that's being put on the computer by background processes.

Also, comparing CPU utilisation between OSes might be misleading, because you can't be certain they measure it in exactly the same way.

Anyway, happy to see that your experience has improved!

Thanks man!

My SSD on the desktop is also +175% better than the one in the mac according to userbenchmarks. Every component on the desktop beats out the Mac by a good factor. It's interesting seeing the subtle differences that each computer displays in performance. Both ran Winows 10 as well, one native, other bootcamp.

I plan on upgrading to a i7-8700 in the future as I do realize my little i3 has it's weaknesses. I have it liquid cooled and can OC to 4.20Ghz, but I have't had the need to push it for gaming sake.

Gotta update those drivers! They are so crucial for properly working gear.

[candybag]

I plan on upgrading to a i7-8700 in the future as I do realize my little i3 has it's weaknesses. I have it liquid cooled and can OC to 4.20Ghz, but I have't had the need to push it for gaming sake.

Gotta update those drivers! They are so crucial for properly working gear.

Or maybe wait for the new i9 series that are due sometime in october. 8 physical cores and HT on a solid platform seems like a big deal. Their new i7 series will not have HT tho, only 8 physical cores.

Personally i'd never buy or recommend an i3 for audio production

[Fotu]

My first thought was that it was the interface causing me trouble. The Focusrite 2i2 (gen 2) I have seems to stutter at times if it isn't set to 96k/1024. I've been eyeing the RME Babyface pro as the upgrade to the 2i2. RME gets a good rep for stable drivers and really low latency speeds.

FWIW I recently replaced my Focusrite 2i2 (gen 1) with an RME and that eliminated stuttering issues I had long thought (apparently incorrectly) were general performance issues with my relatively old PC.

[sublunar]

My first thought was that it was the interface causing me trouble. The Focusrite 2i2 (gen 2) I have seems to stutter at times if it isn't set to 96k/1024. I've been eyeing the RME Babyface pro as the upgrade to the 2i2. RME gets a good rep for stable drivers and really low latency speeds.

FWIW I recently replaced my Focusrite 2i2 (gen 1) with an RME and that eliminated stuttering issues I had long thought (apparently incorrectly) were general performance issues with my relatively old PC.

I used to own two 18i20's. They ran like shit in Windows 10. I updated the drivers, I used the old drivers and I tried it out with the factory shipped firmware and updated firmware. They just couldn't handle high res/low latency, despite being awesome on paper. I bought an M-Audio interface instead and right out of the box there were no issues. Running strong even at the lowest latency settings. The experience really soured me on Focusrite. Your mileage may vary.