Reason 9.5 & above CPU Stress Tests (2017, two different songfiles included)!

This forum is for discussing Reason. Questions, answers, ideas, and opinions... all apply.
User avatar
Kategra
Posts: 327
Joined: 18 Jan 2015

01 Jun 2017

devilfish wrote:Windows 10 x64
Intel Xeon E5-2683v3 (QFQK - 250€ eBay)
16GB RAM ECC
Mackie Onyx Blackjack USB Audiointerface
44.1kHz / 95% / 256 Samples (7ms/9ms)
Ati Radeon EX 460
Super Silent PC, you can´t hear anything :D

Complex stop @ 33.2.2.37 (0:01:04:644)
Glad to hear!

What about the Simple FX Chain, can you please test that also? I have feeling that the Simple song favors frequency more than core count :)

User avatar
AttenuationHz
Posts: 2048
Joined: 20 Mar 2015
Location: Back of the Rack-1

01 Jun 2017

Kategra wrote:
devilfish wrote:Windows 10 x64
Intel Xeon E5-2683v3 (QFQK - 250€ eBay)
16GB RAM ECC
Mackie Onyx Blackjack USB Audiointerface
44.1kHz / 95% / 256 Samples (7ms/9ms)
Ati Radeon EX 460
Super Silent PC, you can´t hear anything :D

Complex stop @ 33.2.2.37 (0:01:04:644)
Glad to hear!

What about the Simple FX Chain, can you please test that also? I have feeling that the Simple song favors frequency more than core count :)
Out of interest I would upload a minimised version the songfile also to test. Minimizing a device lowers its dsp usage. That is clicking on the little arrow to make it as small as possible. Took about 3-4 minutes to minimize all the open racks.
It is not too much of an ask for people or things to be the best version of itself!

User avatar
selig
RE Developer
Posts: 11685
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

01 Jun 2017

AttenuationHz wrote:
Kategra wrote:
devilfish wrote:Windows 10 x64
Intel Xeon E5-2683v3 (QFQK - 250€ eBay)
16GB RAM ECC
Mackie Onyx Blackjack USB Audiointerface
44.1kHz / 95% / 256 Samples (7ms/9ms)
Ati Radeon EX 460
Super Silent PC, you can´t hear anything :D

Complex stop @ 33.2.2.37 (0:01:04:644)
Glad to hear!

What about the Simple FX Chain, can you please test that also? I have feeling that the Simple song favors frequency more than core count :)
Out of interest I would upload a minimised version the songfile also to test. Minimizing a device lowers its dsp usage. That is clicking on the little arrow to make it as small as possible. Took about 3-4 minutes to minimize all the open racks.

Two things: 1, FIY option/click on one "minimize triangle" and ALL in the same rack will collapse, and 2, collapsing devices does not have any effect on CPU usage that I've ever seen (including with this test file).
Selig Audio, LLC

User avatar
AttenuationHz
Posts: 2048
Joined: 20 Mar 2015
Location: Back of the Rack-1

01 Jun 2017

selig wrote:
AttenuationHz wrote:
Kategra wrote:
devilfish wrote:Windows 10 x64
Intel Xeon E5-2683v3 (QFQK - 250€ eBay)
16GB RAM ECC
Mackie Onyx Blackjack USB Audiointerface
44.1kHz / 95% / 256 Samples (7ms/9ms)
Ati Radeon EX 460
Super Silent PC, you can´t hear anything :D

Complex stop @ 33.2.2.37 (0:01:04:644)
Glad to hear!

What about the Simple FX Chain, can you please test that also? I have feeling that the Simple song favors frequency more than core count :)
Out of interest I would upload a minimised version the songfile also to test. Minimizing a device lowers its dsp usage. That is clicking on the little arrow to make it as small as possible. Took about 3-4 minutes to minimize all the open racks.

Two things: 1, FIY option/click on one "minimize triangle" and ALL in the same rack will collapse, and 2, collapsing devices does not have any effect on CPU usage that I've ever seen (including with this test file).
Damn son I knew there was a button for it couldn't remember what it was! Alt click. It frees up space in RAM when you minimize. I have been doing it for years on even worse machines than I currently have the post I made shows how much of a difference it makes. It does make a difference granted it might be negligible for some tracks like the benchmark here. But if your project is just barely tipping the too slow alert minimising will get you through the rest of the track if it is a small project that should not have a massive impact on performance.
It is not too much of an ask for people or things to be the best version of itself!

User avatar
devilfish
Posts: 183
Joined: 20 Jan 2015

02 Jun 2017

We need just one Testsong. A longer one.. let it Play 20-30min!! First: It is more accurate, Second it's possible to Test with all CPU's. Single CPU up to 20 Multicore CPU's..
The Signal Chain to copy should be very light on CPU, so you have a much more accurate result.. i will do a Test with the simple fx Song today..
2 Test Songs are not very Handy ...

Sorry for my english :(

User avatar
Kategra
Posts: 327
Joined: 18 Jan 2015

02 Jun 2017

devilfish wrote:We need just one Testsong. A longer one.. let it Play 20-30min!! First: It is more accurate, Second it's possible to Test with all CPU's. Single CPU up to 20 Multicore CPU's..
The Signal Chain to copy should be very light on CPU, so you have a much more accurate result.. i will do a Test with the simple fx Song today..
2 Test Songs are not very Handy ...

Sorry for my english :(
That is what I want, you need to increase the load every few seconds (every sequencer bar), this means more devices need to have sound going trough them. But devices consume considerable amounts of DSP for just being in the rack, even if they are turned off!
Especially for RE type devices like Pulverizer, The Echo, Audiomatic etc; Even 100 Substractors not playing anything, just created in the rack with no midi on their track add 2 DSP bars on my 6800K with song not playing! So how could an i3 CPU type computer play that more than a few seconds file?

The Simple FX Chain is 6 minute long at 130 BPM with 3 different FX tracks:

A) 200+ RV7000 devices loaded with Impulse Response;
B) 200+ Mclass Comp;
C) 200+Mclass EQ with all bands enabled.

The devices are all in Bypass mode and each time a new bar is played, one RV7000+ one MclassComp + Mclass EQ are beeing enabled.
I wanted to add another track

The only 20-30 minute benchmark song in Reason would be made up only by Mclass and old stock FX Devices. But I don't think that would represent any meaningful test for Reason as most of use Synths, Samplers, FX like Pulveriser, The Echo and other REs of all kinds! That is why I made the Complex RE version, to simulate an interactive monster song with all kinda of stock Instruments and FX including the stock REs + automation. We all use automation or LFO don't we? :)

PS. Don't be sorry for your English, I understand you perfectly well and I'm not a native speaker myself! We may speak / write bland and dry but we get the core message across. Do you agree Ostermilk? :P

User avatar
devilfish
Posts: 183
Joined: 20 Jan 2015

02 Jun 2017

Simple Stop @ 34.1.4.65 (0:01:01:301)

edit

complex V2 stop @ 33.3.3.23 (0:01:05:387)


Windowsupdate, Kaspersky and a download is running.. :D
also the browser .. have more than enough power ;)

User avatar
Kategra
Posts: 327
Joined: 18 Jan 2015

02 Jun 2017

devilfish wrote:Simple Stop @ 34.1.4.65 (0:01:01:301)

edit

complex V2 stop @ 33.3.3.23 (0:01:05:387)


Windowsupdate, Kaspersky and a download is running.. :D
also the browser .. have more than enough power ;)



Windows 10 x64
Intel Xeon E5-2683v3 (QFQK - 250€ eBay)
16GB RAM ECC
Mackie Onyx Blackjack USB Audiointerface
44.1kHz / 95% / 256 Samples (7ms/9ms)
Ati Radeon EX 460
Super Silent PC, you can´t hear anything :D
AttenuationHz wrote: Your CPU model : Intel i7 4790k @4.0Ghz no OC
16BG Ram 2x8
Mbox 2 mini will be upgrading that soon.
sample rate 44100 Hz
95%
COMPLEX RE ----> @ 2048 buffer everything is maximised 0:00:02:043 ???
Simple FX Chain ----> @ 2048 buffer 0:01:32:741 @4096 buffer 0:01:36:131



Thanks for posting your results! :puf_smile: :puf_smile: :puf_smile: :puf_smile: :puf_smile:



My results again for comparison:

Complex RE = 24 seconds
Simple FX Chain = 1 min 10 seconds


CPU usage limit set in Reason 8 = 95%
Sample rate = 44,100 Hz
Buffer length = 1024 samples (max )

Hardware:
CPU i7 6800K @ 3.98 Ghz
RAM 16 GB Quad Kit 2400 Mhz, CL14
SSD Samsung 250 GB (Windows 10 Creators edition 64 bit + Reason installed on SSD)
RME ASIO Fireface 400 audio interface


It's interesting, that I get more on the SIMPLE FX, 9 seconds more than the E5 2683, but less than the 4790K which even tough it has no OC, it has a higher Turbo Boost clock than my i7 6800K 4Ghz OC.

Of course, on the Complex, the 14 core beast Xeon E5-2683v3 can handle a lot more than the 6 core i-6800K and the quad core i7 4790K can't play a full bar of the benchmark, mainly because all the stock REs present in song consume DSP just because they were added to the project, they don't have anything to do just be ready to process audio :) Even if they were put to Bypass or OFF that does not drop the DSP they consume by any meaningful amount. Off course, they don't consume as much DSP for being in the rack as they do when sound actually passes trough them, but it's not negligible and that is why most Quad cores can't play the Complex at all.

This means that you can have more RV7000 Convolution Reverbs playing in series on the the higher GHz CPUs! But when you do a complex projects with lots of RE on different tracks it seems that high core count will perform a lot better and that I will upgrade to when the time comes (hope I keep the 6800k at least 3 years)
Last edited by Kategra on 02 Jun 2017, edited 2 times in total.

User avatar
devilfish
Posts: 183
Joined: 20 Jan 2015

02 Jun 2017

Dude, change your Buffer to 256 Samples.. and we talk again.. :)

User avatar
selig
RE Developer
Posts: 11685
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

02 Jun 2017

AttenuationHz wrote:
selig wrote:
AttenuationHz wrote:
Kategra wrote:
devilfish wrote:Windows 10 x64
Intel Xeon E5-2683v3 (QFQK - 250€ eBay)
16GB RAM ECC
Mackie Onyx Blackjack USB Audiointerface
44.1kHz / 95% / 256 Samples (7ms/9ms)
Ati Radeon EX 460
Super Silent PC, you can´t hear anything :D

Complex stop @ 33.2.2.37 (0:01:04:644)
Glad to hear!

What about the Simple FX Chain, can you please test that also? I have feeling that the Simple song favors frequency more than core count :)
Out of interest I would upload a minimised version the songfile also to test. Minimizing a device lowers its dsp usage. That is clicking on the little arrow to make it as small as possible. Took about 3-4 minutes to minimize all the open racks.

Two things: 1, FIY option/click on one "minimize triangle" and ALL in the same rack will collapse, and 2, collapsing devices does not have any effect on CPU usage that I've ever seen (including with this test file).
Damn son I knew there was a button for it couldn't remember what it was! Alt click. It frees up space in RAM when you minimize. I have been doing it for years on even worse machines than I currently have the post I made shows how much of a difference it makes. It does make a difference granted it might be negligible for some tracks like the benchmark here. But if your project is just barely tipping the too slow alert minimising will get you through the rest of the track if it is a small project that should not have a massive impact on performance.
Minimizing the devices on screen cannot free up RAM AFAIK, or the song would not play correctly. Also, this would be an important, useful, and therefore documented tip if true, but I cannot find any mention of it anywhere.

As previously mentioned, these tests to no indicate ANY CPU savings when comparing minimized vs non minimized rack windows with the test songs, nor have I ever noticed this in my work - maybe because I play the test song mostly from the sequencer view (with the rack view hidden)?

Anyone else measuring different CPU hit with minimized vs non-minimized rack views? What test would reveal this advantage if these song files do not?
(Hoping it's true, but not seeing any evidence to support this 'hope')
:)


Sent from some crappy device using Tapatalk
Selig Audio, LLC

User avatar
AttenuationHz
Posts: 2048
Joined: 20 Mar 2015
Location: Back of the Rack-1

02 Jun 2017

selig wrote:
AttenuationHz wrote:
selig wrote:
AttenuationHz wrote:
Kategra wrote:
devilfish wrote:Windows 10 x64
Intel Xeon E5-2683v3 (QFQK - 250€ eBay)
16GB RAM ECC
Mackie Onyx Blackjack USB Audiointerface
44.1kHz / 95% / 256 Samples (7ms/9ms)
Ati Radeon EX 460
Super Silent PC, you can´t hear anything :D

Complex stop @ 33.2.2.37 (0:01:04:644)
Glad to hear!

What about the Simple FX Chain, can you please test that also? I have feeling that the Simple song favors frequency more than core count :)
Out of interest I would upload a minimised version the songfile also to test. Minimizing a device lowers its dsp usage. That is clicking on the little arrow to make it as small as possible. Took about 3-4 minutes to minimize all the open racks.

Two things: 1, FIY option/click on one "minimize triangle" and ALL in the same rack will collapse, and 2, collapsing devices does not have any effect on CPU usage that I've ever seen (including with this test file).
Damn son I knew there was a button for it couldn't remember what it was! Alt click. It frees up space in RAM when you minimize. I have been doing it for years on even worse machines than I currently have the post I made shows how much of a difference it makes. It does make a difference granted it might be negligible for some tracks like the benchmark here. But if your project is just barely tipping the too slow alert minimising will get you through the rest of the track if it is a small project that should not have a massive impact on performance.
Minimizing the devices on screen cannot free up RAM AFAIK, or the song would not play correctly. Also, this would be an important, useful, and therefore documented tip if true, but I cannot find any mention of it anywhere.

As previously mentioned, these tests to no indicate ANY CPU savings when comparing minimized vs non minimized rack windows with the test songs, nor have I ever noticed this in my work - maybe because I play the test song mostly from the sequencer view (with the rack view hidden)?

Anyone else measuring different CPU hit with minimized vs non-minimized rack views? What test would reveal this advantage if these song files do not?
(Hoping it's true, but not seeing any evidence to support this 'hope')
:)


Sent from some crappy device using Tapatalk
https://music.tutsplus.com/tutorials/10 ... audio-1075 - see suggestion #2
It is not too much of an ask for people or things to be the best version of itself!

User avatar
selig
RE Developer
Posts: 11685
Joined: 15 Jan 2015
Location: The NorthWoods, CT, USA

02 Jun 2017

AttenuationHz wrote:
selig wrote:
AttenuationHz wrote:
selig wrote:
AttenuationHz wrote:
Kategra wrote:
devilfish wrote:Windows 10 x64
Intel Xeon E5-2683v3 (QFQK - 250€ eBay)
16GB RAM ECC
Mackie Onyx Blackjack USB Audiointerface
44.1kHz / 95% / 256 Samples (7ms/9ms)
Ati Radeon EX 460
Super Silent PC, you can´t hear anything :D

Complex stop @ 33.2.2.37 (0:01:04:644)
Glad to hear!

What about the Simple FX Chain, can you please test that also? I have feeling that the Simple song favors frequency more than core count :)
Out of interest I would upload a minimised version the songfile also to test. Minimizing a device lowers its dsp usage. That is clicking on the little arrow to make it as small as possible. Took about 3-4 minutes to minimize all the open racks.

Two things: 1, FIY option/click on one "minimize triangle" and ALL in the same rack will collapse, and 2, collapsing devices does not have any effect on CPU usage that I've ever seen (including with this test file).
Damn son I knew there was a button for it couldn't remember what it was! Alt click. It frees up space in RAM when you minimize. I have been doing it for years on even worse machines than I currently have the post I made shows how much of a difference it makes. It does make a difference granted it might be negligible for some tracks like the benchmark here. But if your project is just barely tipping the too slow alert minimising will get you through the rest of the track if it is a small project that should not have a massive impact on performance.
Minimizing the devices on screen cannot free up RAM AFAIK, or the song would not play correctly. Also, this would be an important, useful, and therefore documented tip if true, but I cannot find any mention of it anywhere.

As previously mentioned, these tests to no indicate ANY CPU savings when comparing minimized vs non minimized rack windows with the test songs, nor have I ever noticed this in my work - maybe because I play the test song mostly from the sequencer view (with the rack view hidden)?

Anyone else measuring different CPU hit with minimized vs non-minimized rack views? What test would reveal this advantage if these song files do not?
(Hoping it's true, but not seeing any evidence to support this 'hope')
:)


Sent from some crappy device using Tapatalk
https://music.tutsplus.com/tutorials/10 ... audio-1075 - see suggestion #2
"Vastly improving Reason's performance"?

Again I'm not seeing ANY difference with any test - what are folks doing differently? I run up against the CPU issue often, and minimizing has never made any difference.

At least this clears up that the claim is not about RAM, but about CPU.

Anyway, thanks for the tip (wish it had come from the horses mouth though). :)


Sent from some crappy device using Tapatalk
Selig Audio, LLC

User avatar
AttenuationHz
Posts: 2048
Joined: 20 Mar 2015
Location: Back of the Rack-1

02 Jun 2017

selig wrote: "Vastly improving Reason's performance"?

Again I'm not seeing ANY difference with any test - what are folks doing differently? I run up against the CPU issue often, and minimizing has never made any difference.

At least this clears up that the claim is not about RAM, but about CPU.

Anyway, thanks for the tip (wish it had come from the horses mouth though). :)


Sent from some crappy device using Tapatalk
That's where I heard it from and tried it a good for years back maybe 8. The song I was working on at the time was just tipping the too "slow alert" crappy dual core laptop! I'm not full sure what it is doing when minimized but It would make sense that freeing up the ram would have less impact on CPU, less ram to read. In this benchmark I said it was negligible but in a smaller project it can make a big difference in my experience.

Here look for yourself keep an eye on the Memory tab in windows task manager.

It is not too much of an ask for people or things to be the best version of itself!

WongoTheSane
Moderator
Posts: 1851
Joined: 14 Sep 2015
Location: Paris, France

02 Jun 2017

selig wrote: "Vastly improving Reason's performance"?

Again I'm not seeing ANY difference with any test - what are folks doing differently? I run up against the CPU issue often, and minimizing has never made any difference.

At least this clears up that the claim is not about RAM, but about CPU.

Anyway, thanks for the tip (wish it had come from the horses mouth though). :)
I think I might have a clue. I did various tests yesterday, one of them trying to determine whether having the rack and mixer open and visible or closed would make a difference. It did, by a factor of 3:1 (i.e. the same benchmark song would play three times longer when the windows were closed). I was rather taken aback because I had never noticed such a discrepancy.

Then today I was benching the new simple benchmark and had weird results (which I will post in next message). So I restarted the computer and made sure I launched only Reason (no Chrome or other CPU hogs in the background). I redid the tests... No difference whatsoever this time, whether the windows are open or closed.

As I like gaming, it is very probable that yesterday I played a game before doing tests with Reason. My hypothesis is that, on Windows at least, some processes (DirectX? Video?) can keep a handle on the graphics system and slow down the drawing in Reason (or any other programs) when more windows are open. That might explain why some people are seeing a difference and others aren't: it would depend on whether they have run something before Reason that would alter the behavior of the graphics system. Pure conjecture at this stage, though, as I don't know which process could have caused that, and can't reproduce it yet...

WongoTheSane
Moderator
Posts: 1851
Joined: 14 Sep 2015
Location: Paris, France

02 Jun 2017

So I have weird results with the new simple benchmark (and the other one too).

For reference, my setup:

CPU i7 6700K @ 4.00 Ghz / 4 cores (8 with hyperthreading)
RAM 16 GB
SSD Samsung 1 TB (Windows 7 64 bit + Reason installed on SSD)
RME Babyface Pro audio interface

I tested both benchmarks on R9.2 and R9.5. Here are the results ("pass" means I can play the whole song):
benchmark results.PNG
benchmark results.PNG (7.16 KiB) Viewed 3501 times
Apparently, the smaller the buffer, the better it plays, which seems counter-intuitive to me. I always thought: when in trouble, increase the buffer size. What gives??

User avatar
Kategra
Posts: 327
Joined: 18 Jan 2015

02 Jun 2017

WongoTheSane wrote:So I have weird results with the new simple benchmark (and the other one too).

For reference, my setup:

CPU i7 6700K @ 4.00 Ghz / 4 cores (8 with hyperthreading)
RAM 16 GB
SSD Samsung 1 TB (Windows 7 64 bit + Reason installed on SSD)
RME Babyface Pro audio interface

I tested both benchmarks on R9.2 and R9.5. Here are the results ("pass" means I can play the whole song):

benchmark results.PNG

Apparently, the smaller the buffer, the better it plays, which seems counter-intuitive to me. I always thought: when in trouble, increase the buffer size. What gives??

On my RME Fireface 400 it's kinda the similar trend. Here is why I think it happens: Reason won't stop the song unless it detects that the CPU cores used for real time audio processing reach 95% load. The problem is that the sample buffer, especially at low settings, is depleted before the cores reach 95% load.
I don't know why this happens, I presume that driver makers or Propellerhead or it's the industry standard decided that it's ok to skip chunks of audio samples instead of stooping the aplication (DAW, audio apps etc) and presenting an error.
So i think that there is no detection mechanism in Reason that can stop playing the sound unless the CPU reaches the load threshold set by the user in the general settings 95 in our case. I tested at 48 samples latency and CPU limit at 80 and crackling starts early and goes trough out the song and I stopped it because I can't stand it that long, it's clearly not something to base performance on other that it say to me: don't use it with heavy load projects :lol:

I am curious if this the trend on other Audio Interface makers? Anyway in the 1st post, I hint that people should test with maximum buffer setting
"Your audiocard driver settings (sample rate 44100 Hz recommended and max buffer settings)".
But selecting the largest buffer setting would surely load the CPU past the 95% limit set into Reason. Hope it will help me to chose the next best bang for buck CPU for Reason next time I need it. So far so good!

WongoTheSane
Moderator
Posts: 1851
Joined: 14 Sep 2015
Location: Paris, France

02 Jun 2017

Kategra wrote:On my RME Fireface 400 it's kinda the similar trend. Here is why I think it happens: Reason won't stop the song unless it detects that the CPU cores used for real time audio processing reach 95% load. The problem is that the sample buffer, especially at low settings, is depleted before the cores reach 95% load.
I don't know why this happens, I presume that driver makers or Propellerhead or it's the industry standard decided that it's ok to skip chunks of audio samples instead of stooping the aplication (DAW, audio apps etc) and presenting an error.
So i think that there is no detection mechanism in Reason that can stop playing the sound unless the CPU reaches the load threshold set by the user in the general settings 95 in our case. I tested at 48 samples latency and CPU limit at 80 and crackling starts early and goes trough out the song and I stopped it because I can't stand it that long, it's clearly not something to base performance on other that it say to me: don't use it with heavy load projects :lol:

I am curious if this the trend on other Audio Interface makers? Anyway in the 1st post, I hint that people should test with maximum buffer setting
"Your audiocard driver settings (sample rate 44100 Hz recommended and max buffer settings)".
But selecting the largest buffer setting would surely load the CPU past the 95% limit set into Reason. Hope it will help me to chose the next best bang for buck CPU for Reason next time I need it. So far so good!
That makes perfect sense! As Ostermilk says, the ideal setup would be to listen to crackles instead of waiting for the "computer too slow" message, but I don't see how it can be done, as those crackles can be really subtle at the beginning...

User avatar
AttenuationHz
Posts: 2048
Joined: 20 Mar 2015
Location: Back of the Rack-1

02 Jun 2017

What's the deal with overclocked cpu. I tried to overclock yesterday and loaded the benchmark afterwards expecting it to play more than 8 seconds. Overclock i7 4790k to 4.4Ghz couldn't even start the heavy benchmark. It was a stable overclock.
It is not too much of an ask for people or things to be the best version of itself!

User avatar
stfual
Posts: 127
Joined: 17 Jan 2015
Contact:

02 Jun 2017

Hardware:
CPU Intel i7 2600K Quad core @ 3.43 Ghz
RAM 16 GB
SSD Samsung 1TB (Windows 10 Creators edition 64 bit + Reason 9.5 installed on SSD)
Focusrite ASIO 214 audio interface

Reason 9.5 2017 COMPLEX RE benchmark song V2 = Loads but wont play

Reason 9.5 2017 Simple FX Chain benchmark song V0 = 99 Seconds

GRIFTY
Posts: 658
Joined: 16 Jan 2015

06 Jun 2017

R7 1700 OCed @ 3.7. In the heavy stress test i didn't make it passed 5 seconds with buffer at 900+ and sampling at 48000

GRIFTY
Posts: 658
Joined: 16 Jan 2015

06 Jun 2017

whats weird is that task manager never shows utilization above 49%, even when i remove the CPU usage limit. I don't think reason 9.5 is optimized for more than 4 cores

User avatar
Kategra
Posts: 327
Joined: 18 Jan 2015

06 Jun 2017

GRIFTY wrote:whats weird is that task manager never shows utilization above 49%, even when i remove the CPU usage limit. I don't think reason 9.5 is optimized for more than 4 cores
R7 1700 has 8 cores physical and 16 threads. Windows task manager reports the load percentage for the 16 threads (logic cores), but Reason only uses the 8 physical cores for DSP. If the 8 real cores are at 100% usage, but the other 8 logic threads are not utilized, then Windows task manager will only show <50% CPU utilization.

I think Reason runs well on Intel architecture with single CPU, so far user: devilfish reported an XEON CPU with 14 real cores (28 logic threads) playing more than 1 minute of the Complex RE song:
devilfish wrote:Windows 10 x64
Intel Xeon E5-2683v3 (QFQK - 250€ eBay)
16GB RAM ECC
Mackie Onyx Blackjack USB Audiointerface
44.1kHz / 95% / 256 Samples (7ms/9ms)
Ati Radeon EX 460
Super Silent PC, you can´t hear anything :D

Complex stop @ 33.2.2.37 (0:01:04:644)

GRIFTY
Posts: 658
Joined: 16 Jan 2015

06 Jun 2017

but, if reason only utilizes half of the available threads, doesn't this mean it only uses half the available processing power? it is not as if logical threads are some magical "other" type of of CPU cores that processor makers conjure up.

i am interested in what devilfish's CPU utilization is during this reason test. is it somewhere around 29%? it appears that reason is either not optimized for SMT (ryzen), or, depending on some other folks' CPU utilization numbers, not optimized for more than 8 threads

(edited to appear less confrontational, my earlier tone was unintended)

WongoTheSane
Moderator
Posts: 1851
Joined: 14 Sep 2015
Location: Paris, France

07 Jun 2017

GRIFTY wrote:but, if reason only utilizes half of the available threads, doesn't this mean it only uses half the available processing power? it is not as if logical threads are some magical "other" type of of CPU cores that processor makers conjure up.
No, it still uses all the available power even if it uses half the threads. Thread are a way to balance all the available power between two logical units; in the case of DSP, whether one unit does 100% of the work or two units do 50% each would be strictly equivalent, except that going back and forth between the two logical units consumes power in itself, so it makes more sense to only use one. Half the threads are inactive, but you CPU's power is fully used.

To use an analogy: a company might be better suited having one full-time employee, or two half-time employees. Some businesses will benefit from the second solution, while most will opt for the first because it's the same number of work hours, with half the paperwork. Same with CPUs!

User avatar
pushedbutton
Posts: 1538
Joined: 16 Jan 2015
Location: Lancashire, UK
Contact:

07 Jun 2017

Intel i7 6700k 4Ghz (stock clock)
32 Gb RAM
Propellerhead Balance
44100 Hz @ 4096 samples
CPU Usage Limit 95%
COMPLEX RE ----> 0.00.08.173
Simple FX Chain ----> 0.01.54.149

Expected better results to be honest.
@pushedbutton on twitter, add me, send me a message, but don't try to sell me stuff cos I'm skint.
Using Reason since version 3 and still never finished a song.

Post Reply
  • Information
  • Who is online

    Users browsing this forum: chimp_spanner and 22 guests