The No.1 Website for Pro Audio
At last: new Mac Pro!
Old 27th January 2020
  #5731
TNM
Lives for gear
Ok, finally got around to testing the 8 core imac pro properly with DIVA RC.

In Logic, with audio track highlighted and sustained looping of the test project, 128 buffer, medium process buffer, imac pro internal audio, 32 tracks.

All Diva's own multicore is disabled of course. But this is the kicker...

Do you know how much REAL cpu is being used to do this? 92.6%.

Yep, I am not kidding. Diva has improved so much that Logic can now use 92.6 of the entire system cpu.. it's basically benching almost as well as cinebench can!

So from 6 instances, to 31. WOW. I have to commend you on that URS.

My quad core laptop gets 15. So Diva is now scaling perfectly on the newer cpu architecture and more cores. It's actually highlighted how many OTHER plugins are broken.. and could be fixed overnight like you did! I really hope some other devs listen.. Diva is now one of the best performing algorithmic VI's! Certainly THE best when the equation of cpu usage vs quality of sound being produced for that cpu usage.

I totally forgot that i deleted the bootcamp partition like 2 weeks ago, as I planned to start fresh and make the disk one whole partition, install mojave fresh, then re install windows, but I have got caught up in other stuff and composing itself. I actually can't pit it against windows on the same machine till my fresh (and final) install is done, as I am not prepared to go through installing windows and drivers and pro tools and diva just to do this test and then erase it again.

In any case, absolutely remarkable results. I am quite literally gobsmacked. I also am not understanding how I got the same result as the 10 core, but I equally don't understand how the 9900K hackintosh is getting 60. My cpu was sitting pretty much at 3.92ghz throughout the test, which surprised me.. I doubt a Catalina pro mode would make any difference in this particular benchmark and I am guessing windows will be the same. The hackintosh's all core turbo is 4.7 I think, so for 800 mhz more per core he is *doubling* instances? There is something afoot here.. perhaps the tester mistakenly had an instrument track highlighted which causes logic to operate akin to cubase/reaper/S1 etc where it doesn't overload and stop playback but slowly and steadily introduces pops and clicks to the audio instead and eventually just hangs or gives the sample rate mismatch error.. If it's not that, then I just do not understand it. If my cpu was dropping to 3.ghz, OK, i could see it somehow.. but.. what we are talking about is an ~18% difference per core giving him 100% more instances. Something is amiss indeed.
Old 27th January 2020
  #5732
TNM
Lives for gear
sorry double post

Last edited by TNM; 27th January 2020 at 10:25 PM..
Old 27th January 2020 | Show parent
  #5733
Gear Nut
 

Quote:
Originally Posted by TNM View Post
wiki chip and cpu world put the stats of these up as they become known.

<snip>

the 28 core is rated to be at 3.2ghz on all 28 cores under load. That means anything above this when all 28 cores are being used would be considered overclocking.
That information is not published by Intel, nor is it on Wiki chip or Cpu World for the W-3275M. What do you mean by "rated"? What is the source of that info?
Old 27th January 2020 | Show parent
  #5734
Lives for gear
 
juiseman's Avatar
https://www.intel.com/content/www/us...s/w-3275m.html

I wouldn't say that the 28 core would do more than the base Frequency in all core boost. That is what the 205W TPD is specified at. But It may very well be able to do a lot more if the limits are taken off. I think that's what TNM is saying.
Old 27th January 2020 | Show parent
  #5735
Lives for gear
 
Chevron's Avatar
 

Quote:
Originally Posted by TNM View Post
In any case, absolutely remarkable results. I am quite literally gobsmacked. I also am not understanding how I got the same result as the 10 core, but I equally don't understand how the 9900K hackintosh is getting 60. My cpu was sitting pretty much at 3.92ghz throughout the test, which surprised me.. I doubt a Catalina pro mode would make any difference in this particular benchmark and I am guessing windows will be the same. The hackintosh's all core turbo is 4.7 I think, so for 800 mhz more per core he is *doubling* instances? There is something afoot here.. perhaps the tester mistakenly had an instrument track highlighted which causes logic to operate akin to cubase/reaper/S1 etc where it doesn't overload and stop playback but slowly and steadily introduces pops and clicks to the audio instead and eventually just hangs or gives the sample rate mismatch error.. If it's not that, then I just do not understand it. If my cpu was dropping to 3.ghz, OK, i could see it somehow.. but.. what we are talking about is an ~18% difference per core giving him 100% more instances. Something is amiss indeed.
Assuming you are talking about my machine TNM, I just posted a screenshot with Power Gadget visible in the Custom OS X benchmark thread. Which I ran again to triple check the track count, and 100% certain to have an audio track selected too. Still 60 tracks solid playback

My processor is a 9900ks though BTW, and I am sure you know already it is 5.0 Ghz turbo.
Old 27th January 2020 | Show parent
  #5736
Lives for gear
 
juiseman's Avatar
https://en.wikichip.org/wiki/intel/x...75#Frequencies

This IS stating that this chip will do 3.2GHZ all cores. I'm surprised.
I'm my experience; Xeons (or any locked CPU) will default to the base speed
when loaded. That is great news. I was a tad worried how 2.5GHZ would perform DAW workload....
Old 27th January 2020 | Show parent
  #5737
Gear Head
 

Quote:
Originally Posted by TNM View Post
In what way are you limiting your core/thread count? In cubase you can do this, on OSX? I have never seen this before.. or do apple literally have a way to dedicated only a certain amount of cores per app now or something?

PS you can do that with an app called CPU setter, as in, choose the amount of cores OSX will use, but that's a global setting for all apps.. nevertheless, I set my quad core to 4 sometimes instead of 8 when I am just watching movies and doing the net or playing emulators, and it helps keep temps down. So I have been doing this for years in my case anyway.. it's nothing new in the 28 core unless apple have included a built in tool to do it.
I'm using the built-in Instruments app (which is either installed with Xcode or comes pre-installed with macOs... I honestly don't know so I just installed Xcode) but it seems to work great.
I limit my core count and I see the clock speed boost. It's like I'm getting every nMP config in one
Old 27th January 2020 | Show parent
  #5738
Tui
Gear Guru
 
Tui's Avatar
Quote:
Originally Posted by BRVLN View Post
I'm using the built-in Instruments app (which is either installed with Xcode or comes pre-installed with macOs... I honestly don't know so I just installed Xcode) but it seems to work great.
It comes with Xcode:

https://en.wikipedia.org/wiki/Instruments_(software)

How do you use it, though?
Old 27th January 2020 | Show parent
  #5739
Gear Nut
 

Quote:
Originally Posted by juiseman View Post
https://en.wikichip.org/wiki/intel/x...75#Frequencies

This IS stating that this chip will do 3.2GHZ all cores. I'm surprised.
I'm my experience; Xeons (or any locked CPU) will default to the base speed
when loaded. That is great news. I was a tad worried how 2.5GHZ would perform DAW workload....
Interesting - it's on that page but not the page for the M version (which is what's in the Mac Pro). They also don't have that table for the W-3245 (16 core). Feels like someone is just running experiments to derive that table.
Old 28th January 2020 | Show parent
  #5740
Gear Head
 

Quote:
Originally Posted by Tui View Post
It comes with Xcode:

https://en.wikipedia.org/wiki/Instruments_(software)

How do you use it, though?
I just open the preferences and there’s a big slider that lets you choose your thread count. I just lower it to 32 or 24 and see the change almost instantly in Cubase and in iStat menu.
Old 28th January 2020 | Show parent
  #5741
Here for the gear
 

Quote:
Originally Posted by BRVLN View Post
I'm using the built-in Instruments app (which is either installed with Xcode or comes pre-installed with macOs... I honestly don't know so I just installed Xcode) but it seems to work great.
I limit my core count and I see the clock speed boost. It's like I'm getting every nMP config in one
yeah, it comes with Xcode
Old 28th January 2020 | Show parent
  #5742
Lives for gear
 

Quote:
Originally Posted by cyborgssc View Post
Interesting - it's on that page but not the page for the M version (which is what's in the Mac Pro). They also don't have that table for the W-3245 (16 core). Feels like someone is just running experiments to derive that table.
They are. These come from sources WikiChip feels are reputable who do their own testing. Intel doesn't report these for every CPU anymore. It's a more official version of what we do here; designing our own benchmarks and testing scenarios and reporting the results for everyone to see.
Old 28th January 2020 | Show parent
  #5743
TNM
Lives for gear
Quote:
Originally Posted by Chevron View Post
Assuming you are talking about my machine TNM, I just posted a screenshot with Power Gadget visible in the Custom OS X benchmark thread. Which I ran again to triple check the track count, and 100% certain to have an audio track selected too. Still 60 tracks solid playback

My processor is a 9900ks though BTW, and I am sure you know already it is 5.0 Ghz turbo.
wow indeed.. Just to think how much of a massive difference a little clock speed can make! I believe you, I am just shocked!
Old 28th January 2020
  #5744
TNM
Lives for gear
or just use this and make life even easier
https://www.whatroute.net/cpusetter.html
Old 28th January 2020 | Show parent
  #5745
Lives for gear
 
juiseman's Avatar
Quote:
Originally Posted by cyborgssc View Post
Interesting - it's on that page but not the page for the M version (which is what's in the Mac Pro). They also don't have that table for the W-3245 (16 core). Feels like someone is just running experiments to derive that table.
The m version of the 28 core is only different in the max memory being 2TB+?? RAM instead of the rest at 1.5TB. That's not worth $2500-$3000....there are programs and situations that need this much ram...it wasn't intended for audio...
Old 28th January 2020 | Show parent
  #5746
Gear Head
Hey FWIW - i have a few vids on the Mac Pro Rack a full workflow runthrough once it's up and running in the studio.

I use VEP, Logic X and Protools HDX all simultaneously on the same machine with independent audio hardware.

as a quick unscientific test, I loaded up my bud's Orch Score mix session from a mega Hollywood film and decided to test it natively using the built-in speaker as a playback engine . This is a real-world working session with a few surprises. The session is 96/24, 7.1 bussing and I set Pro Tool's buffer to 64 for no reason other than to hammer the CPU for kicks.

Take a look:

Old 28th January 2020 | Show parent
  #5747
Lives for gear
 

Quote:
Originally Posted by Neil Parfitt View Post
Hey FWIW - i have a few vids on the Mac Pro Rack a full workflow runthrough once it's up and running in the studio.

I use VEP, Logic X and Protools HDX all simultaneously on the same machine with independent audio hardware.

as a quick unscientific test, I loaded up my bud's Orch Score mix session from a mega Hollywood film and decided to test it natively using the built-in speaker as a playback engine . This is a real-world working session with a few surprises. The session is 96/24, 7.1 bussing and I set Pro Tool's buffer to 64 for no reason other than to hammer the CPU for kicks.

Take a look:

I just posted your unboxing video here yesterday! Glad to see one that's from an audio guy like all of us, and also one that gives us a good view of the differences between the rackmount and tower versions

How long have you been running the Mac Pro in your studio at this point?
Old 28th January 2020 | Show parent
  #5748
Gear Head
Quote:
Originally Posted by MattyJoe View Post
How long have you been running the Mac Pro in your studio at this point?
I'll be doing the deep dive tomorrow get everything rigged up and finally put through the real paces.
Old 28th January 2020 | Show parent
  #5749
Tui
Gear Guru
 
Tui's Avatar
Quote:
Originally Posted by BRVLN View Post
I just open the preferences and there’s a big slider that lets you choose your thread count. I just lower it to 32 or 24 and see the change almost instantly in Cubase and in iStat menu.
I see, thanks.
Old 28th January 2020 | Show parent
  #5750
TNM
Lives for gear
Quote:
Originally Posted by Neil Parfitt View Post
Hey FWIW - i have a few vids on the Mac Pro Rack a full workflow runthrough once it's up and running in the studio.

I use VEP, Logic X and Protools HDX all simultaneously on the same machine with independent audio hardware.

as a quick unscientific test, I loaded up my bud's Orch Score mix session from a mega Hollywood film and decided to test it natively using the built-in speaker as a playback engine . This is a real-world working session with a few surprises. The session is 96/24, 7.1 bussing and I set Pro Tool's buffer to 64 for no reason other than to hammer the CPU for kicks.

Take a look:

Neil, thanks for the video. In case you wondered who jumped in about the buffer size in the YT comments, that was me..
Basically would love to see a test with a bunch of tracks armed at 64 buffer or even 32, as this is the sort of computer that's supposed to make interfaces with DSP for low latency monitoring redundant. Pro tools has a fixed playback buffer size of 1024 samples (in other words, like logic's medium choice in the process buffer settings).
Seeing hoe many tracks you could monitor with effects would really be something to throw for or against the HDX argument. I personally still believe strongly in DSP for LLM but many don't.
Congrats on the beast of a machine, and if the rumored pro mode does come out and does what we hope it will do, it will be even more beastly. That was a humungous session which pretty much proves VE Pro slave machines are redundant, even with massive sessions. One dead silent machine CAN do it all.
Old 28th January 2020
  #5751
Gear Head
 

No sure if anyone posted or updated Diva but here is some good news . I do work a lot with Zebra and if Diva is at the same level, that it's important :D

"Diva 1.4.4 now has improved multicore support, and new compilers allowed us to further reduce the CPU hit. We also improved compatibility with MPE controllers and implemented the new voice stealing algorithm from Hive – which is just that extra tad more musical! Altogether we think Diva 1.4.4 feels like a whole new synth: Better performance, better playability, better everything. "
Old 28th January 2020 | Show parent
  #5752
TNM
Lives for gear
Quote:
Originally Posted by RoboOiuo View Post
No sure if anyone posted or updated Diva but here is some good news . I do work a lot with Zebra and if Diva is at the same level, that it's important :D

"Diva 1.4.4 now has improved multicore support, and new compilers allowed us to further reduce the CPU hit. We also improved compatibility with MPE controllers and implemented the new voice stealing algorithm from Hive – which is just that extra tad more musical! Altogether we think Diva 1.4.4 feels like a whole new synth: Better performance, better playability, better everything. "
That's all we've been talking about the last few pages

my iMac pro has gone from 6 to 30 and others are seeing similar massive gains.
Old 28th January 2020 | Show parent
  #5753
JGM
Lives for gear
 

Quote:
Originally Posted by Neil Parfitt View Post
Hey FWIW - i have a few vids on the Mac Pro Rack a full workflow runthrough once it's up and running in the studio.

I use VEP, Logic X and Protools HDX all simultaneously on the same machine with independent audio hardware.

as a quick unscientific test, I loaded up my bud's Orch Score mix session from a mega Hollywood film and decided to test it natively using the built-in speaker as a playback engine . This is a real-world working session with a few surprises. The session is 96/24, 7.1 bussing and I set Pro Tool's buffer to 64 for no reason other than to hammer the CPU for kicks.

Take a look:

Great video.
Not sure if you mentioned it, but what’s the specs of your nMP?
Old 28th January 2020 | Show parent
  #5754
Lives for gear
 

Quote:
Originally Posted by JGM View Post
Great video.
Not sure if you mentioned it, but what’s the specs of your nMP?
He's got the 28-core, 384 GB of RAM (third party, not Apple's), and the stock AMD Radeon Pro 580X.
Old 28th January 2020 | Show parent
  #5755
Gear Head
 

Jus thought I would update, I’m about to install 2 more sticks of 32GB of RAM today for a total of 6 channel memory. Should give me a performance boost, so I’ll se how much and let you guys know.
Old 28th January 2020 | Show parent
  #5756
Gear Maniac
Quote:
Originally Posted by Neil Parfitt View Post
Hey FWIW - i have a few vids on the Mac Pro Rack a full workflow runthrough once it's up and running in the studio.

I use VEP, Logic X and Protools HDX all simultaneously on the same machine with independent audio hardware.

as a quick unscientific test, I loaded up my bud's Orch Score mix session from a mega Hollywood film and decided to test it natively using the built-in speaker as a playback engine . This is a real-world working session with a few surprises. The session is 96/24, 7.1 bussing and I set Pro Tool's buffer to 64 for no reason other than to hammer the CPU for kicks.

Take a look:
Neil - good videos, love the cat review
Old 28th January 2020 | Show parent
  #5757
JGM
Lives for gear
 

Quote:
Originally Posted by MattyJoe View Post
He's got the 28-core, 384 GB of RAM (third party, not Apple's), and the stock AMD Radeon Pro 580X.
Got you.
Old 28th January 2020
  #5758
JGM
Lives for gear
 

I seem to be reading that the 16 core is the sweet spot for audio guys, is that a fair assessment? And can someone breakdown why that is if it’s indeed true?
I know there’s tons of variables, but speaking generally, if that’s allowed.
Thanks
Old 28th January 2020 | Show parent
  #5759
Lives for gear
 
uptheoctave's Avatar
Quote:
Originally Posted by JGM View Post
I seem to be reading that the 16 core is the sweet spot for audio guys, is that a fair assessment? And can someone breakdown why that is if it’s indeed true?
I know there’s tons of variables, but speaking generally, if that’s allowed.
Thanks
It is because the next jump up in processor is a lot of money.
Old 28th January 2020 | Show parent
  #5760
Lives for gear
 
charlieclouser's Avatar
 

Quote:
Originally Posted by uptheoctave View Post
It is because the next jump up in processor is a lot of money.
And the higher core units have a lower base clock speed, which may result in less single-core headroom for those hungry synth plugins on Logic's "live core" situation. Jury still out.
📝 Reply

Similar Threads

Thread / Thread Starter Replies / Views Last Post
replies: 4594 views: 587161
Avatar for Arseny
Arseny 3 weeks ago
replies: 97 views: 27989
Avatar for mowmow
mowmow 15th September 2010
replies: 835 views: 170859
Avatar for jlgrimes11
jlgrimes11 30th August 2020
replies: 56 views: 8823
Avatar for Aiyn Zahav
Aiyn Zahav 10th August 2012
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
🖨️ Show Printable Version
✉️ Email this Page
🔍 Search thread
🎙️ View mentioned gear
Forum Jump
Forum Jump