Is my MONSTER 12-core 3,33GHz Mac Pro underperforming on low buffer sizes?

[m4tune]

Originally Posted by Kroon

So if I understand you correctly, you tried many things but what fixed the problem for you in the end was installing a PCI SSD adapter card?

Yes and maybe.
The thing is that each version of Pro Tools was a little different. It has been gradually improved - for example 12.6.1 has been a better version than anything before. but I kept the OS updated, too. So there might be reasons which interact between OS and PT versions, too.

But one day I spoke to the CEO of Angelbird and he told me that the pure existence of spinning drives might be the problem at my system. This reminded me to a problem I've had many years before with my MacBook Pro which became very slow when I updated to 10.7.. The only thing which helped back in these days was exchanging the spinning drive to a ssd after I already maxed out the RAM.

In my personal opinion the most important reason for weak performing Macs is something Apple changed in their OS since 10.7 that conflicts with spinning drives. It doesn't seem so important to get the latest ssd with max speed rather than having a ssd at the right slot instead of a spinning drive.

Since I have the Angelbird Card I didn't see any 9071 anymore. I have still 2 spinning drives (against the recommendation) in the system.

[musicman691]

Originally Posted by Southsidemusic

Why are we talking about the OP's Mac not beeing up tp scratch because he use Samsung 850's

I can run the same sessions he posted that his machine can't run properly on a 15 y/o PowerMac so let stop chasing the wrong dog here (SSD's not the Best) one's and focus on why this is happening to a FULLY Capable Mac Pro!

This computer should be able to run 100's of tracks and plugins without issues, not as he say "a few tracks and minimal plugins" so the problem obviously lay elsewhere.

Agreed. Cheesegraters are plenty capable when set up right and maintained correctly. Clean out the dust bunnies from time to time, don't move the machine around like it's a laptop and feed it with good ac power and one should be golden.

[audiobob]

Originally Posted by TNM

what video card? the standard low grade nvidia? or do you mean a 680 like the OP? I presume you have tried all three pcie slots?

Also Bob, you say in sig you have a video pcie editing solution, as well as HDX.

so i am a little confused.. i thought HDX was independent of buffer...

Could you tell me the specs of the machine you can't record at lower than 256 buffer in? Even if it's the 2.4ghz 12 core, it still should be powerful enough to do what the OP is asking, and his is much more powerful than that.
Also what video card, the 520? And whatever other pcie devices are also being used when the overload occurs.

I moved all cards around in every conceivable slot combination.

I said in my original post that I upgraded to HDX after I could not make the native card work. Have no issues with HDX.

Did not change out the video card, but my opinion is that if a video card can stop a 12 core 2.66 machine with 32 gigs of ram from recording at a low buffer, then either the machine or Pro Tools software or hardware is effed up. I've moved on to HDX now so I'm done trying to figure it out.

[Kroon]

UPDATE:

I tried recording using my X32 as a USB interface.

No improvement over the HD Native.

Then I rebooted the computer from a 10.8.5 Mountain Lion drive - a 7200 spinner in a regular bay - launched PTHD 10, and run the test again.

Different ballgame. Milestone discovery!

I was able to record 24 tracks just fine with a 64 buffer.

I tried to put strain on the system by going ape with zooming and scrolling while recording, but I was unable to interrupt it. The GUI felt snappier too.

The CPU meter was stable at 13-14%. The disk activity meter was at 2% when recording to a PCI SSD, and 3-4% when recording to a 7200 spinner.

I tried recording 48 tracks too, and the CPU meter only increased to 17-18% but unfortunately recording was interrupted after a few seconds, so I guess there were some CPU spikes that the meter didn't catch.

No worries, I never record more than 24 tracks simultaneously anyway, so I believe I could easily live with this performance if I could have it in PTHD 2018.1 on Sierra.

I'm not so keen on going back to PTHD 10 for good.

I need to find a way to make things work in 2018.1 or at least 12.

At this point I suspect the graphics card, but am reluctant to buy a different one on chance off eBay as I have no way of knowing which card to go for.

I'll try a few other things before I go down the graphics card route.

[TNM]

Originally Posted by audiobob

I moved all cards around in every conceivable slot combination.

I said in my original post that I upgraded to HDX after I could not make the native card work. Have no issues with HDX.

Did not change out the video card, but my opinion is that if a video card can stop a 12 core 2.66 machine with 32 gigs of ram from recording at a low buffer, then either the machine or Pro Tools software or hardware is effed up. I've moved on to HDX now so I'm done trying to figure it out.

well sorry, i did not go back and search your original post nor did you link to it..I didn't know there *was* more info to find on that, as I simply quoted and answered what you told me in context there on the spot.But I apologise if you think I was purposely ignoring your situation, as i promise i wasn't. Just crossed wires.

Re the vid card, yes, a video card even stops some windows machines at low buffer and people switch to the generic microsoft driver to alleviate it. I have seen people have issues with the most up to date intel chips, like the 7820X, and an nvidia 1080, where they can't go under 256 buffer.. and they are the only two pcie devices in the system..

Anything and everything can go wrong with multiple pcie devices on any computer.. like Christopher said, the machine is plenty powerful to do what the OP is asking of it.
Something is not right.. i suspect it may be a dual processor issue with the pcie bus and the hdn card.. that's all i can think of after a process of elimination. Just saw pro tools expert and their cheese grater at 64 samples performing like ca champ.. it's on youtube.. but once again, a single hexa not a dual.

Perhaps it's just not possible for a 12 core cheese grater to work at low buffer with PT HDN.

What i think would be a great test, i hope the OP agrees, is to see whether the issue is purely the HDN card..

for that i suggest a usb interface via core audio and see if it works at 64 buffer to record.. if it does, then the machine is not the issue, right? And the processor is plenty powerful to do what the OP wants. I believe since the usual suspects don't seem to be at play this time, that it is a specific issue with the cheese grater motherboard with the dual processor upgrade, and the HDN card.. something is not allowing priority to the HDN which is why it only manifests at low latencies.

Note, as i said prior, i can say with certainty that RME pcie devices work at 64 samples in a 12 core cheese grater in other DAWs via core audio. I've seen this myself with cubase and RME.. we also tested it with asio guard disabled so the card was at 64 samples all the time, not just on record enabled tracks.

[TNM]

Originally Posted by Kroon

UPDATE:

I tried recording using my X32 as a USB interface.

No improvement over the HD Native.

Then I rebooted the computer from a 10.8.5 Mountain Lion drive - a 7200 spinner in a regular bay - launched PTHD 10, and run the test again.

Different ballgame. Milestone discovery!

I was able to record 24 tracks just fine with a 64 buffer.

I tried to put strain on the system by going ape with zooming and scrolling while recording, but I was unable to interrupt it. The GUI felt snappier too.

The CPU meter was stable at 13-14%. The disk activity meter was at 2% when recording to a PCI SSD, and 3-4% when recording to a 7200 spinner.

I tried recording 48 tracks too, and the CPU meter only increased to 17-18% but unfortunately recording was interrupted after a few seconds, so I guess there were some CPU spikes that the meter didn't catch.

No worries, I never record more than 24 tracks simultaneously anyway, so I believe I could easily live with this performance if I could have it in PTHD 2018.1 on Sierra.

I'm not so keen on going back to PTHD 10 for good.

I need to find a way to make things work in 2018.1 or at least 12.

At this point I suspect the graphics card, but am reluctant to buy a different one on chance off eBay as I have no way of knowing which card to go for.

I'll try a few other things before I go down the graphics card route.

So in a nutshell, when that user quoted me and said i spoke nonsense and what a moron I basically was, by suggesting to use a drive in the sata bay, it did exactly what it said I would. Sorry about my post just above, i wrote it already before i saw your post with this result. Hmm.. i wonder if i will get an apology.

So the pcie SSDx2 was stealing too much bandwidth and being given priority over the audio, just as i thought.

PS, the 7200 will be much slower overall as an OS drive, i suggest to put an SSD in the sata bay.. PPS..aah, i see you retained one of the PCIE ssd's. Humor me and remove that too and see if performance gets even better to do those 48 tracks? I think the OS was giving the main boot pcieSSD priority over everything basically.

2 SSD's in the sata bays are all that you need.. you do NOT need a pcie SSD to record 24 tracks of audio to.. not under any circumstances.. trust me on this. You do not need a pcie SSD to record even the max hdn 64 tracks at 96k of audio to.. A sata ssd will not even break a sweat doing that. Not even a little.

[Sardi]

You missed the part about him switching OS and PT version. That is more significant than the SSD at this point in time.

I only just moved from ML to Sierra because it was so damn stable. Certain plug-ins forced my hand, but I had a dual boot for 3 months before I was satisfied enough to permanently switch.

OP, install PT12 on that 10.8.5 partition and run the same tests please. Even if it’s on the 7200 RPM drive for the the time being.


Sent from my iPhone using Tapatalk

[TNM]

Originally Posted by Sardi

You missed the part about him switching OS and PT version. That is more significant than the SSD at this point in time.

I only just moved from ML to Sierra because it was so damn stable. Certain plug-ins forced my hand, but I had a dual boot for 3 months before I was satisfied enough to permanently switch.

OP, install PT12 on that 10.8.5 partition and run the same tests please. Even if it’s on the 7200 RPM drive for the the time being.


Sent from my iPhone using Tapatalk

Well PT 12/2018 has totally changed the way low buffers work.. as i have detailed before in my own topics.. it engages all cores for even just one track at a 32 or 64 buffer.. it makes no sense at all and is completely different to other DAWs.. PT 10 did not do this.. So it's basically the way they changed the engine for 64 bit aax i think (guessing).

yes i did miss the ML (my apologies to you Sardi and to the OP) bit but i'd be curious to see how it works with both PCIE SSD's engaged.. as in OS 10.8 on the pcie SSD and recording to the second pcie SSD.

Regardless, it is sounding like there is something simply not right with his Sierra install...

Yesterday I tried my mbox 3 USB and 64 buffer on my 2011 imac and it recorded fine for 10 minutes, 4 tracks.

remember, the OP can not record a single track at 64 buffer..

I am on sierra 12.12.6 on both machines.

YES, pt has definitely changed the way it works at low buffers, and this is the same on windows 7 and 10 also believe me, with the latest PT versions and processors.. It's not just mac nor just with "older" processors.

BUT.. what the OP is trying to do, record 24 tracks at 64 buffer without effects, can be done even at *32* samples on my crappy imac, and even with Pro tools' new, inferior low latency native engine.

After this topic i got so intrigued that I tried every interface i have here under core audio and they all worked.. I maxed out the tracks to whatever inputs the interface had..

I also did research as i said, on cheese graters and low latency.. and it IS possible to have them working perfectly well at 64 samples..

My original post was still valid advice.. it wasn't concrete, it was suggestions of maybes. After it seemed others had the issue with the 12 core cheese graters only, it started making sense that it's likely a combo of the dual processor with the HDN card, so my focus has shifted. My intention was only to ever try help the poor guy because believe me i know what it's like.. I myself on my "not old" macbook pro wouldn't even dream of working under 128 buffer at 44k or 256 at 88k+, if i was recording audio and using VI's. Even though i have tested it recording 32 audio tracks with NO insert fx, at 32 buffer and it worked, the cores were still at tip point.. one effect and it was over.. (just to clarify, all my tests on this topic have been done with NO effects on the recording audio tracks).

PS when the OP says X32 usb i presume he means behringer.. i wouldn't think that would have a good low latency driver.. i haven't tried one myself but i have never read good things about behringer drivers..

And re the video card, that's still a possibility as the source of issue. Possibly.

Regardless, even without stellar drivers, any USB interface should be able to record 24 tracks at at least 128 buffer.. which the OP couldn't do.. So this situation in particular is a bit of a mystery.. as the mod said also.. that machine SHOULD be able to do what the OP wants.. maybe just not with PT...(i go back to the RME and cubase story at 64 samples on an upgraded 12 core cheese).

[musicman691]

Originally Posted by Sardi

You missed the part about him switching OS and PT version. That is more significant than the SSD at this point in time.

I only just moved from ML to Sierra because it was so damn stable. Certain plug-ins forced my hand, but I had a dual boot for 3 months before I was satisfied enough to permanently switch.

OP, install PT12 on that 10.8.5 partition and run the same tests please. Even if it’s on the 7200 RPM drive for the the time being.


Sent from my iPhone using Tapatalk

That would be interesting considering they aren't even remotely considered compatible with each other. Personally I think this talk about the ssd's in a pcie slot being an issue is a red herring. I would never go back to a drive bay ssd if I already had the drives in a pcie slot especially for sample streaming. Recording could be a different animal though.

I'd love to see Darryl Ramm weigh in on this thread as he's the ssd guru around here.

[Sardi]

Originally Posted by musicman691

That would be interesting considering they aren't even remotely considered compatible with each other.

??

Pro Tools 12 up to v12.5 was fully supported under Mountain Lion.


Sent from my iPhone using Tapatalk