The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
Audio Interface - Low Latency Performance Data Base Audio Interfaces
Old 17th January 2012
  #211
Lives for gear
 
valis's Avatar
Quote:
Originally Posted by RebornOnline View Post
Hi TAFKAT - thanks for taking the time to look into that for me! I really appreciate it.

Do you have a set of results for the PCI Pulsar/Scope PCI platform? If not, I can try and help get some? (assuming the project is trying to accumulate as much data as possible that is)
I would go ahead and provide him with results, my Scope cards are still housed in an old P4 era Xeon rig (and I have a gen1 card as well) and so can't do it myself but think it's a worthy addition to the tables.
Old 17th January 2012
  #212
Lives for gear
 
TAFKAT's Avatar
 

Quote:
Originally Posted by dubrichie View Post
Just a thought, but is this true for the C20X (Sandybridge) Xeon server chipsets? I thought that they still had native PCI, and probably will indeed be the last series that do.

Well I bloody hope they do, because I'm going to build a rig based on one to squeeze the last few years out of my PCI PT TDM rig!
Hey D,

It is true for the H67/P67/Z68/X79 , haven't personally had any time in front of a C20x chipset, but having a quick browse across the specs at the Intel site it does look to have an older Intel® 82801CA I/O Controller , which has native PCI . With it being designed more specifically for server application were legacy compatibility is more of a focus , the native PCI support seems to have been maintained.

Come to think of it , I do remember reading in on some threads where PTHD PCI users were qualifying systems with them, not sure of the outcome or where I read it now, sorry.. :-(

Good Luck with the PTHD PCI rig, let me know how you go.. :-)

I digress,

I won't be qualifying the chipset myself , so my PCIe only testing moving forward still stands... :-)

@ Valis, Reborn,

Sorry I missed that last post in the mix, no problem forwarding me the results but for consistency all the offical testing has to be done on identical computer hardware to remain valid. My U.K test partner has mirrored my exact test/benching system and was able to duplicate my posted results on interfaces that I had tested , just to ensure we were on the same page. It would be difficult for Reborn to do that , but his system is close to spec - same chipset/CPU, so it would be interesting what he actually achieves at his end.

Old 17th January 2012
  #213
Lives for gear
 
mutilatedlip's Avatar
Holy ****!!?!

Does anyone actually record any music around here, or have I accidentally visited a mathematics forum?
Old 17th January 2012
  #214
Lives for gear
 

Quote:
Originally Posted by mutilatedlip View Post
Holy ****!!?!

Does anyone actually record any music around here, or have I accidentally visited a mathematics forum?
That's why it's called gearslutz. If you want to discuss songs, you'll have to visit songslutz.

I kinda like this thread, I have about 10 Echo interfaces and apparently from the measurements, they suck.
Well to be fair, I don't have their firewire stuff, but this is still an interesting thread to read. No lie.

Only at gs can you come in and find out that you have problems.... even if you didn't know you had problems.
Old 18th January 2012
  #215
Gear Maniac
 
seagull's Avatar
 

Quote:
Originally Posted by thenoodle View Post
Only at gs can you come in and find out that you have problems.... even if you didn't know you had problems.
THAT has to be the quote of the day!!!
Old 18th January 2012
  #216
Lives for gear
 
TAFKAT's Avatar
 

Cool

Quote:
Originally Posted by thenoodle View Post

I kinda like this thread, I have about 10 Echo interfaces and apparently from the measurements, they suck.

Well to be fair, I don't have their firewire stuff......


I have not made any such claims , and you are comparing Oranges and Pears Mate , just because the PCI units perform in one manner, doesn't guarantee anything on the FW/USB2 side. Echo are not an an island there. As I noted earlier, I have personally owned and used Echo PCI interfaces over the years, and they were fine , so lets keep the eye on the ball , or the abacus, which ever the case maybe.. ;-)

Enjoy using your Echo's...

Old 18th January 2012
  #217
Lives for gear
 
dubrichie's Avatar
Quote:
Originally Posted by TAFKAT View Post
Hey D,

It is true for the H67/P67/Z68/X79 , haven't personally had any time in front of a C20x chipset, but having a quick browse across the specs at the Intel site it does look to have an older Intel® 82801CA I/O Controller , which has native PCI . With it being designed more specifically for server application were legacy compatibility is more of a focus , the native PCI support seems to have been maintained.

Come to think of it , I do remember reading in on some threads where PTHD PCI users were qualifying systems with them, not sure of the outcome or where I read it now, sorry.. :-(

Good Luck with the PTHD PCI rig, let me know how you go.. :-)

I digress,

I won't be qualifying the chipset myself , so my PCIe only testing moving forward still stands... :-)
Yes indeed, in the course of my exhaustive research of this I have come across several reports by other PTHD PCI users of solid Sandybridge Xeon systems using motherboards with the C20X chipsets, so I should be good to go.

I haven't come across anyone using the particular motherboard that I have chosen, but if PT doesn't like it I will swap it out for another one that is confirmed as working. Not a big deal. I chose this one because the complement of PCI and PCI-E slots is the best fit that I could find for my requirements, but there are similar options from other manufacturers.

I will certainly keep you posted on how i get on with it, but it will be at least a couple of weeks before I can put the new system together.

The best bit about my approach is that I can install a Lynx or RME AES PCI-E card in the same machine, connected to the same Lynx Aurora converters as the PTHD cards, and simply switch between the TDM and native (ASIO) playback engines. This will allow me to use TDM for live multitracking at the start of a project, then switch to native for mixing and overdubs, which will allow me to do everything at 88.2/96KHz without the crippling voice limitation of the TDM system. (Plus I get the shiny new floating point mix engine, longer ADC, Session Disk Cache, etc.)

The best of both worlds, and I won't even have to plug in a patch lead! This system should see me good for another 3 or 4 years of happy record making, at which point the TDM cards will probably be put out to pasture, but I will still have my PTHD licence, Lynx/RME interface and will have settled into a solid workflow for recording and mixing using PTHD without AVID PCI-E cards. Who knows what my requirements will be at that stage?! Whatever's clever, happy out!

All the best,
Old 4th February 2012
  #218
Gear Addict
guys to clear this out to me , are you stating that the rme hdspe aio is the card which has the lowest rtl latency in the market , is that why you are taking it as a baseline reference in all your tests ? and does that mean its the best PCIe card in the rme series in term of latency ? does it out perform the rme 9632 PCI card or they are both the same with the AIO is only PCIe card ? help please
Old 4th February 2012
  #219
Baseline means best of assembled test cards,so Iv'e bin told.

With the "ADI8qs hdspe is 1st",AIO is 2nd

AIO being a newer revision fer pcie I'd hazard a guess,however if you happen to notice an earlier driver gets the lowest:baseline[with the AIO] so there's yer answer.



Don't mind me......I'm going spare trying to find a pcie for a sandy bridge W7 x64 and nothin under $1000![in AU]

ANYONE KNOW OF ALTERNATIVES??[no USB NO Firewire!]
Old 4th February 2012
  #220
Gear Addict
Quote:
Originally Posted by rksguit View Post
Baseline means best of assembled test cards,so Iv'e bin told.

With the "ADI8qs hdspe is 1st",AIO is 2nd

AIO being a newer revision fer pcie I'd hazard a guess,however if you happen to notice an earlier driver gets the lowest:baseline[with the AIO] so there's yer answer.



Don't mind me......I'm going spare trying to find a pcie for a sandy bridge W7 x64 and nothin under $1000![in AU]

ANYONE KNOW OF ALTERNATIVES??[no USB NO Firewire!]
im guessing thats why they'r choosing the aio for most of their tests as the baseline reference , so it must be the best for rtl then (i am guessing)

still duno if theres much difference in terms of latency between PCI and PCIe cards to justify the huge jump in prices .. maybe some1 could help me with that..

for pci cards ur bet would be the rme hdspe,lynix or motu 424 as far as i know , where rme cards score the best as always
Old 4th February 2012
  #221
Lives for gear
 
TAFKAT's Avatar
 

Cool

Quote:
Originally Posted by fito_88 View Post

still duno if theres much difference in terms of latency between PCI and PCIe cards to justify the huge jump in prices .. maybe some1 could help me with that..
Actually the baseline is the RME HDSPe AES / ADI8 QS , it was originally the AIO but was superseded by the AES/ADI combo due to the tighter AD/DA and subsequent lower RTL.

You need to also balance RTL with the overall results at respective latencies, i.e - some cards have comparatively good RTL, but do not perform well at respective lower latency.

To answer the question re PCI and PCIe in regards to latency , one isn't any better than the other as latency is dependent on the driver and AD/DA , performance may vary depending on the respective card tho.

In my testing the HDSP 9632 and the HDSPe AIO performed pretty much on par , I haven't tested a MOTU 424 PCIe but I would suggest the performance would be close if not on par to the PCIx.

Having said that, using PCI cards on the bridged PCI slots on the current Intel chipsets is a crap shoot, so if you are weighing up the HDSP 9632 and HDSPe AIO for for a current chipset rig, go the AIO, IMO.

@ rk,

AIO's are under $1000.00 AU even with the optional XLR cabling.

Old 5th February 2012
  #222
Quote:
Originally Posted by TAFKAT View Post

Having said that, using PCI cards on the bridged PCI slots on the current Intel chipsets is a crap shoot, so if you are weighing up the HDSP 9632 and HDSPe AIO for for a current chipset rig, go the AIO, IMO.

@ rk,

AIO's are under $1000.00 AU even with the optional XLR cabling.

Cheers Vin,it's just from googling the older pci was/is almost half,compared to the pcie AIO.

Im sure its worth it,it's just when your only recording yourself, anything over $700 is a lot.

I should just bite it and go Babyface,at least I can use it on the laptop then.......but.

This is too embarrasing,with the xlr option[AIO],this means you can plug a mic in without a pre?,and can TOTAL MIX then control "phantom power" if a condensor is plugged into said xlr?

I realise the AIO is obviously supposed to be part of a much bigger signal chain,but if elderly digital noob, wanted to just use it[AIO] in this way,is it even possible?[even if it's not allowed?]

THANKS
Old 5th February 2012
  #223
Lives for gear
 
TAFKAT's Avatar
 

Quote:
Originally Posted by rksguit View Post
...with the xlr option[AIO],this means you can plug a mic in without a pre?,and can TOTAL MIX then control "phantom power" if a condensor is plugged into said xlr?
Unfortunately No, the XLR option simply allows +4 balanced interconnect over the standard unbalanced RCA connections. I have never understood why they shipped the cards with the RCA only, I always recommend the XLR option.

Old 13th February 2012
  #224
Lives for gear
 
TAFKAT's Avatar
 

Quick Update,

I managed to finally run a Lynx AES16e / Aurora combo thru its paces.

The earlier Lynx 2 card performed very well so I was interested in getting the latest PCIe offering on the bench to see if the performance could be maintained , especially with the ultra tight AD/DA latency of the Aurora.



Overall performance of the PCIe unit did not match the older Lynx 2 card unfortunately , I also experienced some very odd loss of buffer sync issues at 256 samples with the DAWbench VI test sessions which resulted in random garbled playback and metallic like artefacts on release tails, etc. This was more often than not triggered by loading a previous maxed out session from the 128 buffer size and trying to incrementally continue to load the session. I had to to start the 256 session from scratch each time. Where this could be a problem in Real World is if you had a heavily loaded session and wanted to bump up a buffer size to get some more headroom, you would in fact not be able to play the previous session without artefacts.

Overall I/O and RTL latency is actually lower than the RME AES/ADI8 reference unit , so potentially could have tipped the base reference interface if it had performed equally to the older Lynx 2 card, but unfortunately it didn't measure up to its older sibling. The results look quite good, but there are a few quirks that left me with the feeling that its not as solid as it could be.

I am currently in contact with Lynx support and am working thru a few angles.

I'll give you guys a heads up if anything fruitful eventuates.

* I have added the Lynx to the most recent database list *

Peace
Old 14th February 2012
  #225
Lives for gear
 
dubrichie's Avatar
Thanks again Vin! You're doing trojan work here for all of us!

Can you comment on the performance of ProTools 10 with the Lynx AES-16e compared to with the HD Native PCIe card and to the RME HDSPe AES?

That is what I'm really interested in!

All the best,
Old 14th February 2012
  #226
Lives for gear
 

Tafkat, your otherwise monumental and much appreciated research has touched on the Mackie Onyx 'i" series . . . but kind of left it hanging. I think you had to return a loaner device or something along those lines.

While I can't contribute hard corps numbers or nth degree testing to the discussion, I can chime in with some "empirical data" and real world experience.

First, let me say I am new to both the Onyx Blackbird and PT10/Win64 Native. I have a lot of flight time in studios -- none of it over the last few years. But I am pretty intimately familiar with PCs.

Cutting to the chase, without even really knowing what I was doing, using an upper middle class PC [i2500kwith 16 GB RAM], I just found I can hardware monitor with no detectable latency whatsoever at the lowest buffer setting of 64 [there's no option for 32]. I don't know what would happen if I was recording 16 tracks simultaneously, but layering one track at a time, syncing layered vocal tracks was right on -- and I have not practiced doubling myself in ages. It was basically a breeze.

I'm not sure, but I sense that hardware monitoring takes the Dice II chipset out of the equation.

So while I acknowledge that RME might be the king of the test lab, there are other ways to skin a cat in latency land which are way more affordable --my Blackbird was $325 through Guitar Center used online! I believe a Babyface is around $800.

I'm not quite ready to make a value judgement about the sound quality as I'm listening through phones till I make a monitors purchase, but it sure sounds at least "good."

btw this is all with a 10-year old SIIG firewire card I bought for $5.

My last comment is it's pretty obvious a lot of people were disenchanted by the original Onyx series' drivers, to the point the new "i" series gets overlooked. I don't want to gush too much, because it could still bite me, but another thing that turns people off is that Mackie is not active on their own forum and takes forever to respond to emails. HOWEVER, you can get a tech on the phone with barely any wait at all! They are fantastic at answering their phone. With Skype, you can call them from Borneo.

So that's my initial experience with the Blackbird. I'll try some virtual instruments and see if that throws it for a loop or not and report back.

Carry on.
Old 15th February 2012
  #227
Lives for gear
 
TAFKAT's Avatar
 

Quote:
Originally Posted by dubrichie View Post
Thanks again Vin! You're doing trojan work here for all of us!

Can you comment on the performance of ProTools 10 with the Lynx AES-16e compared to with the HD Native PCIe card and to the RME HDSPe AES?

That is what I'm really interested in!

All the best,
Hey D,

On the to do list for 2012 , I am hoping to get a PTHD Native rig back on the bench shortly, and I can give you a definitive on the HD Native v higher end ASIO, at least in regards to LLP.


Quote:
Originally Posted by troggg View Post
Tafkat, your otherwise monumental and much appreciated research has touched on the Mackie Onyx 'i" series . . . but kind of left it hanging. I think you had to return a loaner device or something along those lines.
Hey T,

I was promised the current rack/desktop Mackie Onyx units by one of the local reps to do some qualification and testing on, but that didn't eventuate unfortunately.

I did do a full round of testing on an i Series mixers, that however is not a Dice controller, its based on the earlier one that is a joint effort of Echo/Mackie I believe. I am needing some accurate RTL numbers before I post the unit into the charts , but just can't seem to get in front of another unit to finalise them.. :-(


Quote:
Cutting to the chase, without even really knowing what I was doing, using an upper middle class PC [i2500kwith 16 GB RAM], I just found I can hardware monitor with no detectable latency whatsoever at the lowest buffer setting of 64 [there's no option for 32]. I don't know what would happen if I was recording 16 tracks simultaneously, but layering one track at a time, syncing layered vocal tracks was right on -- and I have not practiced doubling myself in ages. It was basically a breeze.
Hardware Direct Monitoring has no correlation to the LLP aspect I am focusing on. With Hardware Direct Monitoring the only latency is that of the AD/DA , the buffer setting has no relevance

Quote:
I'm not sure, but I sense that hardware monitoring takes the Dice II chipset out of the equation.
Correct... :-)

Quote:
So while I acknowledge that RME might be the king of the test lab, there are other ways to skin a cat in latency land which are way more affordable --my Blackbird was $325 through Guitar Center used online! I believe a Babyface is around $800.
Horses for Courses.

Where the LLP aspect becomes more significant is the overall scaling performance at the respective buffer settings.

@ 64 Samples for example, most Dice based units will collapse at around 50% ASIO load in Cubase , whereas the better units will scale cleanly into the 90+% range , that is quite a significant difference for those that require low latencies for playback of virtual instruments and monitoring thru FX -i.e Guitar Amp Sims .

YMMV depending on individual demands / working environments

Old 15th February 2012
  #228
Gear Maniac
 

Great thread, thanks.

Have had so many frustrating experiences with firewire interfaces - the Aurora is already mentioned - so this thread is greatly appreciated.
Old 15th February 2012
  #229
Lives for gear
 

wonder how UA Apollo wants to stay under 2ms
Old 15th February 2012
  #230
Lives for gear
 
TAFKAT's Avatar
 

Quote:
Originally Posted by lllubi View Post
wonder how UA Apollo wants to stay under 2ms
That is only in reference to hardware direct monitoring , granted that is also with their own plugins as well running in their proprietary external mixer, so it is different to most other units in that respect. YMMV.

However , no one has any idea how the unit will actually scale at low latencies , which will be totally dependent on the controller /drivers being implemented , so we will have to wait and see what they actually deliver.

Old 15th February 2012
  #231
Lives for gear
 

Speaking of horses for courses and YMMV according to individual demands/working environment, you are probably correct in implying [I'm reading between the lines, not quoting you directly] that for mission critical applications in a professional environment, RME or Lynx are more battle-tested solutions [I've owned a Lync Two C PCI and it performed flawlessly; weirdly, it's the same price now it was in 2003].

But I reckon there are just as many guys like me on GS who are more artists than engineers.

So, if you don't run a mission-critical commercial studio, and you're recording material that is short of a full film score, that makes you a candidate for the Onyx "i" series which could free significant funds for other key purchases. A Stellar CM-6 mic with a NOS tube springs to mind . . .

Regarding the drivers [which I believe are the same for the rack and the mixers, though I'm not sure] and tie-in with Echo, that arrangement was certainly present for the previous Onyx series which PO'ed a lot of people. The driver for the new series, which is not new and no updates are threatened, for whatever reason Mackie seem to have gotten right from the get-go. Echo is no longer mentioned, so I'm guessing drivers are handled completely in house.

I don't know everything, but I know enough to say that Onyx 'i" is a contender worthy of inclusion in your testing once you get your hands on one.
Old 15th February 2012
  #232
Lives for gear
 
TAFKAT's Avatar
 

Quote:
Originally Posted by troggg View Post
Regarding the drivers [which I believe are the same for the rack and the mixers, though I'm not sure] and tie-in with Echo, that arrangement was certainly present for the previous Onyx series which PO'ed a lot of people. The driver for the new series, which is not new and no updates are threatened, for whatever reason Mackie seem to have gotten right from the get-go. Echo is no longer mentioned, so I'm guessing drivers are handled completely in house.
Early Onyx rack units were using the jointly developed Echo controller/driver , the current Blackbird and Blackjack are using the Dice controllers/drivers that Presonus/Focusrite, etc, are using , the i Series is not using either from what I can make out. The i Series has a very stripped down control of the FW interconnect, no hardware monitoring via control panel , no ASIO direct monitoring , buffer sizes only accessible via DAW host , driver reporting nominal figures. None of that gels with either the earlier Echo based controller/driver or the current Dice , so it may be an alternate offshoot from the earlier joint development.

Quote:
I don't know everything, but I know enough to say that Onyx 'i" is a contender worthy of inclusion in your testing once you get your hands on one.
I have tested it and all the results are done, the LLP rating and final inclusion is pending due to the unit - as mentioned above- reporting nominal figures for I / O latencies. I need to get an accurate RTL value before I can include the unit in the charts.

Old 16th February 2012
  #233
Lives for gear
 

Quote:
Originally Posted by TAFKAT View Post
Early Onyx rack units were using the jointly developed Echo controller/driver,

---I believe you're referring to the Onyx 400F, the Onyx 800F, and the Onyx 1200f -- which make up the previous series, as opposed to the Blackbird which is the only rack-mounted offering in the new series.

the current Blackbird and Blackjack are using the Dice controllers/drivers that Presonus/Focusrite, etc, are using , the i Series is not using either from what I can make out.

---I have no direct info to dispute this, other than to say most people with the patience and experience to work through a few difficulties here and there instead of throwing blame around seem to be having an equally smooth experience whether we're talking about the Blackbird, or the mixer versions of the current Onyx line including the 820i, the 1220i, the 1620i, and the 1640i. Which is why I assumed the driver was the same or very similar. The definite similarity between the current Blackbird rack and the i mixers is the presence of the same preamps and converters.

The i Series has a very stripped down control of the FW interconnect, no hardware monitoring via control panel ,

---Now I'm really glad I didn't get the 1220i I was considering, cause the hardware monitoring works surprisingly great.

no ASIO direct monitoring , buffer sizes only accessible via DAW host , driver reporting nominal figures.

---I don't know what "nominal figures" are.

None of that gels with either the earlier Echo based controller/driver or the current Dice , so it may be an alternate offshoot from the earlier joint development.

---All I can say is that the current Blackbird driver is just not flaky at all. It's given me zero grief, and I'm running a used unit. Every surprise has been pleasant.

I have tested it and all the results are done, the LLP rating and final inclusion is pending due to the unit - as mentioned above- reporting nominal figures for I / O latencies. I need to get an accurate RTL value before I can include the unit in the charts.

---Sorry to interject subjective values into a numbers-oriented thread, but I can't help but like the unit since a divorce left me in a different financial position than I was in when I was running a Lynx Two and preamps like Avalon 2022 and Groove Tubes Vipers; in other words $5,000 worth of gear. The Blackbird is giving every indication it's capable of replacing that for $325. Sure, I could bitch that most of the gain comes from 3 o'clock and up on the gain knob, but this device has meant I can record again. You can't measure that. There may or may not be a Lynx Hilo in my future, but I'm OK with the Blackbird as my present.
Old 16th February 2012
  #234
Lives for gear
 
TAFKAT's Avatar
 

Quote:
Originally Posted by troggg View Post

-I don't know what "nominal figures" are
Nominal figures are the theoretical values for each respective buffer setting , they do not include the actual safety buffers or the AD/DA , many interfaces do not report AD/DA as I have noted but most do at least report the safety buffers.

If the i Series has high additional buffers that are not being reported to the host DAW, that could cause possible issues with synchronisation.

Quote:
All I can say is that the current Blackbird driver is just not flaky at all. It's given me zero grief, and I'm running a used unit. Every surprise has been pleasant.
I haven't reported any of the Dice interfaces as being flaky, so thats not surprising.

Quote:
The Blackbird is giving every indication it's capable of replacing that for $325. Sure, I could bitch that most of the gain comes from 3 o'clock and up on the gain knob, but this device has meant I can record again. You can't measure that. There may or may not be a Lynx Hilo in my future, but I'm OK with the Blackbird as my present.
What I am reporting needs to be taken into perspective of individual working environments, if you are doing basic live tracking and not relying on large plugin counts , heavy sample based instrument usage or amp simulators , then the Dice units are fine, its why I keep repeating YMMV and horses for courses.

Enjoy your Blackbird.. ;-)

Old 16th February 2012
  #235
Lives for gear
 

Thanks. Keep up the great work!
Old 18th February 2012
  #236
Here for the gear
 

Hi Vin!

Thanks for your amazing efforts & work. It's really great to have this kind of data in the public domain. I would also like to commend you on the fact that you share this kind of data as a commercial DAW builder - you might have kept it all to yourself in the name of "trade secrets" too!

I've been following your work for a long while. I've actually conducted your VI polyphony tests on my current hardware too. I'm still on a C2D 2.6 MBP (non-unibody) and over the years I've "collected" an MA Firewire 1814, an MA Fast Track Ultra, an MA Transit USB (darned PT-MP). Along with C6, I ported your test to Logic 9 and PT-MP 8 too and I've done all the tests on a Boot-camped W7 SP1 x64 partition as well - including the built in card via ASIO4ALL. I actually have made a chart of the test data too but I don't want to make this post about that. The only thing I'd like to note is that the built in sound card is currently my best LLP interface. This is in the case of a MBP relevant because it's 1/8" ports double as mini TOSLINK. So theoretically one could add 2 channels of AD/DA (say the new Lynx HILO) and have stellar audio quality too. I think the same would go for some PC notebooks and most desktops with S/PDIF port(s). I need not mention the fact that these internal cards are PCI/e - in the case of the Mobile 965 chipset, there's a direct connect between "Intel HD Audio" and the ICH so I can't tell what type of bus it actually is.

I'm planning to get my hands on an MBox 3 Pro, an FF UCX and a UA Apollo when it's released. I'm also intending to get a new MBP which will boast similar "Geekbench" power to your i7-920 it seems. I'll then run the tests cross-platform like before. Then my data may be worth sharing and more relevant...

Anyhow, a couple of things I wanted to note / ask;

- The way you've set up the VI test, when you say "0" notes polyphony, it may be kind of misleading because the first K4 instance is already doing anywhere between 60-80 voices of polyphony. I think this is still significant as there is hardware / software / buffer setting combos that can't even do playback on the "0" polyphony version of the test. It might make sense to somehow incorporate this info into the test results - say a "1" or "PB" for playback and a "0" for no playback at all at the given buffer setting with no added voices. "PB" would mean all those bad boys with "0"s are actually doing 60-80 voices of polyphony at 32 samples!

- Cubendo is the king of the hill in terms of voice count. BUT, one of the variables we have no control over is: is K4 equally optimized across VST, RTAS and AU? I tested this idea like so; in my PTMP port of your test, I've basically kept the first K4 instance as is but loaded the K3 synth patch into "Structure Free" of PTMP (anyone can do that easily as Structure comes with built in K2 and K3 patch support). And bingo, PTMP did many, many more voices of polyphony without skipping a beat*, compared to what it did with the RTAS K4. Now, I haven't been able to run the whole test with Structure because I don't have the full version and I'm limited to 4 Structure Free instances (i.e. 16 patches) per session. But this is on my to-do list. I'll do this once I get my hands on the AIR Instrument Collection package. I suspect the same might be true for Logic's EXS. I have a plan for that too through patch translation. Finally, that brings up the question; could HALion perform even better than K4 on the Cubendo side - the trend being DAW makers have the best access to their own code and optimize their samplers accordingly. Of course, we will never know whether it's AVID that reserves certain types of functions to its own VIs and does not let NI or other VI developers use those, or whether it's NI that's not giving RTAS their best...

- I've been crunching your last data set into an Excel file for further analysis. I calculated the differences between measured RTL and reported RTL. In some cases it seemed to be a fixed amount across buffer sizes: Lynx AES16e/Aurora: 0.567 | MOTU424/24IO: 1.202 | MOTU 828mk3: 0.605. But for other combos it varied across buffer sizes: AVID Mbox 3 Pro: 0.861, 0.839, 0.969, 0.860 || RME HDSP AES/ADI8QS: 0.634, 0.635, 0.688, 0.668, 0.628. This may be caused by; a) AD/DA behaving differently across buffers, b) Interface using different safety buffers for different buffer settings, and c) those RTL measurements might have been less then exact while still in the ball park. What do you think the cause may be?

- When you don't report a "measured RTL", is this because you've done the measurement and it matches the reported RTL, or is it because you trust builders like RME and Lynx in this area?

- One last mundane note; I think there is no "Fast Track Pro 8R" - it's either FT Ultra 8R or FT Pro ;)

Many, many thanks for all your efforts!

K.

p.s. If you think parts of this post are more relevant elsewhere, let me know and I'll edit those out and move them over...


* A quick example; with Structure Free, PTMP 805 did 120 voices where it did 10 with K4 on the poly patches.
Old 23rd February 2012
  #237
Lives for gear
 
TAFKAT's Avatar
 

Quote:
Originally Posted by krofa View Post
Hi Vin!

Thanks for your amazing efforts & work. It's really great to have this kind of data in the public domain. I would also like to commend you on the fact that you share this kind of data as a commercial DAW builder - you might have kept it all to yourself in the name of "trade secrets" too!
Thanks for the support , much appreciated.

Sorry for being so slack in replying, I somehow missed your post earlier this week.
Quote:
I've been following your work for a long while. I've actually conducted your VI polyphony tests on my current hardware too. I'm still on a C2D 2.6 MBP (non-unibody) and over the years I've "collected" an MA Firewire 1814, an MA Fast Track Ultra, an MA Transit USB (darned PT-MP). Along with C6, I ported your test to Logic 9 and PT-MP 8 too and I've done all the tests on a Boot-camped W7 SP1 x64 partition as well - including the built in card via ASIO4ALL. I actually have made a chart of the test data too but I don't want to make this post about that. The only thing I'd like to note is that the built in sound card is currently my best LLP interface. This is in the case of a MBP relevant because it's 1/8" ports double as mini TOSLINK. So theoretically one could add 2 channels of AD/DA (say the new Lynx HILO) and have stellar audio quality too. I think the same would go for some PC notebooks and most desktops with S/PDIF port(s). I need not mention the fact that these internal cards are PCI/e - in the case of the Mobile 965 chipset, there's a direct connect between "Intel HD Audio" and the ICH so I can't tell what type of bus it actually is.
Definitely some interesting areas there , especially the performance of ASIO4All on the MBP's inbuilt audio card. You need to make note of the actual RTL under ASIO4All, my understanding is that it adds some hefty safety buffers.

Quote:
I'm planning to get my hands on an MBox 3 Pro, an FF UCX and a UA Apollo when it's released. I'm also intending to get a new MBP which will boast similar "Geekbench" power to your i7-920 it seems. I'll then run the tests cross-platform like before. Then my data may be worth sharing and more relevant...
Cool, keep us up to date..

Quote:
Anyhow, a couple of things I wanted to note / ask;

- The way you've set up the VI test, when you say "0" notes polyphony, it may be kind of misleading because the first K4 instance is already doing anywhere between 60-80 voices of polyphony. I think this is still significant as there is hardware / software / buffer setting combos that can't even do playback on the "0" polyphony version of the test. It might make sense to somehow incorporate this info into the test results - say a "1" or "PB" for playback and a "0" for no playback at all at the given buffer setting with no added voices. "PB" would mean all those bad boys with "0"s are actually doing 60-80 voices of polyphony at 32 samples!
I hear ya, I should make a more pronounced note that its added polyphony over the template.

When I report a 0 value for the VI test, its indicating that the template was not playing back cleanly.

Quote:
- Cubendo is the king of the hill in terms of voice count. BUT, one of the variables we have no control over is: is K4 equally optimized across VST, RTAS and AU?
That is a good question and not one that I can easily answer.

Quote:
I tested this idea like so; in my PTMP port of your test, I've basically kept the first K4 instance as is but loaded the K3 synth patch into "Structure Free" of PTMP (anyone can do that easily as Structure comes with built in K2 and K3 patch support). And bingo, PTMP did many, many more voices of polyphony without skipping a beat*, compared to what it did with the RTAS K4.
Interestingly when I first approached a contact at AVID about the poor performance of PT in the VI test , his first suggestion was to do a test in Structure. My response was that I needed a VI that was cross platform and multi protocol- that was also widely used and supported , Structure does not fit into any of those boxes.

The issue with the RTAS VI performance is not reserved to K4 tho, its also very evident in say VSL for example , so there is a little more meat to it.


Quote:
Now, I haven't been able to run the whole test with Structure because I don't have the full version and I'm limited to 4 Structure Free instances (i.e. 16 patches) per session. But this is on my to-do list. I'll do this once I get my hands on the AIR Instrument Collection package. I suspect the same might be true for Logic's EXS. I have a plan for that too through patch translation.
All very interesting and would be great to get some comparative figures across the VI's using the translated libraries.

Quote:
Finally, that brings up the question; could HALion perform even better than K4 on the Cubendo side - the trend being DAW makers have the best access to their own code and optimize their samplers accordingly. Of course, we will never know whether it's AVID that reserves certain types of functions to its own VIs and does not let NI or other VI developers use those, or whether it's NI that's not giving RTAS their best...
I think Steinberg totally dropped the ball on HALion unfortunately , it would be interesting to see how HALion 4 performs comparatively, I just need to weigh up the time and energy against the care factor for the wider community.

The simple fact that Structure performed so much better in your initial test is hinting that the devs could be holding their cards closer to their chests in some aspects of the optimizations , but it would be to their own detriment.

Quote:
- I've been crunching your last data set into an Excel file for further analysis. I calculated the differences between measured RTL and reported RTL. In some cases it seemed to be a fixed amount across buffer sizes: Lynx AES16e/Aurora: 0.567 | MOTU424/24IO: 1.202 | MOTU 828mk3: 0.605. But for other combos it varied across buffer sizes: AVID Mbox 3 Pro: 0.861, 0.839, 0.969, 0.860 || RME HDSP AES/ADI8QS: 0.634, 0.635, 0.688, 0.668, 0.628. This may be caused by; a) AD/DA behaving differently across buffers, b) Interface using different safety buffers for different buffer settings, and c) those RTL measurements might have been less then exact while still in the ball park. What do you think the cause may be?
More than likely slight discrepancies in the RTL calculation by the Utility.

We are still working out some minor kinks to give a tighter and more consistent result.

Quote:
- When you don't report a "measured RTL", is this because you've done the measurement and it matches the reported RTL, or is it because you trust builders like RME and Lynx in this area?
Both, RME and Lynx are very accurate in what they report to the ASIO host including the AD/A on the respective interfaces that have them onboard. We even uncovered a minor glitch where the RTL utility was under reporting the RTL compared to the reported with RME FF800/UFX for example and when we resolved the bug , it was right on what RME were reporting to the host.

Quote:
- One last mundane note; I think there is no "Fast Track Pro 8R" - it's either FT Ultra 8R or FT Pro
You are right, thanks for the heads up, I'll amend that when I next update the charts.

BTW: It would be great to get the Logic ports of the VI test if you don't mind sharing them and have them included in the suite, full credit given of course.

Drop me a P.M


Old 28th February 2012
  #238
Lives for gear
I wonder how most of the newer "DJ oriented" interfaces stack up. They're usually straight analog in/out boxes that tend to prioritize stable latency for live use over features like mic pres, digital in/out, internal dsp mixing, etc. The goal for reading timecode vinyl is to have it feel as close to a real record as possible.

The new Native Instruments interfaces (Traktor 6 and 10) supposedly have improved drivers (although the previous generation had few complaints) with even better latency.

The same with the newer Rane/Serato boxes. They've been bumped up to 24bit/96khz, USB 2.0, and claim better drivers.

I guess my reason for asking is that the price seems to lower for these interfaces (for the amount of analog ins/outs) and should be of interest to folks that don't need all the bell and whistles.
Old 2nd March 2012
  #239
Gear Maniac
 

Apogee Symphony @ 1.6-1.8ms
Lynx AES 16e / Aurora 16 (2.19e) @ 2.132ms
ESI U46 XL : Driver 3.0 @ 2.902ms
RME HDSPe AES/ADI8QS : Driver 3.24 - MC Off @ 2.947ms
RME HDSP Express/Multiface (3.083) @ 3.628ms
MOTU 424 : 24 IO @ 3.673ms
Lynx TWO-C @ 3.855ms
RME HDSPe AIO @ 4.263ms

We need more real data like this. I've contacted Vin to see if he has any more RTLs, but I wanted to put it to the forum as well. Where else can we start collecting this data?
Old 2nd March 2012
  #240
Where are you getting 2.902ms for the ESI U46XL?
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…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Similar Threads
Thread
Thread Starter / Forum
Replies
mattneale / Music Computers
15
gtxn / Music Computers
12
piotrekz / Music Computers
5

Forum Jump
Forum Jump