Results 1 to 3 of 3

Thread: Blofelds Stage II/ L-Factor Report - Results :

  1. #1
    Join Date
    Aug 2005
    Location
    Melbourne : Australia :
    Posts
    4,249

    Default Blofelds Stage II/ L-Factor Report - Results :

    Blofelds DSP40 Stage II results for Core2Duo/ Woodcrest -

    -----------------------------------------------------------------------------------------------------------------------------------

    Blofelds DSP 40R5 : Core2 Duo E6600 @ 2.4 GHZ : 1066 FSB : Nuendo 3.2

    Motherboard : i975

    Memory : 1 GB - XMS 6400 @ PC6400 - 800 MHZ : Standard Timings by SPD :

    O.S : XP SP2 :

    Lynx 2 : Driver : V2 Build 13l

    256 Samples - 71 Magneto's / Save-ReOpen 71 : CPU 95+%

    128 Samples - 64 Magneto's / Save-ReOpen 64 : CPU 95+%

    064 Samples - 55 Magneto's / Save-ReOpen 50 : CPU 95+%

    032 Samples - No Go From Get Go :

    ---------------------------------------------------------------------------------

    Blofelds DSP 40R5 : Core2 Duo E6600 @ 2.4 GHZ : 1066 FSB :Nuendo 3.2

    Motherboard : i975

    Memory : 1 GB - XMS 6400 @ PC6400 - 800 MHZ : Standard Timings by SPD :

    O.S : XP SP2 :

    RME HDSP 9632 : Driver : V3 Beta 10 / V3 Firmware

    256 Samples - 66 Magneto's / Save-ReOpen 57 : CPU 95+%

    128 Samples - 61 Magneto's / Save-ReOpen 40 : CPU 95+%

    064 Samples - No Go from get Go

    ---------------------------------------------------------------------------------

    Blofelds DSP 40R5 : Core2 Duo E6600 @ 2.4 GHZ : 1066 FSB : Nuendo 3.2

    Motherboard : i975

    Memory : 1 GB - XMS 6400 @ PC6400 - 800 MHZ : Standard Timings by SPD :

    O.S : XP SP2 :

    RME HDSP 9632 : Driver : V2.94 / V3 Firmware

    256 Samples - 69 Magneto's / Save-ReOpen 58 : CPU 95+%

    128 Samples - 64 Magneto's / Save-ReOpen 41 : CPU 95+%

    064 Samples - No Go from get Go

    ---------------------------------------------------------------------------------------

    The Lynx results are right on par proportionally to Scotts results with the 2.67 E6700 - slight pauses when enabling Magneto's , smooth overall scrolling , metering, etc..only a slight sluggishness at close to the break point. The RME results however are confusing to say the least.

    Firstly , the V3 results , although good - don't tell the whole story, as the behaviour running the session was frustrating and tedious .., extremely long pauses when enabling the Magneto's - accompanied by constant HD overloading , overall sluggishness, glitches when scrolling across the mixer, in short a bloody nightmare navigating thru 60+ PlugIns.. :-(

    The V2 drivers were forced onto the newer firmware that has been flashed onto the card for the V3 Beta Drivers. The performance is nothing short of amazing compared to the earlier experience I had with the V2x drivers / firmware on the i975 / Pentium D Systems. What is really bizarre is that they not only outperformed the new V3 drivers , but the overall behaviour of the system was far smoother - slight pauses when enabling the Magneto's , smooth scrolling, no glitching related to GUI, etc., etc.. Obviously I need this to be confirmed , but in short, If I was using an RME card on an Intel C2D system right now, I'd be using a combination of the latest Beta Firmware, and the existing 2.94 Driver.. Go figure.. ??

    So much for moving away from the RME v Lynx Debate.. Sigh.. :-(

    O.K time for some Fun..,

    I wanted to get an idea of what these chips would deliver at 3.0GHZ + , so I clocked the FSB up 25%, I used only standard voltages, Air Cooling via a Heatpipe Cooler - running near silent.., system never even broke into a sweat under full load.. LOL !

    -------------------------------------------------------------------------------------------

    Blofelds DSP 40R5 : Core2 Duo E6600 @ 3.0 GHZ : 1333 FSB : Nuendo 3.2

    Motherboard : i975

    Memory : 1 GB - XMS 6400 @ PC6400 - 833 MHZ : Standard Timings by SPD :

    O.S : XP SP2 :

    Lynx 2 : Driver : V2 Build 13l

    256 Samples - 92 Magneto's / Save-ReOpen 92 : CPU 95+%

    128 Samples - 84 Magneto's / Save-ReOpen 84 : CPU 95+%

    064 Samples - 74 Magneto's / Save-ReOpen 68 : CPU 95+%

    032 Samples - No Go From Get Go :

    ---------------------------------------------------------------------------------
    Shite !!!!

    That just about summs it up.... ;-)

    The system was not 100% stable at this setting, I did experience a few freezes on re-opening the session at the loaded setting, but considering I spent all of 30 seconds overclocking the system , I am not overly concerned.


    ------------------------------------------------------------------------------------------

    Blofelds DSP 40R5 : Core2 Duo E6400 @ 2.13 GHZ : 1066 FSB : Nuendo 3.2

    Motherboard : i975

    Memory : 1 GB - XMS 6400 @ PC6400 - 800 MHZ : Standard Timings by SPD :

    O.S : XP SP2 :

    Lynx 2 : Driver : V2 Build 13l

    256 Samples - 56 Magneto's / Save-ReOpen 56 : CPU 95+%

    128 Samples - 47 Magneto's / Save-ReOpen 47 : CPU 95+%

    064 Samples - 04 Magneto's / Save-ReOpen 04 : CPU ?+%

    032 Samples - No Go From Get Go :

    --------------------------------------------------------------------------------

    A quick browse over the results clearly indicates that the 4MB L2 Cache is the ace for lower latency performance below 128 Samples.

    Still not a bad overal results, considering it has just easily accounted for the AMD FX60 Chip, which would be substantially more expensive, even with the price gut that AMD have recently slashed thru. Where some believed AMD would compete on Price / Performance , these results have just put a dampener on that as well.. :-(


    -----------------------------------------------------------------------------------

    Blofelds DSP 40R5 : Dual Xeon Core2 Duo 5150 @ 2.66 GHZ : 1333 FSB : Nuendo 3.2

    Motherboard : i5000

    Memory : 2 GB - FBDIMM DDR2 667 - 4 Channel Interleaved : Standard Timings by SPD :

    O.S : XP SP2 :

    Lynx 2 : Driver : V2 Build 13l

    256 Samples - 159 Magneto's / Save-ReOpen 159 : CPU 95+%

    128 Samples - 130 Magneto's / Save-ReOpen 127 : CPU 95+%

    064 Samples - 101 Magneto's / Save-ReOpen 98 : CPU 95+%

    032 Samples - 65 Magneto's / Save-ReOpen 55 : CPU 75+%

    ----------------------------------------------------------------------------------------------------

    Well that just about ends any and all argument .. Simply astounding performance , I was able to push the system at each latency setting except 32 Samples to 95+ % CPU, and all was smooth as silk right to the break point. Save / Reopen results were well within what I would consider acceptable.

    The performance at 32 samples still has me shaking my head in disbelief.. !!

    For those wondering about the RME Results .., Well heres the rub, after much consideration I have decided to hold off until the final release of the V3 drivers, then I will post a full report.

    After being recently target for daring to test and then report a less than glowing summary of my personal experiences with the Beta drivers on my Conroe Rig, I have decided that I have far more important things to do than get into pointless arm wrestles with certain individual here that seem to think that while its fine for them to give glowing reports on their personal experiences with said unfinished drivers, the same courtesy does not extend to those of us having less than stellar results. Being labeled an Attack Dog was something that at first pissed me off, but after giving it some thought.., I came to the conclusion that its far better to be an Attack Dog, than a Lap Dog.. LOL !!

    In the meantime we can use Pals results for his Dual 3.0 GHZ Woodcrest / RME Rig as a reference..

    --------------------------------------------------------------------------------------------------------------

    Blofelds DSP 40R2: Dell P650 Dual Xeon 5160 @ 3 GHZ: Nuendo 3.2 .1
    C/O Psvennevig:

    Motherboard : Dell P690, Intel 5000 based.

    Memory: 4 GB, FB-DIMM DDR2 667 MHz (original RAM)

    O.S: WinXP SP2:

    RME HDSP MADI: Driver: 3.0 Beta10

    512 Samples – 161 Magneto's (task manager shows 88% load during playback)

    256 Samples – 131 Magneto’s (task manager shows 74% load during playback)

    64 Samples - 71 Magneto's (task manager shows 68% load during playback)

    32 Samples - 18 Magneto's (task manager shows 56% load during playback)

    I had to switch off between 2 and 6 magnetos to get crackle free playback after reload of Nuendo or the project.
    ----------------------------------------------------------------------------------------------------------------

    A quick comparison with my results on a system clocked substantially lower clearly shows just how much variable there is with respective audio cards at the lower latencies , even on Uber Systems such as these.., I'll let the figures speak for themselves.. :-)

    Peace:

    V:

  2. #2
    Join Date
    Aug 2005
    Location
    Melbourne : Australia :
    Posts
    4,249

    Default

    Hey All,

    After numerous twists , turns, and a few false starts...

    Firstly I would like to thank those who graciously gave their time and energy to help me thru the Beta Stages, much appreciated.

    Credits: Original Music composed by Rob John : Independent Music Productions : Melbourne : Australia :

    Session development, arrangement, methodology, production, synth programming : All by yours truly..:-)

    L-Factor II : Specifically designed for Nuendo / SX 3.0 , but can be run on C4 :

    Note : Results my vary on C4 :

    http://www.aavimt.com.au/downloads/bench2.zip

    There are 17 Stereo Audio Tracks - 16 Sine Waves, 1 Drum Loop , all with some EQ/Dynamics..

    16 Midi Tracks in the musical template with melodic and sustained parts , all playing individual VSTi's.

    A mixture of most of the VSTi's in Nuendo/SX 3.0 .. :-)

    There are an extra 48 A1's , the first 25 all active , but with the Midi parts muted, each part having 10 notes of polyphony in sustained sections that correlate to the musical context. The parts are all zeroed out volume wise tho, as it would turn into a mess pretty quickly..

    The session as it stands allows me an extra 480 voices of polyphony above the template, as that is the maximum Nuendo /SX will allow me.., I'll explain below..

    There are a maximum 64 available VSTi slots, so I had an extra 48 slots left after the original 16 used in the sequence , that allowed me an extra 48 A1's, each assigned to a track using 10 voices to give an extra 480 voices.

    Polyphony tracks 260- 480 are placed in a separate folder, each track is muted, + the accompanying A1 is also off, that way we can scale it progressively for both Single and Dual DualCore/Quadcore rigs , so you unmute the track, and then turn on the appropriate A1 in the VSTi rack, a bit of a pain, but once you get a rhythm going its easy..: Note: Tracks for Polyphony 490 and 500 are assigned back to A1 11/12 respectively.., I had another template right up to 600 Voices, re-assigning the tracks to the A1's already loaded , but that was not really scaling correctly as we need an individual VSTi per track, so I dumped that idea.. just ignore 490/500

    BTW: The Woodcrest sailed past 600 notes of polyphony using the method of recycling the earlier A1's, but I decided it was not reflective, so I canned it.. :-)

    Instructions for running the test is very straight forward.

    While the session is running, you unmute each track in turn, each time adding 10 notes of polyphony.

    Note when you hear some crackles or drop outs, I would then move back to the previous track , mute and unmute that track again just to ensure that the previous setting had played back cleanly..

    Thats it.., It'll take about 5-10 minutes to run the test across all latency settings.

    A quick guideline: Older and/or lower end DualCores will be within the first 25 tracks, faster single DualCore Rigs will exhaust the first 25 tracks easily, Dual Dualcore Opterons are still well within the available headroom of the 480 voices , the Dual DualCore and Quadcore Intel Rigs will exhaust all the available voices in some instances:

    Once the Dual Quadcores hit, I will need to have another look at ways to scale this test further.

    I have collated preliminary results from some of the guys that have been Beta Testing this session, as well as adding a whole bunch of my own figures.

    O.K,

    Here are some numbers..,

    ---------------------------------------------------------------------------------

    L-Factor II:R9 : AMD X2 4200+ @ 2.2 GHz: SX 3.1
    C/O Franjo :

    Motherboard : ASUS A8N-E

    Memory : 2 GB – PC3200 , 1T :

    O.S : XP SP2 :

    Presonus Firebox : Driver : ??

    1024 Samples - 80 Added Notes of Polyphony:

    512 Samples - 60 Added Notes of Polyphony:

    256 Samples - 30 Added Notes of Polyphony:

    128 Samples - 00 Added Notes of Polyphony:

    --------------------------------------------------------

    L-Factor II:R9 : AMD X2 4200+ @ 2.6 GHz: SX 3.1
    C/O Franjo :

    Motherboard : ASUS A8N-E

    Memory : 2 GB – PC3200 , 2T :

    O.S : XP SP2 :

    Presonus Firebox : Driver : ??

    1024 Samples - 130 Added Notes of Polyphony:

    512 Samples - 120 Added Notes of Polyphony:

    256 Samples - 80 Added Notes of Polyphony:

    128 Samples - 00 Added Notes of Polyphony:

    ---------------------------------------------------------------------------
    L-Factor II:R9 : Core2 Duo E6600 @ 2.4 GHZ : 1066 FSB : Nuendo 3.2
    C/O Captain Caveman

    Motherboard : i975

    Memory : 2 GB - XMS 6400

    O.S : XP SP2 :

    RME Fireface 400 : Driver : ??

    256 Samples - 130 Added Notes of Polyphony:

    128 Samples - 40 Added Notes of Polyphony:

    096 Samples - 00 Added Notes of Polyphony:

    064 Samples - 00 Added Notes of Polyphony:


    ----------------------------------------------------------------------------------

    L-Factor II:R9 : Core2 Duo E6600 @ 2.99 GHZ : 1333 FSB : Nuendo 3.2
    C/O Captain Caveman

    Motherboard : i975

    Memory : 2 GB - XMS 6400

    O.S : XP SP2 :

    RME Fireface 400 : Driver : ??

    256 Samples - 190 Added Notes of Polyphony:

    128 Samples - 110 Added Notes of Polyphony:

    096 Samples - 40 Added Notes of Polyphony:

    064 Samples - 00 Added Notes of Polyphony:

    -------------------------------------------------------------------------------------

    L-Factor II:R9 : Core2 Duo E6700 @ 2.66 GHZ : Stock speed : * Cubase 4.01*

    C/O ionecube:

    Motherboard : i965

    Memory : 2 GB - DDR2-800 : Standard Timings by SPD :

    O.S : XP SP2 :

    Lynx 2 (L22) : Driver : V2 Build 13n

    256 Samples - 220 Added Notes of Polyphony:

    128 Samples - 200 Added Notes of Polyphony:

    064 Samples - 170 Added Notes of Polyphony:

    032 Samples - 00 Added Notes of Polyphony:

    ----------------------------------------------------------------

    L-Factor II:R9 : Opteron 280 @ 2.4 GHZ Nuendo 3.2
    C/O SimonL

    Motherboard : H8DCE

    Memory : 4 GB - PC3200 ECC

    O.S : XP SP2 :

    Lynx2 : Driver : V2 Build 13n

    256 Samples - ?? Added Notes of Polyphony:

    128 Samples - 280 Added Notes of Polyphony:

    064 Samples - 250 Added Notes of Polyphony:

    032 Samples - 90 Added Notes of Polyphony:

    -------------------------------------------------------------------------------

    L-Factor II:R9 : Core2 Duo E6600 @ 2.4 GHZ : 1066 FSB : Nuendo 3.2
    C/O TAFKAT:

    Motherboard : i975

    Memory : 1 GB - XMS 6400 @ PC6400 - 800 MHZ : Standard Timings by SPD :

    O.S : XP SP2 :

    Lynx 2 : Driver : V2 Build 13n

    256 Samples - 200 Added Notes of Polyphony:

    128 Samples - 190 Added Notes of Polyphony:

    064 Samples - 170 Added Notes of Polyphony:

    032 Samples - 60 Added Notes of Polyphony:

    ----------------------------------------------------------------------------------

    L-Factor II:R9 : Core2 Duo E6600 @ 2.66 GHZ : 1180 FSB : Nuendo 3.2
    C/O TAFKAT:

    Motherboard : i975

    Memory : 1 GB - XMS 6400 @ PC5700 - 780 MHZ : Standard Timings by SPD :

    O.S : XP SP2 :

    Lynx 2 : Driver : V2 Build 13n

    256 Samples - 250 Added Notes of Polyphony:

    128 Samples - 230 Added Notes of Polyphony:

    064 Samples - 210 Added Notes of Polyphony:

    032 Samples - 110 Added Notes of Polyphony:

    ---------------------------------------------------------------------------------------

    L-Factor II:R9 : Core2 Duo E6600 @ 3.0 GHZ : 1333 FSB : Nuendo 3.2
    C/O TAFKAT:

    Motherboard : i975

    Memory : 1 GB - XMS 6400 @ PC5400 - 667 MHZ : Standard Timings by SPD :

    O.S : XP SP2 :

    Lynx 2 : Driver : V2 Build 13n

    256 Samples - 290 Added Notes of Polyphony:

    128 Samples - 280 Added Notes of Polyphony:

    064 Samples - 260 Added Notes of Polyphony:

    032 Samples - 180 Added Notes of Polyphony:

    ----------------------------------------------------------------


    L-Factor II:R9 : Dual Xeon Core2 Duo 5150 @ 2.66 GHZ : 1333 FSB : Nuendo 3.2
    C/O TAFKAT:

    Motherboard : i5000

    Memory : 2 GB - FBDIMM DDR2 667 - 4 Channel Interleaved : Standard Timings by SPD :

    O.S : XP SP2 :

    Lynx 2 : Driver : V2 Build 13n

    256 Samples - 480+* Added Notes of Polyphony:

    128 Samples - 480+* Added Notes of Polyphony:

    064 Samples - 440 Added Notes of Polyphony:

    032 Samples - 290 Added Notes of Polyphony:

    * At both 256 and 128 , there was still plenty left in the tank, unfortunately I can't get anymore VSTi's into the rack
    as the maiximum VSTi's available are 64, so I had an extra 48 slots left, thats an extra 480 voices + / 10 per the
    extra 48 A1's..

    --------------------------------------------------------------------------------------

    L-Factor II:R9 : Core2 Duo E6600 @ 3.46 GHZ : 1540 FSB : Cubase SX3
    Phatriff:

    Motherboard : i975

    Memory : 2 GB – DDR2 @ 770 MHZ : Standard Timings by SPD :

    O.S : XP SP2 :

    FireFace 800 : Driver : 2.5.7.1

    256 Samples - 300 Added Notes of Polyphony:

    128 Samples - 230 Added Notes of Polyphony:

    064 Samples - 140 Added Notes of Polyphony:

    048 Samples - 70 Added Notes of Polyphony:

    ------------------------------------------------------------------------------


    Well , I'll let the numbers do the talking , too early to do any in depth analysis just yet, but there are clearly variables already becoming apparent with the respective audio cards....

    I really have no idea what to do for the Clovertown systems to be honest, these chips are just so far ahead of the applications at the moment its frightening.., I started hitting some major resource allocation issues within Nuendo even with Blofelds on the Woodcrest rig, where Nuendo did not have enough resources available to actually assign to the GUI's of the plugins, i.e: After a certain number of Magnetos, it would no longer bring up the GUI of the PlugIn.., very strange..

    Anyhow,

    The above report should give you a good idea of how the systems/audio cards are shaping up on this test..


    So Dig In..

    Peace:

    V:

  3. #3
    Join Date
    Aug 2005
    Location
    Melbourne : Australia :
    Posts
    4,249

    Default

    Hey All,

    O.K, here are the Kentsfield numbers for both Blofelds DSP40 and L-Factor II :

    ----------------------------------------------------------------------------------------------------------

    Blofelds DSP 40R5 : Core2Quad QX6700 @ 2.66 GHZ : 1066 FSB : Nuendo 3.2

    Motherboard : i975

    Memory : 1 GB - XMS 6400 @ PC6400 - 800 MHZ : Standard Timings by SPD :

    O.S : XP SP2 :

    Lynx 2 : Driver : V2 Build 13n

    256 Samples - 159 Magneto's / Save-ReOpen 159 :

    128 Samples - 131 Magneto's / Save-ReOpen 131 :

    064 Samples - 101 Magneto's / Save-ReOpen 101 :

    032 Samples - 62 Magneto's / Save-ReOpen 54 :

    Basically identical to the Dual Woodcrest at the same clockspeed except a slight dip at 032 samples, being down 3 Magnetos. Interestingly the Performance Droop on Save /Reopen - although slight on Woodcrest at 128 and below, didn't appear until 032 samples on the Kentsfield.

    Some issues I encountered that should be noted. I experienced the GUI resource allocaction issue more severely on the Kentsfield system, which suggests to me that it maybe memory related as this system has only 1GB of memory , whereas the Woodcrest had 2 GB. I actually had Debug Errors popping up and crashing when closing what actually came up as a GUI on some of the runs, very frustrating.., if I left the GUI-or lack of it to be precise - active, I could continue without issue, mind you it is hinting that we are exceeding the capabilities of the application in some areas.

    ---------------------------------------------------------------------------------------------------------------------------------------------------------

    L-Factor II:R9 : Core2Quad QX6700 @ 2.66 GHZ : 1066 FSB : Nuendo 3.2

    Motherboard : i975

    Memory : 1 GB - XMS 6400 @ PC6400 - 800 MHZ : Standard Timings by SPD :

    O.S : XP SP2 :

    Lynx 2 : Driver : V2 Build 13n

    128 Samples - 480+* Added Notes of Polyphony:

    064 Samples - 450 Added Notes of Polyphony:

    032 Samples - 310 Added Notes of Polyphony:

    Again, almost identical with the Woodcrest at the same clockspeed.., a touch better across the board actually..

    I am not seeing any FSB saturation issues across either test, so we can put that one to bed right here and now.. !!

    ---------------------------------------------------------------------------------------------------------------------------------------

    Now having an unlocked chip for the first time has allowed me to overclock this puppy without the usual scaling of the FSB /juggling of the memory timings.. that has been the norm in the past , I easily cranked this system to 3.2 GHZ on standard voltage, simply by raising the multiplier from 10 to 12.., temps rose about 2-3 degrees only.., I kid you not.. :-)

    Results were rather mixed tho, on L-Factor @32 Samples I managed another 60 voices of polyphony, while on Blofelds at 064 and 032, the results were basically identical to the standard 2.66 GHZ results.., which suggests that simply raising the clock, without raising the FSB is not giving me the desired scaling.., very interesting, and something that I will be looking into further when I have the time.

    O.K, Thats it from me for a while.., I'll stick my head back up when the faster clocked Clovertowns are in the Wild.

    In the meantime, please run and post results for the L-Factor II bench, it would be great to get as much of a cross reference with as many systems/audio cards as we can..

    I would like to make one small note .., the recent and ongoing attempts to cast dispersions towards the Blofelds DSP session being flawed/ subjective / unscientific, etc, has really drained me, I have answered all of the queries multiple times , and no-one has been able to provide any evidence to the contrary. Simply claiming that the variables presented were due to a "flaw" in the test session, without presenting the evidence to prove it, holds about as much water as a siv . While the Save/Reopen seems to be the most controversial issue, there is also the performance variables that were evident simply running the session, but of course no mention has been made of that. Also the simple fact that the Save/Reopen issue has been repeated across multiple DAW applications, using 3rd part plugins totally negates the argument that the variables were caused by a "flaw" in the session, something that is conveniently ignored by those that continue to posture that opinion.

    I am way past needing to defend the work that we have all put into this project over the last 18 months, I will continue to test and report on the newer architectures as they come online , the information given can be weighed up by the end users in respect to how it translates to their Real World working environments, I am done arguing with the select sector that seem to think that anything I do is somehow weighted one way or the other..

    As always I'll let the numbers do the talking..

    Peace:

    V:

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts