jcschild
Max Output Level: -41 dBFS
- Total Posts : 3409
- Joined: 2003/11/08 00:20:10
- Location: Kentucky y'all
- Status: offline
Cores and VSTi the facts UPDATED/DDR3/nehalem
Hey Guys, over time i have had numerous people debate me on this. my comments have always been that for samples (regardless of how the host does load balancing) are far more effected by Front side bus and GHZ than more cores. in other words a 3.0Ghz dual core 1333FSB will kill the 2.4GHz quad core @ 1066FSB so i thought we would post some benchmarks. using the same exact system with Komplete 5 and Sonar. we loaded up a huge amount of samples well over 2gig. 4gig of ram RME Fireface 400. XP Pro SP3 AMD Phenom 9950 @.6GHz ddr2 800 256 buffer clean dual core 4.0GHz 1900 FSB DDR2 1066 128 buffer clean dual core 3.0Ghz 1333 FSB DDR2 800 512 buffer clean Quad Core 2.4GHz 1066 FSB DDR2 800 768 clean Quad Core 3.0GHz 1600 FSB DDR2 800 512 buffer clean, almost did the next lower buffer (384) clean but not there. the 1600 FSB helped. Quad core 3.2GHz 1600 FSB DDR3 1600 CL7 128 buffer clean i7 3.2GHz 6G DDR3 1600 Vista 64 48 buffer clean ***** a new winner this is a huge performance gain. compared to the 3.0GHz dual or quad... remember if this was a dual core 3.2 it would be the same.. so this clearly shows GHZ and FSB are king for samples. and quad does not do squat for it. mind you this is strictly for sampling, a composer would benefit. where quad/octo core most certainly help with effects, large projects and 96K. having wrote off AMD, after we completed these tests knowing sampling is all about memory bandwidth we got to thinking.... Hmmm Intel 7000, AMD nearly 10k, Opteron 14k.... so we decided to test an AMD Phenom the only 2 things that beat the Phenon was the overclocked to 4GHz and the DDR3 system i also need to add this is on an older Asus M2 board we had laying around next we try a newer chipset. we also ran some benchmarks using our older Sonar test to compare to the older Intel quads. it beat the Q6700 but a hair. we are testing with newer procs in a few days. (we have not updated the benchmarks on our site for awhile)
post edited by jcschild - 2008/11/04 09:30:56
Scott ADK Home of the Kentucky Fried DAW!
|
John
Forum Host
- Total Posts : 30467
- Joined: 2003/11/06 11:53:17
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 19:24:43
(permalink)
Did you try more then one instance of the VSTi?
|
Lanceindastudio
Max Output Level: -29 dBFS
- Total Posts : 4604
- Joined: 2004/01/22 02:28:30
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 19:44:20
(permalink)
So Im upgrading my rig soon. My board is listed below. I would like to get the best chip for the best price to throw in this same board as to not have to install a new MB and all. Its FSB goes up to 1333, so that is limiting with the new chips FSB capabilities, correct? What chip should I get Scott? Im thinking I should go quad core because I use a lot of effects and softsynths at the same time, and my projects get pretty big and complex, so this seems like the quad would be better. Again, I lay a lot of effects down as I mix and master the projects all in Sonar 7.0.2 Should I get the Q9300 and Overclock it? If I get a good air cooler for it, what kind of clock speed is a no brainer for that chip? Or, should I go with the Q9450? Is it really worth the extra 70 bux or so? I dont want to spend extra money if it isnt going to make much of a difference. Thanx, Lance
Asus P8Z77-V LE PLUS Motherboard i7 3770k CPU 32 gigs RAM Presonus AudioBox iTwo Windows 10 64 bit, SONAR PLATINUM 64 bit Lots of plugins and softsynths and one shot samples, loops Gauge ECM-87, MCA SP-1, Alesis AM51 Presonus Eureka Mackie HR824's and matching subwoofer
|
John
Forum Host
- Total Posts : 30467
- Joined: 2003/11/06 11:53:17
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 19:56:33
(permalink)
should I go with the Q9450? Is it really worth the extra 70 bux or so? Yes!
|
plectrumpusher
Max Output Level: -81 dBFS
- Total Posts : 479
- Joined: 2007/10/22 04:29:27
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 20:11:37
(permalink)
Were you disk streaming the samples or loading the whole set up in ram ???
If you haven't got a smile on your face and laughter in your heart.......Then you are just an old sour fart!!
|
SilkTone
Max Output Level: -59.5 dBFS
- Total Posts : 1566
- Joined: 2003/11/10 17:41:28
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 20:15:05
(permalink)
ORIGINAL: John Did you try more then one instance of the VSTi? Yea, seriously. Without loading up multiple VST/VSti this test is worthless. Multiple cores shine when you have complex songs. How many VSTs, VSTis and tracks did you have total in the project? EDIT: I guess your point about large amounts of samples (implying low FX plugin count = low number crunching) might be valid. But how many people use Sonar with heavy sample loading but light processor loading? SilkTone
post edited by SilkTone - 2008/07/23 20:41:44
Windows 10 Pro x64, SONAR Platinum 64-bitFocusrite Scarlett 18i8 USB, ASRock Z97 Pro4, Haswell 4790k @ 4.4GHz32GB DDR3/1600, 500GB SSD (OS) + 256 GB SSD + 3TB MDNVIDIA GTX-1070, 40" 4K Monitor + 1 Monitor in ISO booth
|
Creator
Max Output Level: -87 dBFS
- Total Posts : 170
- Joined: 2004/01/14 20:54:23
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 20:28:33
(permalink)
Ah, found it. Sorry Scott, I didn't realize this was going to be a new thread. To anyone that was complaining about the applicability of this test, it actually addressed a very specific issue that I and other had wanted more information about. Thanks Scott!
|
billruys
Max Output Level: -73 dBFS
- Total Posts : 852
- Joined: 2003/11/05 00:18:31
- Location: Outer Space
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 20:52:56
(permalink)
ORIGINAL: SilkTone ORIGINAL: John Did you try more then one instance of the VSTi? Yea, seriously. Without loading up multiple VST/VSti this test is worthless. Multiple cores shine when you have complex songs. How many VSTs, VSTis and tracks did you have total in the project? Agreed, this post is very light on facts. With only one VSTi instance, only one core will really get loaded in Sonar, as one VSTi will only run under a single thread. There would need to be at least 4 VSTi instances before this was a fair contest, but we don't know any details.
|
RE: Cores and VSTi the facts
2008/07/23 21:11:51
(permalink)
Hi Scott, Great info once again. Doean't it just make the most sense then to buy a quad-core with the faster fsb and OC it like mad like the QXxxxx series of cpus? Or at least the Q9xxx?
|
Lanceindastudio
Max Output Level: -29 dBFS
- Total Posts : 4604
- Joined: 2004/01/22 02:28:30
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 21:20:10
(permalink)
Scott made a valid point, but it just isnt useful if you are willing to get a GOOD Quad
Asus P8Z77-V LE PLUS Motherboard i7 3770k CPU 32 gigs RAM Presonus AudioBox iTwo Windows 10 64 bit, SONAR PLATINUM 64 bit Lots of plugins and softsynths and one shot samples, loops Gauge ECM-87, MCA SP-1, Alesis AM51 Presonus Eureka Mackie HR824's and matching subwoofer
|
RE: Cores and VSTi the facts
2008/07/23 21:21:42
(permalink)
Exactly, that is what I ws aluding to. Get a good Quad and there is no argument, although Scott's point directly to that.
|
Lanceindastudio
Max Output Level: -29 dBFS
- Total Posts : 4604
- Joined: 2004/01/22 02:28:30
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 21:32:57
(permalink)
yeah people argued with him a lot about a fact and thats only why he posted this, So n e ways lol... q9450 is the best bang fo buck right?
Asus P8Z77-V LE PLUS Motherboard i7 3770k CPU 32 gigs RAM Presonus AudioBox iTwo Windows 10 64 bit, SONAR PLATINUM 64 bit Lots of plugins and softsynths and one shot samples, loops Gauge ECM-87, MCA SP-1, Alesis AM51 Presonus Eureka Mackie HR824's and matching subwoofer
|
RE: Cores and VSTi the facts
2008/07/23 21:43:41
(permalink)
As far as I can tell Q9450 is the shot right now. From what I have read you can OC it without upping the voltage to 3.2 or 3.4 ghz ... also can up the fsb to 1600 ... You could always get the QXxxxx for a grand and OC it to 4ghz plus, but that just costs too dang much for me.
|
Lanceindastudio
Max Output Level: -29 dBFS
- Total Posts : 4604
- Joined: 2004/01/22 02:28:30
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 21:47:53
(permalink)
I have to change my board to get 1600 FSB though right?
Asus P8Z77-V LE PLUS Motherboard i7 3770k CPU 32 gigs RAM Presonus AudioBox iTwo Windows 10 64 bit, SONAR PLATINUM 64 bit Lots of plugins and softsynths and one shot samples, loops Gauge ECM-87, MCA SP-1, Alesis AM51 Presonus Eureka Mackie HR824's and matching subwoofer
|
RE: Cores and VSTi the facts
2008/07/23 21:52:14
(permalink)
Hmm, I think Scott could answer that better. Some MoBos support 1600 fsb via OC, but I dont know if yours does ... I have seen other Gigabyte boards that do. I'd either check with Scott or check the Gigabyte forums to see how far you can push your board. It probably depends on your cpu also as the cpu must be able to handle a 1600 fsb ... so maybe with your board plus a Q9450 you could do it, but thats just a guess.
|
Marah Mag
Max Output Level: -71 dBFS
- Total Posts : 1000
- Joined: 2008/07/12 18:27:12
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 21:54:07
(permalink)
Does anyone know where in the life cycle of these things we're in? The Q6600 is $200. The Q9450 is $329. That's USD at Newegg. Any informed speculation on when the Q9450 might approach the mid 200's?
|
John
Forum Host
- Total Posts : 30467
- Joined: 2003/11/06 11:53:17
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 22:06:18
(permalink)
At MicroCenter its $279. Here
|
Marah Mag
Max Output Level: -71 dBFS
- Total Posts : 1000
- Joined: 2008/07/12 18:27:12
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 22:17:18
(permalink)
ORIGINAL: John At MicroCenter its $279. Here Ooo nice. That's getting close, isn't it? Thanks for posting that! Funny when I went to the usual price comparison sites, I didn't find it anywhere for ,uch less than at Newegg. Do you think MicroCenter is just a fluke, or the front end of a trend?
|
John
Forum Host
- Total Posts : 30467
- Joined: 2003/11/06 11:53:17
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 22:26:44
(permalink)
I bought my q6600 from them and they were the cheapest. At the time $199. Newegg was $219. MC is I think a bigger outlet for Intel and can get them at a cheaper price. But that is speculation. All CPUs have always been cheaper at MC. I do expect them to come down more in the next two months, or at least by the late fall. The hyperthreading quads should be coming out early next year.
|
space_cowboy
Max Output Level: 0 dBFS
- Total Posts : 9813
- Joined: 2007/07/20 14:49:31
- Location: Front and center behind these monitors
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 22:29:05
(permalink)
I am waiting for the 16 cores. I was an early adopter of the Q6600. I think based on a purely unscientific analysis that my Q6600 runs about 10x what my pentium d did. That is based on freezing a track of say 3 minutes with effects applied and watching how long it took before and after to do the freeze. By the way, the DAW Gurus at Sonica tell me there is an issue with the 8 cores and the current version of Sonar. I cant seem to figure where it is on their website tonight (long day) but two days ago Guy was pointing me towards an article showing that with the 8 cores, Sonar is having issues in maxing out one core, leaving one apparently idle and using the other 6 about 20-30%. CAKE - is this something that will be addressed in the next versions? Wait a minute, wait a minute, I think the article is coming through now. -->XXXXXXX<--
Some people call me Maurice SPLAT Pro lifetime, ADK 6 core 3.6Ghz with 32 GB RAM, SSD 1TB system drive, 3 3TB regular drives for samples, recordings and misc. Behringer X Touch, UAD Apollo Quad. 2 UAD2 Quads PCI (i think - inside the box whatever that is), Console 1. More guitars (40??) and synths (hard and soft) than talent. Zendrum!!!
|
RE: Cores and VSTi the facts
2008/07/23 22:29:20
(permalink)
ORIGINAL: Marah Mag Does anyone know where in the life cycle of these things we're in? The Q6600 is $200. The Q9450 is $329. That's USD at Newegg. Any informed speculation on when the Q9450 might approach the mid 200's? We're at the end of the Socket 775 life cycle. Intel will be releasing their Nehalem series of chips in Q4 of this year which are designed around an entirely new architecture. The chips will feature an on-board memory controller which is linked directly to the RAM, eliminating the need for a Front Side Bus. AMD introduced this technology several years ago with their Opteron processors, and Intel will apply it to its new offerings. Because of the radical design change the new Nehalem cpus will not be compatible with today's motherboards and will most likely require DDR-3. This means that the Q9450 is presumably the last of its kind as far as Socket 775 chips are concerned. I have heard rumors on other forums that the 9450 will drop below 200 USD this fall as the Nehalems are released, but who knows, maybe a new Nehalem system will be worth saving up for.
|
John
Forum Host
- Total Posts : 30467
- Joined: 2003/11/06 11:53:17
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 22:34:45
(permalink)
By the way, the DAW Gurus at Sonica tell me there is an issue with the 8 cores and the current version of Sonar. I cant seem to figure where it is on their website tonight (long day) but two days ago Guy was pointing me towards an article showing that with the 8 cores, Sonar is having issues in maxing out one core, leaving one apparently idle and using the other 6 about 20-30%. CAKE - is this something that will be addressed in the next versions? We know about this and its well documented on the forum. Its not just 8 cores it effects the 4 core chips as well. I wouldn't worry about it. CW knows too.
|
John
Forum Host
- Total Posts : 30467
- Joined: 2003/11/06 11:53:17
- Status: offline
RE: Cores and VSTi the facts
2008/07/23 22:40:07
(permalink)
Seth Hi! Good to see you here and welcome to the forum. BTW thanks for the information.
|
RE: Cores and VSTi the facts
2008/07/23 22:46:47
(permalink)
John, Thanks! It's great to be here, although I've been here before for a long time, just not as a part of Cakewalk. You know, about the quad-core updating and "when should I upgrade" question ... Scott has said the best advice many times ... its all about what you need to do and when you need to do it. If you HAVE TO get work done now and your current machine is not cutting it (i.e. dropouts, having to bounce to disk or freeze a lot), then it is time to upgrade. I think that kind of sums it up in a nutshell.
|
jcschild
Max Output Level: -41 dBFS
- Total Posts : 3409
- Joined: 2003/11/08 00:20:10
- Location: Kentucky y'all
- Status: offline
RE: Cores and VSTi the facts
2008/07/24 09:30:11
(permalink)
HI Guys, Hard2hear will be posting the exact project there were numerous instances.... Seth, thanks for the nice comments.. and congrats on your new job! yes it makes the most sense to get a quad and OC it... however a Q6600 overclocked to 3.0G 1333 is not going to yeild any more samples than a dual core E 8400 it will yeild more overhead horsepower (effects, large track counts) the Q6600 OCed is the "budget" OC and is a nice way to go for those on a tight budget. a non OCed Q6600 is fine for managing a large well balanced project with not to much VSTi. Ideally a Q9450 OCed is better for those with a good amount of samples and effects. i posted this so people can get a better ideal about CPU selection before laying down hard cash. those who run mostly VSTi would be best off with the highest dual OC. (4GHz) or the fastest GHz they can afford without OCing. most dual cores higher GHz can be had for the price of lower GHz quads.... or if you have the dosh... higher quad (we have them running @ 4gig as well, but these extreme OC's (out of spec) require a good knowledge of OCing and really good ram a bad investment would be buying a 1600 FSB and trying to OC those... you will get better results from a 1333. while i highly advocate overclocking within spec, i dont recommend the extreme OCing for most. with in spec means its OCed to what Intel actually sells, and all your timings are syncronous EG: intel sells the QX9770 3.2G 1600 FSB so a Q9450 runs @ 2.66 1333 (333 real fsb) OCing it to 1600 FSB makes it a QX9770... all with in Intel spec.
Scott ADK Home of the Kentucky Fried DAW!
|
Hard2Hear
Max Output Level: -86 dBFS
- Total Posts : 210
- Joined: 2005/01/21 16:09:12
- Location: KY
- Status: offline
RE: Cores and VSTi the facts
2008/07/24 09:42:00
(permalink)
6 Instances of Kontakt3 in rack All using K3 included Vienna instruments Default dfd settings All running live, no freeze, no bounce MIDI tracks running to every patch 1 Violin ens14 (all) Viola ens10 (all) Cello ens8 (all) Bass ens6 (all) 2 Piccolo (all) Flute (all) Oboe (all) FrenchOboe (all) EnglHorn (all) Clarinet (all) 3 TrumpEns3 (all) FrnHornEns4 (all) Trombens3 (all) Tuba (all) 4 Violin ens14 (all) Viola ens10 (all) Cello ens8 (all) Bass ens6 (all) 5 Tenor Recorder (all) Piccolo Recorder (all) Low Whistle (all) 6 Timpani Hits
post edited by Hard2Hear - 2008/07/24 09:45:33
|
jcschild
Max Output Level: -41 dBFS
- Total Posts : 3409
- Joined: 2003/11/08 00:20:10
- Location: Kentucky y'all
- Status: offline
RE: Cores and VSTi the facts
2008/07/24 10:20:44
(permalink)
for the pessimists... and the perfectionists... (of which i am as well) notes* cpu was never even close to redline dropouts/pops clicks occured at 30-40% on all tests. this further proves the quad wont help. obviously its not the hard drive (low usage read out) as if it was bound by HDD we would not have seen the increase with the 4.0GHz its not that we are maxing ram total usage as again the increase with the 4gig would not have occured. what it IS, is memory bandwidth. memory bandwidth is effected by CPU speed and Front side bus. period. (and CL rating to a point)
Scott ADK Home of the Kentucky Fried DAW!
|
jcschild
Max Output Level: -41 dBFS
- Total Posts : 3409
- Joined: 2003/11/08 00:20:10
- Location: Kentucky y'all
- Status: offline
RE: Cores and VSTi the facts
2008/07/24 12:46:38
(permalink)
for those who missed the updated part in first post Quad core 3.2GHz 1600 FSB DDR3 1600 CL7 128 buffer clean this is a huge performance gain. compared to the 3.0GHz dual or quad... remember if this was a dual core 3.2 it would be the same..
Scott ADK Home of the Kentucky Fried DAW!
|
inmazevo
Max Output Level: -42.5 dBFS
- Total Posts : 3276
- Joined: 2006/01/03 18:30:38
- Location: Pacific Northwest
- Status: offline
RE: Cores and VSTi the facts
2008/08/26 11:33:28
(permalink)
Looking in to getting a new system pretty soon, so I'm bringing this up again in case anyone's got any numbers for me. Given the focus on memory bandwidth, I have a question: So, what's the situation for those of us looking into getting a Xeon system... 1600MHz FSB with screaming procs, BUT: 800MZh DDR2 FB-DIMM memory. I'd be curious as to how well something like that stacks up against the test machines (say, a Mac Pro or equiv.). I'd also like to get some ideas, academically at this point, unless someone's got a test rig already (doubtful), for whether or not those in-the-know think that the upcoming Intel architecture switch to more direct memory control would be worth waiting for (Nehalem: integrated memory controller, a-la AMD). The results of this test suggest that it SHOULD be a score for sample folks, once the kinks are worked out. Any thoughts? - zevo
|
Muziekschuur at home
Max Output Level: -62 dBFS
- Total Posts : 1442
- Joined: 2006/03/01 03:30:22
- Status: offline
RE: Cores and VSTi the facts
2008/08/26 17:00:33
(permalink)
In another thread I just posted in this was said: droddey: quote: No 32 bit version of any OS will support over 4GB of RAM. In fact, do to page memory allocation, you will only see about 2.8-3.2GB on a Win XP 32 bit or Vista 32bit machine. Actually, that's not technically true. Any *single process* will only see 4GB of RAM, because the virtual address space will be 32 bits due to register size. But on an Intel CPU anyway, a 32 bit OS can make use of much more RAM than that, and spread it out among the individual processes. It's basically the same segmented architcture that has been there since the 186/286 years, but the segments are now 32 bit. The full theoretical memory capacity of a 32 bit Intel CPU is quite huge for that reason, and it doesn't have any of the complexities (for the programs I mean) that the original (16 bit) memory segments caused since each process will just see a single, flat 32 bit address space. It's only the OS that has to deal with that extra layer of virtual address mapping. It's just that Windows, I assume for reasons of compabitility with other CPUs that don't provide that capability, doesn't choose to make use of this feature. It's kind of a bummer since in the end Windows' multi-platform capabilties came to naught and it really only is ever deployed on CPUs that do provide this segmented architecture, AFAIK. Of course it would also make the OS memory management more comlex, but the CPU provides a lot of support for that stuff. So anyway, it wouldn't give SONAR any more *virtual* memory space, because of the issue of mapping things into each process' high GB of virtual address space. But it could mean that SONAR could have a full 4GB of physical memory to itself because the overall memory capacity could be 64MB or whatever and each process that required it could have that full 4GB, with none taken by other processes. That wouldn't make a huge difference for some folks, but for others it could very much help if they are running other large apps at the same time. So this could mean I can rewire Tascam Gigastudio into Cakewalk Sonar 7. So a system with 10 GB memory and 8 cores on WINXPPRO should show 3.2 free memory. But when the two apps are started and setup both should get their own space of 4GB and thus the system benefits of cores and memory on a dual quad. Offcourse the system should have three disks. One OS one SAMPLES and one audio. JCCHILD have you tested this and is there added benefit? Muziekschuur
Cakewalk Sonar Platinum Windows 7 32bit & 64bit (dualboot) Gigabyte mobo Intel dual quad 9650 & 4GB Ram RME DIGI9636 & Tascam DM24. M-audio Rbus & SI-24 Alesis Pro active 5.1 & Radford 90 transmissionline monitors. Roland RD-150 piano Edirol UM-880 & alesis fireport. Remote recording Alesis HD-24 & Phonic MRS 1-20. P.A. D&R Dayner 29-8-2 & behringer MX8000 (& racks) Rackpc Sonar Platinum with win10 AMD X6 1055T, 16GB Ram Dell inspiron 17R 6gb ram W10 two SSD's Sonar Plat.
|