I realize this isn't a forum for hardware discussion, but I'd be interested to know if server class chips are a good choice for a DAW. In general it seems that clock speeds are slower in the server chips but the parallelism is greater.
Diva raises the hardware uograde eyebrow....
-
- KVRAF
- 1895 posts since 13 Oct, 2002
After some huffing and puffing and observing the Diva's rather extravagant demands on my systems' hardware, I've begun to consider upgrading my main computer. It's something I've been wanting to do anyway cause it's about 4 years old now. It looks like the kind of real-time rendering Diva likes is dependent on processor speed AND on the number of cores and the number of threads they can handle. Looking the Intel comparison site at desktop chips here, i7's seem to cap at 6 cores / 12 threads with clock speeds reaching 3.46 gHz. But their best server core here, does 10/20 (!) but only 2.4 gHz (for the moderate sum of $4616
).
I realize this isn't a forum for hardware discussion, but I'd be interested to know if server class chips are a good choice for a DAW. In general it seems that clock speeds are slower in the server chips but the parallelism is greater.
I realize this isn't a forum for hardware discussion, but I'd be interested to know if server class chips are a good choice for a DAW. In general it seems that clock speeds are slower in the server chips but the parallelism is greater.
-
- KVRAF
- Topic Starter
- 1895 posts since 13 Oct, 2002
Just found this interesting site...
-
- KVRAF
- 6323 posts since 30 Dec, 2004 from London uk
This thread may be of interest :
http://www.kvraudio.com/forum/viewtopic ... sc&start=0
I would read to the end before making any conclusions.
http://www.kvraudio.com/forum/viewtopic ... sc&start=0
I would read to the end before making any conclusions.
- KVRAF
- 2083 posts since 28 Feb, 2011
I know Urs has said he intends to make Diva 64-bit, but I'm not sure when multi-core support will happen.
For the record, I have a 2-year-old i7 (first-generation) that is rated at 3.2 GHz, but I have always run it at 3.6 GHz and it's very stable. I can increase the clock speed up to 4.0 GHz, but 3.6 GHz is the highest speed at which it is completely rock-solid.
So a second-generation i7 rated at 3.4 GHz could very well run very solidly at around 3.8 GHz or possibly higher.
For the record, I have a 2-year-old i7 (first-generation) that is rated at 3.2 GHz, but I have always run it at 3.6 GHz and it's very stable. I can increase the clock speed up to 4.0 GHz, but 3.6 GHz is the highest speed at which it is completely rock-solid.
So a second-generation i7 rated at 3.4 GHz could very well run very solidly at around 3.8 GHz or possibly higher.
-
- KVRist
- 260 posts since 5 Jun, 2011
I've been shopping around too. Been thinking about the E5s that will replace the current Xeon 5600s.
http://www.pcworld.com/article/239986/s ... _2012.html
Price range:
http://www.electronista.com/articles/11 ... explained/
Dual LGA 2011 Socket Motherboard:
It would be nice to have 16 cores.
http://wccftech.com/evga-teases-x79-bas ... ement-sr2/
Or the MAC Pro way:
http://techbend.com/2011/10/30/new-mac- ... estimates/
Club Ho
http://www.pcworld.com/article/239986/s ... _2012.html
Price range:
http://www.electronista.com/articles/11 ... explained/
Dual LGA 2011 Socket Motherboard:
It would be nice to have 16 cores.
http://wccftech.com/evga-teases-x79-bas ... ement-sr2/
Or the MAC Pro way:
http://techbend.com/2011/10/30/new-mac- ... estimates/
Club Ho
-
- KVRAF
- Topic Starter
- 1895 posts since 13 Oct, 2002
Thanks for the links, guys. That's just what I was looking for. When I'm not in active "buy mode" I like to stay away from the hardware sites to keep my head clear of all that data, but then I lose track of new developments and announcements. And then finding the best price/performance ratio for one's budget in the glut of gear available always seems to get complicated...
It's funny: I used to be more of a sample lib user but in recent years my work has taken me more and more into real-time sound generation and manipulation. It's interesting the way the work one does can completely redefine the hardware requirements: for me, the single monster box is starting to make more sense than multiple machines. These days I need to keep my softsynths and processing as on-line as possible as the mixing process has become inexorably intertwined with sound design.
It's funny: I used to be more of a sample lib user but in recent years my work has taken me more and more into real-time sound generation and manipulation. It's interesting the way the work one does can completely redefine the hardware requirements: for me, the single monster box is starting to make more sense than multiple machines. These days I need to keep my softsynths and processing as on-line as possible as the mixing process has become inexorably intertwined with sound design.
-
- KVRAF
- 1890 posts since 23 Dec, 2003
i can play between 5-8 voices on my 3,3 ghz i7 before it maxes out a single core. that leaves me with 11 more to play with.
Finally a plugin that makes my overall cpu meter jump above the 50% mark!
Finally a plugin that makes my overall cpu meter jump above the 50% mark!
-
- KVRAF
- 5511 posts since 6 May, 2002
Xils Synthix will do it too.Dr.Wu wrote:i can play between 5-8 voices on my 3,3 ghz i7 before it maxes out a single core. that leaves me with 11 more to play with.
Finally a plugin that makes my overall cpu meter jump above the 50% mark!
Intel Core2 Quad CPU + 4 GIG RAM
-
- KVRian
- 504 posts since 8 Aug, 2009
It's really the classic question - scale out, or scale up? More cores, or faster cores? You could get a dual CPU workstation like a Mac Pro (due for a refresh soon), but the discussions I've had with programmers regarding audio soft synths are that multiple cores aren't really an advantage, because of latency and the way audio is processed in real time. I think Urs may have even said something, or perhaps it was one of the guys at FXpansion.
I imagine for a string synth where the keyboard is split into two different sounds multiple cores could be used. Perhaps Urs has found a way around the latency, or the improved speed of QPI and other methods for the cores to speak to one another aren't as big an issue now.
And of course DAWs like Logic will put plugins into different cores if you essentially set them up as different tracks: Track 1 (soft synth) feeds into track 2 (EQ) feeds into track 3 (compressor) into track 4 (etc).
I don't know what will be optimal for Diva, but I imagine Urs is the guy to ask. But based on all I know, clock speed is far superior to multiple cores if you're looking to get the most bang for your buck with soft synths and analog modeling.
I imagine for a string synth where the keyboard is split into two different sounds multiple cores could be used. Perhaps Urs has found a way around the latency, or the improved speed of QPI and other methods for the cores to speak to one another aren't as big an issue now.
And of course DAWs like Logic will put plugins into different cores if you essentially set them up as different tracks: Track 1 (soft synth) feeds into track 2 (EQ) feeds into track 3 (compressor) into track 4 (etc).
I don't know what will be optimal for Diva, but I imagine Urs is the guy to ask. But based on all I know, clock speed is far superior to multiple cores if you're looking to get the most bang for your buck with soft synths and analog modeling.
THIS SIGNATURE DOES NOT VIOLATE THE KVR FORUM RULES.
THIS SIGNATURE DOES NOT VIOLATE THE KVR FORUM RULES.
THIS SIGNATURE DOES NOT VIOLATE THE KVR FORUM RULES.
THIS SIGNATURE DOES NOT VIOLATE THE KVR FORUM RULES.
THIS SIGNATURE DOES NOT VIOLATE THE KVR FORUM RULES.
THIS SIGNATURE DOES NOT VIOLATE THE KVR FORUM RULES.
THIS SIGNATURE DOES NOT VIOLATE THE KVR FORUM RULES.
- KVRAF
- 1617 posts since 11 Dec, 2008 from Minneapolis
The real bottleneck is how often and how much cores need to sync up. Diva will spread basically a new voice onto new cores, which should be good - patch state, midi data, buffer/timing location and logic shouldn't require much communication and there shouldn't be any cases where one core relies on results of calculations from another core halfway through rendering a voice.lagavulin16 wrote:It's really the classic question - scale out, or scale up? More cores, or faster cores? You could get a dual CPU workstation like a Mac Pro (due for a refresh soon), but the discussions I've had with programmers regarding audio soft synths are that multiple cores aren't really an advantage, because of latency
-
- KVRist
- 420 posts since 29 May, 2005 from Central Ohio, USA
If you're going to build a PC, might I suggest waiting for Ivy-Bridge, due spring 2012.
Fig Newton: The force required to accelerate a fig 39.37 inches per sec.
-
- KVRAF
- 1548 posts since 12 Jan, 2010 from Copenhagen
I too am looking to upgrade my hardware...
I have been thinking of making a mac mini modular..
For size and portability..
This will be hooked up together using an etherenet hub, Vienna Pro would be the enabeling software.
I will get one new mini quad core i7 to start, combine that with my dual core imac mid 2007, and then every six months or so add a new computer.
But on the other hand the comment by Lagavulin16;
"I've had with programmers regarding audio soft synths are that multiple cores aren't really an advantage, because of latency and the way audio is processed in real time. I think Urs may have even said something, or perhaps it was one of the guys at FXpansion. "
Perhaps Dual core minis would be a better more economical choice..
Back to the main thrust; A Mac mini modular has the advantage that I don't have to worry about 64bit multithreading I have taken the discreet path..
Running 12 instances of multi voice Diva in Devine mode spread out over three maschines would be easily managed...
With the Mac mini's I get an extremely compact system, that would have the added benefit of a certain redundancy, It would be far better to lose a portion of the resources on a system failure than the whole rig going down at once, in a performance situation.
I can repurpose the minis for demo, teaching, Vj graphics, dedicated workstation "maschine" server or whatever,
Also the mac minis have the least sleep mode power consumption of any desktop, and when not doing big projects or just surfing I'll do just that put the unused machines to sleep..
The drawbacks of course are obvious, price is rather high if I went out and bought 4 mac minis, then there's the complicated setup where there's additional software needed. But I just don't see this multithreading thing coming to fruition, promises promises, but still no delivery. I'd be rather disappointed if I went with a big expensive multicore system and it's resources couldn't be used efficiently..
I also wonder whatever happend to the concept of offloading processes to the GPU? I won't need graphics from all 3 or 4 mini's it would be quite a boost if these resources could be used for audio as well...
Wade
I have been thinking of making a mac mini modular..
For size and portability..
This will be hooked up together using an etherenet hub, Vienna Pro would be the enabeling software.
I will get one new mini quad core i7 to start, combine that with my dual core imac mid 2007, and then every six months or so add a new computer.
But on the other hand the comment by Lagavulin16;
"I've had with programmers regarding audio soft synths are that multiple cores aren't really an advantage, because of latency and the way audio is processed in real time. I think Urs may have even said something, or perhaps it was one of the guys at FXpansion. "
Perhaps Dual core minis would be a better more economical choice..
Back to the main thrust; A Mac mini modular has the advantage that I don't have to worry about 64bit multithreading I have taken the discreet path..
Running 12 instances of multi voice Diva in Devine mode spread out over three maschines would be easily managed...
With the Mac mini's I get an extremely compact system, that would have the added benefit of a certain redundancy, It would be far better to lose a portion of the resources on a system failure than the whole rig going down at once, in a performance situation.
I can repurpose the minis for demo, teaching, Vj graphics, dedicated workstation "maschine" server or whatever,
Also the mac minis have the least sleep mode power consumption of any desktop, and when not doing big projects or just surfing I'll do just that put the unused machines to sleep..
The drawbacks of course are obvious, price is rather high if I went out and bought 4 mac minis, then there's the complicated setup where there's additional software needed. But I just don't see this multithreading thing coming to fruition, promises promises, but still no delivery. I'd be rather disappointed if I went with a big expensive multicore system and it's resources couldn't be used efficiently..
I also wonder whatever happend to the concept of offloading processes to the GPU? I won't need graphics from all 3 or 4 mini's it would be quite a boost if these resources could be used for audio as well...
Wade
waves break, but somehow it all makes sense.
