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
DualCore = Twice as Fast? DAW Software
Old 19th August 2007
  #1
Lives for gear
 
TheRealRoach's Avatar
DualCore = Twice as Fast?

Hey everyone,

Does DualCore = Twice as Fast? (i.e. a Dual 2.0gHz is like having a 4gHz single-core)

I consider myself to be well versed in computer tech stuff but the above argument makes no sense to me even though I hear people constantly spewing this "fact".

My understanding of dualcore (and multi-processors in general) is that instead of four cars traveling at 100km/h on a one lane road, it's like four cars traveling at 100km/h on a 2 lane road. Thus still processing information at the same speed, but on parallel threads? But then does this actually mean that the system is processing twice as much data in a given time because there are more flood-gates open? And so my head explodes...

Roach
Old 19th August 2007
  #2
Lives for gear
 

Everything you said was right.

Another metaphor would be delivery trucks. If you have two packages to deliver in opposite directions, two drivers gets you there faster. With one package, it doesn't matter.

But how often do you only need to deliver one package?

You have plugins, midi messages, file retrieval, summing, etc. Either the software has to stop one, start another, and so on, or they can run in parallel and come back for their next package of instructions. So, dual cores come in at nearly twice as fast. Dual procs usually would get you a max of 60% increase. So, two dual cores will be slower than one quadcore. A quadcore would be capable of getting nearly 4 times the work of a single core done in the same time.

This all assumes the shipping center's dispatcher is not ******** and can tell 4 drivers where to go. Otherwise, three drivers are sitting around eating lunch while the other one does all the work himself.
Old 19th August 2007
  #3
To the original question: No.

You will get enhanced performance on many but not all tasks. But you will never get full double performance.


Here's more than you may want to know about multi-core architecture computing: Multi-core (computing) - Wikipedia, the free encyclopedia

[Please note that the article doesn't currently meet Wikipedia source citation standards.]
Old 19th August 2007
  #4
Lives for gear
 

Quote:
Originally Posted by theblue1 View Post
To the original question: No.

You will get enhanced performance on many but not all tasks. But you will never get full double performance.
Not true, in fact in some cases you can get over twice the performance. CPU1 isn't a whole CPU. It has tons of O/S stuff fighting with whatever a given program is trying to achieve. If you're just adding numbers from memory, CPU2 can get more work done than CPU1 can, even if CPU1 was playing by itself.

It just depends on the software and what it's inherent bottlenecks are. Most well written software easily gets a 90% scale, damn near linear.
Old 20th August 2007
  #5
In a shootout between the Core Duo and its single core version, the highest performance benchmark differential Anandtech found was under 60% (on a benchmark that used multithreading heavily) -- and the overwhelming majority of apps delivered much smaller performance benefits. And the single core was faster on some benchmarks. But Anandtech was pleased and not a little surprised by the performance gains, even as they were.

Now, is it possible that by picking and choosing very specific processes you might set up a non-real-world comparison that would deliver better numbers than a balanced benchmark? Of course. But that's not a measure of real world performance or the ability to get real work done.

Welcome to AnandTech.com [ Article: Intel's Core Duo Launch - Notebook Performance Revealed]

I stand by what I said.

You're not going to get full double performance over the single core version of a given dual core chip.


But if you've got contrary information from an equally reliablesource -- I'll certainly take a look at it.
Old 20th August 2007
  #6
Lives for gear
 
gsilbers's Avatar
 

u also have to realize that not all apps are written for dual cores. like gigastudio g3. it will only use one core. but the goood thing is that the rest of the crap that runs oin the background will be on the other core.

now u can get motherboards that support dual and quad cores so u can always upgrade your cpu.

but no, dual core doent mean twice as fast.. it will mean u have to look up in the web the exact model of intel or amd and then look for benchmarks to see the performance against other single and dual cpus.
Old 20th August 2007
  #7
Gear Maniac
 
hurd300403's Avatar
 

listen to theblue1. he is correct.
Old 20th August 2007
  #8
Lives for gear
 
allencollins's Avatar
 

dual core is obsolete quad is almost twice as fast
as dual providing the software is multi-threaded
Old 20th August 2007
  #9
Lives for gear
 

What is the reason for the limit in speed on the cpu´s? Seems like 3ghz is what it´s been at for the last five years almost. Are there technical problems with heat etc?
Old 20th August 2007
  #10
I came within two percent or so of having to eat my words heh -- on one very specific benchmark handpicked by Intel to show off performance of the Kentfield QX6700quad Core 2 Extreme vs an otherwise very similar Dual Core.

Anandtech kvetched about the conditions of the benchmarking (set up and controlled by Intel) and the by and large highly selective nature of tests -- but they were impressed by the gains in most areas.

Quote:
Using the POV-Ray multi-threaded beta we ran the application's built in benchmark and noticed a tremendous increase in performance. The QX6700 is over 80% faster than the X6800 thanks to a doubling in the number of cores. Estimating performance scaling at equal clock speeds again, we should get nearly twice the performance with four cores as with two cores, a 98% performance increase.
[UPDATE: forgot the link the first time around: AnandTech: Quad Cores: Intel Kentsfield Preview ]


HOWEVER...

That was on one specific test. Nothing else was nearly that high. Still, on other heavily multi-threaded apps, the performance gains were in the 60% - 80% range. And that's really impressive, I think.

On some business app type benchmarks, which are representative of most applications that don't make much use of multithreading, the quad showed a much smaller advantage, 8% and 26% in two different measures.

It's worth checking out, even if it was a set of benches selected to show off the chip by Intel. Anandtech follow through explaining things pretty well. Whenever I need to get the scoop from people who've torn apart things to see why they're the way they are -- I go to Anandtech.


Anyhow, they look like great processors and I'd love to have a machine based on one.

For apps that make use of multi-threading (workstation type apps like video editors, DAWs and other stuff near and dear) they make a lot of sense.


I thought I was pretty safe saying, "you will never get full double performance." I'm pretty shocked how very close I came to being wrong.

You can bet I'll be hedging my bets a little harder next time.

___________

Fredrik

I saw an article while I was fishing around for the one above that addressed that very issue -- unfortunately, I just skimmed it to see if it had anything that applied to this issue and couldn't find it just now when I looked. IIRC, the gist was much as you suggest, the advances in chip technology that had fueled the big leaps in computing power are fairly well exploited at this point. So in order to deliver the kinds of performance boost needed to keep the market excited, they've had to find new ways to exploit existing or more slowly evolving technology. Done wrong and multi core processors can consume way too much power while delivering marginal or even inverted benefits (ie, runs hotter, gets less work done). So the news from Intel with the their Pentium M-descended family of super fast, efficient chips is pretty darn good.
Old 20th August 2007
  #11
Lives for gear
 
Farshad's Avatar
 

Quote:
Originally Posted by TheRealRoach View Post
Does DualCore = Twice as Fast?
NO!
Old 20th August 2007
  #12
Lives for gear
 
macgee's Avatar
i'd put it like this

not double the speed but pretty much double the power!
Old 20th August 2007
  #13
Gear Guru
 
UnderTow's Avatar
theblue1, one of the things I miss each time on sites like Anandtech are proper DAW benchmarks. Their so called audio benchmarks usually use stuff like MP3 encoding. Pretty useless to give an indication of how DAWs will perform. The reason I mention this is that I think that DAWs (can) have much more parallelism than most other applications. In other words, I really don't think that typical "real world" tests reflect DAW performances well at all.

I think the POVRay test is closer to the way DAWs (can) perform due to the way DAWs need continuous streams of number crunching that is very well fit to parallel processing implementations.

I havn't been paying much attention to CPU tech these days. Are Intel dual and quad core processors CPUs still just several separate CPUs etched on the same die or are they like AMD CPUs that share the on-die cache memory? If so, AMD CPUs (although slower than current Intel chips) will show a bigger increase in dual-cores compared to Intel CPUs.

I also suspect that CPUs with on-board memory controllers will also give more benefit to dual-processor systems due to NUMA compared to those that don't.

Alistair
Old 20th August 2007
  #14
I'm not sure I'd really agree on the POV-Ray bench -- that's a test of a very specific process which splits nicely into multi-threading (obviously). I mean... most of us don't spend ALL of our time doing ray-tracing. (But -- if you do -- your path should be clear! heh )

BTW: I forgot to link that quad-core vs. dual article: AnandTech: Quad Cores: Intel Kentsfield Preview

I would say that an audio editor is probably analogous to a video editor in many ways -- but, of course, different apps will have different efficiencies. Clearly not all DAW apps are equally efficient.

But, you're certainly right that a DAW is not at all much like an old-fashioned office application.


I didn't mean to suggest for a second (and I'm sure those who are familiar with my line of thinking understand that) that these dual core processors aren't a big jump in performance -- they clearly are for multi-threading as well as multi-tasking. (And I have to say a bigger jump that I would have expected... at least based on these Intel-selected benchmarks.)

As a database and web developer, it's not at all uncommon for me to have five or more content creation apps and a bunch of browser and file windows open at a time... My HyperThreading Pentium helps a little, giving me a small but not insignificant boost but another core would certainly give me a leg up at times.

I stumbled into the Pentium M family with my now 3-1/2 year old Dell laptop and I'm a big fan of the chip family. I was leery of a 1.4 gHz chip -- but as I hurriedly read more (with the machine sitting in my 15 minute hold in the Dell refurbish store) it became clear that the Pentium M had some really nice aspects.

And it has proved to be a little champ. I still get entirely respectable performance, I feel, out of it (of course, my 2.8 gHz P4HT isn't exactly a rocket ship -- though, like I said, the HT does seem to make a significant difference in performance when I get into my DAW, with its Virtual Instruments and banks of plugs).
Old 20th August 2007
  #15
The limit on processor speeds is due to heat vs power consumption problems. Thus the increase in number of processors rather than speed. The Pentium M family is a great line of chips built off the Pentium 3 platform and my Pentium M 1.6 is much faster than my non HT P4 2.0. The Intel Core line of processors is descended from the Pentium M rather than the P4.
Old 20th August 2007
  #16
Lives for gear
 
TheRealRoach's Avatar
Dual Core vs. Dual Single Core Processor

Very interesting discussion indeed. Thanks for all of the feedback. Here's the next logical question then:

How would a Dual Core perform against a Dual Single-Core processor of the same rated speed?
Old 20th August 2007
  #17
Gear Nut
There are lots of various benchmarks out there but what really matters is how the CPU performs with YOUR applications. Single threaded apps will show very little improvement, where multi-threaded apps will perform significantly better. There are a number of posts on the Digidesign DUC where users have posted performance benchmarks with Pro Tools and various process processors. The test they’re doing is loading a Pro Tools session with D-Verb reverb plugins to see how many it will run before obtaining 90% CPU utilization. This is not any sort of scientific test, but it will give you a relative idea of the capabilities of the various processors. To answer your question more specifically, here are the results of my test:

Intel D975XBX MB w/4GB RAM
Pro Tools 7.3.1CS3
Approx. 90% Processor Utilization (per Task Manager)

Single core - 48 D-Verbs (I don't recall the processor #)
Dual Core – 122 D-Verbs (E6600)
Quad Core – 240 D-Verbs (Q6600)

Of course YMMV. I hope this helps.
Old 20th August 2007
  #18
Quote:
Originally Posted by TheRealRoach View Post
Very interesting discussion indeed. Thanks for all of the feedback. Here's the next logical question then:

How would a Dual Core perform against a Dual Single-Core processor of the same rated speed?
It depends on how much multi-threading there is in the apps you're using and how well it's implemented. If there's a lot you could see a big boost. But if they're typical productivity tools (web browsers, word processors, etc) you might not see much of any.

In this benchmark comparison, Anandtech found as much as 60% improvement from a Core Duo and a single core based on the same processor/speed:

Welcome to AnandTech.com [ Article: Intel's Core Duo Launch - Notebook Performance Revealed]

Quote:
Let multithreading apps have their way with the Core Duo and the performance differential jumps significantly, with the Core Duo based laptop outperforming the Dothan offering by almost 60%.
As DavidJ suggests -- it all really comes down to the application you're running -- if it makes good use of multithreading -- you'll see a big gain but if it makes poor use or doesn't use it at all, you won't.

Not all DAWs will presumably handle multi-threading as well. (I'm not even positive they all will even use it at all!)


Now, even if you're running all non-multithreading tasks, you'll still see a potential performance boost of varying degrees when you're multitasking a number of apps.
Old 20th August 2007
  #19
Lives for gear
 

There's lots of "it depends" issues here, not to mention some sloppiness in what is meant by "performance."

Widgetninja is correct in that it is possible to get better than twice the performance out of a *specific* application if there's already a lot going on in the background (such as kernel activity.) For example, if the system was already 75% busy, an application could theoretically get five times as much CPU (125% instead of 25%.) Even if the app were single-threaded in this example, it still could get four times as much CPU (100% versus 25%.) It is a bit contrived, but it *is* possible. Note that you're not getting better than twice the performance at a *system* level, but for a specific application.

As far as whether it is better or worse to have the same number of cores in more chips or less (say, one dual-core versus two equivalent single-core) it's very hard to predict. On the one hand, separate chips reduce contention for some resources (since they are duplicated), but increases contention for other resources (since multiple chips are vying for them.) In the real world, it probably depends on exactly what's happening in your system at any particular instant, or put another way, don't worry about it.

Even with single-threaded applications (or, more generally, applications with less parallel threads than you have cores) there will still be some systemic performance improvement with more cores, all other things being equal, since there's lots of other stuff (including the kernel and other o/s functionality) that can run in parallel with the application. Thus, eight cores with Logic ought to have more thrust than four cores, even though Logic itself appears incapable of using more than four.

Short version: as usual, if you need a machine *now*, it's usually best to buy as much machine as you can afford, as it will put off having to buy the next one, and any unutilized capacity is likely to be overtaken by software updates. If you don't need a machine *now*, it's usually best to wait, since they only get faster and wider and cheaper.
Old 20th August 2007
  #20
That works for me.

I should never have said never...


heh
Old 20th August 2007
  #21
Lives for gear
 
rackdude's Avatar
 

duel core is twice as fast if your program lets it be twice as fast

Old 21st August 2007
  #22
Lives for gear
 
santibanks's Avatar
to make things short as regarding to the gHz number,

if we have a core duo 2 processor, 2gHz does this mean that every core is 2 gHz or that the total sum of cores is 2gHz and that the performance is just better balanced?
Old 21st August 2007
  #23
Lives for gear
 
jupiter8's Avatar
 

Quote:
Originally Posted by rainy-taxi View Post
to make things short as regarding to the gHz number,

if we have a core duo 2 processor, 2gHz does this mean that every core is 2 gHz or that the total sum of cores is 2gHz and that the performance is just better balanced?
Both cores runs on 2.0 gHz.
Old 23rd August 2007
  #24
Lives for gear
 
rackdude's Avatar
 

the thing is, the program can usually only run one processor at a time. A duel core would only help to have one processor run the background data while the program is on the other, causing some increase in speed. If your program allows it, it could have both processors running it, so that would be basically double speed. It's really dependent on the program, and if your program doesn't allow it, having the second processor running the background won't make your computer twice as fast, but it will make the background run smoothly and have some needed background functions *like audio latency* run a bit faster.
Old 23rd August 2007
  #25
Gear Head
 
miditus's Avatar
 

Speedwise, nothing is different. The that counts is on handling of loads or the tasks... The more the core, the greater the load it can handle... but the speed is still the same...

Hope this will clarify the confusion....
Old 23rd August 2007
  #26
Again: it all depends on what you're doing.

Your application needs to have been written to make optimal use of multi-threading (breaking the program flow up into multiple execution threads -- which can also help on some single core processors, such as the Hyper-Threading P4's -- P4-HT, as it were) to get much out of a single app running on a multi-core machine. And that will vary, even within a given app -- different processes lend themselves to multi-threading differently. (That's why there's such a variation in benchmarks as discussed aobve.)

However, you should see some benefit when you're multi-tasking multiple apps, to some extent in simultaneous execution but also in switching the focus from one task to another.


Take a look at dkatz' answer above.
Old 23rd August 2007
  #27
Lives for gear
 
crypticglobe's Avatar
I agree that it is not double the speed.

I have done a lot of benchmark testing here on Nuendo and Cubase rigs. You are only going to see about a 60% to 65% increase. Still worth doing though!!

p.s. The two "cores" still have to communicate with each other and that takes a little time/speed as well. This is a very simplistic explanation of what is going on, but accurate.

Dual Core is potentially far more efficient than dual processors, but it is correct that operating systems and host application software still don't take full advantage of what is available yet.

If you don't believe us, try it yourself and report back to us.
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
T_R_S / Music Computers
30
11413 / Music Computers
6
jimmyz / Music Computers
1
jimmyz / Music Computers
1
Dog_Chao_Chao / Music Computers
4

Forum Jump
Forum Jump