64bit apps

[dawman]

Agreed.

Let us proceed anyway, since I am a pawn being manipulated.

[astroman]

...http://vsl.co.at/en/65/71/512/336.vsl

The above application allows 35,000 such RAM paths on 1 MIDI channel. I wouldn't call that bad programming, but actually taking advantage of more memory addressing. Maybe this is extreme, but film scoring folks need this to pull off elaborate orchestrations, as opposed to using an entire orchestra.
...
Let's say that 64k is all that is needed for a RAM path. In a 4GB address, that would be 4GB's divided by 64kb, if my math is correct, is 65,536. It sounds large but the above analogy demonstrates how quickly RAM gets chewed up when multiple velocity layers ( dynamics ), is added w/ multiple articulations ( sample level ).
So more is better, as in life I suppose. ...

well I hope not to annoy, but this a good example to illustrate the less obvious side of disk streaming ...

what was so smart about the very first Gigasampler ?

Strategy - they divided all samples into an attack phase and the rest
attack was stored in memory, the 'rest' was loaded from disk while 'attack' was playing.
Obviously the small lag was sufficient to feed the rest in time from the buffers.
At the days of the 400MHZ Pentium that was pretty smart programming indeed

anyway, if this mechanism is implemented (more or less) perfectly, your 'slots' for attacks are 172 per Megabyte at a 44ms attack phase, meaning 172k attacks per Gigabyte (calculated with 132k for 1 sec of 24bit audio at 44k)

a regular disk based database can easily look up any of a few million records in far less than that amount of time.
The smart move of the original developer (Nemesys) was to combine the 2 different technologies
Roland cound have done it even earlier if they... well, had thought about their LA stuff in a broader context

cheers, Tom

[garyb]

Agreed.

Let us proceed anyway, since I am a pawn being manipulated.

no sir. i am your pawn...

[garyb]

astroman, so this is another good example of lazy programming wasting the real resources of these amazingly powerful machines?

[husker]

just because they won't.

Well .. I'm really clueless in this area. I never had more than 2GB installed in a XP machine and I never cared about more. That's why I asked.
However this article indicates it should be possible in some way:
http://technet.microsoft.com/de-de/libr ... 57155.aspx :
"PAE mode enables processors to address greater than 4 gigabytes (GB) of memory. "

There you go...you found your own proof for my argument

XP PAE is what is needed to get access to more than 4GB...however XP PAE is probably on about 0.1% of XP machines, which leaves the other 99.9% of machines with no ability to access more than 4GB. So for XP there is no 64 bit myth...honestly...truly. No amount of magic in an application can avoid that

[astroman]

astroman, so this is another good example of lazy programming wasting the real resources of these amazingly powerful machines?

oops - I can't tell.
Didn't even know it was about Vienna Symphony.
I just referred to Jimmy's numeric example and forgot about the link.
Should have written 'opportunity'

cheers, Tom

[astroman]

...XP PAE is what is needed to get access to more than 4GB...however XP PAE is probably on about 0.1% of XP machines, which leaves the other 99.9% of machines with no ability to access more than 4GB. ...

I considered it (more or less) an option for custom designed and highly specialized systems. Completely confusing in 'when, if and but...'

cheers, Tom

[wolf]

XP PAE is what is needed to get access to more than 4GB...however XP PAE is probably on about 0.1% of XP machines, which leaves the other 99.9% of machines with no ability to access more than 4GB. So for XP there is no 64 bit myth...honestly...truly.

Well, those numbers are highly questionable .. ...honestly .. truly
Btw, the myth is that 32bit applications can't handle >4GB, which is definitely *not* true as I'm working daily with such an app (incl. Vienna Library). Indeed there is no 64bit myth .. yet, hehe.

[dawman]

Brotha' Man Wolf,

Please tell me about your rig, you might save me an unnecessary amount of money spent.

I am basing my purchases on a Vista machine since Gigastudio 4 and GVI 4 are being optimised for it. However they have recently gone with a Mac based compatability as well.

I believe it is a Mac O.S. w/ the Intel CPU's only.........................................................
http://www.tascam.com/details;8,7,1055.html

I do not like the way that I am being thrown into this, but when I have to start using a new app. where it's designed to run w/ Vista 32 / 64, or XP 32 / 64 or Mac OSX, things start getting confusing. I finally got a grip w/ Scope and now all of this shit's happening.

Damn it I just want to play. I wish they'd leave me alone for more than one year at a time.

[pollux]

Hi,

Just for the records: I have servers running 32 bit Windows 2003 server on boxes with 16 and even 32 GB of RAM...
There is an option called "PAE" (standing for Physical Address Extension) that allows the kernel to use more than 4 GB of RAM.

Still the applications are 32 bit, and have their own memory addressing limitations, like the Java Virtual Machines that in 32 bit cannot spawn over 2GB.

So back to this thread, it is possible for a 32 bit application to use more than 4 GB of RAM, as long as it has the right addressing logic...
Now, most commercial vendors will simply not do it because it takes lots of efforts and it's rarely justified (economically speaking), unless developing products such as database or application servers...
It's simpler and cheaper to overcome this by using native 64 bit kernel memory addressing.

[garyb]

jimmy, i'm sure that xp64 runs just as fast as vista(if the program is written correctly it should be able to use all the ram you can give it while in xp32!). vista may be just fine, but it may also have it's own problems(protection schemes, VERY heavy resource usage). if you KNOW that vista will really run more samples smoother than xp 32 OR 64, then vista, by all means, no point being scared about it. it's possible that tascam, in league with m$, has decided to not fully enable the product for anything other than vista, since it's the new os, maybe that's where all the money was spent. as far as i can tell, however, although there are differences and improvements(mostly product "backdoors"), vista and xp share a lot in common.....all this "vista ready" has to do with product protection, anti piracy "features" that must be hardware supported.

[dawman]

Wise words indeed.

GS4 is XP 32 / 64bit capable, and also Vista 64bit capable.

Check This Out.

http://www.dawbench.com/blofelds-xp-v-vista.htm


I might just re think this thing through a little more.

We must definately upgrade the hardware though. I want to use more RAM, that is the main concern. The 3GB switch works fine in GVI standalone , but only supports the crappy wav drivers, not the ASIO. We needed a host to get Bidule w/ GVI working, remember ?

[garyb]

vista looks to be inferior at this point according to that piece.

once again, the way i read the references, if the app is written properly, no switch should be needed.
i'd double check with tascam to see if gs4 can handle more than 3.6gb of ram in 32bit....

even though vista has improvements in the handling of certain tasks, from what i've seen, the new security and graphics features make the improvements a wash. if you are playing online games, there may be a significant improvement(i don't know), but i haven't seen evidence of a significant improvement in the daw world, unless it's an ad from m$ or cakewalk.

adding ram and a faster processor doesn't seem to be an issue....it all depends on whether the app was written to make use of such things....

[dawman]

Here was an explanation that made some sense to me, but if Logic's EX24 sampler can address more than 4GB's or RAM, I am once again confused. Maybe this will spark further debate, which I am soaking up like a sponge.



_____________________________________________________________

Hi, I work as computer analyst so just for fun, I will try to explain why 32bit OS can usually only manage 4GB of RAM. The explanation is pretty simple…

4GB means you have something around 4 000 000 000 bytes right? Let’s be a little bit more precise…

32bits is the size of the processor stacks. Because of that, the biggest number the CPU can handle in one cycle is a 32bits binary number. Now, what is that maximum binary number of 32bits you can create? If you align 32 times “1” on your calculator and convert binary to decimal, the number you get is: 4 294 967 296. Divide that number by 1024, (bytes to kylo), then divide the result by 1024 again (kylo to megs) then one last time by 1024 (megs to gigs). What do you get? 4!

In simple terms, the largest amount of addresses the cpu can handle on his memory bus has 32bits so the maximum number is 4294967296, or “4Gigs” or 4294967296 “bytes”. That’s why 32bits OS can handle only 4G of ram.

So now, the new Intel CPU are 64 bits right? SO let's do the same math, align 64 times “1” on your calculator and transfer it to decimal. You will get the maximum number of addresses the CPU could handle in one cycle. You will quickly notice that the 128GB limit is only a practical limit and not a mathematical limit. J

So on the programmer’s point of view, handling more that 4 Gigs on a 32bits CPU means that all address management has to be “tweaked” and will probably need more than one CPU cycle to handle the addresses so it becomes extremely difficult!

That’s why only Enterprise version of OS like Microsoft Windows Server 2003 can have more than 4GB or RAM installed on 32bits CPU. They do the tweak at the Kernel level of the OS. But the license cost is around 4K$. J So that is why it is useless to hope that DAW applications will ever be able to handle more than 4GB of RAM on a 32bits CPU.

But now, since Intel CPU have reach 64bits, everything becomes a lot more easier.

Regards
__________________
Nicolas Roy


_____________________________________________________________


This makes me wish to stay where I am at w/ 32bit. I will simply add a second DAW back to my live rig w/ GS4 XP/32bit version. If I see greater productivity and huge advantages next summer, no problemo. Just add the new O.S.

I think a 2 DAW approach is proven and safer, I had 2 DAW's until a few months ago.

Solution: Build another DAW for GS4 / Forte, and use another Scope AES card.

I KNOW this works. Scope never fails me, ever.

[MD69]

Hi S4L,

This is an oversimplified view of memory addressing, but it give you a part of the problem. This explanation doesn't take into account the virtual memory management. Globally you can think a program is tied to 32bits adressing (it cannot access a memory cell which is farther than 4G due to 32 bit program register limitation), but this is true for the program .... not the computer. You could have several programs having access to 4Gb of separate memory areas. So if you think EXS24 as a program implementing a monotimbral sampler, you could have several instances having each 4Gb of separate memories .... So reallity is, limitations lies in the OS implementation of the virtual memory management which use (in Intel case) the MMU inside the Chipset (AMD have the MMU - Memory Management Unit inside the processor).
Depending on the chipset AND how it is used by the mainboard manufacturer,
the OS can access 4Gb...or more.

cheers

Michel

(and this is still a simplified view!)

[dawman]

Thanks for the info, I am grateful.

[hubird]

so...you stay where you were, S4L...32 bit XP?
if so, i take back my 'monster' qualification

[wolf]

Please tell me about your rig, you might save me an unnecessary amount of money spent.
I believe it is a Mac O.S. w/ the Intel CPU's only.........................................................
http://www.tascam.com/details;8,7,1055.html

I don't know if GS is capable of addressing >4GB on a Mac.
I'm using Logic8's EXS24 on a quad core (ppc&intel) and octa core (intel only), OSX 10.4 & now partly OSX 10.5.1.

Here was an explanation that made some sense to me, but if Logic's EX24 sampler can address more than 4GB's or RAM, I am once again confused.

EXS24 can address as much RAM as is installed and availalabe (=not used by other apps). Michels explanation goes in the correct direction (beside it's also possible with the "older" ppc cpus).

I do not like the way that I am being thrown into this, but when I have to start using a new app. where it's designed to run w/ Vista 32 / 64, or XP 32 / 64 or Mac OSX, things start getting confusing. I finally got a grip w/ Scope and now all of this shit's happening.

yeah, the world is turning .. hehe

To follow up to Gary: Don't let yourself confuse by options, which might be possible somewhen in the future. Concentrate on what is available now.
Also I wouldn't switch to another system without some safety belt, i.e. keeping a running system, which you're confident with.

Jimmy, try to get a picture of how many instruments you are planning to use, add a fair amount on top of it. Then you can estimate how much RAM you will use/need and therefor the hardware needs. Then compare prices & convinience and there you go.

One last word to the switch 32bit to 64bit:

Now, most commercial vendors will simply not do it because it takes lots of efforts and it's rarely justified (economically speaking), unless developing products such as database or application servers...
It's simpler and cheaper to overcome this by using native 64 bit kernel memory addressing.

No, it's not, sorry.
Implementing the >4GB thingie into Logic8 took 4 weeks of one manpower. Rewriting (again) the whole codebase of Logic will at least take a year with a manpower of 70 guys. Going 64bit is not simply a switch in the compiler.

[dawman]

Thanks Wolf,

My dual DAW approach is more sensible.

This is what I use to use, and I could load tons of libraries. I am leaning towards the VSTi's now that I have heard some in use, and hear no difference in quality of the audio, versus the sampler apps. A local guy here uses the Broadway Big Band VSTi on a Mac, and it sounds amazingly realistic. But his whole DAW is used for that instrument only. I don't do many Nelson Riddle arrangements anyway, Elvis and Rat Pack were cool, but I love FAT BASTARDS, and Classic Rock better.

The last 2 years have convinced me that this is a better approach. Libraries made for Giga can still be used, and the VSTi's will fill in the holes.

I am happy w/ DSP and hardware analog based synths over VSTi synths still. However I heard and demo'd the VSTi synths that Sonar has and the sound is getting closer, but still no cigars. I am confident that Bowens Solaris in hardware w/ 16 voices will rule supreme. But VSTi developers have so much power to play with these days. Maybe in 2008 someone will release some decent sounding VSTi emulations of analog synths. But I have what I need, and have no desire to audition anymore of those. I do look forward to new synthesis techniques which hardware, and Scope cannot achieve.


Octa Core !!?? You are very sick my friend. But I love it so.

[garyb]

well vsti's of that sort are not sound generators, they are sample players with nice interfaces. the trick is in the library and how it's implemented. i would expect a vsti rompler to sound as good as giga and vice-versa depending on the library....