Apple explains benefits of 64 bit mode

[papageno]

I dont find purging manually a big problem. It is actually not a problem at all: by default, Kontakt manages samples, hosted by memory server, automatically. I have chosen the manual approach, I could as well have it all unloaded automatically.

The other advantage of the memory server:
* with 64-bit logic, one would need to have plugins compiled for 64 bit app too (which won't generally happen in a very near future). I can use more ram now, hosting 32-bit plugins in 32-bit host.
BTW. The memory server process itself is 32-bit too. There is one forntend application, which creates a menu at the right side of the menu bar. Server processes are opened silently on the background. In case one 32-bit process gets filled, the next one will be opened.

[michkhol]

Engadget has learned -- thanks to an extremely reliable source -- that not only is Apple complicit in the development of Light Peak, but the company actually brought the concept to Intel and asked them to create it. More to the point, the new standard will play a hugely important role in upcoming products from Cupertino.

Not that I'm skeptical about press sources but FireWire was also originated by Apple and it is dying now.

AND it can be implemented for any type of data and it has been shown http://www.youtube.com/watch?v=khPx1dEIPnA copying huge files... so you do record to a FW drive right? This would make it look like a floppy drive.

It's a marketing trick, you are shown a tremendous data transfer speed which is true, but the sustained data transfer rate of a hard drive is as fast as it can rotate and currently it is much slower. Same thing is with SSD, the mass marketed ones (and not necessarily those that Intel used in its demo) have the sustained read rate that is on par with a fast hard drive, the sustained write rate is much slower. What makes SSD generally faster, it is almost zero seek time. But it does not help much with big video and HD audio data files. The primary goal of the Light Peak is to eliminate the multitude of different peripheral protocols and the substantial throughput is necessary to accommodate them all at the same time. In other words instead of having FW, USB, video, etc. flowing in different cables in different directions we shove them all in a single cable and make it tick faster.

But the reference to QPI is no more FSB to/from anything... it runs 1:1 with the clock... so when speed step kicks in, it's faster there too...

There is the common confusion here between the transfer rate and the speed of a certain peripheral can operate at. Generally they are not related. For instance no Light Peak would help if your RAM or PCI module operate at the same clock rate as before.

I don't know if on your system when you change the buffers the processor have to work harder... so wouldn't the DSP performance be effected in the DAW?

A CPU can handle only a certain number of operations per time unit. Each DSP unit (like a plugin) requires a certain number of operations per sample to complete. If you halve your buffer it means that it has to be updated twice as fast. So you have only half the time for the update but in the DSP unit the number of operations per sample remains unchanged. If the CPU still can process them in time - fine, if not - you hear a dropout.

The faster the drive bandwidth the easier it is on the processor for those tasks.

Actually in modern technology those two are not related anymore. Almost all peripheral I/O is done using some kind of Direct Memory Access, the actual data transfer is done without the CPU involved. A good video card has the GPU to transfer data to the monitor. The only exception is USB.

[newrigel]

AND it can be implemented for any type of data and it has been shown http://www.youtube.com/watch?v=khPx1dEIPnA copying huge files... so you do record to a FW drive right? This would make it look like a floppy drive.

It's a marketing trick, you are shown a tremendous data transfer speed which is true, but the sustained data transfer rate of a hard drive is as fast as it can rotate and currently it is much slower. Same thing is with SSD, the mass marketed ones (and not necessarily those that Intel used in its demo) have the sustained read rate that is on par with a fast hard drive, the sustained write rate is much slower. What makes SSD generally faster, it is almost zero seek time. But it does not help much with big video and HD audio data files. The primary goal of the Light Peak is to eliminate the multitude of different peripheral protocols and the substantial throughput is necessary to accommodate them all at the same time. In other words instead of having FW, USB, video, etc. flowing in different cables in different directions we shove them all in a single cable and make it tick faster.

Here's another marketing trick: http://www.maxupgrades.com/istore/index ... N=53379791
if you use a RAID, (in which that's what they used) the only drawback is how much $$ you have for the drives and the pipe the data is traveling down. The controller is where the true bottleneck is in that axiom...

[michkhol]

Here's another marketing trick: http://www.maxupgrades.com/istore/index ... N=53379791
if you use a RAID, (in which that's what they used) the only drawback is how much $$ you have for the drives and the pipe the data is traveling down. The controller is where the true bottleneck is in that axiom...

The SATA3 (if it is what they used) maximum data rate is about 500MB/s, all SATA controllers are independent so the total maximum data rate for 8 controllers is 4GB/s. So SATA3 still has some spare bandwidth and it is still well below the theoretical QPI limit of 25GB/s.

BTW

Although the history is not yet well recorded, shortly after the IDF presentation technology blog Engadget reported that Light Peak had been developed by Intel in collaboration with Apple. The article claimed that Apple had been working on the technology since 2007, around the time that ClearCurve was introduced, and that Apple CEO Steve Jobs personally asked Intel CEO Paul Otellini to take up development of the system as a new standard, stating that an all-optical interconnect was the only way to proceed.[18]

However, cnet later reported that other "industry sources" dispute this claim. This report states that Apple was contacted by Intel as part of an ongoing effort to introduce its industry partners and garner additional feedback. Sony was also mentioned at the IDF in this context, Apple was not mentioned at all.[12][18]

http://en.wikipedia.org/wiki/Light_Peak

[newrigel]

Here's another marketing trick: http://www.maxupgrades.com/istore/index ... N=53379791
if you use a RAID, (in which that's what they used) the only drawback is how much $$ you have for the drives and the pipe the data is traveling down. The controller is where the true bottleneck is in that axiom...

The SATA3 (if it is what they used) maximum data rate is about 500MB/s, all SATA controllers are independent so the total maximum data rate for 8 controllers is 4GB/s. So SATA3 still has some spare bandwidth and it is still well below the theoretical QPI limit of 25GB/s.

BTW

Although the history is not yet well recorded, shortly after the IDF presentation technology blog Engadget reported that Light Peak had been developed by Intel in collaboration with Apple. The article claimed that Apple had been working on the technology since 2007, around the time that ClearCurve was introduced, and that Apple CEO Steve Jobs personally asked Intel CEO Paul Otellini to take up development of the system as a new standard, stating that an all-optical interconnect was the only way to proceed.[18]

However, cnet later reported that other "industry sources" dispute this claim. This report states that Apple was contacted by Intel as part of an ongoing effort to introduce its industry partners and garner additional feedback. Sony was also mentioned at the IDF in this context, Apple was not mentioned at all.[12][18]

http://en.wikipedia.org/wiki/Light_Peak

It's running on OS X... oh ••••... that's fake too!
Anyone can edit wiki... big deal!