Anyone using Aperature with MacPro 2.6GH/stock video card?

I am considering getting a Mac Pro 2.6 W IG RAM and the standard video card. I will be using the 23 inch Mac display... How well will Aperture work with this setup?

Thanks

Yup, that is exactly the system I now use. Aperture has crashed a couple of times for unknown reasons (the latest version) while importing bmp.'s, otherwise it is pretty fast. AM

Leicaman said:

I am considering getting a Mac Pro 2.6 W IG RAM and the standard
video card. I will be using the 23 inch Mac display... How well
will Aperture work with this setup?

Click to expand...

Well, the Mac Pro 2.66 and the 23" display are good. What W IG RAM is... typo? Anyhow I would consider 1 GB minimum and prefer 2 GB (in 4 x 512 modules). I have this machine here for testing (with 4 x 1 GB RAM) and the standard card... I am extremely impressed by the machine. Universal applications that do not care about the graphics card (the vast majority) flies, some applications seem to be about twice as fast as on the Dual 2.7 G5. Photoshop runs OK and After Effects is equal or faster than on the Dual G5 (except for start-up) . And: It is quiet and cool! No comparison to the G5. Just... whatever requires a good graphics card (out of the applications I use mainly Aperture and Motion...): Unimpressed. Motion's preview is slower than on the MacBook and Aperture is an adventure in beach balling... adjustments take longer than on the MacBook Pro, previews of 12 and 16 MP images (out of sequence) take a long time. I have ordered our MacPro's with the X1900XT and told my dealer I will only keep the machine after testing if I get the GPU upgrade for the 350 USD BTO price iso the full 500 USD for the later upgrade. He agreed. Until the card arrives Aperture remains on the MacBook Pro and the Dual G5 with the upgraded GPU.

Cheers,
Uwe

Uwe,

I wonder why Aperture would be so much more GPU resource intensive than PS CS2? Many of my images are high res cans of 6X6 chromes and the basic image files will be upwards of 650-700MB. That is why the Mac Pro is so aluring.. I can expand to all the RAM I am likely to need... but the need to upgrade the GPU kind of throws me for a loop.

Rich

Aperture uses the video card to render the RAW files and associated processing 'live', instead of using the cpu like other photography software. Hence the speedup with a faster video card. Other programs (such as photoshop) would see little to no speed up from the faster video card.

So Aperature uses the GPU instead of the CPU... how curious.

Any idea how expensive to step the Mac Pro up to the next level of video card... I am not exactly sure which card that would be,,, Exactly HOW much GPU is needed to run the program smoothly?

Leicaman said:

So Aperature uses the GPU instead of the CPU... how curious.

Any idea how expensive to step the Mac Pro up to the next level of
video card... I am not exactly sure which card that would be,,,
Exactly HOW much GPU is needed to run the program smoothly?

Click to expand...

You would need the ATI X1900XT. If you order the Mac Pro built-to-order from Apple the upgrade is USD 350. And yes, Aperture uses the GPU for computations.

Cheers,
Uwe

Leicaman said:

Uwe,

I wonder why Aperture would be so much more GPU resource intensive
than PS CS2? Many of my images are high res cans of 6X6 chromes
and the basic image files will be upwards of 650-700MB.

Click to expand...

See posts below. The GPU does a lot of the computations for Aperture.

leicaman said:

That is
why the Mac Pro is so aluring.. I can expand to all the RAM I am
likely to need... but the need to upgrade the GPU kind of throws me
for a loop.

Click to expand...

Compared to the FB DIMMS a decent graphic card is a bargain and Aperture is actually no big RAM consumer (never seen it using more than 2 GB - not even when processing TIFFs from 25 MP Leaf backs supplied by free-lancers, so adding RAM above 2-4 GB will certainly not make it faster...).

Of course you an try with the stock card and see for yourself how it does, just, BTO is USD 350 and getting the card separately later will be USD 500...

Cheers,
Uwe

Uwe,

This seems like a silly question.... but why did they task the the Aperature processing to the GPU... to sell upgrade video cards?

Leicaman said:

Uwe,

This seems like a silly question.... but why did they task the the
Aperature processing to the GPU... to sell upgrade video cards?

Click to expand...

Hehe... one could think so. But then - they normally do not have any for sale... try to get a GPU for a Quad G5...

I really cannot explain the low-level stuff going on in CoreImage. As I understood it, some of the screen drawing magic going on there can only be brought to have the lowest latency if the processing is done in the GPU. Also it takes load off the CPUs and makes (if you have all the right components) for better use of the system. If I look at Motion 2 it makes sense I have to say. It uses all cores, all RAM and of course giving some processing to the GPU makes for a faster program. With Aperture I do not really get it yet (maybe with future versions). If I look at my CPU meters... it seems Aperture does not give them that much to do and maybe giving some of the work back to the CPUs would make the program more convenient for people with less advanced GPUs... but then, I really do not understand the very details.

Cheers,
Uwe

My thoughts exactly: I just don't quite get it. I mean, I do understand the GPU is immensely faster for some of the image processing, but like you, the CPU seems to loaf. I would love to see Aperture use it just a bit more--especially on the Mac Pro, with all those cores sitting around.

Jeff
--
http://www.pbase.com/jhapeman

To put it as simply as possible...

A CPU and a GPU have massively different basic architectures. A CPU is designed to process a list of instructions. A GPU is designed to process, display and manipulate pixels. CPU designers have to concern themselves with making a processor that will perform all manner of different tasks. A GPU can be (and generally is) designed to do nothing more than push pixels. On top of this, a GPU gets its own dedicated memory, and the bandwidth it has to that memory is phenomenal. Part of the reason for this is how close the actual GPU is to the memory on the board. There isn't really a good way of getting memory bandwidth that high for a CPU, and when dealing with large image files, having a lot of memory bandwidth is a big help (since there is simply so much data that the CPU/GPU needs to transfer between its onboard cache/registers to it's memory).

Having CoreImage use the GPU for graphics processing tasks does make sense, when the GPU is designed for exactly that (although motivated by things like gaming etc...). A company called (I think) Ageia has released a physics processor on a PCI card, specifically designed for processing physics data in games. It's rather curious that things are going this way, since the Amiga, introduced in 1985, used a very similar setup, with custom chips for graphics, sound etc... They were considered a long way ahead of their time, and PCs and Macs took a very long time to catch up. It looks like things are starting to turn full circle, and move back to a custom chip like setup.

Anyway, back to the point, memory bandwidth and the specialised nature of a GPU are the two main reasons CoreImage uses them for it's dirty work. Putting those tasks onto a normal processor would result in a real slowdown. To sum it all up, the processing load is put onto the GPU because it's the best place for it to go.