Which Mac Pro Processor?

[TimothyHyde]

I am configuring a 8-core Mac Pro for use with high-end scanning, PS3, and a variety of other photo-related tasks.  I'm moving from a PC environment.  I am trying to build something that will sing for several years, so it's okay to pay for features that may not be necessary today.  Question:  is there any advantage in getting the 3.2 GHz processor over the 3.0 GHz, or over hte 2.8 for that matter?  Thanks

[Jack Flesher]

I am configuring a 8-core Mac Pro for use with high-end scanning, PS3, and a variety of other photo-related tasks.  I'm moving from a PC environment.  I am trying to build something that will sing for several years, so it's okay to pay for features that may not be necessary today.  Question:  is there any advantage in getting the 3.2 GHz processor over the 3.0 GHz, or over hte 2.8 for that matter?  Thanks
[a href=\"index.php?act=findpost&pid=171652\"][{POST_SNAPBACK}][/a]

Well for one, it will be about 7% faster than the 3.0 and 15% faster than the 2.8 on most tasks...  

While that perhaps sounds like I'm being sarcastic, it is true: many processes related to images are limited simply by processor clock speed.  I set up a dedicated CS3 benchmark action and image and sent it out to friends with different computers.   My new Mac Pro has the 3.2's and the bottom line is it took 7% longer to run on a last generation Mac Pro 3.0 and 20% longer to run on a previous generation MP 2.66.  All of these machines had at least 4 cores (the max CS3 can really access) and 8 or more Gigs of Ram. (CS3 will only use 3G RAM max, then tags the scratch disk --- contrary to popular belief, CS3 does NOT utilize excess system Mac system RAM. Hopefully this and processor core utilization will be addressed in CS4.)

Now for scanning, I suspect you won't see any difference, since the port you're scanning through will be your bottleneck.

[luong]

Well for one, it will be about 7% faster than the 3.0 and 15% faster than the 2.8 on most tasks... 

(CS3 will only use 3G RAM max, then tags the scratch disk --- contrary to popular belief, CS3 does NOT utilize excess system Mac system RAM


[a href=\"index.php?act=findpost&pid=171656\"][{POST_SNAPBACK}][/a]

I thought it would do so if you enable the Virtual Memory Buffering plug-in. In my test, this made a difference.

In optimizing PS performance, note by using a RAID0 2x 10K RPM scratch disk, I gained about 70% speed over a single 7.2K scratch disk for a mere $350.

[Jack Flesher]

I thought it would do so if you enable the Virtual Memory Buffering plug-in. In my test, this made a difference.

In optimizing PS performance, note by using a RAID0 2x 10K RPM scratch disk, I gained about 70% speed over a single 7.2K scratch disk for a mere $350.
[a href=\"index.php?act=findpost&pid=171663\"][{POST_SNAPBACK}][/a]

Tuan:  Your 10K raptors are SATA 1, or maximum burst rate 1.5 GB/s and about 50 MB/s sustained, while the newest 7200 drives are SATA2 with 32 MB buffers, 3 GB/s burst and over 100 MB/s sustained.  But yes, striping them does improve read/write performance some, but not a lot --- CS3 still uses some archaic programming to access scratch and that is the limitation...  This is something that hopefully gets addressed in CS4 --- how about a new solid-state drive for scratch?

,

[BernardLanguillier]

how about a new solid-state drive for scratch?

,
[a href=\"index.php?act=findpost&pid=171708\"][{POST_SNAPBACK}][/a]

Bad idea, at least on the old Mac Pro 8 cores.

The intel chipset of the MB cannot take advantage of the full bandwidth of the SSDs... this is coming from first hand experience with a 32GB MTRON unit.

I am seeing higher transfer rates with a single Raptor than with the MTRON unit...

Cheers,
Bernard

[mistybreeze]

is there any advantage in getting the 3.2 GHz processor over the 3.0 GHz, or over hte 2.8 for that matter

It seems the biggest advantage, according to BareFeats.com, is resale value. From experience I can tell you, it's always a little easier to sell the fastest model of the series four years from now. But, unless you're a speed-freak and you have an intimate relationship with a stopwatch, 3.0 is equal to its top-of-of-the-line predecessor from the previous series. How can you not love that? One must always question the numbers game in big business and make your investment choices wisely.

[CatOne]

It seems the biggest advantage, according to BareFeats.com, is resale value. From experience I can tell you, it's always a little easier to sell the fastest model of the series four years from now. But, unless you're a speed-freak and you have an intimate relationship with a stopwatch, 3.0 is equal to its top-of-of-the-line predecessor from the previous series. How can you not love that? One must always question the numbers game in big business and make your investment choices wisely.
[a href=\"index.php?act=findpost&pid=171755\"][{POST_SNAPBACK}][/a]

I doubt the increased resale value could make up for the difference in acquisition price.

I would say -- DEFINITELY get the dual quad cores.  Whether it's 2.8, 3.0, or 3.2 it's not going to be a HUGE difference (15% is not so easy to notice).  Make sure you have a lot of RAM (at LEAST 512 MB per core, so we're talking about 4 GB minimum) and all should be quite fast.

[luong]

Tuan:  Your 10K raptors are SATA 1, or maximum burst rate 1.5 GB/s and about 50 MB/s sustained, while the newest 7200 drives are SATA2 with 32 MB buffers, 3 GB/s burst and over 100 MB/s sustained.  But yes, striping them does improve read/write performance some, but not a lot

[a href=\"index.php?act=findpost&pid=171708\"][{POST_SNAPBACK}][/a]

Jack, this 50 MB/s figure doesn't sound right, but the proof is in the pudding anyways. Why don't you run the Retouch artists test on CS3  (http://retouchartists.com/pages/speedtest.html) with 100 history states, 4 cache levels,  70% memory ?

[Jack Flesher]

Jack, this 50 MB/s figure doesn't sound right, but the proof is in the pudding anyways. Why don't you run the Retouch artists test on CS3  (http://retouchartists.com/pages/speedtest.html) with 100 history states, 4 cache levels,  70% memory ?
[a href=\"index.php?act=findpost&pid=171945\"][{POST_SNAPBACK}][/a]

Because it really doesn't stress a newer machine with over 3G ram and does not force scratch.  It runs on my laptop in about a minute with only 2G RAM allocated to CS, 50 history states and 6 cache levels,  and I'm sure it would only be in the 20 second range with my new desktop.  I wanted something I know forces CS to scratch.

[jerryrock]

(CS3 will only use 3G RAM max, then tags the scratch disk --- contrary to popular belief, CS3 does NOT utilize excess system Mac system RAM. Hopefully this and processor core utilization will be addressed in CS4.)
[{POST_SNAPBACK}][/a]

This statement is just incorrect.  The following excerpt is from an Adobe Tech note:

"When you run Photoshop CS3 on a 64-bit operating system, such as Mac OS X v10.4 and later, Photoshop can access up to 8 GB of RAM. You can see the actual amount of RAM Photoshop can use in the Let Photoshop Use number when you set the Let Photoshop Use slider in the Performance preference to 100%. The RAM above the 100% used by Photoshop, which is from approximately 3 GB to 3.7 GB, can be used directly by Photoshop plug-ins (some plug-ins need large chunks of contiguous RAM), filters, and actions. If you have more than 4 GB (to 8 GB), the RAM above 4 GB is used by the operating system as a cache for the Photoshop scratch disk data. Data that previously was written directly to the hard disk by Photoshop is now cached in this high RAM before being written to the hard disk by the operating system. If you are working with files large enough to take advantage of these extra 2 GB of RAM, the RAM cache can increase performance of Photoshop."

The full tech note is available here:

[a href=\"http://www.adobe.com/go/kb401089]http://www.adobe.com/go/kb401089[/url]


For this reason, my MacPro is equipped with 12 gigs of ram.  This is a more efficient option for speeding up your system. I have no need for striped drives.

Jerry

[Jack Flesher]

This statement is just incorrect.  The following excerpt is from an Adobe Tech note:

"When you run Photoshop CS3 on a 64-bit operating system, such as Mac OS X v10.4 and later, Photoshop can access up to 8 GB of RAM. You can see the actual amount of RAM Photoshop can use in the Let Photoshop Use number when you set the Let Photoshop Use slider in the Performance preference to 100%. The RAM above the 100% used by Photoshop, which is from approximately 3 GB to 3.7 GB, can be used directly by Photoshop plug-ins (some plug-ins need large chunks of contiguous RAM), filters, and actions. If you have more than 4 GB (to 8 GB), the RAM above 4 GB is used by the operating system as a cache for the Photoshop scratch disk data. Data that previously was written directly to the hard disk by Photoshop is now cached in this high RAM before being written to the hard disk by the operating system. If you are working with files large enough to take advantage of these extra 2 GB of RAM, the RAM cache can increase performance of Photoshop."

The full tech note is available here:

http://www.adobe.com/go/kb401089
For this reason, my MacPro is equipped with 12 gigs of ram.  This is a more efficient option for speeding up your system. I have no need for striped drives.

Jerry
[a href=\"index.php?act=findpost&pid=171981\"][{POST_SNAPBACK}][/a]

Uh Jerry, we know that's what they say, but we've proven it simply doesn't work that way in practice with Leopard/CS3.  First, the preference dialog will only allow you to 3G max for Photoshop (3072MB to be exact, NOT anything close to 3.7GB), and if you then monitor system RAM in a heavy processing situation with no other applications running, you'll see that most of our RAM sits dormant while your scratch drive is running...   Moreover, the scratch dialog Adobe has written seems to peak out at about 50 MB/s max read/write which is below the capabilities of many current single drives let alone a striped array.  We confirmed this by attaching an 5-disk striped array as scratch and seeing that I/O never exceeded 50 MB/s sustained. This array is proven capable of over 380 MB/s sustained read/write with other applications...

It is true that other applications including plug-ins can access the remaining free RAM though, and this is of some help.

Sorry to burst your bubble,

[luong]

Because it really doesn't stress a newer machine with over 3G ram and does not force scratch.  It runs on my laptop in about a minute with only 2G RAM allocated to CS, 50 history states and 6 cache levels,  and I'm sure it would only be in the 20 second range with my new desktop.  I wanted something I know forces CS to scratch.
[a href=\"index.php?act=findpost&pid=171965\"][{POST_SNAPBACK}][/a]

Want to bet a beer ? I am wondering what's so different about our systems. I have an early 2008 mac pro with 8GB RAM, and with the settings I listed, scratch *is* used. The activity monitor also frequently indicates 5+GB in use.

[Jack Flesher]

Want to bet a beer ?
[a href=\"index.php?act=findpost&pid=172036\"][{POST_SNAPBACK}][/a]

LOLOL!  I am in Moab prepping for a workshop, so let me get back to my desktop next week and actually run it there and monitor it.   Regardless, I'll be happy to buy you a beer and share the results --- Heck, I'll even buy you lunch if you drive!

[jerryrock]

Uh Jerry, we know that's what they say, but we've proven it simply doesn't work that way in practice with Leopard/CS3.  First, the preference dialog will only allow you to 3G max for Photoshop (3072MB to be exact, NOT anything close to 3.7GB), and if you then monitor system RAM in a heavy processing situation with no other applications running, you'll see that most of our RAM sits dormant while your scratch drive is running...   Moreover, the scratch dialog Adobe has written seems to peak out at about 50 MB/s max read/write which is below the capabilities of many current single drives let alone a striped array.  We confirmed this by attaching an 5-disk striped array as scratch and seeing that I/O never exceeded 50 MB/s sustained. This array is proven capable of over 380 MB/s sustained read/write with other applications...

It is true that other applications including plug-ins can access the remaining free RAM though, and this is of some help.

Sorry to burst your bubble,
[a href=\"index.php?act=findpost&pid=172011\"][{POST_SNAPBACK}][/a]

My bubble is still intact, thank you.

Who is this ubiquitous "we" that you keep referring to?  Have your findings been published?  
Sorry, but until I see a published study concerning this issue, the Adobe report is factual and anything else is just conjecture.

Jerry

[Jack Flesher]

My bubble is still intact, thank you.
[a href=\"index.php?act=findpost&pid=172220\"][{POST_SNAPBACK}][/a]

Run an actual test for yourself and it won't be --- Open only CS3 and activity monitor.  Run even the simple benchmark test QT linked to above and watch your RAM on the activity monitor during the benchmark.  For more definitive proof, take a 10MP image at 16bit and create an action that increases its size by 1600% (4x in each direction), then rotates the image 90 degrees, then runs USM at any simple setting like 200/2/2.  This will take a minute or two on a fast machine like yours depending on configuration.  Watch activity monitor RAM usage and scratch while this runs and you'll see what I'm talking about for yourself...  

PS: The results are part of a larger review on system performance that will be published soon.

Best,

[jerryrock]

Run an actual test for yourself and it won't be --- Open only CS3 and activity monitor.  Run even the simple benchmark test QT linked to above and watch your RAM on the activity monitor during the benchmark.  For more definitive proof, take a 10MP image at 16bit and create an action that increases its size by 1600% (4x in each direction), then rotates the image 90 degrees, then runs USM at any simple setting like 200/2/2.  This will take a minute or two on a fast machine like yours depending on configuration.  Watch activity monitor RAM usage and scratch while this runs and you'll see what I'm talking about for yourself... 

PS: The results are part of a larger review on system performance that will be published soon.

Best,
[a href=\"index.php?act=findpost&pid=172308\"][{POST_SNAPBACK}][/a]

I do look forward to reading the review.  If Apple's activity monitor was the only measure used, the study could be very flawed.  The only real number in the Activity Monitor is the page out number and swap number.

[gdeliz]

I do look forward to reading the review.  If Apple's activity monitor was the only measure used, the study could be very flawed.  The only real number in the Activity Monitor is the page out number and swap number.
[{POST_SNAPBACK}][/a]

See [a href=\"http://homepage.mac.com/boots911/.Public/PhotoshopAccelerationBasics2.4W.pdf]http://homepage.mac.com/boots911/.Public/P...nBasics2.4W.pdf[/url]
page 9 for how to use any ram beyond 3.7GB  for photoshop scratch space. The document is a good guide for optimizing a mac for photoshop use.

George Deliz

[jerryrock]

See http://homepage.mac.com/boots911/.Public/P...nBasics2.4W.pdf
page 9 for how to use any ram beyond 3.7GB  for photoshop scratch space. The document is a good guide for optimizing a mac for photoshop use.

George Deliz
[a href=\"index.php?act=findpost&pid=172491\"][{POST_SNAPBACK}][/a]

George,

I agree with you wholeheartedly and have utilized all of the suggestions in your article (which are the same as the ones mentioned in the Adobe Tech Note that I previously posted).

My reply was written for the benefit of Jack Fletcher who claims he has done testing that proves Photoshop can only use 3 gigs of ram and then goes directly to the hard drive for scratch disk.
Mr. Fletcher is the non-believer!

Jerry

[silvergc]

Jack:

FYI, SATA 1 runs at 1.5Gb/s (Gigabits per second) not 1.5GB/s (Gigabytes per second) the encoding on the bus uses 10bits for every byte of real data so the maximum burst rate on a SATA 1 link is 150 MB/s.  Your sustained rate is probably correct.  SATA 2 is twice as fast as SATA 1.