Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
'Real World' Photoshop Benchmark - Speed Test
- Thread starter BakedBeans
- Start date
- Sort by reaction score
Unfortunately yes
If you run the test again the program has items stored in the cache that will benefit the speed next time round, if only by a little. That's why steps are in the readme to restart / Open photoshop and nothing else / open file / run action.10 rupees says the new Nahalem Mac Pro's don't do this test as fast as a Nahalem PC for 1/3rd the price.
Well, I think Nehalem Xeons and Nehalem i7s at the same clock will perform the same in this particular test. I don't see any reason to think that a 2009 quad core Mac Pro at 2.66GHz or 2.93GHz will perform this test any slower than an i7 PC at the same clocks respectively. Of course, the i7 PC will cost considerably less. So, the Mac would be performing the same as a PC that might cost 1/3 less.
Now, if you overclock that i7 PC, it'll perform significantly better than the Apple offerings without costing any more. In that sense, a quad 2009 Mac Pro won't do the test as fast as a Nehalem PC. And that'll be just as true once DP motherboards are released. Although, I think if this test has shown us anything, it's that Photoshop isn't using more than 4 cores. Evidenced by the fact that lots of people's 4-core Hacks and Macs perform the same or better than other people 8-core Hacks and Macs.
So, if Photoshop is your thing, you won't get any benefit in that application with more than 4 cores. Now if you want to run Photoshop while encoding something or while running another processor intensive application then you might have a good reason for going for a DP 8-core setup.
Oh nothing to do with the CPU, it'l be the memory that pulls the PC ahead. Same scenario as the 08 Octo cores vs LGA755 Quad PC's. Both used DDR2 (Or DDR3 in some PC's cases), but the slower buffered memory in the Mac Pro meant the Quad PC's could consistently pull faster results. It's been very surprising to see some of the Hackintoshes Cinebench scores too, both non and overclocked (Especially when you put the price tags next to each system). Getting harder and harder to justify OSX that's for sure!
Maybe there's a small difference at the same clocks due to faster memory, but I don't ever recall any benchmarks that indicated that, at the same clocks, the faster memory of a Hack was making that much of a difference. I'm specifically thinking of Geekbench and Cinebench.
Edit:
Check out these Geekbench results.
Hack Quad 2.66GHz
Mac Quad 2.66GHz
And awulf's Cinebench results pretty much tell the same story.
Cinebench
Hack Quad 2.66GHz: 1CPU = 3480; 4CPU = 14825
Mac Quad 2.66GHz: 1CPU = 3573; 4CPU = 14753
So, even though, in both cases, the Hack is using faster memory, the machines perform almost identically at the same clocks. That isn't to say that there are no situations in which the faster RAM might pay off more than what's being shown here. I just have yet to see something that indicates that the faster RAM makes as big a different as you're suggesting.
Sorry should have been more clear, I'm was talking about the benchmark this topic was created for, The Retouch Artists one.. If you scan back over the pages most of the Quad PC's were pulling circa 18sec times with the 2.8 Octos coming in around 22. It's not LOADS faster, but I just said it would be faster, which I still think it will be. I think this topic has become a little lost with all the different benchmarks being thrown into it!
The Cinebench results are quite amusing though!
I see. Are the 18 second results your talking about on machines running at 2.8GHz? I know I posted some 18 second results with my Hack quad but mine is running at 3.7GHz; almost a full GHz higher. Maybe I'll downclock my system to 2.8GHz and 800MHz on the RAM to see how it compares to the average 2.8GHz Pro.
Yeah, a OC'd 4-core i7 performs better in Cinebench than a 2.26GHz 8-core machine. That's impressive, even if a bit unsurprising.
Unsure as to the extent of the overclocking but most chips were Q6600s (What i'm still running now) which are only 2.4ghz at stock. I'm sure someone many pages back did a test essentially summarising zero gains for increased cpu speed over a certain barrier (most likely at the point the action was bottlenecked by memory speed/latency) or the memory controller, or any number of other thing. I know on my 790i system i had a Q6600 @ 3.6ghz with DDR31800mhz memory that couldn't do it any faster than my Q6600 @ 3.4ghz with DDR2850mhz memory. But then all results were compared to 8-core Mac Pro systems, so it's hard to equalise in that scenario (Could have 4 2.8ghz cores for photoshop whilst the other 4 handled core system for example, which may equate in real world speed to say, a 3.2ghz quad rather than 2.8). Now we have essentially the same systems with different memory in i7 PC and Mac Pro flavours it will be easier to remove the variables.
What do you get on your Hack out of interest?
The last result I posted in this thread was 18.8 seconds at history 1, cache 4, memory 100% at 3.7GHz.
Very interesting.. Reconfirms my desicion to stay with my Q6600 setup untill i7 has matured a little, doesn't seem like I would gain much real world performance moving. Good job, I was beggining to get a little tempted!
It seems 17/18seconds is about the floor limit for this test, I wonder why?
Yeah, I don't feel compelled to move off my Q6600 yet. I'm amazed at the performance I can get out of this dusty old thing.
Seems like it. Somebody did post a 15 second result in 64bit CS4 under Vista I believe; it was with a OC'd C2Q at ~4GHz I think. There's also a ~15 second result with an OC'd i7 on the last page too.
PS CS 3
Retouch artists :
History levels 1,Cache 4
MP 4x3.00
Raid10
CS3
10Gb
10.4.11
=27.8s
So,basically my 2 1/2years old MP still hangs out with the new bigboys?
Good investment...
But have to say...shame adobe...
Retouch artists :
History levels 1,Cache 4
MP 4x3.00
Raid10
CS3
10Gb
10.4.11
=27.8s
So,basically my 2 1/2years old MP still hangs out with the new bigboys?
Good investment...
But have to say...shame adobe...
That is comprehensively trounced by my Quadcore Hack too.
CINEBENCH R10
****************************************************
Tester :
Processor : Intel® Core2 Quad CPU Q9550 @ 2.83GHz
MHz :
Number of CPUs : 4
Operating System : WINDOWS 64 BIT 6.0.6001
Graphics Card : GeForce 8800 GS/PCI/SSE2
Resolution : 800x600
Color Depth : 8bit
****************************************************
Rendering (Single CPU): 5105 CB-CPU
Rendering (Multiple CPU): 18142 CB-CPU
Multiprocessor Speedup: 3.55
Shading (OpenGL Standard) : 6546 CB-GFX
Same result from the same basic machine equipped with 12GB and the Nvidia GT120. I didn't expect the different graphic card to have any effect on the result and I wanted to confirm that. I wasn't disappointed.
Just checking it out : did adobe promise some improvement to the performance with the gpu´s in CS4 vs. CS 3?
Remember some talk about it but havent seen any hard evidence.
Just curious.
Adobe has just been so full of ***** the last few years with the PS performance it is starting to be hillarous.
If I wouldnt be working with it daily,that is..
Edit: Marsil, CS 3 or CS 4?
I believe using the GPU as a co-processor is a Snow Leopard enhancement rather than an application level enhancement. Adobe is promising that the rewrite from Carbon to Cocoa will be accomplished with CS5 and that release will bring PS into architectural parity with the Windows version - i.e.: a 64 bit app.
I ran the benchmark using CS4.
Speedtest has a limit
Out of kicks, decided to humor everyone including myself, and realized that 22seconds seems to be the max. 18seconds if you reload the image and run the test again. I used a Mac Pro (2009) 2.93 8 core, 12gb ram, Photoshop CS4, with no other apps, at the settings requested. Repeated the test and each time from scratch got 22 seconds. Hope this helps, but seems this test is useless for the newer machines.
Peace,
Noushy
Out of kicks, decided to humor everyone including myself, and realized that 22seconds seems to be the max. 18seconds if you reload the image and run the test again. I used a Mac Pro (2009) 2.93 8 core, 12gb ram, Photoshop CS4, with no other apps, at the settings requested. Repeated the test and each time from scratch got 22 seconds. Hope this helps, but seems this test is useless for the newer machines.
Peace,
Noushy
22 seconds isn't the limit. I can do it faster than that in 32bit and 64bit versions of Photoshop.
Seems I was right about the speed of the new Mac Pro's though. So once again, much like the old 8-core, if you are buying a Mac Pro purely to run Photoshop, you're wasting your money.
Photoshop Speedtest
My point was that it does not run any faster in my 2.93 8 core than a 2.66 8 core. It can be due to CS4 not being 64bit, although that has a lot to do with memory allocation as well. I could not get photoshop to use more than 4gb of ram, and that has to do with the 32bit version. Lets see what it can do when Adobe releases CS5 running 64bit native. I heard that is going to be timed with the release of snow leopard. Should be interesting. I made this point only to show the lack of a difference between all of the fastest macs. This does not mean the new machines are no faster than the previous 8 core machines. Ran Geekbench 64bit on my Mac Pro 2008 2.8 8 core with 16gb ram, Quadro 5600 FX card, and received 9004. On my Mac Pro 2009 2.93 8 core running 12gb ram, Radeon 4870, received 17716. My 2008 Mac Pro has the WD 500gb drive, my 2009 has the WD 1tb drive. Other than that, running 10.5.6 on both (although build on Nehalem machine is newer).
Peace,
Noushy
My point was that it does not run any faster in my 2.93 8 core than a 2.66 8 core. It can be due to CS4 not being 64bit, although that has a lot to do with memory allocation as well. I could not get photoshop to use more than 4gb of ram, and that has to do with the 32bit version. Lets see what it can do when Adobe releases CS5 running 64bit native. I heard that is going to be timed with the release of snow leopard. Should be interesting. I made this point only to show the lack of a difference between all of the fastest macs. This does not mean the new machines are no faster than the previous 8 core machines. Ran Geekbench 64bit on my Mac Pro 2008 2.8 8 core with 16gb ram, Quadro 5600 FX card, and received 9004. On my Mac Pro 2009 2.93 8 core running 12gb ram, Radeon 4870, received 17716. My 2008 Mac Pro has the WD 500gb drive, my 2009 has the WD 1tb drive. Other than that, running 10.5.6 on both (although build on Nehalem machine is newer).
Peace,
Noushy
Where did you hear that? Usually Photoshop is on an 18 month development cycle which should place the next release between March and June of next year.
Jim
Uuh,duude..
It is from april 2008,and the "next" version they refer to is CS4..
So the über performance they rave about can be seen at works in CS4...hehhe..
.
So,it is just about to sit back and wait for the CS5 that is propably arriving in agust 2010. Maybe then again we will se whopping 5% boost in speed again.
OR,adobe might manage to engage a second core!!! That would double the performance of the photoshop!!! Imagine that!
http://blogs.adobe.com/jnack/2008/04/photoshop_lr_64.html said:
If you're in college, stay there. You need what they have to offer.