Sign in with
Sign up | Sign in
Your question
Solved

Why do photoshop run better with Intel than AMD?

Last response: in CPUs
Share
December 16, 2009 1:45:17 AM

Here's the benchmark I've found after some searching:

http://www.tomshardware.com/reviews/athlon-ii-propus,24...

As you can see, the performance of

AMD athlon X4 620 (4 core, 2.6Ghz) = Intel E8600 (2 core, 3.33Ghz)

AMD Phenom II X4 965 (4 core, 3.4Ghz) = Intel Q8200 (4 core, 2.3Ghz)

How is this possible? I know the athlon X4 lacks L3 cache but still it's shocking. Can anyone explain this result? Is photoshop optimized for Intel processors? Cuz I saw the benchmarks for Intel and AMD in gaming were pretty close...
December 16, 2009 2:18:22 AM

its all about architecture and app coding...and intel is faster clock for clock than AMD right now. Gaming benches will minimize the differences because they are more gpu dependent, whereas Adobe programs are cpu and memory dependent.
December 16, 2009 2:29:52 AM

student_sol said:
Here's the benchmark I've found after some searching:

http://www.tomshardware.com/reviews/athlon-ii-propus,24...

As you can see, the performance of

AMD athlon X4 620 (4 core, 2.6Ghz) = Intel E8600 (2 core, 3.33Ghz)

AMD Phenom II X4 965 (4 core, 3.4Ghz) = Intel Q8200 (4 core, 2.3Ghz)

How is this possible? I know the athlon X4 lacks L3 cache but still it's shocking. Can anyone explain this result? Is photoshop optimized for Intel processors? Cuz I saw the benchmarks for Intel and AMD in gaming were pretty close...


Intel always lead in PS like the way AMD leads in WINRAR
Related resources
December 16, 2009 2:55:45 AM

Thanks for the reply. I was surprised to see such a big difference in benchmarks, could it have something to do with their L1 cache policies? here's an interesting article I've found: http://www.tomshardware.com/reviews/athlon-l3-cache,241...

it says AMD has larger 2-way associative L1 cache with higher chance of misses whereas Intel has smaller 8-way associative L1 cache but with higher chance of hits.
a b à CPUs
December 16, 2009 3:09:57 AM

soshow2009 said:
Intel always lead in PS like the way AMD leads in WINRAR

errr

a c 83 à CPUs
a b À AMD
December 16, 2009 3:43:00 PM

I think it's due to the instruction sets, the SSE or whatever it's called, Intel usually has the newest version out before AMD. Even back in the P4 days Intel did really well with Photoshop while the Athlon 64 was able to win many other benchmarks.
December 16, 2009 10:46:45 PM

Is there any other application like photoshop that performs significantly differently with Intel and AMD processors?
a b à CPUs
December 17, 2009 12:17:19 AM

Quote:
With Photoshop it has nothing to do with Intel being faster clock for clock as per say because the Phenom II X4 965 is clocked 1.1 GHz higher than the Intel Q8200 yet they perform the same in Photoshop, shockingly clock for clock the Q8200 should be slower than a Phenom II due to the Q8200 being a "cut down" Core 2 with only 4 MBs of cache.

I think the only expanation in this situation is that Photoshop has been optimized to run better on an Intel architecture.


No, as belial2k said, Intel has a higher IPC (instructions executed per clock). Modern CPUs are superscalar, in that they can execute more than one instruction per clock cycle. Intel just happens to have 4 decoders vs. AMD's 3 (for K8 - K10.5 anyway, Bulldozer will have more). Plus Intel has better prefetch, Out of Order execution, and several other advantages in their Core2 and Nehalem designs, which basically means their CPUs do more in the same number of clock cycles than AMD CPUs do.

Of course, the above is dependent on the type of instructions being executed, and especially if they're SSE instructions.
a b à CPUs
December 17, 2009 12:22:53 AM

student_sol said:
Is there any other application like photoshop that performs significantly differently with Intel and AMD processors?


Have you looked at the THG CPU performance charts HERE ? Looks like the acrobat doc creation test significantly Intel CPUs, as does Fritz chess and a bunch others..
December 17, 2009 1:56:50 AM

Interesting. And thanks for the link fazers_on_stun :) 
December 17, 2009 9:04:57 AM

Come to think of it, Athlon II x4 620 costs less than E8600 and can be excused, but Phenom II x4 965 really should perform better than Q8200. Hope AMD will come up with new architecture that has improved performance in this area... I'm looking forward to how $1.25billion will make a difference next year.

Best solution

a c 97 à CPUs
December 17, 2009 12:16:26 PM
Share

student_sol said:
Here's the benchmark I've found after some searching:

http://www.tomshardware.com/reviews/athlon-ii-propus,24...

As you can see, the performance of

AMD athlon X4 620 (4 core, 2.6Ghz) = Intel E8600 (2 core, 3.33Ghz)

AMD Phenom II X4 965 (4 core, 3.4Ghz) = Intel Q8200 (4 core, 2.3Ghz)

How is this possible? I know the athlon X4 lacks L3 cache but still it's shocking. Can anyone explain this result? Is photoshop optimized for Intel processors? Cuz I saw the benchmarks for Intel and AMD in gaming were pretty close...


I would agree that Photoshop is optimized at the code and/or compiler level, based on how those CPUs stack up against each other in other applications. Photoshop is a proprietary application, so we can't see the code to see if the code itself might run better on Intel CPUs vs. AMD CPUs. I don't know what compiler Adobe used or what compiler flags they used, but those can greatly affect how even the same code performs on two different CPUs. I would bet that they probably used Intel's ICC as the compiler as ICC is a relatively common compiler and it tends to produce binaries that run much faster on Intel CPUs vs. AMD CPUs compared to using other compilers (MS VC++, GCC, Open64) to compile that same exact code. ICC also has a CPUID() function that can turn certain optimizations on and off at runtime depending on what CPUID processor/model/family flags are returned. Some specific programs do this as well. This was demonstrated pretty clearly with an evaluation of VIA's Nano processor with PCMark 2005. VIA allows people to change the CPUID fields, AMD and Intel do not. Ars ran PCMark 2005 on the Nano with the default CPUID flags (CentaurHauls) and got one score. They then changed the flags to GenuineIntel (Intel's flag) and the I/O scores improved significantly. For kicks, they tested the Nano with the AuthenticAMD flag (AMD's flag) and watched the score drop back to pretty much the original score with the native VIA flags.

If you really want to know how well particular CPUs stack up in raw performance, benchmark them with open-source code compiled with multiple compilers given known compile-time options. Then average the scores for the different compilers. This allows eventually isolate the CPU as the source of the performance difference, not the code or the compiler. If that is too much work, then benchmark on open-source code compiled with an open-source compiler such as GCC. That way you can at least see how the compiler would optimize for different CPUs and better tell what may or may not be happening during execution that if you used a closed-source ("black box") compiler such as ICC. This is the way that most benchmarking is done on the Unixes (notably Michael Larabel's Phoronix Test Suite) and the results for what certain CPUs can do can often be a bit different than what you see with the completely proprietary, black-box benchmarks and programs you see here and at many other review sites. Now don't get me wrong, there is some utility in doing proprietary benchmarks, particularly if that benchmark is a program that you use frequently like Photoshop. The Phenom II X4 965BE may be faster than the C2Q Q8200 in just about everything else other than Photoshop, but if you're mostly just running Photoshop, you'd want to know that the Q8200 is about as fast as the 965BE.
a c 172 à CPUs
a b å Intel
December 17, 2009 3:52:41 PM

student_sol said:
I'm looking forward to how $1.25billion will make a difference next year.

I suspect that most of the $1.25 billion will go to pay off debt.
December 17, 2009 8:36:38 PM

Wow that's a very technical explanation MU_engineer. Actually too technical for a ordinary guy like me :)  But I get your point, you're saying that the difference between Intel and AMD may not come from "raw performance" but from the type of compilers used for each processors? Hmm this is interesting stuff, I'll definitely do some research after my exams are over.


Quote:
I suspect that most of the $1.25 billion will go to pay off debt.


Oh no! Too bad :( 
!