Table of Contents
Introduction
At Puget Systems have always held the belief that a computer's hardware should be specifically tailored to match what the computer will be used for in order to maximize its effectiveness. Different programs make use of hardware in different ways, and knowing how the program behaves is important to determine what pieces of hardware need to be more powerful, and which are not as critical. Most programs can be grouped into categories that have similar hardware requirements, but there are some programs that warrant a bit of extra in-depth examination to discover the fine nuances for that specific program.
Last September, we looked at how video cards affect Photoshop CS6 performance in our Adobe Photoshop CS6 GPU Acceleration article; and due to its popularity we wanted to expand on the topic of hardware performance in Photoshop to take a look at RAM. Specifically, in this article we want to determine how the speed and configuration of RAM affects the performance of Photoshop CS6.
What Affects Memory Performance
The main factor when it comes to RAM performance is simply the speed of the RAM. Obviously, faster RAM is indeed faster, but often the difference is so small that it does not have a measurable impact on system performance. We've looked at this in the past in our Breaking the Hype of High Frequency RAM article, but the focused of that article was mostly on gaming and basic computer usage and not specifically Photoshop. It is entirely possible that Photoshop CS6 can benefit from using higher frequency RAM, so that is one question we want to answer.
The configuration of the RAM is a bit more complicated as it is a multi-part question. The first part is whether it is better to have fewer, larger sticks of RAM or more, smaller sticks of RAM. More physical sticks of RAM allow you to utilize more RAM channels, but you lose some potential upgradeability. The secondary part of this question is if the size of RAM has an impact on performance. 8GB sticks of RAM are much more convienent to use, but if you get better performance out of 4GB sticks you may opt to use those smaller sticks. Usually you would want to use the fewest number of sticks to give you more room to add additional sticks in the future, but this completely changes if performance is at all affected.
The last thing we will cover does not relate directly to performance, but is the estimated amount of RAM you will want in your system. Performance can be drastically reduced if you do not have enough RAM, so when you are in the market for either a new computer or an upgrade, it is useful to have a reference to help you decide how much RAM is right for you. This heavily depends on the sizes and types of files you are working with, but we will be giving some estimations that you can use as a baseline.
Test Setup
To ensure consistent benchmark results, we created a series of “Actions” in Photoshop CS6 64-bit to apply the effects we chose to use for our benchmark run. Each effect was applied with its default options to an image of various sizes to determine if RAM has a greater or lesser effect on larger images. Most of our testing will be done with a 109MB and 250MB image, but we will also do some testing with images as large as 1024MB. To actually measure how long each effect took to apply, we simply enabled the integrated “Timing” feature in Photoshop CS6 which displays how long Photoshop took to perform an action. This method is based off of HardwareHeaven.com’s popular Photoshop Benchmark V3 benchmark, but adapted for our uses. The image we used is also from that benchmark, and was simply resized to achieve the larger image sizes.
Photoshop was left at its default settings with the RAM allocation set to 60% and GPU acceleration fully enabled. We ran each benchmark loop four times with the average between each run being our final result. Before changing the test image size (and when changing the installed RAM), the system was rebooted to ensure that there was no errant data being stored in the RAM.
The latest (13.0.1) version of Photoshop CS6 64-bit was used, and to see how the chipset affects our results we used two separate testing platforms consisting of the following hardware:
Testing Hardware | ||
Motherboard: | Asus P8Z77-V Pro | Asus P9X79 Deluxe |
CPU: | Intel Core i7 3770K 3.5GHz | Intel Core i7 3960X 3.3GHz |
PSU: | Antec HCP-1000 1000W Power Supply | |
GPU: | EVGA NVIDIA GTX 680 2GB | |
Hard Drive: | Intel 320 120GB SATA II 2.5inch SSD | |
OS: | Windows 7 Ultimate 64-bit |
For our test RAM, we used a variety of RAM at different frequencies. The only RAM that is not at a different frequency is the 8GB version of the Kingston HyperX Low Voltage RAM which we used in our memory configuration testing. For our RAM speed testing we will be using a single pair of RAM installed in dual channel mode.
Model | Frequency | Timing | Voltage |
Patriot Viper Xtreme (2x4GB) PC3-17000 Enhanced Latency Kit | 2133MHz | 11-11-11-30 | 1.65V |
Patriot Viper Xtreme (2x4GB) PC3-15000 Enhanced Latency Kit | 1866MHz | 9-11-9-27 | 1.65V |
Kingston HyperX DDR3-1600 8GB Low Voltage | 1600MHz | 9-9-9 | 1.35V |
Kingston HyperX DDR3-1600 4GB Low Voltage | 1600MHz | 9-9-9 | 1.35V |
Kingston DDR3-1333 4GB | 1333MHz | 11-11-11 | 1.5V |
Memory Speed Performance
To see how much of an impact the speed of the RAM has on Photoshop CS6, we simply ran our benchmark series each set of RAM (1333, 1600, 1866 and 2133 MHz) and compared them to each other. Half of the effects we are using utilize GPU acceleration, while the other half do not. So in addition to looking at the overall performance changes, we will also take a closer look at both of these types of effects individually.
As we stated in the test setup, we will also be using both a 109MB image as well as a 250MB version of our test image. This will let us see if any changes in performance become more pronounced as the image size is increased.
109MB Image
Starting with the Z77 test system, we saw a slight overall performance increase moving from 1333MHz to 1600MHz (about 1.3%), but very little performance increase after that. The X79 system on the other hand showed an increase in performance all the way up to 2133MHz. These performance gains are very small, however, with only about a .6% difference between the 1600MHz and 2133MHz RAM.
When we separate the GPU accelerated effects from the standard effects, we get some really interesting results. The GPU accelerated effects clearly show the advantages of using faster RAM, but the standard effects show virtually no difference at all. This is really good to know since it tells us that for medium-sized images (in this case 109MB) faster RAM only affects GPU accelerated effects. Even then, the difference is small (about 1-2%) as long as you are using RAM faster than 1333Mhz.
250MB Image
For both the Z77 and X79 test systems, we see almost exactly the same performance differences for the 250MB image that we saw previously for the 109MB image. The measured time difference is larger since the effects simply take longer to run on larger images, but the percentage difference is within a few tenths of a percent. This applies to not only the overall benchmark times, but the times for both the GPU Accelerated and Standard effects as well.
Our testing has shown that 1866 and 2133MHz RAM does not provide a significant increase in Photoshop CS6 performance, and since most Intel CPUs natively support RAM up to DDR3-1600MHz, that is the speed of RAM we would recommend using. RAM that operates at faster speeds than the CPU natively supports tend to have an increased risk of failure, and in our opinion the small performance improvements are not worth the risk.
Memory Configuration Performance
Testing how the memory configuration affects the performance of Photoshop CS6 is not as easy as testing how the speed of the RAM affects performance, but it is certainly possible. Unfortunately, the results are a little bit messy and require some effort to see exactly what is going on.
To test the memory configuration, we decided to focus on both the number of memory channels being used and size of each stick of RAM. By testing with two and four 4GB sticks of RAM in our Z77 system and two, four, six, and eights 4GB sticks of RAM in our X79 system, we can see if there is any advantage of using more or less memory channels. Also, by testing two and four 8GB sticks of RAM in both of the test systems, we can see if there is any advantage to using more 4GB sticks or fewer 8GB sticks to achieve the same total amount of memory in a system.
109MB Image
Starting with the Z77 system, we saw pretty much no change in performance moving from two to four sticks of RAM. With X79, however, there was a .6% increase in performance moving from two to four sticks, but only a .25% performance increase moving from both four to six sticks and six to eight sticks. Interestingly, this once again appears to mostly affect GPU accelerated effects and leaving the standard effects almost completely unchanged. So while the number of RAM sticks doesn't appear to matter for Z77 systems, X79 systems do get a small performance advantage by using four sticks of RAM versus two sticks. After that, the increase in performance is so small that it is not worth worrying about.
As for the size of the memory, our testing shows that there is no change in performance by using larger physical sticks of RAM. So if you are going for the absolute best performance possible for this size of images, you are slightly better off going with more 4GB sticks than fewer 8GB sticks. Whether a performance increase that is smaller than 1% on GPU accelerated effects is worth the more limited upgradability is something that each user will have to decide for themselves.
250MB Image
Upping the image size to 250MB shows similar results to the benchmarks with the 109MB image. The Z77 system in particular shows pretty much no difference going from two to four sticks of RAM. The results for the X79 system on the other hand, are slightly different. This time, moving from four to six sticks does not show any performance increase at all. Going up to eight sticks, however, does show a small .3% increase in performance. So for X79 systems, there is once again a small performance increase going with four sticks over two sticks, but after that the increase in performance is so small that it is not worth worrying about.
As for the size of the memory, our testing once again shows that there is no change in performance by using larger sticks of RAM. So if you are going for the absolute best performance possible, you are again slightly better off going with more 4GB sticks rather than fewer 8GB sticks. Again, a performance increase that is smaller than 1% is likely not worth the more limited upgradability, but that is something that each user will have to decide on their own.
Recommended Memory Size for Photoshop
Perhaps the most important aspect when it comes to memory for Photoshop CS6 is that you simply have enough in the first place. If Photoshop needs to use more RAM than there is available, you may suffer a huge performance hit and possibly even errors that stop your work completely. While the exact amount of RAM you will need depends on the images you are working with, we can provide you with an estimation to at least get you started.
To see how much RAM Photoshop CS6 uses for different sizes of images, we took our original benchmark image and simply resized it to 250MB, 500MB, 750MB and 1024MB. Then, while logging the total amount of RAM being used by the system, we ran the same benchmark series with four passes that we used in our previous testing. By installing 32GB of RAM in the system (which is more than Photoshop needed for even the 1024MB image), we can ensure that Photoshop has all the RAM it could possibly want.
Note that the data in the chart below has been compressed so that the benchmark loops for each image size are scaled to line up with one another. The larger images took much, much longer to benchmark, but scaling the results makes it much easier to compare the RAM usage between the different image sizes.
By default, Photoshop CS6 wants to allocate 60% of the system's RAM for itself. By taking the peak value and using this default allocation of 60% of the system's RAM, we can come up with the following recommended amount of RAM for each image size:
Peak Usage | Minimum RAM | Recommended RAM | |
109MB Image | 4778 MB | 6689 MB | 8GB |
250MB Image | 7836 MB | 10970 MB | 16GB |
500MB Image | 12304 MB | 17225 MB | 16GB |
750MB Image | 17589 MB | 24624 MB | 24GB |
1024MB Image | 22480 MB | 31472 MB | 32GB |
Keep in mind that this is for a single image with the default settings. While you will not run into too many performance issues if you are a bit under these amounts, if you get too far below you will likely see a huge drop in performance as Photoshop will have to write to the hard drive rather than the RAM. If you expect to have multiple images open, you will likely want to have a more RAM than our recommended amounts.
If more RAM is not an option, you can adjust the default settings a bit to make more efficient use of the RAM you already have. The easiest setting to change would be to turn up the allocated RAM amount above the default of 60%. However, if you do this you need to make sure you leave enough RAM free for the OS and other running programs or you will run into a multitude of other problems. If you want to see what else you can tweak in Photoshop CS6 to maximize performance, we recommend reading Adobe's How to Tune Photoshop CS6 for Peak Performance blog post.
Conclusion
While the speed and configuration of a computer's memory does make a small impact on Photoshop CS6 performance, there are likely many more important factors to consider before worrying about memory. The only time we saw a noticeable change in performance was when we used 1333MHz RAM and even then the difference between it and 1600MHz RAM was just over 1%. On top of that, we only observed this performance difference in GPU accelerated effects, which are only a fraction of the effects available in Photoshop.
Our testing shows that while the physical size of the RAM has no measurable impact on performance, using more memory channels did provide a very small performance boost. However, even that performance gain was under 1%, which is arguably small enough to be considered within a margin of error. It may seem like a waste to write a four page article just to say that it doesn't matter, but knowing for a fact that you do not have to worry about the speed of a system's RAM is very beneficial when configuring a system for Photoshop CS6.
From our results, we can conclude that for Photoshop CS6 the only major factor you should worry about in regards to memory is simply having enough RAM. This is going to vary from user to user, so it is important to have a good idea of how the system will be used and what sizes of images you will be working with before you decide on how much RAM you need. Most users will only need 8-16GB, but professionals who work with very large images may need 32GB or more to satisfy their needs. As for the speed of the RAM, we would suggest simply matching whatever speed your CPU natively supports (most likely 1600MHz). Faster RAM might give you a very slight performance increase, but in our opinion is not worth the additional risk of failure.
Puget Systems offers a range of powerful and reliable systems that are tailor-made for your unique workflow.