-
Rule of Thumb for amount of ram for a given system
Often times editors are spending a significant budget on an editing workstation and do not account for the correct amount of ram such a system should have. The concept to remember now especially with GPU acceleration is the processing dataflow is a pipeline now and not just limited to the CPU. This means an editor can buy Dual 8 core Xeons to maximize the workflow experience but may use a only a percentage of that processing capability because hardware such as ram is to limited. The amount of ram required now to edit with the current GPU acceleration applications is massive compared to applications in the past. This is because data that has to transit to the GPU’s for processing must first be put into buffer in the system ram to transmit down to the GPU ram. This exponentially increases as frame size and data rate per frame increases. Once the data is processed at the GPU then it hass to transmit to the CPU for final encoding. Once again this entire pipeline transmits through the ram and often times such device buffering is not tracked by the OS as ram in use. Think of system ram as a Grand Central Station of Data. You have data going in any number of directions whether device specific or application data and it all centers around the system ram. To help editors configure their editing workstations so as to eliminate very common but easy to resolve bottlenecks causing wasted editing processing ability, I will give a few generalized rules here. Keep in mind workflow and media codec/resolutions used will alter this some but this is a good place to start.
1. Account for 1GB of ram per thread available on the system. This is often the most critical. having a 32 Thread system with 16GB or less of ram is seriously inhibiting the processing capability of the system. Applications that are not realtime priority will shift their entire processing data rate down to adjust for the ram available. This is what you want to avoid. Also remember many Intel CPU’s have Hyperthreading which means 2 threads per actual core. Still account for each thread whether physical or logical. A 8 Core Xeon chip is 16 actual threads. Higher end workflows that include 4K+ media need to account for 2GB per thread for ideal performance.
2. Add on a block of ram that is the same amount of ram as the Vram on your video card. If your video card has 2GB then make sure you have 2GB of system ram besides the 1GB per thread.
3. Add 2GB for standard OS load. When the operating system loads you normally use 1 to 2GB of ram for all of the processes and services. This obviously can be adjusted by startup configurations but that is what you plan for.
4. Multitasking add’s to your ram requirement. If you plan on utilizing more than 1 Adobe application at a time then add 4 to 6GB of ram to the total of 1 through 3. This can get even higher when 1 of the applications is AE since the Multithreading settings in AE will grab ram based on those settings and lock it down. However 6GB is a good place to start.
5. Optional – AE multithreading ram assignment can be adjusted between .75 to 4GB per thread. The greater this allotment is the faster the Ram preview will be. Keep this in mind if your workflow centers around AE versus Premiere.
Those are the general rules I would go by when looking at configuring an editing workstation that will give you the performance you expect. These rules obviously are subject to changes such as workflow and hardware config ie 2x GPU’s but as a general rule will keep the current editing systems working close to peak capability outside of Disk performance. There are many devices such as Raid controllers, Lan adapters, along with expansion devices that use ram caching for performance. Ensuring the system has more than enough ram increases the over all user and editing experience.