- Processor (or clock) speed (GHz) is THE most important factor for faster calculations! The faster the processor, the faster the calculations. Most processors have two speeds: the stated clock speed and the "turbo" or maximum speed. AGi32 will utilize everything the processor has available (see Windows Power Management).
- Physical core count is a secondary consideration but can also be very important. Physical core count is the actual number of cores on the processor. Many Intel processors use a proprietary technology called hyper-threading to "virtually" double the number of cores. While hyper-threading may benefit some programs, it is the physical core count that matters for AGi32. A processor with 8-10 physical cores is a good choice. A higher core count might be more beneficial if the models being calculated are extremely heavy with calculation points (though a faster processor speed will also help with this). Processor speeds tend to drop when the core-count goes up (8-10 cores seems to be the sweet spot). There is only so much heat that can be dissipated from a processor. Since cores and speed produce heat, either cores or speed needs to go down as the other is increased.
- Cache sizes can also affect processing speed. Larger cache sizes will help speed up calculation times. This is the third thing to consider after speed and cores. If it comes down to two "like processors" but one has larger cache sizes, go with the processor with larger caches.
- Multiple processors may or may not be of benefit and are extremely expensive. The reason one might go with a multi-processor machine is that it provides a way to keep the processor speed high but also double-up on the physical core count. It is important to mention that multiple processors other than Xeon processors will be treated as separate processors and will not impact the AGi32 calculations. Xeon processors can be treated as a single processor if core count is a concern. Core count really only benefits AGi32 when calculating the calculation points. Thus, unless one typically works with enormous roadway or site models with tens of thousands of calculation points, multiple Xeon processors may not make any difference aside from lightening the wallet.
- Intel versus AMD. AGi32 and specifically the AGi32 rendering calculation engine, uses C++ code which has been optimized to run with Intel processors. This means an Intel processor will calculate an AGi32 model more efficiently and faster than an equivalent AMD processor. This may not make much difference with smaller AGi32 models but it can add up with larger files.
- Windows Power Management settings can also be a factor. To get the most out of the processor, make sure the power settings are set to High or Ultimate Performance. In both cases, make sure both the "Minimum processor state" and "Maximum processor state" are set to "100%". This will ensure that the processor runs at full capacity from the start to the end of the AGi32 calculation process. The image below shows these settings.
AGi32 is 32-bit application that can only address 2GB of RAM. This makes RAM a lesser consideration but is still important especially given all of the background Windows processes. 8GB is generally sufficient, 16GB to 32GB is recommended.
A standalone graphics card is not necessary for AGi32 as graphics and rendering information can be processed by the graphics chipset onboard the motherboard. A graphics card can, however, still be beneficial when in Render Mode. Some drivers used with high-end graphics cards can have problems with the AGi32 graphics information which can lead to rendered "artifacts" remaining on the screen and even an inability to enter Render Mode. In cases like this, if updating the graphics driver does fix the problems it may be necessary to disable Hardware Acceleration for Render Mode (a setting in AGi32) and force the graphics to be handled by the software. This can negate any benefits that might otherwise be gained by a high-end graphics card (the money spent on the graphics card would have been better spent on a better processor). However, if other programs benefit from a high-end graphics card, it would still make sense to consider this when purchasing a machine. In our experience, NVIDIA graphics card drivers also tend to encounter fewer over-all problems than AMD/ATI drivers and although NVIDIA cards are slightly more expensive, it may be an extra expense that is well-spent.
Hard drives are not going to impact the speed of the actual calculation process other than the time it takes to read data to calculate and write calculated data. They will also impact the time it takes to load and open all programs (including boot time). In AGi32, hard disk speed directly affects the time to open, edit and save files as well as how long it takes to initiate and complete the calculation process. Solid State Drive (SSD) devices will perform drive-related activities faster than standard hard drives. SSD technology and pricing is constantly changing. Be sure to do some research before purchasing an SSD to make sure you are getting the best bang for the buck.
Docking Stations for Laptops
There can be hardware conflicts between the docking station and the laptop, particularly when it comes to video drivers, which can cause AGi32 and potentially other programs not to function correctly. When such problems arise, the only solution can be to remove the docking station from the configuration. Docking stations are not recommended.
Everything comes down to performance versus cost. When on a budget, this can be quite a balancing act but as the information above makes clear, the focus should be on the processor when it comes to AGi32 and everything else is secondary. If other programs require or benefit from more RAM and/or a better graphics card then those, too, need to be weighed alongside the processor. As of the writing of this article, the Intel Xeon Gold processor running at a base speed of 3.5Ghz with a "turbo" speed of 4.5Ghz and 8 cores is about as good as it gets (at least today). Of course, if budget is not an issue, a dual-processor machine using this same processor would even be better!