You can really just go for the single thread rating?
It is probably still the case, although not proven, that a small enough fort with little enough history, few enough items and dwarves, etc. is CPU bound, and the readily accessible measure that comes closest to representing DF performance in this case is the single thread rating.
In cases of larger forts, it is almost certainly I/O bound; the limiting factor is sustained (not burst) transfer from the main memory to the actual core (through various levels of cache, which don't help as much in this case). The single number that comes closest to representing this is probably the
transfer rate of your memory (as long as the motherboard and supporting chipsets support it). The speed of the memory directly affects this within a family, but can be difficult to compare between generations; you need to look at the transfer rate in bits or bytes per unit time.
In other programs I/O bound processes can also be dependent on networking or interconnect speed, which is irrelevant for DF, or on either burst or sustained hard drive transfer rates, which has an effect on startup speed for DF, but on a system with reasonable modern memory is largely irrelevant once everything has been loaded. (The DF program code and all of its data together, can't exceed 4 GB total due to 32-bit limitations; and is usually smaller, less than 2 GB. On a modern system with at least 3 GB of free memory, it should only need to read from disk once, and be fully resident in main memory thereafter.)
This is unlike almost all other gaming, and much more like a scientific workstation running older single-threaded code with a large data set for a simulation. Which, when you think about it carefully, isn't that surprising.
Much of modern computing is about multi-core performance, and much of that is about cache size and optimization. DF's data set is simply too large to fit in even an on-package cache, let alone the smaller on-die ones; so it's the transfer rate into the CPU package from the motherboard that limits everything.
My recommendation for a DF system varies somewhat as we find out more info on DF and technology changes, but is usually close to:
1) Figure out the best cooling system that you are willing to put up with (in the case of desktops), listen to, and/or carry (in the case of laptops). Laptop performance is almost entirely cooling dominated, and single-core sustained turbo ratings can be significantly different on a desktop with better cooling.
2) Pick a motherboard and memory that offers the best GB/sec sustained transfer rate from the main memory to the CPU that you can budget, and can be adequately cooled by the solution from step 1. You want at least 4 GB of memory, and preferably a bit more; more than 8 GB isn't going to help current-generation DF any, but might someday.
3) Pick the best current-generation CPU for single-thread performance under sustained load, that has at least two real cores, while cooled by 1), that is compatible with the motherboard and memory from 2), and that fits your budget. (While DF is single core, you want at least one more, so that the OS and so on doesn't cut into DF's time.)
4) Give it a good power supply with a high efficiency rating and enough headroom to handle everything cleanly.
5) Accessorize with everything else you need; most of the details are of low importance for DF's performance, although considerations like input devices and monitors can be important for the user experience.
Author
Topic: Intel vs AMD, or: share your PC and framerate! (Read 8687 times)