Hi !
I know this question is very outside the box but I think you guys can answer it.
My old cpu is very weak and during long compilation my pc freezes.
8 Core FX AMD CPU for 200$ or 4 core I5 intel CPU for the same price.
And before you say (optimize your maps), I do, and I know how to.
Sincerely, Doctor_GLaDOS

Since those compilers are using multiple threads, the 8 core CPU would be theoretically better. But I recommend to get the 4 core Intel, since this will give you overall a better performance for other progams, too. Intel has far better CPUs (it's hard to see from the outside or technical data, it's a question of the internal structure) and most programs work better on those.

Ye, I'm with Skotty. Also choose ALWAYS Intel before AMD (I'm running an i7-3770K).

Maybe Intel out-perform AMD but with 8 core CPU I could allocate 6 or 7 cores to the compilator and 1 or 2 to watch youtube videos without 5 minute freeze.
Right now I have intel core i7 920 so any CPU for me would be an improvement.
Also for "Other programs" argument you probably mean games and yes, right now Intel blows AMD out of the water, but when DirectX 12 and Vulcan and all that good stuff that supports multi-threading, don't you think 8 cores will win ?
But thats topic not meant for thinking with portals community.

Doctor_GLaDOS wrote:
[...] but when DirectX 12 and Vulcan and all that good stuff that supports multi-threading, don't you think 8 cores will win ?

I wonder what this changes for CPUs? DX12 and Vulcan are APIs for your GPU. And if they need much CPU power as in Minecraft, then your program is just programmed very bad.

Btw: Usually Intel cores support so called Hyperthreading. This means a 4 core CPU could become a theoretical 8 core CPU. But please don't ask me if they really increased speed. And if you want to do things while compiling, just add the -low parameter to the compiler. This makes it run automatically on low priority and you can do whatever you want in the meantime.
In the end, if you take a 4 core Intel or 8 core AMD won't change that much. The compile time will be usually (for Portal 2 maps) under 2 min if you compile full. And besides compiling, 8 cores would be just wasted. Nearly no program ever uses that.

For the same prize, AMD was always far faster than Intel (but also eats more power). However after a certain point there are no better AMD processors (in speed) but Intel still have faster and faster ones. So Ye, the 8 core AMD is better than the same prized Intel, but its better to buy a faster Intel. i7 only 5-10% faster then the same speed i5, but I dont know how the i7's hyperthreading works on map compiling.

Skotty wrote:
Btw: Usually Intel cores support so called Hyperthreading. This means a 4 core CPU could become a theoretical 8 core CPU. But please don't ask me if they really increased speed.

Ye its 8 thread and not 4, but its still 4 CPU core that working so I guess since it just virtually splits performance to 8 thread, just a minor speed increase fpr more $. so if a single core does 2 virtual thread, then it can only do it at half speed on each thread. ON processes where lot of thread needed, and with only 4 thread the process just waits for the threads to finish to start another calculation, it can speed up things, but with vrad all my 4 core runs at 100% load, so I guess with an i7 hyperthreading it would be the same, and the extra technology involved could just resultin a minor speed increase at the same GHz.
Of course if money is not a problem then the fastest i7 overclocked is the best!

Personally I have an i5-4690k 4core (overclockable) and with an average prized (not cheap) cooler it runs at 4.5GHz (0.8GHz idle, with turbo boost and energy saving). MY maps are huge, I only stop mapping when I reach hammer's and the compiler's limits and they starting to give me errors (too many brush, too big spaces) and lot of detail but even my biggest maps are compile in 5 minute with full compile and static prop lighting (but no HDR). If a map compiles for more than 10 min then its bad mapping (unless its the prop lighting that takes the most time, that cant be optimized if there are a lot of prop).

Even my old 6 year old DDR2 LGA socketed computer could compile anything in 10-12 min, so... ye, a good cpu can double or maxbe 4-5x or more compile times but a good optimization can do it as well

Hint: light compile (vrad) takes the most time. With huge surfaces/areas, if the surface for example has the same brightness (same shadow amount) then changing lightmap scale from 16 to 64, 128 or even 256 can boost the compile on that surface 100x times (and the lighting still look the same)! With wheatley chambers where the fog makes far walls not visible and the huge blocks have almost the same lighting on their sides, 256 lightmap size usually not noticable. Actually running the map and checking the lighting in Portal2 and then alt*tabbing to hammer and changing the lightmap scale on surfaces where the lighting is homogen or where it does not matter if its not sharp, can insanely speed up compile time! Along with good visleaf-optimizations of course...