GW2 and FX CPUs (eight cores).

It's well know that GW2 is a game that rely too much on the CPU, the reasons vary, but it has a lot to do with multithreading. Basically: it sucks at it.

Our FX CPUS are notables for the poor single-threaded performance and, for that reason, they are literally the worst modern "high end" CPU you can have to play GW2 as the game cannot utilize all of our cores properly.

But why is that? A FX CPU with eight cores shares each CU (controller unit) with two of them: Each CU controls two cores, thus limiting it's performance.

BUT. There's a hope for us, mere mortals:

Most motherboards (I have a Gigabyte with a 970 chipset) have a option to force each CU to operate fully in conjunction with only a single core (or any other ratio we tell it to).

On mine it's called "ONE processor per CU". Selecting it disables 4 cores on my FX-8320 but makes Each CU control only one of the remaining cores.

After doing that, the CPU usage in-game boosted from ~35% up to ~80% and resolved part of the bottleneck it was causing on my GPU (R9 280X), wich again have boosted it's usage from ~40% to ~75%. My framerate (right after creating a new character) got about 10 FPS higher while doing the starting zone, and even hit 60 in some areas. And the sudden frame drops reduced a little too!

That being said, modern games or programs that can utilize all our cores properly will suffer from that, a lot. Streaming, video rendering, compacting files etc..

Why does it does not work with my four-cores FX CPU? GW2 can make use of four cores already, reducing it's amount will probably kill it's performance even more.

If you have a six core CPU, you can try another ratio. Your goal is to disable two cores and have your three CUs controlling four cores. I don't know if it's possible as I only have a eight core CPU.

I hope it helps someone. Have a nice day, Tyrians.

It doesn't matter which MMO you talk about but pretty much all of them have some part that require sequential processing which stalls the rest of the threads. That is primarily single-threaded in nature which means instructions per cycle is important (not just MHz).

The problem is Bulldozer architecture is using clustered multi-threading (CMT) so it's technically 4 cores not 8. There's actually a lawsuit going on about this , claiming it was misleading due to 8 integer cores but only 4 FPU. As a Phenom II CPU (as well as Ryzen) owner I passed over the Bulldozer generation (Vishera included) in favor of Sandy Bridge from Intel. I'm pretty sure there's a Windows patch for Bulldozer to improve performance in general for Bulldozer.

Ryzen 1st gen upgraded IPC by 50% and 2nd gen is very competitive in lightly threaded workloads due to the improved boost algorithm, so it's the recommended CPU if you are buying AMD CPUs and by all indications the successor Ryzen 3rd gen will be competitive with i9-9900k/i7-9700k with 30% less power too. A 2nd gen Ryzen 5 or Ryzen 7 is competitive with Intel's i5 and i7 CPUs and if you want to optimize performance you can lock it to one CCX (Ryzen suffers latency when shifting threads to second CCX) using processlasso or Windows' core affinity.

For anyone that doesn't already have a system, an unlocked i5 8th or 9th gen CPU (i5-8600k,i5-9600k), unlocked i7 7th through 9th gen CPU (i7-7700k, i7-8700k, i7-9700k), Ryzen 5 or 7 CPU are good bets. Naturally Ryzen & 9th gen i5/i7 Intel CPUs require less handywork with delidding to allow for lower temperatures and most motherboard manufacturers enable MCE (multicore enhancement) making all cores boost to turbo if you're too lazy to overclock manually. There's no reason to get a quad core CPU in 2019 unless it's a thin 15W laptop of some sort.

As far as CPU intensive settings go, you will want to turn down reflections, shadows, and model limit.

Would this be possible to solve by setting affinity?

This shouldn't be necessary at all if your Windows has been hotfixed per Bitsum (makers of the ProcessLasso tool)

Case 4: Intel HyperThreading and AMD Bulldozer

Another important issue is avoiding placing a load on a HyperThreaded (non-physical) core. These cores offer a small fraction of the performance of a real core. The Windows scheduler is aware of this and will swap to them only if needed. As of mid Jan 2012 the Windows 7 and Windows 2008 R2 schedulers have a hotfix for AMD Bulldozer CPUs that see them as HyperThreaded, cutting them down from 8 physical cores to 4 physical cores, 8 logical cores. This is for two reasons: The AMD Bulldozer platform uses pairs of cores called Bulldozer Modules. Each pair shares some computation units, such as an L2 cache and FPU. To spread out the load and prevent too much load being placed on two cores that have shared computational units, the Windows patch was released, boosting performance in lightly threaded scenarios.

When I say Bulldozer+, I refer to PileDriver, SteamRoller, and all later incarnations we’ve seen, or are rumored to be in the works.

Some types of calculations that the game needs and does by the CPU are not parallelizable, and therefore non-threadable. That you have 30 cores or 1, some type of these calculations will be done only by one unit, and that's what kind of calculation that is a part of the bottleneck