Okay say I go FX8350

So which motherboard do you guys recommend?

I want to be able to sli or crossfire and I might do some overclocking when I learn how.

Asus M5A99FX Pro R2.0

Asrock 990FX Killer

Gigabyte 990fxa-ud3

Thank you

Sabertooth 990fx R2.0

High budget: Asus Sabertooth 990fx r2

Low budget (still good quality): Asrock 990FX Killer or asrock extreme4

As a side note, if you are looking to SLI or crossfire, then why not an i5-4670k? it outperforms the 8350 for a similar price.

I'm actually still deciding what I want I'm looking at i5 4670k paired with an Asus z87-a or an 8320\8350 with Asus m5a99fx pro r2.0 

I have Asrock 990FX Killer and some of the utilities on it are very useful i am very happy with mine and recommend it .

I've got a MSI GD80 V2, and I really like it. Supports the 220w CPU's as well for down the road.



if things were coded for the architecture it wouldn't. dx12 is coded for the multi threaded apu onboard the xbox one, so we may see that carry over to PC. Lets put it this way the 8350 has the potential to out perform every cpu at its price point, but the hardware isn't being utilized correctly.

I have the i5-3570k with an asrock extreme4 and I am very happy with the quality of the board and the features. I got a stable 4.4 GHZ overclock without bumping up the voltage at all. I got up to 4.6 on the smallest increment of voltage. My friend had similar results on his extreme 3.

Highly recommend the asrock motherboards for budget and low overclock builds. I like asus for extreme overclocking and for those that want the premium.

I would recommend one of these two options:




The sabertooth board is more premium, but costs a lot more. The asrock boards have great quality as well as software included.

The 4670k is superior to the fx 8350 in an architecture view.


Haswell is the one currently not been utilized enough.

An i5 4670k with an Asus or Asrock board is the better option overall than the 8350 imo :)

I agree with AMD and Nvidia, the way we are doing things right now is at a dead end. AMD and Nvidia are trying things that intel seems to be late to the party on.


Anyways http://www.phoronix.com/scan.php?page=article&item=intel_4770k_linux&num=1 the 8350 at stock goes toe to toe with the 4770k (year newer and $140 more expensive) in applications and in an os that take advantage of the hardware.

Okay, I'll break it down to you, as that benchmark doesn't take any kind of advantage of the hardware and supported ISA.


First lets see priority;

The software itself, needs to be coded for a certain amount of cores (If multi-threaded ofc), some task (like rendering and that kind), lose little to nothing splitting it over more cores, as it's doing the same work over and over again, and latency isn't an issue.

Then there is the kernel, the kernel can split up the different processes, to the different cores. Most kernels prefer single but stronger cores, because working on cores with active registers is always in favor, and latency will become an non-issue.

Then there is the actually core, this is the ultimatum.


Lets break it down to what is inside each cores backend;

Piledriver CMT core (What AMD calls a module);

4xALUs (2xALUs per core)(this is doing the integer work)

4x128-bit SIMD (this is shared between the two cores in the module)(This is doing the SIMD work)


Haswell core;




What is most effective?

Clearly haswell is easy to work with, example;

It is easier to schedule work to 4 stronger workers than 8 weakers workers.


In the end; Well organized multi-threaded supported heavy predictive integer instruction stream should be around the same for haswell and piledriver. (Piledriver might take the lead due to higher clockspeed, but shouldn't be huge, as there are a ton of other things pulling it down).

But in SIMD heavy work, haswell is totally crushing piledriver, in 9 out of 10 ways, it isn't even funny.

Have an unpredictive workload? Piledrivers weak branch predictor, will certainly have a hard time predicting for both cores, and will end with a branch miss.



Haswell i5 is superior, I can go way more indepth if that is what you wish.



What is it you are agreeing with AMD and Nvidia?

What is it Intel just joined?

Where do you even learn this stuff?

A combination of a ton of research, asking stupid question and been interesting in how computing actually works.


And alot of it was an requirement for my current job.


I would like to say, it is MUCH easier getting information about AMDs processor than Intel.


Anandtech to make some rather informative post on every architecture update.



Can you give examples of SIMD heavy workloads? Or applications that are being in development that use SIMD?

SIMD is basically floating point, which can be very effective (an example) on some game physics.

Are more heavy SIMD-workload you would be looking at AVX or AVX2 workload.


Note AVX2 actually implemented support for 256bit integer-arrays, which is an heavy upgrade from the standard SSE4.X


There are not many application using only SIMD workload, but some may have critical workload (where latency generally is the main issue, so it need to be executed as fast as possible, this is certainly a critical point with game physics).

Thanks, I was trying to gauge its usefulness for me. For me, it's not worth the price difference, nor using an Intel CPU.

AMD and Nvidia both agree that we are quickly reaching a dead end on the way we typically do processing. The Intel cpus and even the AMD cpus (or at least the way we think of them) will be a thing of the past and new better ways of computing are going to take their place.