Why are amd proccesors so cheap?

Solution
There are other differences, such as the more expensive i7 5960x running on 2011v3 platform (x99) which has support for larger amounts of memory, ddr4, increased 40 pcie lanes etc on top of the cpu architecture differences. It's hard to compare single specs from different platforms since multiple parts of the environment play together to provide different options.

Another reason for developers/programmers not opting to jump on the bandwagon to rewrite their software to take full advantage of gpgpu parallel processing is it's a major headache. For the same reason when 64bit computing was being introduced, even today many programs are still 32bit. It's not as easy as just saving their program in a different format, it has to be...

BrandonYoung

Reputable
Oct 13, 2014
1,114
1
5,960
Simply, because AMD is struggling to compete with the much faster, much more efficient Intel CPU's. However, "Normal" Intel CPU's are no where near that price. $180-$350 is the "Normal" range for Intel (i5/i7) CPU's unless you are going ultra-premium, or server based.
 

njxc500

Distinguished
Apr 30, 2008
181
0
18,710
First of all, it always helps to get the facts correct when you post a serious question like this....

AMD FX-9590 4.0Ghz $239
Core i7 5960X 3.0 Ghz $1,049

But you can't really compare these two....if you refer to the heirarchy chart in the gaming cpu review....

---------NOTE please don't burn me for this, I'm only pointing out the flaw in the question------------------

If you look in the heirarchy chart, Tom's expects comparable gaming performance from an AMD FX-9590 and Core i3 4370

So if you want to buy a CPU with comparable performance -

Core i3 4370 - $160
AMD FX-9590 - $240

The point is.....there is much more to consider than price-Ghz-brand.
 

Intel999

Honorable
Jan 13, 2015
17
2
10,525
AMD assumed four years ago that the software industry would migrate to distributing the workload more efficiently by utilizing the GPU instead of pushing everything through the CPU. Doing this allows the GPU to process parallel data much more efficiently than having the CPU perform the same tasks. At the same time, the CPU's burden is reduced and it has the ability to concentrate on tasks it performs best on.

However, even though by taking this approach would have had impressive gains on most software, the software industry, for the most part, opted to still develop their wares by pushing everything through the CPU. They most likely stayed the course because Intel was and still is the most dominant processor used around the world. Intel has a very strong single threaded core that can process data in a sufficient manner but in many cases not the optimal manner.

So Intel has had market power and pricing power thanks to the software industry designing towards their CPU architecture.

However, more and more software firms have just started to move toward multi-core GPU intensive processing. The most obvious being DirectX 12 by Microsoft. The beta version of this pending release shows CPU usage dropping dramatically.

So by 2017when all game releases will be launched using DirectX 12 it would be very possible to use a very weak CPU aligned with a top tier GPU to max out all future games.

The GPU will be the only factor in determining what level you want to game at. Even more so than now. And the same migration will happen in mainstream software, as well, only at a slower pace.

So, over the next few years, Intel will finally start a serious effort to get their GPUs up to industry standards which are currently set by AMD on the integrated GPU front.

Due to the lessened need for powerful CPU cores on IPUs you will see more die space being allocated to GPUs which will lead to IPUs having upper mid range gaming capabilities. The discrete GPU market for PCs may shrink to only the very high end. However, the ability to crossfire a dedicated mid range GPU with the integrated GPU will give you high end gaming with a very modest budget.

The GPU will migrate to the server market as it is next frontier to benefit from a GPU intensive approach.
 
There are other differences, such as the more expensive i7 5960x running on 2011v3 platform (x99) which has support for larger amounts of memory, ddr4, increased 40 pcie lanes etc on top of the cpu architecture differences. It's hard to compare single specs from different platforms since multiple parts of the environment play together to provide different options.

Another reason for developers/programmers not opting to jump on the bandwagon to rewrite their software to take full advantage of gpgpu parallel processing is it's a major headache. For the same reason when 64bit computing was being introduced, even today many programs are still 32bit. It's not as easy as just saving their program in a different format, it has to be rewritten. Some programs won't really benefit from 64bit, even the ones that do - thousands upon thousands of lines of code have to be rewritten from the ground up. It's much easier just to tweak what already exists which for the most part is 32bit. Same with programs or games being more cpu centric. It's easier to keep the base framework for a game or the same engine and build a new title on top of it then go from scratch. Not to say it won't happen, moving to more gpu based processing for games and other applications - but it won't happen overnight either.

Imagine trying to use a whole new type of fuel rather than gasoline/petrol (depending where you live). Much easier to build a new car body over the top of an internal combustion engine with an entire framework of petroleum based fueling stations already in place then to tear it all out and do it again from scratch.
 
Solution

BrandonYoung

Reputable
Oct 13, 2014
1,114
1
5,960
Nice post Intel999. Modern OS's (and compilers) should adopt the GPU as a computational device. This would require some work on the GPU side as their drivers are proprietary. But could boost performance vastly (100x) in systems that include a dedicated GPU on applications that are highly parallelable.

CPU for high power single/dual/quad core (threaded) applications, with scheduleable GPU for highly parallel data computation.

I hope we see this shift in the near future. Many apps could greatly benefit from it.