Page 3 of 3

Re: Upcoming new hardware.

PostPosted: Wed Sep 03, 2014 11:11 pm
by pam
The a10 7850k has a powerful graphics chip fused with the CPU. Its equal to the Radeon HD 7750. You will require dual channel DDR3 memory from 2133Mhz to 2400Mhz. The RAM acts as graphics memory and is UMA. UMA will use as much graphics memory depending on how much RAM is available. The a10 7850k can drive 4k displays. You dont need an external graphics card and you also get ITX motherboards which support upto 16gb DDR3 2400Mhz.

The biggest feature of the A10 is HSA. HSA is not yet implemented and will get mainlined in kernel 3.17. HSA is a GPGPU solution from AMD. HSA will make the GPU and CPU work together to solve tasks. Unlike NVIDIA CUDA which only uses the graphics card, HSA will make the APU share cache/memory and data. Programs like Blender/gimp will benefit the most.

But if you really want muscle, then the FX 8320 is the best and can be safely overclocked to 4.5Ghz for everyday and heavy tasks, also. The 8 core FX chips have 16MB total cache. You will need a graphics card to drive displays. There are only 2 models of micro-ATX boards for the FX cpus but you cannot overclock them on any of these boards. If you are looking at 6 years into the future, then you will need ATX boards...

The FX 8320 is 32nm and uses more power than the A10 7850k(28nm).

You can also take a look at the newly released AMD processors... ... s-too.html ... 2sep&num=1

So there is a price cut for AMD 8 core processors. AMD has realized that 6/4core chips made by them are no longer a viable purchase option.

If you are a CPU junkie, get the FX, if not then get the A10.

All in all, this is the 'dying' phase for Intel and AMD in making processors and marks the end of an era. Both of them are focussed on lower power, high performance parts. The future is intelligent multicore computing. Intel has plans for making Atom chips as powerful as an AMD x6 1090t. Wireless AC will also be included on motherboards when DDR4 becomes mainstream at the end of 2016. There will also be PCie lanes dedicated for storage as current SATA 3.0 is too slow. HDD's wont last another 2 years....this is not a prediction. Seagate has spent heavily on R&D for SSD's.

Re: Upcoming new hardware.

PostPosted: Wed Sep 03, 2014 11:58 pm
by pam
Good choice!
550ti is plenty powerful.

Re: Upcoming new hardware.

PostPosted: Fri Sep 19, 2014 11:37 pm
by pam
Nvidia Maxwell GTX 970 released...

Re: Upcoming new hardware.

PostPosted: Thu May 21, 2015 2:46 am
by pam
AMD is readying to release its HBM Graphics cards.
HBM is a successor to GDDR5 and even Nvidia is working on it with its Pascal architecture for a 2016 release. ... -explained ... nd_memory/

GDDR5 is finally hitting a wall. It had to happen someday and that day has arrived. Technology has finally hit a wall of physical limitations.
Intel banks heavily on the materialistic aspect of business and customer service from both the enterprise and home users, which is very
good in getting products to the consumers but nature pays a heavy price.

There are many alternative and ready to deploy solutions that can make money for large businesses. AMD and ARM(and its partners) are at the forefront.

What will happen if the next batch of consumer devices sport 3GB DDR4 with intelligent octacore chips. Intel probably best knows
the myriad permutation and combinations. One being the desktop and laptop computer become obsolete. Game making suddenly become
an apex business model as minimal hardware performance far exceeds mediocrity.....ENOUGH OF HARDWARE!

HBM offers high bandwidth to drive 4k and 8k panels. Its a solution to a problem not a comparative increment to the last phasing technology.
Most people think that the processor determines the performance of a system but what if your interconnects/networking/storage are faster
than your age servers are already experiencing this. HBM will catapult graphics and compute performance.

Surely and steadily single core performance will be joke. The new quad core Raspberry pi are aimed to prevent reinvention of the wheel
by universities/institutes globally and adresses real world issues like building multicore code and algorithms... which work unconventionally.

Apart from photonic and quantum computers, when they take centrestage, by then we will have algorithms and software that build better
software the way machines build machines today.