Samsung begins mass production of 1GB GDDR5 chips #SDC

Going from 16 512mb to 8 1gb would mean games coded under former wouldn't work.......no?

Doesn't change anything other than the number of space taken up by RAM in the box. It's just like a chip die shrink but for memory.

Thats not what i mean. Ill try to simplify:

Right now, games are coded for the 16-512MB modules, so changing.......reducing.........the number of modules is like changing hardware no?

Console games arent coded like PC games, the hardware is fixed.

I would think that.........without heavy patches...........that older games would be presented with glitches, crashes, bugs, etc..........associated with the change in RAM modules.

If i programmed my game specifically for an AMD processor then switched to Nvidia, there would be problems...........no?

Or programmed my game expecting a certain bandwitdth.....and certain setup..........the same applies (with fixed hardware).

Bingo

So cheapers PS4 coming late 2015..

329.99 or 299.99 sounds like the ideal price for a Indie/Bloodborne machine..

I was reading the OP, trying to figure out where TC was going with it. Got to the end and proceeded to ROFLMMFAO!

No.. They can change to 8 1 GB modules. They can switch to a single 8GB module. It doesn't make a difference on the developer level. Memory is addressed by the processor all the same

read some about compilers, API's and memory model.

When you code for CPU AMD or Nvidia people usually do so on a high level language like C. If you want to get closer you use something like Assembler with the x86's instruction set on 32 or 64 bit, and if you are crazy then on binary. TLOU has code optimized on a binary level, you can't get closer than that since you're literally talking to the HW face to face.

But what is the purpose of compilers? It takes high level code like C (mostly words) or slightly lower code like Assembler (mostly hexadecimals with some letters) and turns it into machine language. High-level language of-course uses the most resources due to the overhead caused by the compiling process and so on.

When you code for GPU you usually do it with DX or openGL through a high level language too. On college I coded openGL (my prof hates MS with a passion) on C++ and I made some snowmen with specular lighting and snow falling. Then a planetary system with rotation and lighting changes. All done manually at low level with my own calculations, no engine used with pre-baked stuff. I ran this code on Nvidia GPU, AMD GPU and Intel IGC and there was no problem at all because well... That is the purpose of the API. Communication.

------

Now talking about that memory model... When you optimize for memory you're optimizing for bandwidth not HW since you can only talk directly to the CPU, GPU or I/O. So reducing memory modules for ones that offer higher capacity will go unnoticed by the code.

But when it comes to performance, the code will still run well regardless, I have read on many forums that 2 sticks of 512MB gives better performance than 1GB stick because of dual channel mode. But then again, when these consoles are redesigned they change stuff, so what prevents them from using slightly wider conductors in favor of cheaper manufacturing cost of the GDDR5 memory if the trade off is worth it?

------

Samsung? You mean that Korean company that basically ended Sony's Electronic division?

It's kinda like how Sega works for Nintendo now.

Well actually if they used just 1 stick that would cost some performance, since memory is faster in dual and quad channel configurations. But since it will have 8 modules it wouldn't hamper anything.

The Xbox one already uses quad channel memory so they can't reduce the amount of modules (4x 2gb modules) without lowering bandwidth. Not like they need to since DDR3 is so cheap.