Ante Meridiem

I know some of you are upgrade fashionistas so this article may concern you.

I read the whole article... but what does this mean for GPUs, in english?

Interesting read, but as far as true upgrades go for us comp nuts, true GPU's and CPU's will not be passed in power by a APU. APU's will become extremely useful in smaller form factors, and maybe would be considered nice for playing WoW on a machine built on the cheap since you wouldn't have to buy a videocard.

meh... it took them decades to overcome the gpu bottleneck issues and it's still not perfected.... the fix a cpu/gpu bottleneck is a hell of a lot more difficult especially in it's smaller form factor.

Cremate wrote:I read the whole article... but what does this mean for GPUs, in english?

Think of it as real estate Cremate. Intel and AMD are developers. Intel filled their "Space" with L2 and L3 cache. Where as AMD put a GPU in the space they could have put more memory.

The way L2 and L3 work is this, speed and performance slow the further away from the cpu you get, increased transit distance = delay. L2 is faster than L3, L3 is faster than system memory.

Placing the GPU on the same chip as the CPU in theory is a great idea, but doing so now will constrain both of their abilities. Especially since they'd be sharing memory space.

Shezapriesta wrote:

Cremate wrote:I read the whole article... but what does this mean for GPUs, in english?

Think of it as real estate Cremate. Intel and AMD are developers. Intel filled their "Space" with L2 and L3 cache. Where as AMD put a GPU in the space they could have put more memory.

The way L2 and L3 work is this, speed and performance slow the further away from the cpu you get, increased transit distance = delay. L2 is faster than L3, L3 is faster than system memory.

Placing the GPU on the same chip as the CPU in theory is a great idea, but doing so now will constrain both of their abilities. Especially since they'd be sharing memory space.

Unless they add back L2 or L3 memory. As the article stated it wouldn't take much 256mb and this practice would be a boon for computing. Thus leaving NViDIA in a corner unless they are working with Intel and there hasn't been any indication of such. I can't imagine that NVIDIA and Intel are not talking or working together already on this. It wouldn't surprise me if in the very near future we see Intel outright buy NVIDIA.

There was a release by Intel about vertical construction. Basically, instead of going smaller and smaller L x W, they will continue to miniaturize but go L X W X H 3 dimensions as well.

Thats fine but NVIDIA's mainstay and business model is still discreet GPUs and Cards. The way AMD is pushing things is to combine both CPU & GPU on die to remove the physical memory bottleneck. Hence the reason for their purchase of ATi. I believe this is the way to go.

Thats only the movement for one portion of the market. AMD/ATi are not moving away from GPU's on cards. Do not expect APU's to replace GPU's, because they won't. Only in small form factors will an APU be the way to go, because of the size constraints. Crunching or Gaming a real GPU is still going to be far superior.

What you got to remember is, Nvidia is a GPU maker, and Intel is a CPU maker (yes they do make some horrible onboard video as well). They both know their respective fields well and are the top sellers for each. AMD makes CPU's and ATi (now also renamed to AMD) makes GPU's. AMD buying ATi gave them an advantage neither of their competitors (on either front) have, they may not sell the most CPU's or the most GPU's, but with the expirence of building both on a competitive front, it should allow them to crush the competition as far as APU's go. But I still hold my opinion that an APU will not be replacing a GPU anytime soon in a desktop or even higher end laptops. It's simply something to conserve space and wattage.

i sense extreme over heating issue's to come. But ill stick with what i have untill bulldozer is released.