![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
This post applies to both CPU/GPU,AMD/INTEL/NVIDIA/etc
There are similar threads in Intel and maybe Video Forum. I post here to avoid ultra technical argument (I am not capable
) I would like to give generalist view.
1. Why no 10GHz CPU. OK, the technical-team will give you law of physics. No argument.
1a. I only present consumer view. Look at most of the things consumer do with computer. You do not need 10GHz CPU now.
1b. Hi Consumer, with this 4.5GHz CPU, you can finish this job in 10sec. If not, you need 45sec. Question: After saving of 35s, can you tell me what will you do with the preciously saved 35sec? Hmm. maybe I will not do anything and just sit there enjoying the view, more gatorade, chip and taco. Congratulation. You have earned the right to appreciate life finer moment. Savor it. A lot of people may not have that luxury. Video encode/decode. (following comment for general consumer) If you have so much money to buy blu-ray movies to encode everyday, yes, IT/Media and especially Seagate/WD/Hitachi/Samsung can sit well not worrying too much about SSD problem.
2. CPU teams know they have problem scaling higher clock, so they scale more core. Problem: They cannot easily scale software development people thinking and skillset.
3. GPU teams know console-port/slow-game-engines that cannot push beauty of their new stuff to limit/developers that want marketing money for anything more than necessary as development becomes more and more expensive/mobile gaming from Apple and Phone-makers that draws revenue away/ARM-team waiting on the horizon. Integrated graphics/chipset progress that will strangle base for survival. Law-of-physics and economic scale is much smaller for some teams' combination, plus lesser cash-on-hand.
4. The key factor here is you need software co-operation to justify higher-faster-better-costlier, and the software people is just not co-operating to the expectation of hardware makers.
WHAT TO DO?
1. GPU teams move into CPU territory
2. CPU teams how to teach vast developers for 4+core, if not you will always have the problem of 4GHz E8400/i3 people spoiling further profit growth. Look at GPU teams seem they are able to parallelize many things.
3. Fab consolidation
Consider : OpenCL for GPU/CPU
1. You don't want to buy my GPU --> OK, then buy my 100-core CPU. You get the same benefits
2. You don't want to buy my CPU --> OK, then buy my trillion-core GPU. You get the same benefits.
3. Software: too much efforts/money needed in every area? --> OK, stick with OpenCL as starting point. Let the advance team settle the architectural issue for you 1st. Rip the reward and apply learning to the rest. Overall good? ok marketing money maybe?
4. The new horsepower : 3D virtual reality, physic simulation. etc.
Consumer: You can continue spending more and money on our products and you can actually see some tangible rewards/eye-candy/etc
Obviously we are still down to software people: we hope OpenCL/CUDA/Physics or other similar things move fast enough, the (6/8/12/16) CPU real-cores are coming very soon and 512-3200 shaders GPUs are already on the ground and more behemoth are coming.
There are similar threads in Intel and maybe Video Forum. I post here to avoid ultra technical argument (I am not capable
) I would like to give generalist view.1. Why no 10GHz CPU. OK, the technical-team will give you law of physics. No argument.
1a. I only present consumer view. Look at most of the things consumer do with computer. You do not need 10GHz CPU now.
1b. Hi Consumer, with this 4.5GHz CPU, you can finish this job in 10sec. If not, you need 45sec. Question: After saving of 35s, can you tell me what will you do with the preciously saved 35sec? Hmm. maybe I will not do anything and just sit there enjoying the view, more gatorade, chip and taco. Congratulation. You have earned the right to appreciate life finer moment. Savor it. A lot of people may not have that luxury. Video encode/decode. (following comment for general consumer) If you have so much money to buy blu-ray movies to encode everyday, yes, IT/Media and especially Seagate/WD/Hitachi/Samsung can sit well not worrying too much about SSD problem.
2. CPU teams know they have problem scaling higher clock, so they scale more core. Problem: They cannot easily scale software development people thinking and skillset.
3. GPU teams know console-port/slow-game-engines that cannot push beauty of their new stuff to limit/developers that want marketing money for anything more than necessary as development becomes more and more expensive/mobile gaming from Apple and Phone-makers that draws revenue away/ARM-team waiting on the horizon. Integrated graphics/chipset progress that will strangle base for survival. Law-of-physics and economic scale is much smaller for some teams' combination, plus lesser cash-on-hand.
4. The key factor here is you need software co-operation to justify higher-faster-better-costlier, and the software people is just not co-operating to the expectation of hardware makers.
WHAT TO DO?
1. GPU teams move into CPU territory
2. CPU teams how to teach vast developers for 4+core, if not you will always have the problem of 4GHz E8400/i3 people spoiling further profit growth. Look at GPU teams seem they are able to parallelize many things.
3. Fab consolidation
Consider : OpenCL for GPU/CPU
1. You don't want to buy my GPU --> OK, then buy my 100-core CPU. You get the same benefits
2. You don't want to buy my CPU --> OK, then buy my trillion-core GPU. You get the same benefits.
3. Software: too much efforts/money needed in every area? --> OK, stick with OpenCL as starting point. Let the advance team settle the architectural issue for you 1st. Rip the reward and apply learning to the rest. Overall good? ok marketing money maybe?
4. The new horsepower : 3D virtual reality, physic simulation. etc.
Consumer: You can continue spending more and money on our products and you can actually see some tangible rewards/eye-candy/etc
Obviously we are still down to software people: we hope OpenCL/CUDA/Physics or other similar things move fast enough, the (6/8/12/16) CPU real-cores are coming very soon and 512-3200 shaders GPUs are already on the ground and more behemoth are coming.
Last edited: