As we move away from simple cpu solutions, and gpgpu usage becomes more prominent, a few things change in this industry.
Some have said the cpu is king, long live the king, while others have pointed out, the cpu is becoming more and more as just 1 more component, and is actually losing its import as we move forwards.
Part of this conclusion is the new arch coming from all 3 corners, 2 of which havnt been taken seriously by some.
Intel introwed it 80 many cores awhiles back, and of course, we all know about LRB.
nVidia has been heading in this direction since G80, and their 3rd gen looks to be a makor change in the gpgpu approach.
AMD?
Well, theyve been accused of not trying, having adding nothing, but their new cpus with IGPs have also been maligned.
Some has gone as far as to say LRB is primarily a gfx card.
Looking at AMDs next cpu/gpu chip debunks alot of this , and will play well within the mobile market, having a 1 chip does all solution will be what the demands of tomorrow is all about.
This cpu/gpu, or rather APU will allow for low power usage, makes things easier such as flash etc, lowering power draw in every day average Joe usage.
Many here said alot of people play flash games, yet contend the AMD solution willo be poor.
Others say that Fermi will be a killer, and ATI is just sitting on their thumbs.
Others say LRB will dominate in both.
Both LRB and Fermi are taking brand new approaches in this regard in their arch, and AMD has done nothing, and what they do have wis woeful.
Id like to point out No.5 in the fastest HPCs on the planet
http://www.top500.org/lists/2009/11
Without changing 1 thing, without having allll this SW build up, no CUDA, no CUDA cores, no x86 compliancy etc, they did this, and an easy upgrade path using a the newer 5870s could easily put this at the top.
So, while everyone else is talking, pushing and bragging, saying this wont work, theyre lazy etc, guess whats happened?
I think its time to rethink a few things, and those myths some have held need to be upgraded
PS I'll add that adding 5970s to this solution could vault this rig up to the fastest around.
And in six months, do so again, while changing cpus isnt an alternative, as their turnover ratio is simply slower.
9 Months from now, we should have both better cpus and gpus, and then an immediately (in cpu terms) gpu refresh increasing potential 20 % again.
A cheap, easy solution, no new socket changes, no new intructions making for major changes etc.
http://www.anandtech.com/video/showdoc.aspx?i=3678&p=1
Some have said the cpu is king, long live the king, while others have pointed out, the cpu is becoming more and more as just 1 more component, and is actually losing its import as we move forwards.
Part of this conclusion is the new arch coming from all 3 corners, 2 of which havnt been taken seriously by some.
Intel introwed it 80 many cores awhiles back, and of course, we all know about LRB.
nVidia has been heading in this direction since G80, and their 3rd gen looks to be a makor change in the gpgpu approach.
AMD?
Well, theyve been accused of not trying, having adding nothing, but their new cpus with IGPs have also been maligned.
Some has gone as far as to say LRB is primarily a gfx card.
Looking at AMDs next cpu/gpu chip debunks alot of this , and will play well within the mobile market, having a 1 chip does all solution will be what the demands of tomorrow is all about.
This cpu/gpu, or rather APU will allow for low power usage, makes things easier such as flash etc, lowering power draw in every day average Joe usage.
Many here said alot of people play flash games, yet contend the AMD solution willo be poor.
Others say that Fermi will be a killer, and ATI is just sitting on their thumbs.
Others say LRB will dominate in both.
Both LRB and Fermi are taking brand new approaches in this regard in their arch, and AMD has done nothing, and what they do have wis woeful.
Id like to point out No.5 in the fastest HPCs on the planet
http://www.top500.org/lists/2009/11
Without changing 1 thing, without having allll this SW build up, no CUDA, no CUDA cores, no x86 compliancy etc, they did this, and an easy upgrade path using a the newer 5870s could easily put this at the top.
So, while everyone else is talking, pushing and bragging, saying this wont work, theyre lazy etc, guess whats happened?
I think its time to rethink a few things, and those myths some have held need to be upgraded
PS I'll add that adding 5970s to this solution could vault this rig up to the fastest around.
And in six months, do so again, while changing cpus isnt an alternative, as their turnover ratio is simply slower.
9 Months from now, we should have both better cpus and gpus, and then an immediately (in cpu terms) gpu refresh increasing potential 20 % again.
A cheap, easy solution, no new socket changes, no new intructions making for major changes etc.
http://www.anandtech.com/video/showdoc.aspx?i=3678&p=1