I wouldn't say they've stopped trying but certainly under emphasized it. You get far more performance from multiple processors.
Additionally, Intel has stated that more cores is the route of the future as it's more cost effective to put more cores on a chip than to invest in faster clock speeds.
concerned about cramming more cores onto a single piece of silicon
I doubt Intel works on upping the frequency until they move to the next microarchitecture. I don't think Nehalem is going to get them there comfortably.
Also consider the movement towards mobile platforms (and just efficiency in general) ... multiple cores (with the potential to shut some "off" when not needed) seem to be a much better performance/battery life combo than clock speed.
Of course, that all assumes the OS and apps are designed to take advantage of that architecture.
Greater energy efficiency also has more value today than it did during the old megahertz race. Superfast processors have enormous TDP and stresses cooling systems all around.
Apat from high power consumption, the memory architecture is holding back the speed. Nahelem shows this clearly and DDR3 in multi channels seems to be the only method of addressing bus contention issues with very fast superscalar processors.
One would expect 4, 6 8 channels ore more for RAM but it seems , that is costly to design and implement. So the trade-off is getting more done within the megahertz power budget one is stuck with. I think the procesors are fast enough, just SSDs need catching up in a hurry!.
AMD and Intel have effectively decided to slow down MHz ramps and go with multi-core and microarchitecture improvements as they have recognized that 150+ Watt CPUs aren't tenable in consumer desktop systems. At least not in the mass quantities they want to sell it. (300 Watt GPU systems on the other hand, well, at least they have a hardcare market of buyers.) 4 GHz CPUs will be shipped, just not as quickly anymore. Maybe in another year.
On the other hand, IBM has been shipping 4.7 GHz Power6 machines for almost a year (I think). They probably will go for 5.5+ GHz or something when they move the architecture to 45 nm. These are workstations and mainframes. Customers of such machines, I presume, can afford the power and cooling costs.
I would say that it's over for speed increases per chip/core. Apple pretty much considers a 3.2 GHz 8-core machine to be running at 25.6 GHz in there Xgrid Configuration Panel. In all actuality it is 25.6 GHz.
That is where I think the race is over. If you have a machine that is basically rated at 25.6 GHz and a 1 chip/core running at 2 GHz, to increase that to be around half the speed of that 25.6 GHz or even 4x the speed to about 8 GHz seems like that would be impossible, by todays standards.
So, basically why try anymore. The industry has decided to goto 64 Bit multi-core and learn to completely re-write software architectures and get the increase in speed.
2 GHz chips have been around for over 7 years and they still sell them as brand new systems today with 2 GHz chips...
Laters...
oh I almost forgot, to answer your question, we will see 4 GHz chips. (in macs right?)
Given CPU power has been consistently increasing, and "following" Moore's Law, I don't have a problem with GHZs being this or that.
On the GPU side, for years, clock speed has increased and decreased between 300mhz to 600mhz and what not but it seems quite unrelated to overall graphics performance.
Ironically The "GHZ myth" has become the new law. GHZ doesn't really mean much nowadays.
You could have an AMD X2 running at 2.8ghz but an Intel Dualcore at 2.5ghz whips it.
If you're obsessed with 4GHZ just get into the nitrogen-cooled overclocking scene. You can take an Intel 2.something ghz chip up to 4GHZ on nitrogen-cooling.
The main reason to buy a new desktop Mac is a faster graphics card as far as I am concerned. The CPU variations from 2.3GHz to 2.7GHz are meaningless. How often do I run at 90%-100% CPU utilization? If I need more power, multiple CPU workstations with optimized programs will be my choice. Two heads are better than one. /nod /nod (in parallel too. )
The main reason to buy a new desktop Mac is a faster graphics card as far as I am concerned. The CPU variations from 2.3GHz to 2.7GHz are meaningless. How often do I run at 90%-100% CPU utilization? If I need more power, multiple CPU workstations with optimized programs will be my choice. Two heads are better than one. /nod /nod (in parallel too. )
YES. FINALLY people are coming round to it. This 1st half of 2009, GPU power is where the real value and need will start growing... Better GPU power, OpenCL and CUDA growth.
In other words, if an iMac has 20" w. LED backlight, dual core (at "just" 2.5ghz or so) and 9600M GT 256MB VRAM... at USD 1,099 ...BOOM! Those would do killer sales, no need any fancy re-engineering, same form factor is fine. Let Apple produce that iMac at that price point... and we'll see if the "desktop market is dead" or whatever.
Comments
Additionally, Intel has stated that more cores is the route of the future as it's more cost effective to put more cores on a chip than to invest in faster clock speeds.
to make chips faster and are only
concerned about cramming more cores onto a single piece of silicon
I doubt Intel works on upping the frequency until they move to the next microarchitecture. I don't think Nehalem is going to get them there comfortably.
http://en.wikipedia.org/wiki/Sandy_B...roarchitecture)
Sandy Bridge's specifications are reported to be as follows, according to a presentation made by Intel in December 2006:
- 4 GHz clock speed.
- 4 to 8 out-of-order cores.
- Without SSE: 8 DP GFLOPS/core (2 DP FP/clock), 32-64 DP GFLOPS/processor.
- With SSE: 28 DP GFLOPS/core (7 DP FP/clock), 112-224 DP GFLOPS/processor.
- 32 KB L1 cache/core, (3 clocks).
- 512 KB L2 cache/core, (9 clocks).
- 2-3 MB L3 cache/core (8-24 MB total) (33 clocks), most likely pooled and dynamically allocated among the cores.
- 64 bytes cache line width.
- 256 bytes/cycle Ring bus bandwidth. The ring bus connects the cores.
- 0-512 MB GDDR / fast DRAM.
- 64 GB/s GDDR / fast DRAM memory bandwidth.
- 17 GB/s memory bandwidth per QuickPath link with 50 ns latency
No that's HOT. Plus Sandybridge will have AVX extensions the next major upgrade of SSEOf course, that all assumes the OS and apps are designed to take advantage of that architecture.
One would expect 4, 6 8 channels ore more for RAM but it seems , that is costly to design and implement. So the trade-off is getting more done within the megahertz power budget one is stuck with. I think the procesors are fast enough, just SSDs need catching up in a hurry!.
AMD and Intel have effectively decided to slow down MHz ramps and go with multi-core and microarchitecture improvements as they have recognized that 150+ Watt CPUs aren't tenable in consumer desktop systems. At least not in the mass quantities they want to sell it. (300 Watt GPU systems on the other hand, well, at least they have a hardcare market of buyers.) 4 GHz CPUs will be shipped, just not as quickly anymore. Maybe in another year.
On the other hand, IBM has been shipping 4.7 GHz Power6 machines for almost a year (I think). They probably will go for 5.5+ GHz or something when they move the architecture to 45 nm. These are workstations and mainframes. Customers of such machines, I presume, can afford the power and cooling costs.
That is where I think the race is over. If you have a machine that is basically rated at 25.6 GHz and a 1 chip/core running at 2 GHz, to increase that to be around half the speed of that 25.6 GHz or even 4x the speed to about 8 GHz seems like that would be impossible, by todays standards.
So, basically why try anymore. The industry has decided to goto 64 Bit multi-core and learn to completely re-write software architectures and get the increase in speed.
2 GHz chips have been around for over 7 years and they still sell them as brand new systems today with 2 GHz chips...
Laters...
oh I almost forgot, to answer your question, we will see 4 GHz chips. (in macs right?)
On the GPU side, for years, clock speed has increased and decreased between 300mhz to 600mhz and what not but it seems quite unrelated to overall graphics performance.
Ironically The "GHZ myth" has become the new law. GHZ doesn't really mean much nowadays.
You could have an AMD X2 running at 2.8ghz but an Intel Dualcore at 2.5ghz whips it.
If you're obsessed with 4GHZ just get into the nitrogen-cooled overclocking scene. You can take an Intel 2.something ghz chip up to 4GHZ on nitrogen-cooling.
The main reason to buy a new desktop Mac is a faster graphics card as far as I am concerned. The CPU variations from 2.3GHz to 2.7GHz are meaningless. How often do I run at 90%-100% CPU utilization? If I need more power, multiple CPU workstations with optimized programs will be my choice. Two heads are better than one. /nod /nod (in parallel too.
YES. FINALLY people are coming round to it. This 1st half of 2009, GPU power is where the real value and need will start growing... Better GPU power, OpenCL and CUDA growth.
In other words, if an iMac has 20" w. LED backlight, dual core (at "just" 2.5ghz or so) and 9600M GT 256MB VRAM... at USD 1,099 ...BOOM! Those would do killer sales, no need any fancy re-engineering, same form factor is fine. Let Apple produce that iMac at that price point... and we'll see if the "desktop market is dead" or whatever.