Intel announces "Core 2 Duo on steroids" - automatic overclocking

[queamin]

Intel today announced new details of its forthcoming Santa Rosa PC platform, including a significant revision of the Core 2 Duo chip.

"We call this processor Core 2 Duo but really it's Core 2 Duo on steroids. This microprocessor is going to be more powerful than the previous generation," said Intel's mobility chief, Mooly Eden.

The new chips will be able to overclock one of the cores if the other core is not being used.

"The idea is the following," explained Eden. "If you are running a single threaded application, one of the cores can go to sleep, and the left over power can be used by the other core - we give it a turbo boost; the ability to run faster than it used to.

"This is not overclocking. Overclocking is when you take a chip and increase its clock speed and run it out of spec. This is not out of spec. Here, it is within the spec of the dual-cores, we just identify when one core is not using the headroom and we give it to the other core.

"This is called Enhanced Dynamic Acceleration Technology. We've had problems implementing it, but we've been able to do it in Santa Rosa," Eden said.

The Santa Rosa platform also includes draft 802.11n for "whole house" high definition video streaming and Intel Turbo Memory - the flash cache memory that was previously codenamed "Robson". Intel demonstrated it in use showing two identical machines - one with Turbo Memory and one without - and the latter took twice as long to complete a photo processing task.

The Intel 965 graphics chipset is able to run Vista's Aero interface smoothly, though it will not support DirectX10 upon initial release - Intel has more work to do to implement it in the drivers.

"I'll make a claim and I'd like people to prove me that I am wrong: Centrino Pro technology is the best machine to run Vista," Eden said as a challenge to tech journalists attending Intel Developer Conference Beijing.

Eden said Santa Rosa would be available before the end of June - and refreshed in the first half of 2008 with the Penrin processor, Intel's upcoming 45nm processor.

http://apcmag.com/5852/intel_announces_core_2_duo_on_steroids_automatic_overclocking

might be good
Q

 

[NathanE]

Yeah they touched upon this at one of the recent IDF's. I'm sceptical about it though. I'm sure it works just fine... The problem is that it flies in the face of Intel's long term commitment to multi threading and concurrent execution.

It's more a power saving feature for laptops.

Nehalem is C2D on steroids, in my books.

 

[FrankJH]

Its one of those ideas that sounds good in theory but in practice wont add up to much - as NathanE said, we want developers to go multithreaded not more single threads faster

 

[Participant]

Its one of those ideas that sounds good in theory but in practice wont add up to much - as NathanE said, we want developers to go multithreaded not more single threads faster

Both can still happen though. How does this prohibit developers?

 

[Psycho Sonny]

Its one of those ideas that sounds good in theory but in practice wont add up to much - as NathanE said, we want developers to go multithreaded not more single threads faster

i dont think all software can be multithreaded, and even if it is multithreaded i think the same software would work faster ina single core of double the speed, for example you could test a dual core conroe at 1.5ghz then turn off 1 core and test it at 3ghz, im pretty sure the 3ghz conroe would be faster, its not always better to be multithreaded you need to have the right combination of speed and cores

 

[eracer2006]

but whats the point when 99% of the users here buy a c2d and overclock both cores to 3.2ghz permanently anyway.

 

[gurusan]

sounds interesting. I wonder if this means the overclocking headroom will be increased a bit?

 

[titaniumx3]

The article states that it is not overclocking, then what is it? I just can't understand how it's possible to redistribute power to one core or the other without overclocking.

 

[NathanE]

The article states that it is not overclocking, then what is it? I just can't understand how it's possible to redistribute power to one core or the other without overclocking.

So basically when the other core goes into a deep sleep state (as controlled by the OS) then the CPU redirects a significant amount of the wattage to the active core. The active core then increases it's multiplier (ala, core clock) and there you have it. It's basically SpeedStep technology but in "reverse"... instead of reducing the multiplier it increases it.

 

[supernova9]

but whats the point when 99% of the users here buy a c2d and overclock both cores to 3.2ghz permanently anyway.

The point is that the users on this board are an incredibly small minority when it comes to computer users. Overclockers are a tiny segment of the market for CPUs. For example, no major company in the world uses overclocking, as it would take just too much time if workstations had to be built, tweaked and then stability-tested, so something like this would benefit them.

It's for Joe Public, which this board isn't.

 

[Mattus]

And you have to bear in mind that what is and isn't overclocking can be pretty arbitrary. With the OC performance of E6300s, for example, I'm willing to bet that there's very little difference in the quality of E6400 and E6300 dies, since the latter are just speedbinned E6400s, or even higher chips with the cache burnt out. (When was the last time you saw an E6300 fail to reach 2.13 or even 2.4?)

The only difference between the chips is the multiplier that's programmed into them. It's overclocking because Intel say it is. Now they're saying that this isn't.

 

[Psycho Sonny]

but whats the point when 99% of the users here buy a c2d and overclock both cores to 3.2ghz permanently anyway.

i can run my e6600 at 3.6ghz but have not put it back to stock and put speedstep on

why you ask?

it uses far too much power at 3.6ghz to justify the speed increase, maybe its my tec cooler which uses 50W constantly if its overclocked as well as the additional cpu power

 

[FrankJH]

i dont think all software can be multithreaded, and even if it is multithreaded i think the same software would work faster ina single core of double the speed, for example you could test a dual core conroe at 1.5ghz then turn off 1 core and test it at 3ghz, im pretty sure the 3ghz conroe would be faster, its not always better to be multithreaded you need to have the right combination of speed and cores

Thats veru true it cant all be double threaded - but with many single threaded apps open on many pc's simultaneously as standard (av, IE, possibly outlook, firewall, msn) windows still performs better with more than one core

Not only that , Intel were trying to persaude people that "proper" dual cores were the answer and then go and do this.

Once the software gets up to scratch I personally think that multi-threaded apps will suit a lot more than they will now (ie new techniques will allow better usage in unfamiliar types of applications) but more than willing to accept defeat if this doesnt happen

 

[Durzel]

The point is that the users on this board are an incredibly small minority when it comes to computer users. Overclockers are a tiny segment of the market for CPUs. For example, no major company in the world uses overclocking, as it would take just too much time if workstations had to be built, tweaked and then stability-tested, so something like this would benefit them.

It's for Joe Public, which this board isn't.

Exactly. Plus this technology would be great for people who do not even really have the option to overclock comfortably (i.e. laptops, etc)

 

[simonnance]

surely it just makes the core more flexible, which is not a "bad" thing now is it? And who's going to complain about a free boost to single threaded app performance?

Anyway, i wonder whether they will move it forward for Core 2 Quad...... imagine getting a boost to 2 or 3 cores when a single one is being under-utilised! Could encourage people to go Quad when otherwise they would see little benefit over C2D.