• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Thuban X6 to get Dynamic Speed Boost

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Kuroimaho

Member
Joined
May 26, 2004
Location
Japan, Tokyo, Ueno.
AMD Phenom II X6 “Thuban” Processors to Get Dynamic Speed Boost Technology. Link

Xbitlabs said:
In an attempt to provide maximum possible performance for enthusiasts, Advanced Micro Devices plans to implement a dynamic performance boost technology into its six-core processors known as Phenom II X6 “Thuban”.

There is a clear trend towards increase of the number of cores inside central processing units (CPUs), just five years ago a dual-core chip was a dream, whereas now we are approaching six-core microprocessors. The software is, unfortunately, seriously behind the hardware and many applications still cannot take advantage of additional cores, but fully depend on clock-speeds. In order to provide maximum possible performance in such programs, developers of CPUs implement special dynamic performance boosting technologies that disable certain cores and overclock the rest.

Seems Thuban is much more than an Istanbul on a new stepping. This would explain why they release quad cores based on Thuban with 2 cores disabled on the desktop.
 
Thanks for the info!

This will be good for your non-enthusiast user. As long as they have a bios option to disable the feature. Otherwise there are some obvious issues if you overclock it on your own :)
 
I think the performance of the 965's high clocks made it impossible to launch the Istanbul on the desktop. This Dynamic scaling makes Thuban more competitive in apps which do not use scale well, it's lower clocks would make it loose to the desktop cpus this might make

Honestly i am really surprised to see AMD answer to Intel's turbo mode so quickly they did it on their existing arch and not wait till bulldozer to implement it.
It is a really good sign to see them add new features so aggressively.


I wonder if this will find it's way into a new Deneb revision, but most likely not and Thuban with disabled cores will cover that for us.

thats a loooooong chip

Just wait until you see the new server cpus. ;)
 
Makes me suspect they already had it in the works.

I'm looking forward to pics of the ~2000pin 12core server cpu, should look something like the P. Pro :D
 
This was posted on the front page yesterday :D

How this will play out is a different story.

I don't know if OCing unused cores would be more beneficial or not. The heat and power change required for this is a bit on the high end, but that all depends on how much the CPU overclocks. Now I could see this being used for a computer like me. Two monitors with a heavy OC, and a nice GPU. I like playing a game, watching a movie, surfing the web, and downloading mass media; all at the same time. Most games can only take up to 4 CPUs and utilize them to the max. So on the days that I'm not using all 6 cores at once, my CPU can kill two of them and boost the overclock on my CPU.

One question that still remains though: Can if you have thes Speed Boost on, will it OC and already OCed CPU?
 
If I recall, doesn't this feature require Windows 7. Windows 7 is the only operating system I am aware of that basically put jobs on one processor until that processor exceed the limits. Then it try to put jobs on the other one. Otherwise, Windows VISTA/XP would just keep all processor evenly occupied. As the result, you do not get any benefit. So all those annoying back ground job prevent your processor from going full speed on your single core.

In any case, I am very happy to see that. This was one of the feature I really would want to see on Phenom II. It really gave Intel some unfair advantage in benchmarks that are single thread heavy.
 
@ Archer, I believe I read somewhere that most games only use up to 4 cores. Unless this has changed with the new dx10 and 11 games that came out.
 
I think the performance of the 965's high clocks made it impossible to launch the Istanbul on the desktop. This Dynamic scaling makes Thuban more competitive in apps which do not use scale well, it's lower clocks would make it loose to the desktop cpus this might make

Honestly i am really surprised to see AMD answer to Intel's turbo mode so quickly they did it on their existing arch and not wait till bulldozer to implement it.
It is a really good sign to see them add new features so aggressively.


I wonder if this will find it's way into a new Deneb revision, but most likely not and Thuban with disabled cores will cover that for us.


Just wait until you see the new server cpus. ;)

To be perfectly honest, AMD has executed things in almost all areas pretty well since loosing Hector Ruiz. I guess it pays to have a tech head at the helm rather than a gas bag.

I wonder, if Ruiz weren't in charge at any time would AMD be better off today?

Hopefully it'll switch off cores not required and up the clockspeed on the remaining cores wehn needed. That'd be sweet.
 
@ Archer, I believe I read somewhere that most games only use up to 4 cores. Unless this has changed with the new dx10 and 11 games that came out.

Well up and most are a little ambiguous:) I did not get what you were saying, I do now. I think most newer titles will take advantage of two with a few being able to use more. So you weren't wrong (you already knew that:)) But as I said it was a little ambiguous, I read it the wrong way.
 
TBH I think Intel and AMD had dynamic frequency scaling in the pipeline for a long time, probably ever since they started thinking about multiple core designs.

Props to AMD for putting it in, if it's adjustable somehow, that would own.
 
1 thing for people who don't understand multi-threading: You can effectively think of a thread as almost a stand alone executable loaded into memory, so more then one is almost equivalent to having 2 separate programs at once running. Since the 2 or more threads need to communicate between each other for the app to work properly, the thread that finishes a block of code first will wait for a message from the other thread so that it knows what the next step is or whatever that thread is designed to do. There's a lot of new methods coming out to help speed up the latency between communication between threads, but you can obviously tell how this would be problematic for games that need the quickest refresh cycles possible. This is why games are usually not multi-threaded, because it would actually hurt performance by breaking up the main processing loop into 'tasks' for threads. Now, some games are multi-threaded because of emerging technologies to utilize multi-threading without killing performance from latency. So don't get all mad because your new game isn't using all your cores, games just wont benefit until multi-threading techniques become more common place.

Anyways... I though I'd just clear that up, since so many people think that every app could so 'easily' utilize multi-threading when it's really a completely different story.

I am glad AMD released an answer to Intel's turbo-boost. Hopefully the technology will trickle down soon into the $100 range quads after Thuban is launched. This is a pleasant surprise since I thought they wait until Bulldozer for this. Yay =)
 
Thank you Dooms for that post. That helps clear things. This is another thing I need to catch up on. :D
 
Yeah, I am forced to keep up with these sort of things in order to stay competitive as a programmer even if I am not paid for my work. There's still a lot more out there I have yet to grasp. As far as how the threads interact in a way that side steps the latency, I have no idea. I just know there are some companies with enough money to develop these technologies that I doubt I'll get my hands on for a while.
 
Yeah, I am forced to keep up with these sort of things in order to stay competitive as a programmer even if I am not paid for my work. There's still a lot more out there I have yet to grasp. As far as how the threads interact in a way that side steps the latency, I have no idea. I just know there are some companies with enough money to develop these technologies that I doubt I'll get my hands on for a while.

Probably different threads for different things that don't need to know as much about eachother.
AI in one thread (or multiple), as it doesn't need to talk to physics.
Physics in another thread.
Lighting in a third thread.
Maps and keeping track of where the player has been in a fourth thread.
And so on and so forth, the trick would be building the AI so it doesn't need too much data from Physics constantly, or splitting physics into two separate parts.
 
Back