ASUS Develops Own Technology to Rival AMD Turbo Core and Intel Turbo Boost

[btarunr]

ASUS designed its own technology comparable to AMD's Turbo Core, the ASUS Turbo Unlocker. The technology is claimed to be more advantageous than Turbo Core, since it sends every core into Boost state, rather than a limited number of cores designated as turbo cores. What's more, it supports several other Phenom II Black Edition series processors other than those based on the six-core Thuban or the quad-core Zosma. With single-threaded loads, the active core is given the maximum boost in clock speeds, while as multi-threaded load increases, even more number of cores enter a boost state. Boost state speeds decrease with increase in number of active cores. The technology is available on ASUS' entire range of motherboards based on the AMD 8-series chipsets. Details follow in a presentation by ASUS. It would be interesting to see if ASUS goes a step further and works out Turbo Unlocker for Intel Core i3 processors.



View at TechPowerUp Main Site

 

[Dyno]

Very interesting, and thank you for doing what you do on the forums.

 

[newtekie1]

I'm guessing this wouldn't be hard with an i3, they just need to overclock the processor via a small FSB bumb when under load. They kind of already offered this type of turbo mode on some of their older boards.

 

[WarEagleAU]

Is it able to be downloaded now?

 

[xaira]

im guessing this throws tdp out the window

 

[TheLostSwede]

If you actually looked at the pictures you would've noticed that it says that they stay within TDP and "electrical" values...

 

[cadaveca]

I wonder what effect, if any, this will have on overclocking...OCP/TDP may get in the way of pushing farther than these sort of technologies offer...may just be that overclocking is seeing it's deathbed.

 

[afw]

Is this some sort of auto overclocking ... or different ... ???

 

[WSP]

I see K10stats and Phenom MSR tweaker influence

 

[Initialised]

Someone I met was testing an ES 1055T and 1090T. x6 1090T overclocks to 3.6GHz LinX 20 stable with no voltage changes by both HT and multiplier on 890GX so there's no reason this wouldn't work. Shame Intel won't give MB manufacturers/users access to the 0.1x and 0.5x multiplier increments that their power-saving/throttling modes use on 1366 & 1156.

 

[aj28]

As long as this isn't a software implementation, I would gladly partake. Now can we get the reverse to improve energy efficiency, plz? Preferably something that won't simultaneously cripple overclocking like CNQ has been known to do in some implementations.

 

[HalfAHertz]

Well power consumption increases with frequency, so unless they're dropping the voltage at the same time as they are raising the frequency, then they go over the 125W power budget.
AMD's solution is smarter and better and Intel's solution remains the best - power gating FTW

 

[Initialised]

As long as this isn't a software implementation, I would gladly partake. Now can we get the reverse to improve energy efficiency, plz? Preferably something that won't simultaneously cripple overclocking like CNQ has been known to do in some implementations.

That's tricky, usually with all CPUs dynamic clocking (Turbo/CnQ/EIST) has to be turned off when pushing the overclock past the 'Stock Voltage Cut-Off point'. A possible way around it would be to only use settings that can be changed in Windows and use Tubro-V or AoD application profiles or macros to OC depending on what you were running so long as you were confident that the changes could be made on the fly without causing a hang under light load.

 

[Pyarn]

looks promising to me..I think I'll just keep my PII 965 for another year..

 

[Initialised]

As much as I've been tempted to splash out on i7 and now Phenom II x 6 I kind of feel the same way about my Q6600. A cooling upgrade for some extra OC headroom, another 4870 or X2, a Q9650 or an SSD would be of more immediate benefit than anything but a full 1366 platform upgrade and even then there's not much point until there are sensibly priced 6-core CPUs.

For people with AM2(+)/3 boards who don't overclock this is great as it may filter down to your board and provide a bonus performance boost.

 

[xtremesv]

As long as this isn't a software implementation, I would gladly partake. Now can we get the reverse to improve energy efficiency, plz? Preferably something that won't simultaneously cripple overclocking like CNQ has been known to do in some implementations.

I'm agree, I'm not seeing how they're gonna make that trick, they'd have to mod the chipset/bios or implement it through software, the latter will be a disappointment. Either way, the idea is promising

 

[Completely Bonkers]

Remember that the TDP envelope is determined by the CPU's execution units running at full blast on the most complex instruction codes, ie: SSEx, FPU and massive memory intensive loops.

When the CPU isnt using these high-power-demand instruction codes, the power consumption is lower; therefore you can increase the speed of the CPU when running "simple" instructions and remain in the TDP so long as you dont start executing complex code across all cores.

So what does this mean in practice? Regular non-optimised code can now run a little faster. OK so far. But also a hotter CPU on average. And therefore you will need improved cooling.

Yes, it is a valid approach. What will you gain? A few percentage points of improvement. Is it worth all the effort? IMO, nope. You will end up with a hotter, noisier, PC on average, for just a small speed bump. And to keep the noise and heat down, you need parts that probably cost a bit more than having bought the next processor up in the line up.

If you are the kind of person that likes to add after-market exhaust pipes, go-faster stripes, redex in your petrol, to your pimped ride, or if you tune your engine for a hotter, but lower efficiency but with slightly improved acceleration, then this might be for you.

If however you prefer to buy a properly specced sports/race car in the first place then forget this and go buy a better machine.