Turbo CORE Blog

I like the Spinal Tap reference (But why not just make 10 faster and have 10 be the top number?)
 
I was amazed at how many kids in the office didn't get that. I am pretty sure that was 1983 because I saw that movie with Liz <last name withheld for my own safety>.

That means the movie is almost 28 years old and I am much older.
 
So from what I'm understanding, the new TurboCore tech will automatically increase the clock speed when needed (under high workloads) as long as the max TDP isn't reached? So if the CPU is running an application that stresses more transistors than normal apps (and thus more power/heat), the TDP will be reached quicker and the processor won't turbo quite as high?

I was amazed at how many kids in the office didn't get that. I am pretty sure that was 1983 because I saw that movie with Liz <last name withheld for my own safety>.

That means the movie is almost 28 years old and I am much older.
LMAO. I was born in 1982 and even I know Spinal Tap.
Just curious, since these processors go "one louder," does that hint that the core is also called K-11?:p
 
Last edited:
AMD Turbo CORE is deterministic, governed by power draw, not temperature as other competing products are. This means that even in warmer climates you&#8217;ll be able to take advantage of that extra headroom if you choose. This helps ensure a max frequency is workload dependent, making it more consistent and repeatable.

Interesting. Does this imply that Bulldozer isn't as heatphobic as Phenom processors?
 
I was amazed at how many kids in the office didn't get that. I am pretty sure that was 1983 because I saw that movie with Liz <last name withheld for my own safety>.

That means the movie is almost 28 years old and I am much older.

The main reason I know it is because I took a Film Literature class. As such I recognize some things from It Happened One Night, which is far older (1934 if I remember correctly because of the Hayes Code)
 
So from what I'm understanding, the new TurboCore tech will automatically increase the clock speed when needed (under high workloads) as long as the max TDP isn't reached? So if the CPU is running an application that stresses more transistors than normal apps (and thus more power/heat), the TDP will be reached quicker and the processor won't turbo quite as high?

LMAO. I was born in 1982 and even I know Spinal Tap.
Just curious, since these processors go "one louder," does that hint that the core is also called K-11?:p

i was born in '92 and i got the reference ;) gues si'm just special :D


Just out of curisoity, isnt this already in the desktop processors? or is it for all cores rather than just the 3(for my x6) ?
 
So from what I'm understanding, the new TurboCore tech will automatically increase the clock speed when needed (under high workloads) as long as the max TDP isn't reached? So if the CPU is running an application that stresses more transistors than normal apps (and thus more power/heat), the TDP will be reached quicker and the processor won't turbo quite as high?

Correct. The idea behind turbo is to alllow you to "lean into" the additional headroom between base and TDP. For instance, back in the day, when we first worked out ACP (average CPU power) we were making 95W processors that rarely ever hit 50W. That meant almost 50% more headroom. We've closed the gap somewhat since then, but the concept is the same. With Turbo CORE you could get all the way up to TDP if your workload needs it.

Interesting. Does this imply that Bulldozer isn't as heatphobic as Phenom processors?

Don't know what you mean by that.

Just out of curisoity, isnt this already in the desktop processors? or is it for all cores rather than just the 3(for my x6) ?

Yes, but this version is very improved from the desktop version.
 
I think what he meant is will they have a 60c wall (thereabouts) like the ones currently out there? I keep mine cool but we know sometimes pushing it's limits can create heat quickly no matter what you do.
 
with turbo CORE limited to TPD and ambient temps, i wonder where this leave enthusiast clocking.

i understand the blog is referring to AMD's sever part but the desktop is essentially the same in most respects.

overclocking by the enthusiast taps into the 'headroom' they refer to in the article. if AMD is maximising or 'recapturing' potential of the cpu, is overclocking now dead.
unless these features or limits can be disabled in the bios, it seems the only way us to get extra performance out of the chip is using expensive cooling (big WC / phase change) to enlarge the 'boost zone' before we hit the railing (to use John Frueh's Grand Canyon analogy).

i understand if JF can't comment or clarify on this. but as an enthusiast myself, i wonder where this leaves that fun tinkering hobby called overclocking.
course i could just have it all arse backwards...
 
Or it could make that aspect dead easy and allow you to focus your attention elsewhere.
 
Can't really speak to overclocking, not my area of expertise.
 
Don't know what you mean by that.

I'm talking about how Phenom processors (both generations) were extremely heatphobic, so the moment they reach their maximum temperature threshold, they'd BSOD or give errors instantly (~55c-60c max temp). From what you explain, it does sound like bulldozer isn't as susceptible to heat as Phenom processors. That's what I mean.
 
with turbo CORE limited to TPD and ambient temps, i wonder where this leave enthusiast clocking.

i understand the blog is referring to AMD's sever part but the desktop is essentially the same in most respects.

overclocking by the enthusiast taps into the 'headroom' they refer to in the article. if AMD is maximising or 'recapturing' potential of the cpu, is overclocking now dead.
unless these features or limits can be disabled in the bios, it seems the only way us to get extra performance out of the chip is using expensive cooling (big WC / phase change) to enlarge the 'boost zone' before we hit the railing (to use John Frueh's Grand Canyon analogy).

i understand if JF can't comment or clarify on this. but as an enthusiast myself, i wonder where this leaves that fun tinkering hobby called overclocking.
course i could just have it all arse backwards...
1) No amount of testing/binning AMD can reasonably do will get the chips to the edge of performance/stability like a single overclocker working on a single chip. Even within the same wafer of silicon there is a variation in the overclocking potential for a chip.

2) What overclockers consider stable isn't necessarily what AMD/Intel consider stable. Some people are willing to get an extra 100 Mhz even if it fails Prime95, as long as it gives them 1 FPS more. If a well cooled, properly installed CPU failed such intensive tasks as Prime95, IBT, etc. at absolute stock settings due to the Turbo/Turbo Core it would be a shitstorm of bad publicity for the company.

3) As far as I know AMD is limited to 1/2 multipliers (100 Mhz) increments with the core clock. We all know that even in a magical world where Turbo Core was able to function at the absolute highest 100 Mhz increment there would still be overclockers out there looking to squeeze an extra 20 Mhz from the chip by lowering the multiplier and raising the HT clock.;)

4) Overclockers generally go (far) beyond the TDP anyway. It isn't feasible for AMD to release a chip with a higher TDP than the 125/140w we've seen already for a number of reasons. With Thuban and Turbo Core (1.0?) the vCore was typically raised to 1.475v. With aftermarket cooling and a lesser concern for the CPU lasting forever an overclocker will always have the option of pulling more power through their chip than it would normally.


What we will see as AMD and Intel compete with these technologies is they will gradually chip at the minor overclocks and make them defunct to the point that eventually; unless you are willing to go balls to the wall you might as well just let Turbo do it's thing and save yourself the hassle.
 
Thanks JF, interesting and promising...

Can you say whether Bulldozer will power gate the modules that have no threads to run? Or will they just not be clocked as high as the modules that are running threads?

Also, can Turbo CORE be switched off? I know people sometimes switch Intel Turbo off in order to oc further, but from my understanding that is kind of stupid: rather let the chip turbo up (to the same level say to 4Ghz) after you have overclocked it to say 3.5Ghz as opposed to overclocking it to 4Ghz and disabling turbo, better balance of performance and efficiency...

Cant wait for bd!
 
I'm talking about how Phenom processors (both generations) were extremely heatphobic, so the moment they reach their maximum temperature threshold, they'd BSOD or give errors instantly (~55c-60c max temp). From what you explain, it does sound like bulldozer isn't as susceptible to heat as Phenom processors. That's what I mean.

CoolSpeed downclocks is the proc is too hot.

Thanks JF, interesting and promising...

Can you say whether Bulldozer will power gate the modules that have no threads to run? Or will they just not be clocked as high as the modules that are running threads?

Also, can Turbo CORE be switched off? I know people sometimes switch Intel Turbo off in order to oc further, but from my understanding that is kind of stupid: rather let the chip turbo up (to the same level say to 4Ghz) after you have overclocked it to say 3.5Ghz as opposed to overclocking it to 4Ghz and disabling turbo, better balance of performance and efficiency...

Cant wait for bd!

Modules with no threads running are power gated to some degree - that is what leads to extra headroom for turbo. It will turbo all cores if all cores are active, but will not turbo cores that are idle as far as I know.

Yes, you can switch Turbo CORE off in the BIOS if you want.
 
AMD Turbo CORE is deterministic, governed by power draw, not temperature as other competing products are. This means that even in warmer climates you’ll be able to take advantage of that extra headroom if you choose. This helps ensure a max frequency is workload dependent, making it more consistent and repeatable.

You're spreading some missinformation here.

The Nehalem power delivery and Turbo features are significantly more advanced than "Turbo Core". On Nehalem it takes into account both power draw and temperature and this is way better than simply power draw. You can even set the maximum TDP the chip can extract and Turbo will adapt to that.

Secondly, not taking temperature into account is plain stupid. There are thermal hotspots on the CPU, significantly warmer than the rest ( FPUs for example ). Then silicon itself has a max temperature of around 105C.
So, if I don't have proper cooling and I'm burning 70w ( with a TDP of 95w ) and my FPUs are already at 95C, what will your Turbo Core do ? Pump up the frequency until it reaches 95w ? By that time, some parts of the chip will resemble KFC.

So you either didn't understood what was told to you ( about your own solution and what the competition has ) or your deliberatly spreading some uninformed FUD.
 
You're spreading some missinformation here.

The Nehalem power delivery and Turbo features are significantly more advanced than "Turbo Core". On Nehalem it takes into account both power draw and temperature and this is way better than simply power draw. You can even set the maximum TDP the chip can extract and Turbo will adapt to that.

Secondly, not taking temperature into account is plain stupid. There are thermal hotspots on the CPU, significantly warmer than the rest ( FPUs for example ). Then silicon itself has a max temperature of around 105C.
So, if I don't have proper cooling and I'm burning 70w ( with a TDP of 95w ) and my FPUs are already at 95C, what will your Turbo Core do ? Pump up the frequency until it reaches 95w ? By that time, some parts of the chip will resemble KFC.

So you either didn't understood what was told to you ( about your own solution and what the competition has ) or your deliberatly spreading some uninformed FUD.

Colour me stupid, but from my understanding, if a chip is going to over heat @ maximum TDP, then it's likely that it would have already overheated before reaching that TDP, chances are if you're ambient temperature is 75degrees C you're going to die before the chip does, and if your cpu fan is pooched, then it will die at pretty much anything, other than idle.

As for determining the clock speed using both temperature and tdp, I agree, it would make more sense to use both, but that really only applies to 2 scenarios, overclocking, and in notebooks, which temps are far out of what we normally expect.
 
The engineering team went through everything that I wrote to ensure what we were saying is accurate.

You are certainly welcome to accuse me of spreading misinformation (as you have in other places) but the reality is that you do not have all of the details around the technology, so you really aren't in a position to determine anything.

It is clear that you aren't really going to believe anything that AMD has to say. That is fine, we can't be everything to everyone.
 
The engineering team went through everything that I wrote to ensure what we were saying is accurate.

You are certainly welcome to accuse me of spreading misinformation (as you have in other places) but the reality is that you do not have all of the details around the technology, so you really aren't in a position to determine anything.

It is clear that you aren't really going to believe anything that AMD has to say. That is fine, we can't be everything to everyone.

If it were only me, I'd have the same doubts that I am not reading correctly what you wrote. But then, I've noticed people at RealWorldTech ( incomparably more qualified than me ) have said the same thing.

http://www.realworldtech.com/forums/index.cfm?action=detail&id=116566&threadid=116563&roomid=2
 
Back
Top