Homework: How Turbo Mode Works

AMD and Intel both figured out the practical maximum power consumption of a desktop CPU. Intel actually discovered it first, through trial and error, in the Prescott days. At the high end that's around 130W, for the upper mainstream market that's 95W. That's why all high end CPUs ship with 120 - 140W TDPs.

Regardless of whether you have one, two, four, six or eight cores - the entire chip has to fit within that power envelope. A single core 95W chip gets to have a one core eating up all of that power budget. This is where we get very high clock speed single core CPUs from. A 95W dual core processor means that individually the cores have to use less than the single 95W processor, so tradeoffs are made: each core runs at a lower clock speed. A 95W quad core processor requires that each core uses less power than both a single or dual core 95W processor, resulting in more tradeoffs. Each core runs at a lower clock speed than the 95W dual core processor.

The diagram below helps illustrate this:

  Single Core Dual Core Quad Core Hex Core
TDP
Tradeoff

 

The TDP is constant, you can't ramp power indefinitely - you eventually run into cooling and thermal density issues. The variables are core count and clock speed (at least today), if you increase one, you have to decrease the other.

Here's the problem: what happens if you're not using all four cores of the 95W quad core processor? You're only consuming a fraction of the 95W TDP because parts of the chip are idle, but your chip ends up being slower than a 95W dual core processor since its clocked lower. The consumer has to thus choose if they should buy a faster dual core or a slower quad core processor.

A smart processor would realize that its cores aren't frequency limited, just TDP limited. Furthermore, if half the chip is idle then the active cores could theoretically run faster.

That smart processor is Lynnfield.

Intel made a very important announcement when Nehalem launched last year. Everyone focused on cache sizes, performance or memory latency, but the most important part of Nehalem was far more subtle: the Power Gate Transistor.

Transistors are supposed to act as light switches - allowing current to flow when they're on, and stopping the flow when they're off. One side effect of constantly reducing transistor feature size and increasing performance is that current continues to flow even when the transistor is switched off. It's called leakage current, and when you've got a few hundred million transistors that are supposed to be off but are still using current, power efficiency suffers. You can reduce leakage current, but you also impact performance when doing so; the processes with the lowest leakage, can't scale as high in clock speed.

Using some clever materials engineering Intel developed a very low resistance, low leakage, transistor that can effectively drop any circuits behind it to near-zero power consumption; a true off switch. This is the Power Gate Transistor.

On a quad-core Phenom II, if two cores are idle, blocks of transistors are placed in the off-state but they still consume power thanks to leakage current. On any Nehalem processor, if two cores are idle, the Power Gate transistors that feed the cores their supply current are turned off and thus the two cores are almost completely turned off - with extremely low leakage current. This is why nothing can touch Nehalem's idle power:

Since Nehalem can effectively turn off idle cores, it can free up some of that precious TDP we were talking about above. The next step then makes perfect sense. After turning off idle cores, let's boost the speed of active cores until we hit our TDP limit.

On every single Nehalem (Lynnfield included) lies around 1 million transistors (about the complexity of a 486) whose sole task is managing power. It turns cores off, underclocks them and is generally charged with the task of making sure that power usage is kept to a minimum. Lynnfield's PCU (Power Control Unit) is largely the same as what was in Bloomfield. The architecture remains the same, although it has a higher sampling rate for monitoring the state of all of the cores and demands on them.

The PCU is responsible for turbo mode.

New Heatsinks and Motherboards Lynnfield's Turbo Mode: Up to 17% More Performance
Comments Locked

343 Comments

View All Comments

  • 7Enigma - Wednesday, September 9, 2009 - link

    "What you are doing is cheating, and people is not stupid."

    Hand meet Face..... Gentleman we have a winner! Could you please post a picture of yourself so I can make a T-shirt?
  • james jwb - Tuesday, September 8, 2009 - link

    i wouldn't go that far at all, but look a few comments above for my feelings on this. The honest way to go about this, Anand, was to do this review as you have done, but have a page or two dedicated to clock for clock, and to state in bold letter that turbo mode was on for the benches in the current review. Do this and all is forgiven

    Have you stopped to think how well Turbo mode in Lynfield has boosted benchmarks compared to bloomsfield?

    As i said, yep, nice feature, but i do not expect a site to fall for stuff like this, i'd do stock (turbo-crap) results like you have, because this feature is there and cannot just be ignored, but my interest would have been to present clock for clock results on other pages, and it should have been yours as well. You've done a disservice to the enthusiast here. I was loving Lynnfield in your review until i realised turbo was on, then i felt this review was 100% useless to what i care and want to know about - clock for clock, and it must be the same for a huge percent of your readers, who at this stage, some probably haven't even realized turbo was on in your results and are still salivating. All enthusiasts who like to overclock themselves and not have turbo trying to trick us in reviews want clock for clock, you left them out, that is serious cause for questioning what the hell is going on at Anandtech. Sure, for those here who run stock, the review suffices, but, well, i've made my point now i think...

    This is the first real blunder i've seen on this site, but imo its a big one, and that blogger who hates you guys and posts up lots of stuff about your contradictions (to which so far i find all his posts false, the server blog guy, you know who i mean), today he'll have a field day on this one.
  • Anand Lal Shimpi - Wednesday, September 9, 2009 - link

    I can add a disclosure saying that turbo mode was enabled, but this has always been the case for our Nehalem reviews. We always try to review products with all of their features enabled as long as they ship that way and they don't reduce quality (e.g. certain GPU driver optimizations in the past).

    While I'll gladly add a clock for clock against Bloomfield in a few tests, flip back to our Lynnfield preview if you want a preview of that (that chip could only turbo up by 1 multiplier - effectively making it a clock-for-clock comparison to Bloomfield).

    Turbo mode in Lynnfield does help it against Bloomfield in the lighter threaded apps, do keep in mind that Bloomfield has turbo as well.

    It seems like what you're saying is that you want the maximum overclock possible for all cores and as such you will definitely disable turbo mode. In that configuration, the only difference between Lynnfield and Bloomfield (from a purely CPU standpoint) is one channel of memory. As I've mentioned before, that one memory channel is not going to make a tremendous difference for the vast majority of users. But if this is what you're asking for, I can definitely provide it. I'm out in California this week for an unrelated product, let me get back and I'll include that data in my Core i7 860 review.

    Take care,
    Anand
  • james jwb - Wednesday, September 9, 2009 - link

    thanks for the response, it's appreciated. I hold anandtech on a pedestal, and was just surprised not to see some clock for clock stuff for the overclockers out there. I haven't used a CPU at stock for 6+ years for my main rig, so clock for clock is very important, though as you've said, there won't be too much difference between bloomsfield and lynnfield, i'd still like to see it, though :)

    As for your comment below about how intel and amd will be changing things soon for the stock users out there, well, i think clock for clock will become very important for overclocker then in these reviews, more so than ever to find out how each cpu performs, or maybe i'm wrong!
  • ClownPuncher - Tuesday, September 8, 2009 - link

    Why would they turn off a product feature when reviewing said product? Because PhII doesn't have it? That is like not including DX10.1 results for AMD in HAWX. Consumers usually buy products based on features, this product has an aggressive turbo boost.
  • Gary Key - Tuesday, September 8, 2009 - link

    "there you go, finally you said it.
    all the benchmarks have at least 600 mhz over the processor's stock speed.
    that is outrageous, then if you want to compare the result with phenom 2 you have to overclock phenom 2 at least 600 mhz over stock speed.
    just to be fair "

    First off, it is fair as that is how Intel ships the CPU, turbo is an integral part of the CPU's operation and is a major feature, end of subject.

    Second, the 965 BE is not going to run stable at 4GHz under a 64-bit OS without near or sub-zero cooling. The Phenom IIs continue to have a problem when approaching 4GHz with a 64-bit OS. AMD has not fixed it yet.
  • james jwb - Tuesday, September 8, 2009 - link

    Sorry, but i disagree. You seem to have swallowed some genious Intel marketing crap here. Yes, it's a nice feature, but any CPU can overclock, all overclockers will overclock, and your results gloss over that benchmarks were using turbo. This is bad, and the first time i've really seen this site do something really biased, or just stupid, not sure which...

    I'm telling you now, a big percentage of your readers will see the results for the 920 and 750 as clock for clock comparisons and think its a very close game here, and it's not.

    I'll forgive you once you do a clock for clock comparison, but i won't forget that you pushed turbo results and nothing else, and didn't put a warning big bold statements like you normally do for important info stating these are with turbo on. Watch all excitement for this review from overclockers dwindle if you do.
  • goinginstyle - Tuesday, September 8, 2009 - link

    They had an entire section discussing turbo mode, comments were made in the results about the turbo advantage,and their system setup page clearly states turbo in turned on for both lynnfield and bloomfield. Read the article.

  • james jwb - Tuesday, September 8, 2009 - link

    the look at turbo is rubbish, we need clock for clock against bloomsfield.

    It's not in bold, its of that importance to state this for those who flick through, which you'll find many do.

    And still, i have no gripe with the data presented so far (aside from it not being very, very clear that turbo was on), my problem is what has been left out, and that is - a lot of important stuff for the enthusiast who control his own overclock.
  • goinginstyle - Wednesday, September 9, 2009 - link

    Look on page 9 under the system setup, it clearly states that turbo is enabled. It is mentioned several times in regards to the performance results. If you could read, you would have understood that TURBO was ON. Several pages of the article discussed the new Turbo mode and how important it is to the performance of the processor.

    I suppose because you have labeled yourself an enthusiast that means you should be treated special. I guess between you, SnakeOil, TA152H, and a couple of others we could hold a special Olympics in your honor. Wouldn't that be special?

    What is amazing, is that after reading several other reviews, it appears only Snakeoil at Tech Report is posting the same crap there as he did here. Everyone used Turbo mode in testing so why are you not posting at the other sites about how mistreated you are as an enthusiast.

    Probably around 98% of computer owners do not overclock so why is it so hard to understand the testing methodology used here and elsewhere. For those of us that do not overclock, this article was perfectly suited for us and shows that we can get similar performance without opening up the box. Even for those who do overclock, they at least showed clock results and I am sure we will see more based on Anand's comments this morning.

Log in

Don't have an account? Sign up now