Intel Kensfield quad-core CPUs - another broadside against AMD

Intel Kensfield quad-core CPUs - another broadside against AMD

Summary: I've just come across the the first independent benchmarks for the quad-core Intel Kensfield CPUs, and if you thought that the results for the dual-core Conroe CPUs were impressive, then wait until you see these!

SHARE:
TOPICS: Processors
10

I've just come across the the first independent benchmarks for the quad-core Intel Kensfield CPUs, and if you thought that the results for the dual-core Conroe CPUs were impressive, then wait until you see these!

Disclaimer: These tests results are based on pre-production Intel hardware.  I've not seen the hardware myself or carried out testing myself, but the independent testing has been carried out by an individual with an excellent track record of delivering unbiased results.

It's a natural evolution for things to double.  For what seemed like forever, Intel will release the quad-core Conroe that's equivalent to or better than the pre-production unit and leap 12 - 18 months ahead of AMDwe had single core desktop CPUs.  Then along came the dual-core chips.  Now quad-core CPUs are on the horizon that are set to change the entire CPU playing field.  Both AMD and Intel have had plans for quad-core desktop CPUs for some time now, but it's Intel's efforts that are the first to fall into the hands of of enthusiasts for testing.

The CPU under scrutiny is the quad-core Conroe E6600 "Kensfield" CPU which runs at a stock speed of 2.4GHz and has 8MB of cache.  The tests were carried out on a system with a MSI 975X Platinum 2.b motherboard and 2 x Corsair DDR2 8000ul 512MB dual channel RAM.  The system was, rather surprisingly, air-cooled, using a Tuniq Tower 120 cooler.

Looking at these numbers (and everything I see suggests to me that they are genuine), I can see that the Conroe quad-core is going to give Intel a massive lead over AMD for some time to come.  It's already been shown that the dual-core Conroe CPUs have the power and overclocking capability to give AMD some headaches for some time to come.  These quad-core CPUs take that much further, virtually guaranteeing that Intel will have the advantage over AMD for the next 18 to 24 months. 

And like the dual-core Conroes I covered a few weeks ago, these quad-core chips lend themselves to being overclocked very well indeed - this test has shown that the CPU is capable of being overclocked to 3.2GHz, again just air-cooled.

Over the past few weeks it has become clear that Intel is again back on top form with their Conroe CPUs - these desktop versions of the Pentium M range of CPUs are now really showing that Intel has been playing its cards right; working on the 65nm technology and getting the power requirements just right.  By going back to basics, Intel have come up with a chip that has excellent stock performance combined with a massive potential for overclocking, pleasing both ends of the market in one stroke while also leap-frogging ahead of AMD.

I know that AMD has a lot of fans and that the previous statement I made won't make them at all happy, but the fact of the matter is that come July, Intel will launch a dual-core Conroe that will totally blow out of the water anything that AMD currently have to offer.  Then, six months later, Intel will release the quad-core Conroe that's equivalent to or better than the pre-production unit and leap 12 - 18 months ahead of AMD.(This is assuming of course that AMD don't have some super-secret, Area 51 project going on behind the scenes. If they do, I doubt they would have played the weak "4x4" card).  Still, AMD fans shouldn't feel too down-hearted about all this, price-cuts are on the way for most AMD chips and this is bound to stoke the fires of innovation over at AMD HQ.

Switching sides for a second, I have to say that I have a few reservations about quad-core desktop CPUs:

  • Heat
    A quad-core is going to put some hefty demands on the PCs cooling system.  From the images I saw on the XtremeSystems forum, I noticed that the motherboard for the Conroe testing was running outside of a case.  Place that CPU inside a case and air-cooling might not be as effective.
    Talking about heat, I'd like to see some data on how hot this CPU gets under load.
  • Scalability
    Going from one core to two didn't double the performance as there is always overhead in organizing the work between the two cores.  Doubling again to four cores and I would imagine that the overhead would also increase.  I'd like to see more figures relating to this.
  • Quad-core support
    When are we going to start to see quad-core support in games and other multimedia apps?  For rendering and encoding video, these quad cores are going to be awesome, but for standard desktop apps and gaming, the full benefits may take a while to be realized.

Reservations aside, the quad-core Conroe is going to make some serious waves when it's released in Q1 next year. 

[Updated: June 22, 2006 @ 11:21 am]

I've just come across prices for dual-core Conroes on the website of a vendor in the UK.  I've converted the prices to USD here.

Topic: Processors

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.

Talkback

10 comments
Log in or register to join the discussion
  • Scalability indeed

    Considering that desktop applications are almost exclusively dominated by a single process at any given time, I have rather severe reservations over the benefit of even dual CPUs.

    Please note that this is from someone who has spent a good bit of the last ten years with dual-CPU boxes doing compute-intensive applications [1]. Even then, the benefit is marginal.

    Servers are another matter, of course. However, with servers (and especially servers that handle lots of tasks) the performance bottleneck tends to shift to memory response. In cases like that, the limits Intel has on both quantity and latency of DRAM put it at an architectural disadvantage. They seem to have a plan in the roadmap for 2008 and beyond, but that's not going to do them much good this year.

    [1] Try running Monte Carlo simulations in the background on a single-CPU box and see what it does to your HMI responses.
    Yagotta B. Kidding
    • Yes and no...

      I agree that if you are running a single application the benefit of dual core is "iffy". But the minute you fire up and run multiple apps the benefit is immediately apparent.

      Given that most users have an IM window running, email mininmized to the task bar, AV software in the background, possibly a print spooler for printer sharing, etc. the effective speed and system responce for the user is very real.
      No_Ax_to_Grind
      • Always assuming

        [i]Given that most users have an IM window running, email mininmized to the task bar, AV software in the background, possibly a print spooler for printer sharing, etc. the effective speed and system responce for the user is very real.[/i]

        I suppose that those applications [b]could[/b] suck lots of cycles at idle [1] -- but my experience is that they don't even show up as a blip on the CPU load. Their main impact is that they are memory resident; when they activate they load down the memory bus and flush other tasks from cache.

        Which means, of course, that your best bet for performance optimization when they're a factor is:
        (a) Lots of cache
        (b) Faster cache fill (low-latency memory.)

        [1] That is, with no user input. A stupidly-coded userspace app [2] can, of course, suck lots of cycles doing nothing.
        [2] Think MSWindows file copy.
        Yagotta B. Kidding
    • You''re out of touch - you don't get it

      "..Considering that desktop applications are almost exclusively dominated by a single process at any given time, I have rather severe reservations over the benefit of even dual CPUs..."

      Here's clue: While most apps may be single process (do not run in concurrent fashion) , they are very frequently multithreaded and the OS does a good job of allocating resources to multiple threads. Your claimed last 10 years running compute intensive apps have no bearing on what multiprocessors can do. You can run Pi all day long and it says nothing about multi-processing scaling opportunity.

      If you are convinced that multi-core, multi-threaded capable systems have no place in mainstream, you are way out of touch with what is happening out there. Multi-core is combined with intelligent OS's and applications that can dispatch multiple threads to multiple resources is here and together brings a HUGE leap of performance.
      I'm afraid your years in the lab running Pi (or Solitaire, ...whatever) has you mislead about what is happening in the broader industry.
      Prognosticator
      • Multithreaded applications

        [i]While most apps may be single process (do not run in concurrent fashion) , they are very frequently multithreaded and the OS does a good job of allocating resources to multiple threads.[/i]

        By all means name a few of those multithreaded applications. The profiling I've done (your attempts at [i]argumentum ad hominem[/i] aside) indicates that the responsiveness of a desktop system almost [1] always comes down to one causal chain that can't be distributed.

        In fact, thanks to one of the lemmas in Amdahl's Law, distributing system load usually results in [u]reduced[/u] responsiveness.

        [1] The Monte Carlo example is, in fact, one of the few exceptions -- and it isn't a very good counterexample.
        Yagotta B. Kidding
        • Pick virtually any photo or video application

          YBK,
          Even your most basic photo and video application is multi-threaded and optimized for multi-core.

          Photoshop elements is fully multithreaded. Most of the filters are multi-threaded. The same for any video encoding and decoding software. These are basic apps that people want sped up - not xcel or word.
          Here is from Adobe.
          http://www.adobe.com/products/photoshop/pdfs/pscs2_wn.pdf

          One of the most popular noise removal sw out there, Noise Ninja was multithreaded 2 years ago. This stuff is just starting to get serious. Where have you been? Surely, IO constrained "application" like you silly example "file copy" will be have the CPU's waiting for that resource. However, many many things that are slow today (audio, video, imaging) processing are prime candidates for threading.
          Prognosticator
          • And embrace memory latency issues

            Of course you get some multi-threaded benefits in the video encoding world but then you're pushing lots of 1's and 0's through the ol' CPU.

            So, you've gone from 1 CPU to 4 CPU's with no increase in the memory bandwidth. Do you see a potential bottle neck there?

            If you really want to see how architecture affects these things then look at the scaling factors for adding CPU's comparing Opterons to the Intel offerrings.
            Robert Crocker
          • Ya-got-a-Link?

            to the scaling comparisons?
            WinnebagoBoy
          • Here

            Granted it's an AMD person explaining it but it's still informative:
            http://www.devx.com/amd/Article/17580/2046
            Robert Crocker
  • Sounds like a good choice for

    Apple's missing Intel pro-desktops.

    A lot of Mac software has been taking advantage of multi-processors for years on the PowerPC. It would seem like a logical choice now with the Intel transition.
    Kid Icarus-21097050858087920245213802267493