Memory Performance - Not Very Nehalem

Let’s start at the obvious place, memory performance. Nehalem moved the memory controller on-die, but Clarkdale pushes it off again and over to an on-package 45nm graphics core.

To make matters worse, the on-package chipset is a derivative of the P45 lineage. It’s optimized for FSB architectures, not the QPI that connects the chipset to Clarkdale. Let’s look at the numbers first:

Processor L1 Latency L2 Latency L3 Latency
Intel Core i7-975 4 clocks 10 clocks 34 clocks
Intel Core i5-750 4 clocks 10 clocks 34 clocks
Intel Core i5-661 4 clocks 10 clocks 39 clocks
AMD Phenom II X4 965 3 clocks 15 clocks 57 clocks
Intel Core 2 Duo E8600 3 clocks 15 clocks  

 

L1 and L2 cache latency is unchanged. Nehalem uses a 4-cycle L1 and a 10-cycle L2, and that’s exactly what we get with Clarkdale. L3 cache is a bit slower than the Core i7 975, which makes sense because the Core i5 661 has a lower un-core clock (2.40GHz vs. 2.66GHz for the high end Core i7s) Intel says that all Clarkdale Core i5s use the same 2.40GHz uncore clock, while the i3s run it at 2.13GHz and the Clarkdale Pentiums run it at 2.0GHz.

Processor Memory Latency Read Bandwidth Write Bandwidth Copy Bandwidth
Intel Core i7-975 45.5 ns 14379 MB/s 15424 MB/s 16291 MB/s
Intel Core i5-750 51.5 ns 15559 MB/s 12432 MB/s 15200 MB/s
Intel Core i5-661 76.4 ns 9796 MB/s 7599 MB/s 9354 MB/s
AMD Phenom II X4 965 52.3 ns 8425 MB/s 6811 MB/s 10145 MB/s
Intel Core 2 Duo E8600 68.6 ns 7975 MB/s 7062 MB/s 7291 MB/s

 

Here’s where things get disgusting. Memory latency is about 76% higher than on Lynnfield. That’s just abysmal. It’s also reflected in the memory bandwidth scores. While Lynnfield can manage over 15GB/s from its dual-channel memory controller, Clarkdale can’t break 10. Granted this is higher than the Core 2 platforms, but it’s not great.

What we’re looking at is a Nehalem-like CPU architecture coupled with a 45nm P45 chipset on-package. And it doesn’t look very good. If anything was going to hurt Clarkdale’s performance, it’d be memory latency.

Index Clarkdale: The Perfect Home Theater PC
POST A COMMENT

93 Comments

View All Comments

  • Paladin1211 - Monday, January 04, 2010 - link

    On page 13, in WoW benchmark, the Core i5 750 outperforms the Core i7 870 by more than 30% (92.3 fps vs 70.6 fps). Anything wrong here? Reply
  • Crimson67 - Tuesday, January 05, 2010 - link

    WoW doesn't seem to like hyperthreading, it's the only explanation Reply
  • ereavis - Wednesday, January 06, 2010 - link

    is it using the hyperthread core on the 870 but a true core on the 750? That would certainly slow it down. If it's using a true core on both it should be better still. Reply
  • Dyzios - Monday, January 04, 2010 - link

    Remember NVIDIA Hybrid idea, to have two GPUs ? - one weak for 2D/light 3D graphics and powerful discrete GPU for gaming? Those CPU makes sense for this approach - however Radeon 5870 also has good point with optimized low idle power consumption. Maybe still there is point as even Radeon cannot go very low as GPU on-die. The only question is to have capability to switch HDMI output between on-die GPU and discrete.I wonder how this works currently - still needed to stick to DVI output from Radeon or can be combined? Reply
  • Zool - Tuesday, January 05, 2010 - link

    If the 40nm 5600 series cards will show the same improvments in power usage than the 5800 series than u can forget the intels GMA graphic.

    I think that for a 2D card/3D discret card u can buy a mobo with a dirty cheap intel GMA on it clocked much lower if u realy need.
    Reply
  • silverblue - Wednesday, January 06, 2010 - link

    The graphics comparison isn't exactly fair but in the end, it's not something AMD should be too worried about. The i5-661, with its 900MHz 45nm GPU, is being compared to the old 790GX's HD 3300 which operates at 700MHz on a 55nm process. I admit, in the benchmarks we're talking the best i5 vs. the best AMD has to offer, but considering...

    a) the relative performance of the i5-750 as compared to the PII X4 965 which is usually manifested as an advantage
    b) the fact that the 32nm i5s can increase their core speed by 133MHz and 266MHz thanks to Turbo depending on the number of active cores
    c) most games still aren't designed to take full advantage of multithreading so four cores may not yield a tangible performance increase

    ...then AMD's still in the lead for IGPs. If the 3300 had been clocked at 900MHz, would it have lost even one of the listed benchmarks? I'm not sure it would have. What's more, we're still talking a 55nm part; we all know of TSMC's issues with the 40nm process and AMD going fablress, so is it unreasonable to expect that AMD could move their IGP production en masse to 40nm with TSMC or 45nm with GlobalFoundries?

    In closing, it's a big step forward for Intel, however if AMD came out with a higher clocked 40/45nm IGP then, Sideport or not, new tech or not, AMD would be far ahead, at least on gaming terms. Sideport does very little for the performance of current AMD IGPs, anyway.

    I just wish AMD were able to release a Clarkdale competitor sooner rather than later.
    Reply
  • ruetheday - Friday, January 08, 2010 - link

    the IGP on Clarkdale isn't maxed out at 900Mhz; It too can be overclocked significantly. Here's an article on techgage showing an OC to 1133, for example.

    Remember that Intel is very conservative on binning parts to ensure no issues with reliability over time (compare vs nvidia mobile gpus).
    Reply
  • silverblue - Friday, January 08, 2010 - link

    Maybe so, however a 40/45nm AMD part could reach similar clockspeeds. I just don't think that, clock for clock, the new Intel IGP on the Clarkdale die is as powerful as anything AMD or nVidia can produce on the same scale. It's a good step forwards, just not the leadership that some may have been expecting. Reply
  • silverblue - Wednesday, January 06, 2010 - link

    Slight mistake... second best 32nm i5 that Intel has to offer. However, I doubt the performance increase over the 661 will be very noticable with the IGP; won't it be clocked the same in both? Reply
  • Zool - Monday, January 04, 2010 - link

    The irony is the biggest drawback of these cpu-s is the 45nm intel graphic on other die with the memmory controler. The die savings from 45nm vs 32nm are quite big.
    If they would make just 32nm dual core nehalem with memory controler on die it would be still much smaller(and only litle bigger than the clarkdale without imc) than the GMA die with memmory controler.
    The whole thing would be solved with everything as one on 32nm.
    I think plenty of people just wait for 32nm quad core nehalems without the useless GMA graphic.
    Actualy what is the cost of dirty cheap GMA in penryn based 3 package boards. Like 5-10 dolars ?
    Reply

Log in

Don't have an account? Sign up now