Articles: CPU
 

Bookmark and Share

(10) 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 ]

Testbed and Testing Methodology

Together D2500HN and DN2800MT mainboards will give us a complete picture of the Cedar Trail platform. Especially since Intel decided not to no longer produce the 2.13 GHz Cedarview modification – Atom D2700. Therefore, it would make sense to select rivals for these mainboards among other existing nettop platforms. There are currently two platforms like that: the popular AMD Brazos and the previous generation Intel platform called Pine Trail. Note that we deliberately didn’t include Nvidia ION2. Being a very expensive solution with pretty high power needs it failed to become popular, so the mainboards based on it have almost vanished from the market by now.

During our today’s tests the AMD Brazos platform will be represented by ASRock E350M1 mainboard based on a dual-core AMD E-350 processor working at 1.6 GHz frequency. Although this is not the today’s fastest modification, it is definitely the most popular one.

The honor of Pine Trail platform will be defended by Gigabyte GA-D525TUD with an Atom D525 processor working at 1.8 GHz. This processor is the fastest Atom model manufactured using the old 45 nm process.

As a result, we ended up using the following hardware and software components for our today’s test session:

  • Mainboards:
    • ASRock E350M1 (AMD E-350 + AMD Hudson M1);
    • Gigabyte GA-D525TUD (Intel Atom D525 + Intel NM10 Express);
    • Intel Desktop Board DN2800MT (Intel Atom N2800 + Intel NM10 Express);
    • Intel Desktop Board D2500HN (Intel Atom D2500 + Intel NM10 Express).
  • Memory:
    • 2 x 2 GB DDR3-800 SDRAM DIMM (Kingston KHX1600C8D3K2/4GX) 6-6-6-18;
    • 2 x 2 GB DDR3-1067 SDRAM SODIMM (Apogee AS2G733-13G) 7-7-7-21.
  • Hard drive: Crucial m4 256 GB (CT256M4SSD2).
  • Power supply unit: Tagan TG880-U33II (880 W).
  • Operating system: Microsoft Windows 7 SP1 Home Premium x86.
  • Drivers:
    • AMD Catalyst 12.4 Driver;
    • AMD Chipset Driver 12.4;
    • Intel Chipset Driver 9.2.2.1034;
    • Intel Graphics Media Accelerator Driver 8.14.8.1077;
    • Intel Graphics Media Accelerator Driver 15.?12.?75.?50.?7.?2230.

Before we move on directly to the test results, we have to make one important point. Although we have switched to 64-bit Windows 7 version long time ago, this time we had to install a 32-bit one on all test systems. This was necessary because the graphics core of the Cedarview processors doesn’t yet have a 62-bit driver for it. It doesn’t exist even in a beta version yet, so currently Cedar Trail platform can only be regarded as a 32 bit one, even though the actual Cedarview processors formally support 64 bit extensions.

Since they licensed the graphics accelerator from Imagination Technologies, the graphics driver for the Cedar Trail platform is different from the standard Intel’s GMA driver. That is why we mentioned two completely different Intel GMA drivers in the description of the testbed above: one of them was used with GMA 3150 and another one – for GMA 3600.

Performance

Synthetic Benchmarks

We don’t often use synthetic benchmarks as they tend to disagree with real-life applications, but today is different. We're dealing with Cedarview, the new implementation of the Bonnell microarchitecture in silicon, and we want to use synthetic benchmarks to see the progress (or lack thereof) in performance over the Pineview series. We used PassMark PerformanceTest 7.0 (Build 1029) which incorporates a certain number of simple tests that perform various kinds of computations.

As expected, the new generation of Atom CPUs cannot ensure a dramatic performance breakthrough. The Atom N2800 is generally 3-4% ahead of the Atom D525 which has the same number of CPU cores and supports Hyper-Threading. This is in fact equal to the difference in their clock rates, which explains the results in most of the tests.

There are a few cases of a larger advantage of the newer CPU, though. Considering that the Bonnell microarchitecture was not revised during its transition to the 32nm tech process, the gap larger than 4% in some tests must be due to the increased performance of the memory controller. The memory bandwidth of the newer CPU is higher by a third.

 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 ]

Discussion

Comments currently: 10
Discussion started: 05/24/12 05:40:11 PM
Latest comment: 05/23/13 11:44:05 AM

View comments

Add your Comment