Intel DP55KG "Kingsberg" Motherboard
This motherboard is based on the Intel P55 chipset (P55 bridge). The motherboard contains several additional controllers to expand its basic functionality:
- Integrated audio, based on the 10-channel (7.1+2) HDA codec Realtek ALC889 (top codec from this company supporting audio output from HD DVD and Blu-ray in full quality), 7.1-ch audio, front line-ins/outs, optical S/PDIF-Out and S/PDIF-In (!) jacks on the rear panel, S/PDIF-Out to output sound via a graphics card with HDMI.
- Gigabit Ethernet (Intel 82578DC PHY and chipset MAC adapter) 10/100/1000 Mbps.
- IDE/SATA-II RAID controller based on Marvell 88SE6145 (PCIe x1) supporting one PATA channel (not used in this motherboard!) and four SATA300 ports (two on the board and two eSATA ports on the rear panel) that can form RAID 0, 1, or 10.
- FireWire controller, based on Texas Instruments TSB43AB22A (PCI), supporting two ports (one of them is installed on the rear panel).
- Bluetooth 2.0 USB (Cambridge Silicon Radio).
Intel is practically the only manufacturer that uses standard networking chipset features in top motherboards. However, it would have been strange to implement gigabit network support in its chipsets and then ignore it :) PHY controllers with the GLCI interface (used in the DX58SO i82567LM and now in the i82578DC) are too expensive for other manufacturers. That's why Realtek RTL8111DL is so popular (sometimes even in pairs). However, these manufacturers ignore the fact that most owners of home computers use only one integrated network controller at 100 Mbps at best (a typical consumer is an ADSL modem or an inexpensive SOHO router), the second controller is not used at all (not separately, not for aggregation).
So, this Intel solution is necessary (as network support is mandatory these days) and sufficient (even a bit excessive). Still there is one 'but': they should have used the 82567LM instead of the 82578DC, because the latter lacks support for vPro, MACsec (IEEE 802.1AE), and even jumbo frames. The former is relevant only for corporate users (who won't use motherboards with skulls and blue illumination anyway), but the latter can hit home enthusiasts, who already entangled their houses and apartments with gigabit networks. However, it won't be a big problem for all users, so this gripe is just grumbling of a perfectionist.
Unfortunately, we are forced to establish a fact that insufficient number of PCIe lanes (or lanes wasted on PCIe x4) does not allow to setup an extra disk system of decent quality. Marvell 88SE6145 is a time-proven chip, but the idea of a four-port PCIe x1 controller does not look sane to us. It's because each of these ports has theoretical throughput of 300 MB/s, while the entire external interface is just 250 MB/s (theoretically). So not only RAIDs must be based on the chipset controller, even single drives can interfere with each other. However, it's not a big problem: 99% of users will use the additional controller only for a single eSATA drive, if they use it at all, so the throughput of a bus interface must be enough. Theoretically, that is.
In order to evaluate throughput of various interfaces, we performed a simple test using with the help of a 32GB Kingston SSDNow SNE125-S2/32GB SSD drive.
Even though ICH10R and P55 are closely related (even their drivers are compatible), performance of the SATA controller in the new chipset is lower than that of the old controller. However, this is relevant only for the fastest SSD drives, not for popular hard drives, which still cannot exceed 150 MB/s (now we understand why Intel is so skeptical about how fast chipset controllers should master SATA600). Performance of discrete SATA controllers does not compare even with P55. Marvell is especially bad at writing. It's so bad that its restrictions will affect even hard drives. (Don't be surprised, when a new external hard drive connected to eSATA won't be much faster than in case of USB -- it's not an HDD problem.) Perhaps, it will do better in the AHCI mode (we traditionally use RAID for such tests). Besides, we don't rule out a possibility of some motherboard-related problems, but it's highly unlikely. What concerns USB, we are again forced to establish a fact that Intel chipsets do not lead in bus speed. However, this 3-5 MB/s difference is insignificant considering the future USB 3.0 and already existing high-speed interfaces (like eSATA).
Write a comment below. No registration needed!