|
Advice Contact Forum SiteMap Sponsors Home PC Hardware from Made-in-China.com |
. | |
|
Prices: |
Resources Controlled by (Auto(ESCD))
ESCD stands for Extended System Configuration Data and describes a feature that has been implemented with the PnP specifications. The ESCD are stored in a dedicated part of the CMOS memory also referred to as ESCD Pool. The ESCD contain information about the system configuration at the last time the system was booted, keeping a record of the peripheral device configuration that was negotiated with the system and resulting in a conflict-free setup. Keeping a record of these data eliminates the need for re-negotiation of resource allocation and, consequently, helps to speed up the boot time, unless changes in the hardware have occurred. A typical sign for the use of ESCD is the display of an "updating ESCD Pool" string during boot-up.
PCI Latency Timer (CLK) (64)
The peripheral component interconnect (PCI) bus is shared between all interrupts. In order to get access to the bus, each device needs to arbitrate and to capture the bus, which excludes other devices from using it at the same time. Different Interrupt Request lines that are assigned to the individual devices have different priorities which can lead to starvation of a lower priority device by a higher priority device. In order to avoid this problem, fairness algorithms have been implemented to define, for example, how long any device can occupy the bus before it has to relinquish it and check whether there are other system requests that need to be serviced. This interval is called the PCI latency.
Each PCI bus access also needs a certain number of cycles, usually around eight, for the arbitration, including the setup of the device on the bus. The number of setup cycles is fairly constant, irrespective of the total latency. Therefore, a higher number of cycles assigned to the PCI latency will only increase the number of transfer cycles and directly translate into better device performance. On the other hand, with more PCI devices in the system, more devices will vie for the bus and, therefore, very high latencies could result in the above mentioned starvation of one or the other device.
In most modern systems, the empirically found best performance across the board is between 64 and 96 cycles, depending on the feature set that is implemented in the PCI configuration. For example, it is possible to run at lower latencies if so-called bus parking is implemented, that is, the PCI device relinquishes the bus according to the prespecified latency but remains in a “parked” position to facilitate an immediate re-assertation if there is no competition from other devices.
Note that the PCI express (PCIe) bus is not shared with the PCI bus. PCIe interfaces are using dedicated lanes for interconnect that are not shared.
|
AMD Athlon64 X2-3800+ (AM2) |
next page: => Miscellaneous: Power Savings, Memtest =>
All advice and educational articles on LostCircuits are free, but if you feel you can, please make a small donation to us!