Novell NETWARE 6-DOCUMENTATION Manual page 2738

Statistic
TxUnderflowError
TXBufferError
RxECBsOver16MegCount
TxECBsOver16MegCount
PacketUsed2ECBs
300 Utilities Reference
Description
(NE2100, NE1500T, or CNEAMD) This counter is incremented
when something else on the bus takes control of the bus while the
LAN driver is putting the data on the wire.
If this occurs, the packet must be retransmitted.
(NE2100, NE1500T, or CNEAMD) This counter is incremented
when there is a problem with the transmit buffer.
This counter is usually incremented when TxUnderflowError is
incremented; it indicates a hardware problem in the server.
(NE2100, NE1500T, or CNEAMD) One of these counters is
incremented when either a transmit or receive occurs and the
driver has double buffered the ECB in the reserved buffers below
16 MB in memory.
These boards require double buffering because they have a
physical limitation that prevents them from accessing memory
above 16 MB.
Therefore, if the operating system issues an Event Control Block
(ECB) with a memory address above 16 MB, the board uses
some of the reserved buffers below 16 MB to queue the request.
These are not errors. This value tracks how many ECBs are
redirected to the buffers below 16 MB.
In many cases, this counter can be as high as the total packets
sent and received. This double buffering decreases performance.
If you have more than 16 MB of RAM and a board that is bus-
mastering or using DMA that is not a 32-bit adapter, performance
might be degraded.
(NE2100, NE1500T, or CNEAMD) This counter is incremented if
the Server Maximum Physical Receive Packet Size is set to 1514
bytes (default for NetWare 3.11 servers), and you need to receive
a near-full-size packet. For NetWare 3.12 and 4.x, the default
Maximum Physical Receive Packet Size is 4202.
In this instance, two ECBs are used instead of one, since the
CRC on the end of the packet requires an extra four bytes.
Using two ECBs instead of one may decrease performance
slightly.