Vol. 3B 19-103

PERFORMANCE-MONITORING EVENTS

03H

02H

UNC_GQ_ALLOC.RT_L3_MISS

Counts the number GQ read tracker entries for which a 

full cache line read has missed the L3. The GQ read 

tracker L3 miss to fill occupancy count is divided by 

this count to obtain the average cache line read L3 

miss latency. The latency represents the time after 

which the L3 has determined that the cache line has 

missed.  The time between a GQ read tracker allocation 

and the L3 determining that the cache line has missed 

is the average L3 hit latency. The total L3 cache line 

read miss latency is the hit latency + L3 miss latency.

03H

04H

UNC_GQ_ALLOC.RT_TO_L3_RE

SP

Counts the number of GQ read tracker entries that are 

allocated in the read tracker queue that hit or miss the 

L3. The GQ read tracker L3 hit occupancy count is 

divided by this count to obtain the average L3 hit 

latency. 

03H

08H

UNC_GQ_ALLOC.RT_TO_RTID_

ACQUIRED

Counts the number of GQ read tracker entries that are 

allocated in the read tracker, have missed in the L3 and 

have not acquired a Request Transaction ID.   The GQ  

read tracker L3 miss to RTID acquired occupancy count 

is divided by this count to obtain the average latency 

for a read L3 miss to acquire an RTID.

03H

10H

UNC_GQ_ALLOC.WT_TO_RTID_

ACQUIRED

Counts the number of GQ write tracker entries that are 

allocated in the write tracker, have missed in the L3 

and have not acquired a Request Transaction ID.   The 

GQ write tracker L3 miss to RTID occupancy count is 

divided by this count to obtain the average latency for 

a write L3 miss to acquire an RTID.

03H

20H

UNC_GQ_ALLOC.WRITE_TRAC

KER

Counts the number of GQ write tracker entries that are 

allocated in the write tracker queue that miss the L3. 

The GQ write tracker occupancy count is divided by 

this count to obtain the average L3 write miss latency. 

03H

40H

UNC_GQ_ALLOC.PEER_PROBE

_TRACKER

Counts the number of GQ peer probe tracker (snoop) 

entries that are allocated in the peer probe tracker 

queue that miss the L3. The GQ peer probe occupancy 

count is divided by this count to obtain the average L3 

peer probe miss latency. 

04H

01H

UNC_GQ_DATA.FROM_QPI

Cycles Global Queue Quickpath Interface input data 

port is busy importing data from the Quickpath 

Interface. Each cycle the input port can transfer 8 or 

16 bytes of data.

04H

02H

UNC_GQ_DATA.FROM_QMC

Cycles Global Queue Quickpath Memory Interface input 

data port is busy importing data from the Quickpath 

Memory Interface. Each cycle the input port can 

transfer 8 or 16 bytes of data.

04H

04H

UNC_GQ_DATA.FROM_L3

Cycles GQ L3 input data port is busy importing data 

from the Last Level Cache. Each cycle the input port 

can transfer 32 bytes of data.

04H

08H

UNC_GQ_DATA.FROM_CORES_

02

Cycles GQ Core 0 and 2 input data port is busy 

importing data from processor cores 0 and 2. Each 

cycle the input port can transfer 32 bytes of data.

Table 19-20.  Non-Architectural Performance Events In the Processor Uncore for 

Processors Based on Intel® Microarchitecture Code Name Westmere (Contd.)

Event

Num.

Umask

Value

Event Mask Mnemonic

Description

Comment