dexx
06-01-2005, 08:11 PM
Ive been using a64tweaker to play with my new corsair ram. I have 2x512mb xms3200 v4.3 which i'm told are either TCCD or TCC5. I'm running dual channel 1:1 280x9 in my DFI ultra D. Base timings are 2.5-3-3-10-1T @ 2.65V. Interestingly if i increase voltage i get errors in memtest.
A64tweaker shows the following
Tcl 2.5 Rd Q bypass 16x
Trc 12 contr mode 128bit
Trfc 16 ECC disable
Trcd 3 2T disable
Trrd 2 bypass max 7x
Tras 10 max async lat 9ns
Trp 3 Rd preamble 6.5ns
Twr 2 idle cyc lim 256clk
Twtr 2 dyn idle cyc disable
Trtw 2
Tref 3120
Twcl 1
skew 0 off
Using Everest to benchmark, i scored a mem read of 7250 and latency 40.6
Some settings that made a difference:
reducing Trtw by 1 added 10 to read speed
reducing read preamble to 5ns added 5 to read speed
The one parameter which made a noticable difference was 'max async latency'. Reducing this from 9ns to 7ns increased read speed from 7250 to 7315 (0.9%) and reduced latency as reported by Everest from 40.6 to 37.9 (6.5%). No other minor parameter made any difference to the scores.
Looking at the major parameters:
reducing Tras from 10 to 8 increased the read score from 7250 to 7315 (
Tras = 7 gave 7325 read
Tras = 6 gave 7353 read
Tras = 5 gave 7305 read
So Tras of 6 was 1.4% faster for me than Tras of 10
A64tweaker shows the following
Tcl 2.5 Rd Q bypass 16x
Trc 12 contr mode 128bit
Trfc 16 ECC disable
Trcd 3 2T disable
Trrd 2 bypass max 7x
Tras 10 max async lat 9ns
Trp 3 Rd preamble 6.5ns
Twr 2 idle cyc lim 256clk
Twtr 2 dyn idle cyc disable
Trtw 2
Tref 3120
Twcl 1
skew 0 off
Using Everest to benchmark, i scored a mem read of 7250 and latency 40.6
Some settings that made a difference:
reducing Trtw by 1 added 10 to read speed
reducing read preamble to 5ns added 5 to read speed
The one parameter which made a noticable difference was 'max async latency'. Reducing this from 9ns to 7ns increased read speed from 7250 to 7315 (0.9%) and reduced latency as reported by Everest from 40.6 to 37.9 (6.5%). No other minor parameter made any difference to the scores.
Looking at the major parameters:
reducing Tras from 10 to 8 increased the read score from 7250 to 7315 (
Tras = 7 gave 7325 read
Tras = 6 gave 7353 read
Tras = 5 gave 7305 read
So Tras of 6 was 1.4% faster for me than Tras of 10