Revisiting Corsair XMS3200C2v1.1 BH-6 @240mhz 2226

eva2000 · Post by **eva2000** » Sat May 29, 2004 6:21 pm

Well with Corsair and OCZ announcing their new ram based on Samsung TCCD 4ns, XMS3200XL and OCZ PC3200 Plat rev2 rated at DDR400 2-2-2-5 and 2-2-2-8 respectively, I decided to pluck out my good old Corsair XMS3200C2v1.1 winbond BH-6 ram to do some comparison benchies

Using same additional ram cooling with 60x38mm delta 50cfm fan, I proceeded to revisit Corsair XMS3200C2v1.1 BH-6 (batch/lot number 024151 = week 41, 2002).

2.8C SL6Z5 M0
Abit AI7 17bios
128MB Gigacube 9800PRO Extreme
80GB WD 800JB
550W Antec TrueControl PSU

According to infra-red thermometer, heatspreaders are a cool 26-29C under memtest #5 looping load (most stress you can put of ram more stressfull that anything you can run in windows!)

Notes:
1. cpuz on AI7 reports wrong cpu/ram ratio when NB Strap 667fsb is used and swaps 1:1 and 5:4 around. Notice ctiaw reports correct bus speed as 240mhz and ratio as 1:1

2.To run unbuffered you need to go to sandra mem module -> options and uncheck 9 boxes listed in my unbuffered screenie link below

-----------------------
Results = 2x 512MB XMS3200C2v1.1 BH-6 @ 240mhz 2-2-2-6

vdimm = 3.2v in bios (3.17-3.19v vdimm)
vAGP = 1.65v

NB Strap 667fsb
GAT = A-A-A-D-D

Sandra 2004 Pro SP1
Buffered
Unbuffered = 3625/3627 (unbuffered real speed of ram)

Pifast = 57.52s

SuperPi
2M = 1min 32s
16M = 16min 15s
32M = 35min 30s

Aida32 3.93
Read = 6101MB/s
Write = 2160MB/s

240mhz 2-2-2-6 at 3.18v 1:1 ratio

this is prime95 v23.8 x2 at 25hr 22min

more benches to come :D

Winbond BH-6 is

Notes:
1. cpuz on AI7 reports wrong cpu/ram ratio when NB Strap 667fsb is used and swaps 1:1 and 5:4 around. Notice ctiaw reports correct bus speed as 240mhz and ratio as 1:1

2.To run unbuffered you need to go to sandra mem module -> options and uncheck 9 boxes listed in my unbuffered screenie link below

Hope you enjoy

eva2000 · Post by **eva2000** » Sat May 29, 2004 6:22 pm

huh what's causing the page to go off like that, my images aren't even stretching out that far ????

Maestro · Post by **Maestro** » Sat May 29, 2004 6:30 pm

Nathan Kirsch compared the new Samsung TCCD parts to the old Winbond BH-5 and concluded they're about equal in performance at 3200, but the Samsung is the winner since it overclocks much better and supports good latencies at PC4000. BH-6 is not as fast as BH-5, but that won't matter in a month because all Winbond will be gone and the new Samsung chips will be king of the hill.

infinitevalence · Post by **infinitevalence** » Sat May 29, 2004 10:43 pm

We sure are going to miss the BH-5 and BH-6. RIP Winbond.