Slow CMX512-2700C2PT Memory?

[geomak]

Ran Memtest86+, all tests passed with no errors. However, I noted that the memory bandwidth was reading 605MB/s, far less than the 2,664MB/s theoretical bandwidth. Here are some of the general results from changing various BIOS parameters from default settings (see end of post):

 

1) Reduced proc multiplier to default: no change

 

2) Reduced FSB to 133MHz: Bandwidth reduced to <600MB/s

 

3) Increased DDR Voltage to 2.7V(max): no change

 

4) Changed timing to 2-2-2-6: Increased bandwidth to ~688MB/s

 

5) Changed cmd rate from 1T to 2T: no change

 

If the bandwidth were say in the range of 80% to 95% of the theoretical bandwidth, I'd say that's normal. The values I'm getting however seems way too slow. I tested out another module of Corsair Value Select RAM PC2100 at 133MHz and that came out with a bandwidth of ~520MB/s,

roughly 1/4 of the theoretical maximum. Perhaps Memtest86+ is incorrect, or perhaps it doesn't run at the full bandwidth? I find that hard to believe...

 

Default Settings:

 

memory

Speed: 166MHz (333MHz DDR)

Timing: 2-3-3-6 1T

Voltage: 2.5V

Bank Interleave: 4

Queue Depth: 4

 

cpu

Mult: 12x (~2.0GHz)

FSB: 166MHz (333MHz DDR)

Volt: 1.65V

 

system

mem: CMX512-2700C2PT

mobo: Shuttle AK-35GTR2

cpu: Athlon XP 2500+ Barton (overclocked)

[RAM GUY]

This system is based on KT333 chipset and that would be about correct for this MB and or chipset.

[geomak]

This system is based on KT333 chipset and that would be ablout correct for this MB and or chipset.

 

If that's the case for this chipset (KT333) what about the nForce 2 chipset on say a MSI K7N2GM2-LSR mobo; using the same type of memory (CMX512-2700C2PT)?

 

Is a factor of 0.25 of the theoretical maximum for the sustained data transfer rate common to all chipsets or just this particular KT333 on my

particular motherboard (Shuttle AK35GTR2)?

 

Thanks!

[RAM GUY]

You really need to take this up with the MB maker, but NForce2 chipset will have better memory bandwidth.