MSI Neo 4 platinum

ohwell78

Member
Aug 10, 2005
52
0
0
I have a ram problem on this board. 2 things.

There are 4 sticks of corsair xms pc 3200 2x256 and 2x512

1. Im running the 180 multip at 230 fsb. for 207 2/3/3/6 (2t) timings. 2.7 volts. In CPUZ only shows 195 fsb. shows 207 during post and in bios.

2. In memory tests on pcmark and sandra I am running only 5gbs when it should be running closer to 6 gbs. This same board/ram on numerous review sites scores the 6gbs.

Ive ran the ram stable at 225 1:1 ratio for the 230 had to do the multi.

My bios version is 1.5. Also I cannot overclock past this amount. Its completley stable at 11 x 230 but wont even posts at 11 x 231. Its a san dieo 3700 chip.

ANY HELP ON ANY OF THESE PROBLEMS WOULD BE APPRECIATED.

Ive ran Memtest works fine, Ran super pi out to 32 m digits heat never got above 47c. Took about a halve hour.
 

the cobbler

Senior member
Mar 8, 2005
643
0
0
using 2x256 sticks + 2x512 sticks= single channel mode, which is likely the problem with your memory benchmarks

also if you are using a Winchester, the memory controller will knock four sticks down to DDR333 and 2T.
 

the cobbler

Senior member
Mar 8, 2005
643
0
0
my bad, that is weird then

try the 1.6official or one of the 1.7xbetas (but you lose 120/140/150/180 dividers)

yeah that is definitely not right, I am getting in the 61xx range on Sandra benchmarks running crappy value ram at 217mhz and worse timings (2.5,3,3,6 1T)


wait a minute...figured out what's going on...

CPU-Z and CPUID can't read the MSI memory dividers properly (both programs think they are DFI dividers)
CPU-Zseems to be applying the DFI 188 divider to the speed you are already running your RAM at 207 on the 180 divider.
I run the MSI 140 divider and RAM speed doesn't even register with CPU-Z or CPUID, shows up as '-1.0' in ClockGen.

Check your Ram speed with Everest, it is the only program that correctly reports my Ramspeed on an even MSI divider.

MSI 180div=0.9
DFI 188div=0.94

Cpuz seems to be taking your actual ramspeed (230x0.9=207mhz) and calculating it with the DFI divider for some reason on top of that (207x0.94=195mhz)

check everest, should show accurate ramspeed
 

cruZZade

Junior Member
Aug 20, 2005
5
0
0
It's because you're using the memories in 2T...if you can't use them in 1T, then you won't get the high scores...
 

ohwell78

Member
Aug 10, 2005
52
0
0
You cant use 1t with 4 dimms. It sucks but is true. So using 2t alone drops me a full gps memory bandwidth?
 

the cobbler

Senior member
Mar 8, 2005
643
0
0
no, 1T vs. 2T is going to make 1-3% difference in performance tops

not a 20% difference

and that doesn't explain why the ramspeed is being reportedly incorrectly either
 

ohwell78

Member
Aug 10, 2005
52
0
0
When I dont use a multi at all it shows up ok. But I cant run my ram at 230 1:1. Do you have the identical board and chip? and if so how high can you get your overclock?
 

the cobbler

Senior member
Mar 8, 2005
643
0
0
yes and no. same exact board, not platinum revision (non-Ultra). NF4 and NF4 ultra are identical for OCing purposes. I have a VNF4Ultra that overclocks basically identically to my MSI.

winnie 3200+, but this board has bugs where ONLY the Winchester and Venice are concerned, SanDiego is unaffected. You shouldn't have any issues with the CPU itself.


I am running 305fsbx8=2440
RAM on 140div ~219mhz, 2.5,3,3,6 1T (2x512 Corsair value)
with better cooling and more voltage, my Winnie goes to ~2.65 and remains stable, 2.7=not stable.

I really don't know what's going on there, but I bet is has to do with running 2x256+2x512

the sticks may not actually be identical, I would take out 2x256 run with just 1GB and see what happens
 

ohwell78

Member
Aug 10, 2005
52
0
0
I know there not identical as they were pruchased at different times, however it does say dual channel everywhere, which is what i thought be identical mattered for.

I ask why you went with the lower multi? 8x305?
 

the cobbler

Senior member
Mar 8, 2005
643
0
0
because these three things influence performance, in order:

1. CPU speed
2. FSB speed
3. RAM speed/latency
people get 2 and 3 mixed up too frequently

I can run 244x10, 271x9, or 305x8.
305x8= fastest of the three assuming RAM is running at same speed/timings
gives me an extra 5+ fps in BF2 and close to 150 points in 3DMark05

so when you factor in memory dividers:
244x10=2440 @180 divider=219mhz RAM
271x9=2439 @166 divider=225mhz RAM
305x8=2440@140 divider=219mhz RAM

the extra 34FSB makes a much bigger difference than extra 6mhz RAM
also have value ram, so I can't change timings much (can't loosen and go higher, can't tighten and go lower)
 

ASK THE COMMUNITY