65nm 3500+ WaterCooled actions...

[AMD-me]

Those SPi times are slow. My 2.8ghz 3000+ does 30.21s.

[funkflix]

Those SPi times are slow. My 2.8ghz 3000+ does 30.21s.

65nm + DDR2 isn't the best for SuperPI.

http://www.xtremesystems.org/forums/...d.php?t=135063

Reached 28.984s, 3000MHz with the 3500+ plus DDR2 and 26.812s with Venice 3000+ 3106MHz.

@Lightman

Not the best clockers, but fun to play with that stuff!

[Lightman]

@VVJ
Many thanks for quick update
It is working much better now but still I found something



On this pic top CBIds are from left:new beta, older version, new Beta and lower one is new again. When I opened first new CBId it showed wrong clock (2933) but another instance showed correct one??

@funkflix and AMD-me
You need to keep in mind I'm on single channel at the moment and 64bit version of Windows . My best time @3GHz was 30.17 for 1M SuperPi (pic above ). Also I found that when my 250GB USB2 HDD is connected and running SuperPi 1M@3Ghz takes >40 seconds!! This is some sort of bug with this mobo.

@funkflix
Yes, great fun to play with G1 revision

[VVJ]

Lightman, well, it seems there is a lack in measurement. I think it is better to check a dynamic CPU frequency at the Mobility tab. Nonetheless, I will think something out to fix that. Thank you for the feedback, Lightman!

[Lightman]

@VVJ
New version is reporting wrong clocks consistently but I forgot to mention that in meantime I upgraded Windows from SP1 to SP2 and it changed something. Now I can't OC using ClockGen!! nVidia Control Panel is working fine (but only to 250HTT ). Luckly I overcome HTT clocking issue from BIOS as you can see . Spread Spectrum was causing that problem .

Small update with higher HTT/memory and lower multiplier. Nice gains in 3D tests and SuperPi. Also latency is getting extremely low .




All other settings like GFX Card clocks SAME as on previous tests!

[high5]

Lightman,

Your PCI lock seems operational now, what was the problem?

[Lightman]

Lightman,

Your PCI lock seems operational now, what was the problem?

As I said above Spread Spectrum was enabled. This option is well hidden in MSI BIOS . Besides most motherboards switch it off automatically when OCing but not this one .

On the side note: good mobo for it's price (picky but good )

[high5]

I read you loud & clear about spread spectrum.

It's just that you mentioned it in HTT clocking context:

Luckly I overcome HTT clocking issue from BIOS as you can see . Spread Spectrum was causing that problem .

BTW I agree, board looks like a decent value.

[Lightman]

Small update after installing 2nd DIMM
Now in Dual Channel mode!



Memory timings untouched (CL4-4-4-12-15 2T)

[NickS]

Very nice. I love MSI boards too, dunno bout you.

[Lightman]

Very nice. I love MSI boards too, dunno bout you.

Yes! It's great fun to play with this combo (BTW this is my 5th MSI mobo ever).

I also did some 3D action again to check impact of second DIMM on 3DMark2001.


On side note, can anyone tell me what are average memory transfers on AM2 single core DC setups in SiSoft?? Is my 49% bandwidth efficiency normal? Seems to be pretty LOW .

[Calmatory]

Sorry to bump this thread up, but I am curious to know what did you do to achieve that FSB. Disable Spread Spectrum and put HTT to 400(X2)?

As there are few reviews which show that the mobo starts getting problems when going for 240+ FSB. Also, have you tested the max. possible FSB for the board?

[Lightman]

Sorry to bump this thread up, but I am curious to know what did you do to achieve that FSB. Disable Spread Spectrum and put HTT to 400(X2)?

As there are few reviews which show that the mobo starts getting problems when going for 240+ FSB. Also, have you tested the max. possible FSB for the board?

Hi! Basically I disabled every SpreadSpectrum BIOS option possible, changed HTT multiplier to 3x and then started to bump HTT clocks (and CPU voltages ). TBH I never tried max HTT on this mobo but I reached 310HTT pretty easily! If you want I can find out what this MSI can do...

PS. FSB-->HTT
PS2. I did couple of quick tests and Hyper-Transport clocked@~1200MHz (effective) isn't a problem on my motherboard.

[ozzimark]

just curious, how low can you get the 1M time at 3ghz and 500mhz memory clock? 29s seems a smidge slow

[Lightman]

just curious, how low can you get the 1M time at 3ghz and 500mhz memory clock? 29s seems a smidge slow

Here is an answer:



Bear in mind I'm benching on Win XP x64 so SuperPi runs in virtual x32 environment and besides system is not tweaked towards SuperPi. Probably I could get another 1s or so with proper tweaks....

[Calmatory]

Hi! Basically I disabled every SpreadSpectrum BIOS option possible, changed HTT multiplier to 3x and then started to bump HTT clocks (and CPU voltages ). TBH I never tried max HTT on this mobo but I reached 310HTT pretty easily! If you want I can find out what this MSI can do...

PS. FSB-->HTT
PS2. I did couple of quick tests and Hyper-Transport clocked@~1200MHz (effective) isn't a problem on my motherboard.

So, the problem in the reviews is the HT bus, as when HTT hits 240, HT goes to 1200MHz and can not go beyond it.

I am still in early Socket A -age, so FSB feels more comfortable for me, at least as long as I have to work with this ancient system.

[elsupremo]

Is Spread Spectrum on by default?

[Lightman]

Is Spread Spectrum on by default?

Yes, for HTT, SATA and PCIe spread spectrum is on by default!!

BTW new BIOS ver.3.5 seems to change HTT Spread Spectrum to disabled after raising frequency, so at least something improved !