The OFFICIAL MSI K8N Neo2-939 tips and tricks guide thread

[sideeffect]

The .5 multis work just fine with athlon 64 all the bios versions before 1.9. They removed them from the 1.9 and over bios's.

The values are completely gone so cant be enabled.

[ElAguila]

I was told in another forum the .5 multis actually used the next highest whole number for all of the settings.

[sideeffect]

Yes i think it did. But that made a big difference because it made options unavailable available.

Some combinations of cpu ram speed cant be got with the 166 133 divider unless you use .5 multis so the motherboard must be setting 150 or 183 divider to keep the spec.

As 183/150 dividers aren't available to be set the .5 dividers are a kinda hack to allow it. I guess msi thought that they would just enable the 183/150 divider and do away with the .5 multis. But they did away with the multis but didnt add the dividers. I hope that i can enable the divider again.

For example 310 x 10.5 = 3150 mhz cpu speed with a 166 divider ram = 242.3
286 x 11 = 3146 mhz cpu speed with a 183 divider ram = 242

So if the multi is raising from 10.5 to 11 then the neo2 must be using the 183 divider to keep the ram in spec.

[sideeffect]

Ok im running my modded bios now. The extra voltage settings dont work. Shame that.

However the 183 divider hack does work and im now running in a 183 divider set with bios

Going to test 150 divider now.

Going to make a few changes then ill post bios.

[Lestat]

yes all the 1/2 multiplier does is use the 183 divider.

thats what all the amd64 boards do. all 1.2 cpu multipliers use a 183 divider.

technically its not even a 183 divider its more than that i believe.

but anyways they are absolute fools for removing the 1/2 multiplier.. it is and dont say im synical its fvcking retarded same as its retarded to remove the "press insert for safe mode" that is a fvcking life savor and they took it away cuz they are bloody retards.

And now they have made it so you cant even use a PCI SATA or IDE card,,, what kind of b.s. is that i mean really.
the NVRaid is crap, its slow and poop compared the the Sil, Promise, Highpoint and other SATA controllers.


BUT i wll say this..... whatever they did i can get better ram freq. but like i said earlier i think it was never MSI i think it was NFM and Murdock locking those 2 and 3 ram settings so we couldnt change them in the bios. it was too tight.. expecially for TCCD ram.

but so far so good.

Lastly im not happy with the NVraid rom thats in this bios. its definately has some speed and cache issues compared to the others. The hard drive isnt pulling up apps and windows as fast as before. not sure what it is as ive never seen an nvraid rom but something is definately not right with my Raptor performance now.

[sideeffect]

OK 150 and 183 dividers now work in my modded bios. Going to make a few changes then ill post bios.

Lestat if you want another nvraid rom in a custom bios that can be done easily. However it might just be a problem with the latancy. Use powerstrip to set latency and use hdtach or atto diskbench to record differences.

[sideeffect]

Murdock never released a bios with modified memory timings and NFM did his mods with a ISA hack and the timings were looser than msi's in all the dividers.

The pre 1.9 bios files didnt have any memory timings for the modders to change so it was all msi's fault that they were to tight.

However the 1.9 / 1.A timings are a touch looser than the 1.8. However i think your performance gain is due to the improved venice support in the newer bios.

[Lestat]

the gain is 100% memory i didnt gain anything on my cpu.

[sideeffect]

No Lestat the memory controller is on the cpu so any improvements to the cpuid will effect support for the venice hence support of the memory controller hence better ram speed.

[sideeffect]

My bios is ready to be used by anyone who may wish to use it. Its been tested by me and works fine but as always the risk is yours to take.

In my knowledge this is the first neo 2 bios to allow 183 and 150 dividers to be set in the bios.

Hope you enjoy it get it from here Download

[ElAguila]

I just downloaded it. I will give a try tomorrow. Great work getting the 2 new dividers added in. On another note, what exactly does the ati/nvidia speedup do that fastwrites doesn't do?

[sideeffect]

ATI/Nvidia speed up overclocks the video card by about 5 percent over stock values. The problem is if you have already found your maximum gpu overclock which most people have then this setting will push the card over the edge and you will get graphics errors.

Fast writes is a setting which allows faster access to the graphics card ram. However in the real world the performance effect is 0 and you can lose stability its widely accepted to disable fast writes on all graphics cards.

I set the default to off for fast writes but you can turn it back on in bios.

[peterle]

Does the memdivider 150+183 really work??? Murdock tried it also but they did not work, u just could select it at the bios but 150 gone 166 and 183 gone 200 ??

[desolator]

Great Work Side effect...

any other specefic features u made for this bios?

and which original bios did u modify?

[Sonic9]

Great ! I can re-test my overclock ... it's possible to tune SideEffect's bios (I think NFM can no ? ) these values :

Read Write Queue Bypass to 16x
Bypass Max to 7x
Idle Cycle Counter to 256clks

and why not Read Preamble to 5.5ns ... or leave default, what you think ?

[Lestat]

ok check this

10x cpu
183 = 188

9x cpu
183 = 185

thats kind of a rough estimate as im still trying to figure out what its doing.

10x cpu 275HTT = 250mhz ram
9x cpu 275 HTT = 248mhz ram
8x cou 275 HTT = 245mhz ram

the cpu multiplier is changing the memory frequency. which MAY be why they never enabled it. They cant get it to hold the freq like 200 166 and 133

so far i dont see anything different with this bios and the original. other than the obvious removal of a few items. (thanks btw for that!!)

SE i thought you had enabled some more memory settings ? or did you say you tried but it wouldnt work.?

and as for my comment about overclocking. I know the memcontroller is ondie i simply meant that this bios is giving me more memory overclock. i havent gained or lost any overclock on the cpu. Other than now im stuck back at 1.65v being the max vcore.

[Lestat]

hhmm ive lost about ohh 300+ on sandra running the same settings as murdocks.

SE is there some setting that you have in the memory that is pretty loose ?

i was running 2.5-4-4-10 @ 275mhz and getting 71XX in sandra
now im getting 6800~6900 give or take.

take a look at my A64 tweaker and tell me what i can do to either keep the same HTT and get more performance orrrr.... get a higher HTT/Mem without running 2T... i just wont run 2T lol.

i can adjust the 2.5 to 3 and the other 3 values ( 4-4-10 ) and not really get alot more, i didnt know if there was something in the bios right now that i can adjust or something with A64 tweaker that i can adjust. to get it higher than 275 2.5-4-4-10

I just havent had a need to know about those other values before thats why i never really learned what adjusting them does or i'd do it myself lol.

[syar2003]

Nice work Sideeffect , been missing the 183 for a long time !

Lestat !
Multiplier 8 gives the following 183 divider : 9
Multiplier 9 gives the following 183 divider : 10
Multiplier 10 gives the following 183 divider : 11

Actual ramspeed is cpufreq/divider
8*275 =2200mhz , ram freq= 2200/9 ~ 244.44mhz
9*275 =2475mhz , ram freq= 2475/10 ~ 247.5mhz
10*275 =2750mhz , ram freq= 2750/11 ~ 250mhz
So it looks like it's okay

Divider tables here :
http://www.xtremesystems.org/forums/...ad.php?t=41595

[Lestat]

hhmm i dont recall my DFI dropping the ram Freq when you dropped the cpu multiplier but then again i probably didnt pay THAT much attention to 3mhz.

i simply commented on that as a curiosity not an issue. but its good to know thats the way its suppose to be.

i need to do more testing BUT....

i lost 1000 points in 3dmark 05
i lost 2000 points in 3dmark 03
i didnt lose anything in Aquamark really maybe 1500 from a high 75k to a low 74k. but thats minimal.
I did install the new nvidia 78.11 drivers which has an insane Image Quality setting and color saturation even when the Nvisia control panel is set the total performance.
Also Triple buffering does not work on the drivers so that maybe causing abig hit.

last i did try 512 meg 256meg and 128meg AGP aperture and it didnt change.
512meg aperture used to increase 3dmark 05 scores but it didnt help.

again it may just be the drivers but i benched yesterday and i was getting my usually crappy score of 57xx with 3dmark 05.
I say crappy cuz with my old 754 3200 clawhammer and a 6800 GT i used to get over 6100 in 3dmark '05.
but anyways.

no issues so far other than wanting to squeeze some more hz out of my ram..

[sideeffect]

Yes ill work on memory timings next.

Yes the 183 150 dividers really do work. They dont show the correct changes in the bios. Load windows with the 183 divider set and load athlon 64 tweaker and you will see it has set the 183 divider. It has no problem holding the frequancy and is exactly the memory frequancy you would expect from the 183 divider if you dont believe me use This tool it will show you what frequancy you will get in a certain motherboard with a certain memory clock set. Tick DFI to allow it to show the 183 and 150 dividers. The speeds it sets corolate perfectly with what the bios sets.

Lestat the memory timings are all available to be changed in bios theres no reason why another 1.A would be faster with the same timings.

Losing 1000 marks in 2005 is loads are you talking about 1.8 vs 1.A or are you talking about my bios vs murdoks.

Or is it because your IRQ is sharing the graphics card with the smbus like you were saying before.

Im going to benchmark now and test this.

Edit - I get about the same results with 3dmark 2003 as i did with the 1.8 bios. If you get lower i think its due to the IRQ. My graphics card is still on its own IRQ in 1.A bios.

[ElAguila]

I won't get to do anything until this afternoon but I will flash it then and see how the dividers are doing for me. Thanks for the work on this!

[sideeffect]

The 150 divider boots with quite tight timings like the 166 im going to try my best to fix that today and release a rev 2 bios.

the 183 divider boots with lovely timings the same as 200.

Ok im going to try to force

write to Read delay (Twtr) = 2
Read/Write Qued Bypass = 16
Bypass Max = 7
Maximum Async Latency = 8
idle cycle limit = 256

In all dividers. Does that sound ok to everyone?

[syar2003]

Whats with everyone and this 256 idle cycle limit myth .

There are trade-offs if the counter is set too high since it may interfere with DRAM refresh. Empirically, we have found that a value between 16 and 64 cycles will give the highest performance in e.g. gaming applications .
(lostcircuits)

I would like to have it left on 16clks .

[sideeffect]

Ill leave that at default then. From what i heard it helps stability in tccd ram at higher clocks. In my experiance it makes no difference in stability on my Crucial ballistix.

The main settings that i really want to change are max async because 7ns can be to tight if you want to run high mhz ram speeds. And Write to read delay (Twtr) because it defaults to 1 in 166 150 divider and that can limit the overclock. The other settings i have seen no benifit or deficit in changing.

Edit - ok i have hacked the new memory timings in going to test bios then release rev2

[2virgule5]

Just had a try of last sideffect modded bios : 150 option is not listed (only 100, 133, 166, 183 and 200), but 183 option did work well, I'm now running my X2-4200+ at 240 x 11, RAM is at 203.1 Mhz. I'll try higher HTT @ x10 later. CPU voltage (as read in bios and with CPU-Z) dropped by 0.05 to 0.1 v when I select 1.55 vcore +10% (it is at 1.55, was about 1.6 to 1.63 with 1.A3 with the same settings).

Another good news : it also seems that the cold bug disappeared, as I now can cold boot without problem at those settings (I had to reset CMOS at those with crappy 1.9 and 1.Ax).

Thanks for the work !