p5n32-e sli plus oc and voltage problem

[Star5790]

hi everyone, this is gonna be a long post, so please bare with me, i have never overclocked before and im stuck. a friend recommended these forums to me so im hopeing your community and its members can give me some suggestions

my system is

asus p5n32-e-sli plus mobo
e6600 conroe @ 2.4 stock
2 gigs corsair dominator ddr2 6400c4d
evga 8800 gtx superclocked
corsair 620 hx psu2 lite on optical drives, one with lightscribe
2 wd sata 320 gigs hd's


if there is anything else u need to know, just ask.

ok, now the problem and how we got to it.

we overclocked the cpu to 3.2 with a 1600 mhz fsb 800 mhz mem with a cpu multi of 8, ran orthos cpu small ffts with no problems. we then ran a blend test and it wont pass 2 min without errors, we changed the mem timings to 444 12 and pushed the voltage on the me up to 2.0, no deal, then 2.1, no deal there either. so we changed teh timings back to 555 18 with the voltage at 2.1 and still didnt get the test to pass 2 min without errors. changed the fsb voltage to 1.5 by small increments but still no deal. now the problem is not just that this test wont run. its that my fsb voltage jumps from 1.5-1.6, a full tenth of a volt, 1.6 isnt even an option for me to manually set it to! which ive been told is very dangerous. i did some checking, made sure speedstep was disabled, and it was, did more checking and made sure everythign else was disabled, still got the fluctuating voltages. i cant afford new parts, and i really would like to oc this proc to a min of 3.2, ive read af lots of ppl that have had stable oc's of up to 3.6 with this board. so i still have some hope. be that as it may, im stuck guys, im new to this and i dont understand a lot of what ive been reading and im looking for some help, any of which would be very much appreciated. thx =]

alos, if anyone is useing this board and has overclocked it, could u possibly post ur voltages so i can compare?

 

[benniebeeker]

subscribed... im the guy trying to help him out here... i own a 680i and im not familiar with his bios... i think there may be a setting somewhere that we are missing to get this thing to run...

 

[Star5790]

also, i failed to mention that with all my settings at default my voltage still fluxuates, from around 1.28 to 1.38, still a tenth of a volt, but normally with default settings it sits at 1.33 to 1.34

 

[JustaGeek]

I have a "little" cousin of you motherboard, the P5N-E SLI, and have been enjoying a mild, but stable overclock of 2.925GHz for the past 6 months.

Here are my settings (all manual, unless specified otherwise):

CPU Voltage 1.3V (if unstable, try 1.325V, up to 1.35V)
RAM Voltage 2.1V (up to 2.2V - check with Corsair specs!)
NB Voltage Auto (if unstable, increase up to 1.56V, but add a fan to NB to cool it better)
CPU "overvoltage" (or whatever it is called) Auto.

The "Manual Overclocking", "Unlinked", and the following:

CPU frequency: 1300MHz
RAM frequency: 800MHz

RAM timings 4-4-4-12-2T (very important - 2T command rate!).

RAM subtimings all Auto, unless you want to read the SPD and set the values yourself (Everest will allow you to read all the RAM timings stored in SPD).

Do not leave the VCore on Auto! One of the BIOS releases for my MB would automatically raise it to 1.4V with the OC! Set it as described above.

Good luck!

 

[Star5790]

well to be honest, i dont really wanna mess with anything until i find out why my voltages fluxuat so much, anyone have any ideas about that? something wrong with the board, or some setting somewhere?

 

[JustaGeek]

Did you manually set the VCore to 1.3V...?

You know that you should set it manually, correct...?

 

[Star5790]

no, we never actually changed the vcore, just the fsb voltages, thats what is jumping so dramatically. do u think manually setting the vcore would fix that?

 

[JustaGeek]

Absolutely!

Set it manually to 1.3V.

 

[JustaGeek]

BTW, it seems like your motherboard has a lot more voltage settings than mine.

If all else fails, reset the CMOS, and start with the following settings, leaving the blanks, or values not specified below, on Auto:

CPU Voltage 1.3V
RAM Voltage 2.1V
NB Voltage Auto
CPU "overvoltage" (or whatever it is called) Auto.

Choose the "Manual Overclocking" (even if you are not overclocking), "Unlinked", and the following:

CPU frequency: 1066MHz
RAM frequency: 800MHz

RAM timings 4-4-4-12-2T (very important - 2T command rate).

RAM subtimings all Auto.

Hope this helps.

 

[Star5790]

well, i tried all of that, and my fsb voltages still jump about a 10th of a volt, which i dont like much at all. and not to mention orthos wont last for more than 2 min on a blend test, so needless to say im wquite discouraged and thinking about getting a diff board. or at least rmaing this one to asus, but a diff board altogether seems like a better idea. i just dont know anymore, ive tried everything my overclocking buddies have told me, and everything u have, and everything otehr forum users have told me. dunno what eklse i can do

=[

 

[Star5790]

this is a compy/paste of bennie beekers post in anotehr forum of what we've been trying and what we've found

here's a summary of the orthos tests that we ran...

400x8 (core on auto) 4-4-4-12 (2.1) small fft's pass for 30 mins (just to test the cpu)
400x8 (core on auto) 4-4-4-12 (2.1) blend test fails after a few seconds
400x8 (core on auto) 5-5-5-18 (2.1) blend test fails after a few seconds
400x8 (core on auto) 5-5-5-18 (2.1) fsb volts @ 1.5 blend test fails after a few seconds

memtestx86 4-4-4-12 (2.1) 1 full pass with zero errors

its strange that the core passes the small fft test just fine but once you incorporate the memory into the equation the test fails... yet those memory settings pass memtest without errors...

Additional Comment:

weird thing is,

355*9 and 800 mem (yeah i know its not 1:1) passes the blend test...

the front side bus is the limiting factor here... i have read around the forums that 400 fsb is a wall for some people and this board... are the voltages just that picky??? is there a sweet spot that needs to be discovered??? i don't know if my friend star has the patience to find it... hahahhaaa

yeah, f-u ben =]

anyways, we now have it stable at 3194 mhz with mem @ 710 with timings of 4-3-3-10

still hopeing i can find a sweetspot so i can get it to run at an 8 multi

any more ideas guys?
__________________
e6600@3.2 (400x8) | 4x1gb gskill HZ's @ 800mhz (4-3-3-10-2t) | bfg680i p30 | bfg 8800gts 640 SLi (630/2000) | corsair hx620 |

 

[benniebeeker]

the front side bus is the limiting factor here... i can't believe that 400 is the wall for this thing... there has to be a way around this... seems like there is a sweet spot for the voltages... anyone else have this board that can relay their bios settings??? i own a 680i and even though the p5n32-e sli plus is a 680 hybrid board, the bios between the 2 shows many differences... mainly in the voltages area...

 

[JustaGeek]

So if I understand correctly... You guys are not happy with the CPU at ~3.2GHz...?

Do you need to take it higher because.... the performance is not satisfactory...?

 

[benniebeeker]

no... thats not the case... we are looking to achieve 3.2 by going the 400x8 road instead of the 355x9 road... when you run 355x9 the memory has to run underclocked at 720 for a true 1:1 ratio... when you run 400x8 your memory should be at 800 for a true 1:1 ratio...

 

[JustaGeek]

And how do you measure the performance gain (if any...?)

Using the memory multiplier is perfectly fine... That's why we use these ratios...

With my 16/13, and the FSB at 325, I get ~6600MB/s SANDRA memory bandwidth.

Overall PCMark05 score is ~8,000.

Why is 1:1 so important to you...?

 

[benniebeeker]

intels like 1:1 ratios... period... the performance gain in benching is faster read write times... in real life performance its probably not noticeable at all...

you must be a converted amd user... intels don't accept memory dividers as nicely as amd's do...

i guess the biggest thing here is WHY won't this board go over 400 fsb???

its a hybrid 680i board... my 680 does 480 fsb without breaking a sweat...

 

[JustaGeek]

LOL No, I am not the AMD "convert".

I understand that it is better to run the FSB:RAM at 1:1, but if you use the RAM multi... what are you losing...?

The CPU runs happily at 4:1 FSB ratio, and we can't even adjust that! It is what it is, IMO.

The clock is ticking at the FSB speed of 325, 355, 400, 450. To bump it up, we use the 4:1 for the CPU FSB, 9:1 for the CPU itself, and... (for example) 16/13 for the memory! I really dont see anything wrong with that.

If you want the really fast memory, you play with this ratio even further, and with the 300MHz FSB, and the 2:1 ratio, you run your memory bus at 600MHz, for the RAM speed of 1200MHz.

I really don't see anything wrong with that. With the 400MHz FSB the Northbridge is running so much warmer, everything is running at "180 miles per hour", when all you need is 120mph.

I just believe that stability and longevity are so much more important than 10% performance gain. It won't show in real life anyway... Unless/until you want to break the SuperPi or PCMark05 score record, which even with 400MHz FSB is not possible.

It's just my opinion, and you obviously don't have to agree...

3.2GHz is a fine OC for the E6600.

 

[benniebeeker]

his sticks won't do 600... they are ProMOS ic's... not micron d9's... i personally don't see anything wrong with it either... as i said, we are stable 1:1 running 355x9 (1420) with mem at 710 4-3-3-10...

cpu is quad pumped from the actual fsb...

the memory is double pumped from the actual fsb...

300mhz fsb = 300*4 = 1200 effective fsb for cpu (2400 for an 8 multi and 2700 for a 9 multi)
300mhz fsb = 300*2 = 600 effective fsb for memory

that is a 1:1 ratio... not 2:1

 

[JustaGeek]

2:1 was just an example, to show that the 2:1 Memory multiplier would be required to run the memory at 1200MHz.

Again, I am not criticizing - I just would like to emphasize that if you "hit" that sweet spot for your configuration, like 355x9, and you are stable, you should "count your blessings" and enjoy your system.

Why don't you run your memory at 800MHz or higher...? You can easily accomplish that by running it "unlinked" from your FSB, and I can bet that your performance will slightly improve!

It might take you awhile to get the system stable, but at least your memory will run closer to it's specifications, and the system will be faster!

Again - stability is the key. System running 1:1 with the CPU at 3.6GHz and errors in Orthos and Memtest is useless.

3.2GHz stable is priceless.

 

[benniebeeker]

it does no good to run your memory higher then the fsb because data gets bottlenecked by the fsb... its like throwing a hotdog down a hallway... your memory will only handle what your fsb can let through... if your fsb is set to 355 and your memory is set to run 400 (800 real) then you are missing 45 in throughput...

and yes we are orthos stable with the current settings... i know allllllll about stability believe me... my post count does not reflect my knowledge of overclocking...

 

[JustaGeek]

I wouldn't even try to imply a lack of knowledge, because I myself am learning something new every day.

The issue of "bottlenecking" is a supposition: if you can raise the memory speed, if the computer is stable, and you run synthetic tests like Memory Bandwidth in Everest or SANDRA, you will see a rise in the bandwidth.

IMO, the FSB throughput even at 266MHz is way higher than memory at 400x2=800MHz, with the 3:2 Memory multiplier.

It is the RAM that can "choke" the FSB, that's why when you raise the FSB speed, you see a "dramatic" increase in the Memory Bandwidth.

I was running my 800MHz memory at its rated speed with the FSB of 266, and the SANDRA showed 5400MB/s. With the FSB of 325MHz, and memory again at 800MHz, the bandwidth is 6600MB/s. If you keep raising the FSB frequency, you will reach a point that the memory will not be able to accomodate the increased flow of data, and you crash.

Do not worry to use the memory multi - the higher memory frequency will allow for the greater throughput, which in turn will allow you to achieve a greater overclock of your CPU.

 

[benniebeeker]

the rise is marginal if any at all... we benched in everest and found very little performance gains at all... between 720mhz and 800 mhz... the 355 actual fsb being the limiting factor...

 

[Star5790]

well, beeker and i have been at this all day, and all i know is we have made progress and im tired, my overclock is stable at 3.240, my fsb is 355, my memory is at 720 mhz, timings are 4-3-3-10, and things are working fine, so im done messing with it, if i ever want 400x8 ill buy a new board, i have nothing to ask for for xmas anyways. thx for ur help geek, i appreciate that u took the time to post your knowladge of the subject very much, and as always thx ben, post closed!

 

[JustaGeek]

And you obviously realize that you are overclocking yor RAM, by the "tighter" timings, correct...?

Try SANDRA and you will see the difference between 4-3-3-7 and 4-4-4-12 at the same 710MHz.

I bet that your SANDRA bandwidth is about 7000MB/s now - not too shabby at all!

 

[benniebeeker]

we already have the timings @ 4-3-3-10

tried 3-3-3-9 and it wouldn't post... i figured it MIGHT post cause it was underclocked so much... no go....

we are already one step ahead....