Sign in with
Sign up | Sign in
Your question

Hitting wall with Core 2, but which freakin wall?

Last response: in CPUs
Share
February 27, 2007 9:13:20 PM

I don't know if I'm hitting FSB wall, Memory wall, Cpu Wall or some other freakin wall.

here's what I have :

Asus p5B dlx
E6300 Conroe (B2)
2x1GB Crucial 10th anniversary edition DDR2 667
Enermax 460W PSU


Current settings:

460x7
Vcore 1.325
memory voltage: 2.25V
FSB Termination: 1.45V
NB Vcore: 1.55V
SB Vcore 1.6V

Timings are 4-4-4-12

System is stable at above settings. Idle is 40C and Full load is 52C, This is after I lapped my TT BT, and remounted is securely. Is the temp ok for 1.325V? But i seem to can't squeeze anything more out of it.


I tried the following and computer didn't even post.

490x7
Vcore 1.42
memory voltage: 2.4V
FSB Termination: 1.45V
NB Vcore: 1.65V
SB Vcore 1.6V

Timings are 5-5-5-18


What could i try?
February 27, 2007 9:33:27 PM

I would raise your north bridge to 1.75
The cpu and NB are doing most of the 'work' in your o/c.

your other voltages seem too high and are probably not your 'wall'
I would lower them all by at least 1 'step' in your bios.
if these changes work lower them some more
February 27, 2007 9:47:56 PM

NB Vcore is at max 1.65V, can't go higher.

How in the world are people hittin 500+ with the p5b dlx :( 
Related resources
February 27, 2007 10:00:38 PM

Quote:
NB Vcore is at max 1.65V, can't go higher.

How in the world are people hittin 500+ with the p5b dlx :( 


Well, to test the FSB wall issue, try lowering the multiplier on the CPU to 6 and upping the FSB some more, if you are still stuck at the 460 MHz then you are FSB bound.

Also, there is the quality of the memory which could also limit the northbridge. What memory are you running?

Jack

2x1GB Crucial 10th anniversary edition DDR2 667 D9-GMH chips.

Memory is running stock volts (2.2), 4-4-4-12 DDR2 920mhz.

SUppose to be one of the best chips (D9)
February 27, 2007 10:01:06 PM

Quote:
I don't know if I'm hitting FSB wall, Memory wall, Cpu Wall or some other freakin wall.

here's what I have :

Asus p5B dlx
E6300 Conroe (B2)
2x1GB Crucial 10th anniversary edition DDR2 667
Enermax 460W PSU


Current settings:

460x7
Vcore 1.325
memory voltage: 2.25V
FSB Termination: 1.45V
NB Vcore: 1.55V
SB Vcore 1.6V

Timings are 4-4-4-12

System is stable at above settings. Idle is 40C and Full load is 52C, This is after I lapped my TT BT, and remounted is securely. Is the temp ok for 1.325V? But i seem to can't squeeze anything more out of it.


I tried the following and computer didn't even post.

490x7
Vcore 1.42
memory voltage: 2.4V
FSB Termination: 1.45V
NB Vcore: 1.65V
SB Vcore 1.6V

Timings are 5-5-5-18


What could i try?


Your DDr2-667 memory is trying to run @ DDR2-980 at your 490 attempt. Ya think that could possibly be a limiting factor??? When I pushed my DDR2-800 to DDR2-900, I had to loosen the timing to 5-6-6-20. I'd check into whether or not that 460w psu is enough for your OC. I had an Emax 600W that probably was overloading one 12v rail and would reset the comp when I video stress tested @ 9x400. I put in an Antec TP3 650w psu and the prob went away.
February 27, 2007 10:14:17 PM

Lowest I can go in the P5B is 1:1 ratio :( . I've seen people hit much higer with the ram, so I figured ram would not be an issue especially being D9-GMH chip.

It's rated STOCK at 3-3-3-12 DDR2 667.

the DDR2 800 of the D9-GMH is basically the same memory with loosened timings 4-4-4-15 and 5-5-5-15.

Take a look at
http://www.xtremesystems.org/forums/showthread.php?t=12...
http://www.xtremesystems.org/forums/showthread.php?t=11...

Crucial anniversary edition is pretty up there. I kinda don't believe that mine is capping out at 460 :( 

Wish someone had this ram that could chime in.
February 27, 2007 10:42:19 PM

have you tried lowering your multi as jack suggested? That's generally the first step in OCing to lower the multi and see how high you can get the FSB.

Step 1 - Lower the multi, lower teh voltages, back down to 400mhz increase cooling* then start working your way back up with voltage bumps only when needed and only when they make the difference between passing and failing your stress tests. If you get a higher FSB it's an issue with voltage too high, heat, or core frequency too high. Higher voltage isn't always better for stability and it does shorten the lifespan of the components.

If you top out at 460 again:

Step 2 - Keep all the same settings but drop FSB down to 266mhz and decrease your FSB:RAM ratio (eh, 2:3? is that an option? That should give you 399mhz and you can work up from there. double check that before you actually do it though xD) and work your way back up again. If your RAM tops out at 460mhz again it's either your ram or your NB/mobo.

If your ram clocks up higher:

Step3 - Cry, you're screwed. Either mobo or chip is stuck at 460 FSB. But go do it someplace else. It's a very decen OC :p 

*more fans, lower ambient temp, whatever you can do. An open window on a cold dry day and a box fan pointed at an open case should do it until you install a permanent solution or decide to just clock lower.

I would say those temps are too high. I've seen other disagree with me and while it they might be right about stability the chip won't last long at high temp. Keep it under 48c (I like to keep my OCs under 40c although that often proves difficult in the summer). North bridge might also be getting too hot also.
February 27, 2007 11:29:01 PM

lol i run my e6300 at 385 fsb - its must be a stinky one!

you hit a wall! no, you hit the top of mount Everest

i think you full of #$#% 0 your voltage seems to low for that speed?????
February 27, 2007 11:46:08 PM

I wouldn't change the multi, it will OC your NB and probably cause you to hit the FSB wall earlier than you want. The techrepository has the article.... I lost the link.

Your NB seems a little under volted, but the question is what is the temp on the NB.... they get pretty darn hot during heavy OCing. Also, do you have active cooling around the mosfets? If they heat up too much the power to the CPU gets a bit unstable and could cause some issues. Why did you just ramp the vcore like that at 490? you can't just throttle up the voltage .1v and expect it to post. Work from 460, up the FSB 5, if its not stable up the vcore by the smallest increment.
February 27, 2007 11:49:05 PM

Quote:
lol i run my e6300 at 385 fsb - its must be a stinky one!

you hit a wall! no, you hit the top of mount Everest

i think you full of #$#% 0 your voltage seems to low for that speed?????


Screen shots coming your way :) , let me figure out my stuff first.



To those trying to help. I changed the mem divider to 2:3, totally forgot about that at first.
I got the ram up to 487Mhz (DDR2 974 with 5-5-5-18 timing).

EDIT2: Does not seem to be RAM issue. Right now im running DDR2 1000 at 5-5-5-18....



EDIT: here you dragon.

3200Mhz, 1.32 Vcore in Bios, 9 hour orthos.





Here is 3150, only 1.30V in bios
February 28, 2007 2:16:43 AM

run stablity test for 24 hours

google stablity test and down load it from pcworld

i bet it locks up in less then 30 mins

i never used pb5 but i just do not believe you can run 460 fsb @ 155v

let me get this right you running ddr667 @ ddr900? too
February 28, 2007 2:27:46 AM

Quote:
I wouldn't change the multi, it will OC your NB and probably cause you to hit the FSB wall earlier than you want. The techrepository has the article.... I lost the link.


Ya mean the one in my SIG? 8)
February 28, 2007 2:34:53 AM

Quote:
I wouldn't change the multi, it will OC your NB and probably cause you to hit the FSB wall earlier than you want. The techrepository has the article.... I lost the link.


Ya mean the one in my SIG? 8)

Your's isn't as big of an issue if you run the numbers. It is really a concern for the 6300 and 6400.

On your's the default is 9 and the set is 8, so 9/8* 450 = 506.25MHz which isn't beyond the 965's ability.

However if you do the same thing to a 6300, that is 7/6 * 450 = 525, now you are getting up there near the FSB wall of the 965 chipset.

If he wants 485 on a 6 multi, it would be an effective 565.83 MHz on the NB.... that's probably unreasonable.

So it is a context dependent limitation. :wink:
February 28, 2007 2:42:21 AM

Quote:
run stablity test for 24 hours

google stablity test and down load it from pcworld

i bet it locks up in less then 30 mins


What are you betting? Why are you having a hard time believing my results? I'm not alone, a lot of people have achieved what I have. Just because you can't doesn't mean I'm lying.

Edit: I got that test you're talking about. It's not even multi-Threaded man. I can run that thing in the back ground for DAYS. Seriously, that test is a joke.

By the way, SP2004 Orthos is one of the most intensive tests you can run for stability.

You need to start running Orthos to test your systems. The majority can't be wrong when it comes to prime.


Quote:


i never used pb5 but i just do not believe you can run 460 fsb @ 155v



You're judging a product you have never used? I think i'm actually running LESS than 1.55V.

Quote:
let me get this right you running ddr667 @ ddr900? too
[

ddr920 to be exact. You realize I said it's a D9-GMH ICs right?
are you familiar with these ICs?


Seriously man, I have no reason to lie. That's exactly why i'm here, to get more out of my system. It seems like i'm not pushing the system and it's doing very good. If it can do so well how I have it, there has to be potential, and I'm trying to reach that.

I really don't need anyone coming in and accusing me of lying.



Back on topic.

It seems I am getting close to that FSB wall :( .

Is it safe to push 2.4V into my D9-GMH?
February 28, 2007 2:46:32 AM

Quote:
I wouldn't change the multi, it will OC your NB and probably cause you to hit the FSB wall earlier than you want. The techrepository has the article.... I lost the link.


Ya mean the one in my SIG? 8)

Your's isn't as big of an issue if you run the numbers. It is really a concern for the 6300 and 6400.

On your's the default is 9 and the set is 8, so 9/8* 450 = 506.25MHz which isn't beyond the 965's ability.

However if you do the same thing to a 6300, that is 7/6 * 450 = 525, now you are getting up there near the FSB wall of the 965 chipset.

If he wants 485 on a 6 multi, it would be an effective 565.83 MHz on the NB.... that's probably unreasonable.

So it is a context dependent limitation. :wink:

I think he was referring to the link in his sig. not the OC. :) 
February 28, 2007 2:53:34 AM

You're right. I totally misread it. My fault.
February 28, 2007 3:18:40 AM

Quote:
You're right. I totally misread it. My fault.


No sweat......it was probably my "Jersey shorthand" that crossed you up 8)

When are they going to add more interesting stuff about the 965 and it's straps? I'm curious to know if there's more hidden capabilities my setup has. Let's just say that your info a while back had this effect.... :idea: :idea: :idea:
February 28, 2007 3:24:35 AM

Quote:
You're right. I totally misread it. My fault.


No sweat......it was probably my "Jersey shorthand" that crossed you up 8)

When are they going to add more interesting stuff about the 965 and it's straps? I'm curious to know if there's more hidden capabilities my setup has. Let's just say that your info a while back had this effect.... :idea: :idea: :idea:

na. I'm used to looking at numbers all day in public accounting, then I come home to words and I just think "WTF is this mess, where are the numbers?" :lol: 

Yeah, that info has come in handy alot. I just wish I was smart enough to have figured it out. :?

Here's to the future! *raises the bottle of tequila*

Shhhh, don't tell any other Texans I don't drink beer.... I will be exiled! 8O
February 28, 2007 5:06:30 AM

Quote:
I wouldn't change the multi, it will OC your NB and probably cause you to hit the FSB wall earlier than you want.


WTF are you talking about? Lowering the multi doesn't change the NB or the FSB, at all. It lowers the internal frequency of the chip.

If you actually READ the thread you'll see that the ram he is using IS DDR2-800, just with different default latencies. He's only overclocking it by 25% at DDR2-1000. He saved money by buying the "DDR-667" modules instead of the "DDR-800" modules because he know the only difference between the two was the SPD info and he could just manually set it however he wanted. That's the sort of thing that is the bread and butter of OCing.

@n19htmare
Well, your ram is good and the memory controller in the NB is good. Go back and do step 1 now :p  Do you have a multi-meter? You might want to actively monitor your voltages as you try to OC it.

Is it safe to push 2.4v to your RAM? Are you asking that because you already did it or because you want to do it? If you already did it you would know a lot better than most of us. If you didn't already raise the voltage that high why would you want to? You ran it at 500mhz already and you can't even get your CPU to 490mhz.

*EDIT* Ok, I went back and saw that you have tried 2.4v on the memory in 1:1 before and it didn't post, but what did you use when you decreased the memory divider and got the ram up to 500mhz?

You are pretty close to the limits of your hardware though (quite possibly AT the limit at 460mhz FSB). I would really recommend extra cooling and increasing frequencies and voltages more slowly with more testing in between or you will fry something.
February 28, 2007 5:47:16 AM

Quote:

WTF are you talking about? Lowering the multi doesn't change the NB or the FSB, at all. It lowers the internal frequency of the chip.


read this

Quote:
Case 1:
E6600 @ 7x500 = 3.5GHz
NBCC = (9/7)*500 = 642MHz <- this NBCC may be unachievable by the NB which would artifically create what would be misdiagnosed as an "FSB wall"...

Case 2:
E6400 @ 7x500 = 3.5GHz
NBCC = (8/7)*500 = 571MHz
February 28, 2007 6:08:35 AM

Quote:
I wouldn't change the multi, it will OC your NB and probably cause you to hit the FSB wall earlier than you want.


WTF are you talking about? Lowering the multi doesn't change the NB or the FSB, at all. It lowers the internal frequency of the chip.


Lowering the multiplier changes the chipset latency when the fsb is raised
There is a theory that this in a sense overclockes the chipset thus requiring higher voltages to compensate.
The CPU internal frequency is set by the front side bus speed not the multiplier.
It’s the same theory as dividers in ram frequency.
When you get to a point where your lowest frequency available at any divider is higher than the rated speed you need to change the voltage applied or change latency timings or both.
In the Op’s case a raise in termination voltage will increase the current flow through the FSB making the 1.65 volts at the chipset more efficient.
Too much voltage will do the opposite (you have to play with the settings)
A raise in Vcore and memory voltage along with the right VTT voltage settings might break 500FSB with the ram he has providing the CPU has the capability.
One way to tell is if your CPU can run 333 FSB stock voltages you have a good chance to hit 500.
Disable anything you don’t need in bios to take a load off the motherboard.
February 28, 2007 11:07:12 AM

Quote:
I wouldn't change the multi, it will OC your NB and probably cause you to hit the FSB wall earlier than you want.


WTF are you talking about? Lowering the multi doesn't change the NB or the FSB, at all. It lowers the internal frequency of the chip.

If you actually READ the thread you'll see that the ram he is using IS DDR2-800, just with different default latencies. He's only overclocking it by 25% at DDR2-1000. He saved money by buying the "DDR-667" modules instead of the "DDR-800" modules because he know the only difference between the two was the SPD info and he could just manually set it however he wanted. That's the sort of thing that is the bread and butter of OCing.

Did you even look at the article?Seriously?


Stock multi
" alt="" class="imgLz frmImg " />
February 28, 2007 1:51:46 PM

Did u try to crank your nb mhc to 1.6v?

I believe you - better put a fan on the 2.4v memory. I am surprised at 1.55v nb, you getting such a high fsb - but with 2.4v the electrons are flying right by the nb! You got a nice chip and chipset!

I missed the 2.4v mem, but those latencies are low. Good job on the set up - those are impressive numbers

As far as the stability test it locks up system totally and what you wrote is totally contradicts other posts by other people then myself. Do you want me to find that post and a link? or did you run the wrong test possibly?

The stability test will run on both cores with 1 test.
February 28, 2007 2:43:31 PM

Quote:
WTF are you talking about? Lowering the multi doesn't change the NB or the FSB, at all. It lowers the internal frequency of the chip.


:tongue: READ THIS MR. DEE-DE-DEE!:tongue:

...and be ready to apologize to SuperFly03 8)
February 28, 2007 4:38:32 PM

An interesting read, but even they call it a theory and call what I said "The "classicial" understanding of North Bridge (NB)/Memory Controller Hub (MCH) overclocking..." and then use only SiSoft Sandra as evidence (and if SiSoft was actually measuring the NB clock why did they label it FSB?) of what the actual NB clock is. Even according to their theory what I recommended should be safe (in fact, it's exactly the same as the example they use).

This was first posted in Oct. 8 2006 and hasn't been updated since Nov. 2nd 2006. Are there some more up to date discussions of this? They say this may aply to all 865/875/975 chipset mobos? I have an IC7-G MaxII Advance with a Northwood ES and I changed that multi all over the place with a few different strappings and didn't notice this. Neither did any of the other 10 bazillion peopole OCing Northwoods with 875 chipsets so excuse me if I'm a bit sceptical. I did notice that changing the strapping affected my OC but never saw a negative corrolation to lowering my multi. Lower multi yeilded higher FSB in every test I did and I eventually settled for a multiplier that was a couple notches lower than my max for 2.88ghz (yeah, not so super impressive. I blame the early stepping ES). Not that I personally tested it exhaustively or anything but between me and everyone else who OCed with that chipset... yeah.

I still have that setup actually (been neglecting it since I stole all it's ram and sold it's GFX card, it's supposed to become an under-clocked storage server sometime when I get around to it) but too many other things on my plate to go back and OC it again :/  Then again it's SNOWING, perfect wether for OCing.


If you want to use this theory do the step 1 I recommended but start off with a lower FSB and work your way up to see if you hit a wall when your theoretical NB clock reaches ~460mhz. If you set your multi to 6 according to this theory you should hit a wall at 394mhz FSB if you are hitting a NB clock wall at 460mhz (which you shouldn't be anyway) so start at 389 or something and work your way back up and if you're worried about OCing your NB then don't go above ~462mhz with the lowered multi and don't raise voltages if you don't have to.

This theory changes the purpose of the test slightly, but the test itself is only more valid now as you can use it to test max NB clock without stressing the FSB, CPU, or RAM beyond what you already know they are capable of. Plz let us know how it goes.
February 28, 2007 11:10:31 PM

Quote:
An interesting read, but even they call it a theory and call what I said "The "classicial" understanding of North Bridge (NB)/Memory Controller Hub (MCH) overclocking..." and then use only SiSoft Sandra as evidence (and if SiSoft was actually measuring the NB clock why did they label it FSB?) of what the actual NB clock is. Even according to their theory what I recommended should be safe (in fact, it's exactly the same as the example they use).


Well of course it is a theory.... some of the best theories in the world explain how the world exists but theories don't explain every contingency nor can the be proven without a doubt. Big Bang.... still a theory (let's not digress into a philosophical discussion).

Quote:
This was first posted in Oct. 8 2006 and hasn't been updated since Nov. 2nd 2006. Are there some more up to date discussions of this? They say this may aply to all 865/875/975 chipset mobos? I have an IC7-G MaxII Advance with a Northwood ES and I changed that multi all over the place with a few different strappings and didn't notice this. Neither did any of the other 10 bazillion peopole OCing Northwoods with 875 chipsets so excuse me if I'm a bit sceptical. I did notice that changing the strapping affected my OC but never saw a negative corrolation to lowering my multi. Lower multi yeilded higher FSB in every test I did and I eventually settled for a multiplier that was a couple notches lower than my max for 2.88ghz (yeah, not so super impressive. I blame the early stepping ES). Not that I personally tested it exhaustively or anything but between me and everyone else who OCed with that chipset... yeah.


3-4 months old is out of date for you? Good grief, come on now. I am not referencing something ancient. The articles isn't 3 or 4 years old... it's 3-4 months.

Well 10 bazillion people aren't freecableguy, the guy is damned good and reads white papers (crap I can't even understand lol). It may be people have misinterpreted the FSB wall due to a lower than stock multi. Also, did you notice the mutli on a Northwood vs Conroe? Northwoods came in at 14-16x, so 16/15 = 1.066667 but 7/6 = 1.166667 and also mobo's were on a older fab process so the FSB couldn't get as high meaning if you couple the smaller incremental speed increase on the FSB (1.06667 vs. 1.166667) and the lower overall achievable FSB on 875 then you get a smaller increase on the NB.

So it is reasonable that you may mistake an unstable PC for a FSB wall when it was due to NB OC'ing and because the FSB is reasonably close to what others get you chalk it up to a bad batch, or just not a great OC'er. It is possible, just think about it. 1.066667 * maybe 280 FSB on Northwood = 298 so an effective OC of the NB of 18MHz. Now fast forward to Conroe. 515FSB * 1.16668 = 600 for an effective OC of the NB of 85MHz, now you are talking bigger numbers and a more noticable difference.

I still have that setup actually (been neglecting it since I stole all it's ram and sold it's GFX card, it's supposed to become an under-clocked storage server sometime when I get around to it) but too many other things on my plate to go back and OC it again :/  Then again it's SNOWING, perfect wether for OCing.


Quote:
If you want to use this theory do the step 1 I recommended but start off with a lower FSB and work your way up to see if you hit a wall when your theoretical NB clock reaches ~460mhz. If you set your multi to 6 according to this theory you should hit a wall at 394mhz FSB if you are hitting a NB clock wall at 460mhz (which you shouldn't be anyway) so start at 389 or something and work your way back up and if you're worried about OCing your NB then don't go above ~462mhz with the lowered multi and don't raise voltages if you don't have to.


If 965's could only achieve a NB clock of 460 you might be right.... but we see FSB's in excess of 475-500 frequently. I would suspect a large number of problems are due to CPU's, cooling, voltage, and just being a newb.

Quote:
This theory changes the purpose of the test slightly, but the test itself is only more valid now as you can use it to test max NB clock without stressing the FSB, CPU, or RAM beyond what you already know they are capable of. Plz let us know how it goes.


I have a Northwood back in my apartment at college, I would be more than willing to give it a go and see what it kicks out. :D 
March 1, 2007 3:34:53 AM

Quote:
WTF are you talking about? Lowering the multi doesn't change the NB or the FSB, at all. It lowers the internal frequency of the chip.


:tongue: READ THIS MR. DEE-DE-DEE!:tongue:

...and be ready to apologize to SuperFly03 8)

Word up. RJ, you're my dog man. Just keeping it real (in the word's of Randy Jackson - American Idol) LOL. 8O
March 1, 2007 4:47:48 PM

Quote:


3-4 months old is out of date for you? Good grief, come on now. I am not referencing something ancient. The articles isn't 3 or 4 years old... it's 3-4 months.


It's both recent AND out of date ;)  If it had first been posted 4 years ago and only updated 3 months ago it would be a really solid article. But being a new concept that goes against years of OCing tribal knowlege needs more than one update 3 months ago. "Don't lower your multi, skip 360-400mhz and go directly to 401mhz FSB" is pretty out there. I'm not being dismissive, I seriously want to see further discussion of this by people who have actually tested it.

http://www.overclock.net/intel-motherboards/145468-inte...
"Get a X6800 or QX6700 (or even a ES chip). To the north bridge, you are always at a default multiplier with a Extreme Edition processor." That doesn't make any sense. All he does it repost stuff from the tech repository and adds that little tidbit without anything to back it up. A lot of talk and evidence about strappings and having to skip to 401mhz FSB for stability, but not so much about the multi.

http://www.overclock.net/intel-motherboards/145468-inte...
This thread talks about it a little more and how lowering your multi will increase your strapping which will increase NB latencies and decrease performance and he gives this quote:
Quote:

516x7=3611MHz, DDR2-1032 --> 13:55.625 400x9=3600MHz, DDR2-1000 --> 13:35.765
(check the screenshots, lost of info there) but according to the theory that would also mean the NB was clocked at 663.42mhz in the first test which seems a bit high and also makes me wonder how the NB can support so many clock dividers as an undocumented feature.

There are a many unanswered questions here and most of these threads link to eachother more than provide additional evidence.


Oh, and 20/15 multi (which I'm pretty sure I did to I did to my Northwood ES 2.66ghz and 875chipset) is a ratio of 1.3334 (and I'm pretty sure it yeiled a higher FSB than the 20/18 that I'm running right now, but lower core frequency, but it has been awhile).

There were a lot of very good OCers who played with the Northwood and lowering the multi to test max FSB and RAM before going for max clock has been a standard OCing proceedure for years.
March 1, 2007 5:24:15 PM

I read that a while back at XS. weird that they didn't update it yet.

Anyways, I think I might just stick with the settings I had because I think there seems to be an FSB wall.

It seems like i have a strong processor but the mobo can't handle above about 475.


Here are the settings again.

Temp: Full Load doesn't break 45C. :) 

460x7 = (3220mhz)

Vcore = Stock 1.325V

Memory = 2.2V

FSB Terminate = 1.3V

NB Voltage = 1.45V

SB Vcore - Auto

ICH Voltage - Auto

Timings: 4-4-4-12


March 1, 2007 5:37:30 PM

You first post said said north bridge 1.55v do you mena mch?

but intel voltage is mch, not nb? nvidia is nb voltage?

i am using a 975x mobo my mch is 1.5-1.65 yours is lower at 1.45?
March 1, 2007 5:46:44 PM

Quote:


3-4 months old is out of date for you? Good grief, come on now. I am not referencing something ancient. The articles isn't 3 or 4 years old... it's 3-4 months.


It's both recent AND out of date ;)  If it had first been posted 4 years ago and only updated 3 months ago it would be a really solid article. But being a new concept that goes against years of OCing tribal knowlege needs more than one update 3 months ago. "Don't lower your multi, skip 360-400mhz and go directly to 401mhz FSB" is pretty out there. I'm not being dismissive, I seriously want to see further discussion of this by people who have actually tested it.

I have suggested this to several people who are on the edge of 400MHz (generally in the ~385 range) and once they kicked it up to 401-405 they were stable. I will have to find the threads, but it has happened on a number of occasions.

Quote:
http://www.overclock.net/intel-motherboards/145468-inte...
"Get a X6800 or QX6700 (or even a ES chip). To the north bridge, you are always at a default multiplier with a Extreme Edition processor." That doesn't make any sense. All he does it repost stuff from the tech repository and adds that little tidbit without anything to back it up. A lot of talk and evidence about strappings and having to skip to 401mhz FSB for stability, but not so much about the multi.


Programmed in each CPU is a default multiplier, but an EE/QX/X6800 CPU has a freely adjustable multiplier. I *think* that since there is no real default multi, because it can be changed at will up or down, is what they are getting at. I honestly don't know here, just venturing a guess.

Quote:
http://www.overclock.net/intel-motherboards/145468-inte...
This thread talks about it a little more and how lowering your multi will increase your strapping which will increase NB latencies and decrease performance and he gives this quote:

516x7=3611MHz, DDR2-1032 --> 13:55.625 400x9=3600MHz, DDR2-1000 --> 13:35.765
(check the screenshots, lost of info there) but according to the theory that would also mean the NB was clocked at 663.42mhz in the first test which seems a bit high and also makes me wonder how the NB can support so many clock dividers as an undocumented feature.

There are a many unanswered questions here and most of these threads link to eachother more than provide additional evidence.

Hmm


Quote:
Oh, and 20/15 multi (which I'm pretty sure I did to I did to my Northwood ES 2.66ghz and 875chipset) is a ratio of 1.3334 (and I'm pretty sure it yeiled a higher FSB than the 20/18 that I'm running right now, but lower core frequency, but it has been awhile).


Going along with the theory I mentioned above, since it is an ES chip, there is no default multi in essence. Yes there is a boot default, but it is freely adjustable. I could be wrong here, but more data is needed. It would have to be an intense white paper reading to understand it, I think.

Quote:
There were a lot of very good OCers who played with the Northwood and lowering the multi to test max FSB and RAM before going for max clock has been a standard OCing proceedure for years.


I agree, it is a standard procedure. However, it may have been artificially limiting the max FSB. 875 didn't hit a terribly high FSB (relative to what we see now). When I get back to my college apt (2 weeks) I will boot up the good ole 875/865 P4C 3.0 and see what happens. It won't be A+ quality crap because my mobo kinda blows, but I will try and put something together. Hind sight is 20/20.... or in this case 20/? lol.
March 1, 2007 7:30:54 PM

Quote:
You first post said said north bridge 1.55v do you mena mch?

but intel voltage is mch, not nb? nvidia is nb voltage?

i am using a 975x mobo my mch is 1.5-1.65 yours is lower at 1.45?


Yeah I thought I had it at that too, but when i look again, it was at 1.45v. I was surprised myself to see it hit 460FSB, but even more surprised that at 1.65V, I can't get past 475mhz. I can't figure out why .20V increase is only netting me 15mhz :( . That's when I was curious on what the heck my wall is.

Orthos is now at 9 hours and 45 minutes. Still going. I've been using my computer for 3 hours now while running orthos in the back ground. and still going strong.
March 1, 2007 8:32:03 PM

Im running my p5b-d & e6300 @ 500fsb all day. do you have a fan on the nb? i was hitting a wall b4 i did that. Your ram should be able to hit around 1100mhz.

Last night when i got home from work, my wife had the heat off & it was 55 in the house. I waas like, time to oc!! Got the rig up to 3.801 & a pi time of 15.34.
e6300 @ 3.5 1.3 vcore
crucial ballistix tracer pc2 8000 2.4 vdimm
nb 1.45
sb auto

Going to do some garage benching ths weekend. :D 
March 1, 2007 8:40:10 PM

Quote:
Lowest I can go in the P5B is 1:1 ratio :( . I've seen people hit much higer with the ram, so I figured ram would not be an issue especially being D9-GMH chip.

It's rated STOCK at 3-3-3-12 DDR2 667.

the DDR2 800 of the D9-GMH is basically the same memory with loosened timings 4-4-4-15 and 5-5-5-15.

Take a look at
http://www.xtremesystems.org/forums/showthread.php?t=12...
http://www.xtremesystems.org/forums/showthread.php?t=11...

Crucial anniversary edition is pretty up there. I kinda don't believe that mine is capping out at 460 :( 

Wish someone had this ram that could chime in.


Ahhhh, so you did your homework.... I am, myself, not sure of the capabilities of this RAM, but would be my first guess as to the wall you are hitting. See Flasher702 above, he is offering up some good suggestions.

An encouraging thing --- your CPU temps are still good so even at 460 I suspect your CPU has some more room.

What GOOD RAM do you suggest (for a p965)? I have Corsair 6400 CL4. Running at CL6 @ 400mhz. Not impressed. I can raise the volts and try to tweek, but what RAM should I get that doesn't need to be tweeked?
March 1, 2007 8:55:04 PM

Anything thats micron D9 based will oc quite well. Check this page out.
ddr2 ram list
March 1, 2007 9:19:27 PM

That happens alot - the voltage increase of the nb or mch results in NO extra fsb or so little its not worth changing - 5 series nvidia is the same.


my p5w-dh hates anything over 400 fsb - i guess i flash the bios, i am too lazy this is the 701 from last fall.



look below its dr Who! can i get ride in tardess?
March 1, 2007 9:21:59 PM

Quote:
Anything thats micron D9 based will oc quite well. Check this page out.
ddr2 ram list


Good list!
But how I know what Rev.# the RAM is (before I buy it)?
Thanks!
March 1, 2007 9:24:17 PM

What, you miss placed your crystal ball? That's how I know.... lol, j/k.

in all seriousness though, I don't know if you can figure it out until after you get them :?
March 1, 2007 10:33:31 PM

Quote:
What, you miss placed your crystal ball? That's how I know.... lol, j/k.

in all seriousness though, I don't know if you can figure it out until after you get them :?


Too bad I already got my 2 gigs or RAM. When I upgrade to 4 gig, I need to get 2 more gigs of the same type of RAM or get 4 megs of expensive RAM.
Oh well, I know better now.
March 2, 2007 11:31:46 AM

All Crucial ram is going to be Micron D9... memory, as micron owns Crucial. :wink:
March 2, 2007 10:45:14 PM

Quote:
All Crucial ram is going to be Micron D9... memory, as micron owns Crucial. :wink:

THANKS, this makes things a LOT easier! :trophy:
March 2, 2007 11:41:44 PM

I would try 5-6-6-18 timings and try to maximize your FSB, then try to tighten timings to 5-5-5 or 4-4-4 after you are stable and cool enough.

My Crucial Ballistix PC6400 will not work at 5-5-5 @ 500FSB, but will work fine at 5-6-6, (albiet with the ram voltage on auto). This ram is rated at 5-5-5-15 @ 500MHz in CPU-Z, but probably with 2.2v.
!