E4400/DDR2-667 on P35-DS3R --> FSB wall?

[nullpointerus]

Here's my setup:


E4400 (stock: 10 x 200 @ 1.325v)
P35-DS3R (stock: 200, 266, or 333)
HP Value RAM DDR2-667 5-5-5-15, 20, 2T @ 1.8v (passes memtest86+ w/ 825 @ 1.8v)
OCZ 700W GameXStream PSU
Tuniq Tower 120 @ 1700 rpm (never breaks 40 C load for this FSB testing)
CoolerMaster CM690 case w/ plenty of airflow to all vital components


For the moment, here's how I'm testing:

- Keep the CPU and RAM at lowest speed while raising FSB to find max. FSB.

- The RAM is set at a 2.0 multiplier (i.e. 1:1). As I raise the FSB, my board automatically raises the RAM timings to compensate, so by 800 MHz it's up to like 5-7-7-19 or something.

- The voltages start at stock.

- I boot Vista x64 and run Prime95 25.6 x64 for 5 min.. If there are no errors/crashes, I bump the FSB another 10 MHz. If there are errors/crashes, I raise voltages to stabilize.

- PCI-E is set to 100 MHz.

- EIST and C1E are off. TM2 is on.

- I use CPU-Z to monitor vCore and SpeedFan to monitor temps. My typical load temps. are < 40 C (or < 32 C per core) for these tests (which is not suprising given the low total CPU freq.).

Here are my results:

6 x 333 = 1998 MHz + 666 MHz RAM (appears stable)
6 x 350 = 2100 MHz + 700 MHz RAM (appears stable in Windows, may reset o/c on reboot)
6 x 360 = 2160 MHz + 720 MHz RAM (appears stable in Windows, may hard-lock on reboot)
6 x 370 = 2220 MHz + 740 MHz RAM (hard-locks in seconds w/ Prime95 SmallFFT)
6 x 400 = 2400 MHz + 800 MHz RAM (no POST -- I thought I might pass a FSB hole, but no such luck!!)

No matter what I do, the above results never vary, even with raised voltages:

CPU: +0.075v (1.4v)
RAM: +0.100v (1.8v)
FSB : +0.100v (???)
MCH: +0.100v (???)

I would have thought that some setting would prevent the crashes/lockups or at least make them happen a little differently, but, after rebooting several times for each error/crash, I've verified that they always happen in exactly the same way.

(But I remember being told this P35-based board should be able to hit 400 MHz FSB on stock voltages...?)

I've also tried manually adjusting the PCI-E speed (to test the PCI-E lock?):

090 MHz --> Vista locks up with fade-in intro
100 MHz --> (see FSB results)
110 MHz --> (see FSB results)

Question #1 --> What is the cause of my FSB wall? Is this the limitation of the CPU itself, or some part of my board? Am I missing something?

FYI, I've done many tests in the past, and here are my stable results with max CPU o/c:

10 x 266 = 2660 MHz @ 1.375v (BIOS)
10 x 275 = 2750 MHz @ 1.387v (BIOS)
10 x 280 = 2800 MHz @ 1.400v (BIOS)
10 x 290 = 2900 MHz @ 1.450v (BIOS)
10 x 300 = 3000 MHz @ 1.550v (BIOS)
> 3000 MHz (not stable at all)

...so this exponential curve implies that I've properly identified the max. CPU frequency for this chip.

(Before anyone posts common suggestions for 3+ GHz, forget it -- this chip just won't go that high. )

But why the FSB problem, where increased voltages appear to have no effect?

The only thing that has affected my FSB problems is removing my video card (HD2900XT) and replacing it with an X1550. But the power/heat difference from this change ought to account for more than making 333 MHz FSB stable (on reboot) vs. 350+ MHz FSB, which still has all the same issues.

What gives?

Question #2 --> Sometimes Prime95 Blend skips certain tests on my cores. For example, at 1024K, core #1 will do 6 tests before succeeding while core #2 will do 7 tests before succeeding. Is this normal?

If not, shouldn't there be some kind of error message? I don't like wierd, unexplained behavior... :thumbsdown:

I have round-off checking enabled, BTW.

 

[Markfw]

Seems like your ram may be the problem. Get some nice PC-6400 800 DDR ram.

 

[Tungsten666]

2nd the ram comment

 

[nullpointerus]

Hmm...

Two things to remember:

- The RAM tests fine (memtest86+ for hours) at up to 825 MHz w/ stock volts.
- Raising memory voltage did not alleviate the FSB problems, not even a little.

But I'm certainly no expert...

In any case, I don't think buying 4 GB DDR2-800 RAM would be cost effective as this point, given that this PC has an E4400 CPU that can't hit 3 GHz at decent voltages. Under such conditions, faster RAM would provide no real-world performance benefit. A CPU upgrade w/ 333 MHz FSB would be a better purchase...for this CPU+MB.

I just want to be sure of the exact cause for my FSB wall!

Originally posted by: nullpointerus

Question #2 --> Sometimes Prime95 Blend skips certain tests on my cores. For example, at 1024K, core #1 will do 6 tests before succeeding while core #2 will do 7 tests before succeeding. Is this normal?

To answer my own question, I checked out Prime95's custom tests. Apparently, the tests run for a certain number of minutes, not a fixed number of calculations, so obviously a faster CPU (or a less loaded core?) will do less tests for a given FFT size.



I should've checked this before posting!

 

[Markfw]

What was the all the settings @825 ? The comment below suggests ram:
6 x 360 = 2160 MHz + 720 MHz RAM (appears stable in Windows, may hard-lock on reboot)

You multiplier should be set to 2, not auto. At that speed, if it was@825 that would be 412 memory/fsb. But you are saying it won;t do 360 at way below cpu speed. (6x)

 

[apoppin]

*thinking* and reassesing

 

[aussiestilgar]

On my DS3R board, I had to add more than 0.1v extra (0.3-0.4v actually) to the RAM to get the higher OCs stable. Though this is with DDR2-800 Ballistix, not sure if you could/should do that with your RAM.

 

[kentsfield]

...your RAM can't handle it at 1:1 with your CPU @ 3.0 Ghz - not even 2.66Ghz -not 1:1 on this planet with budget RAM *period*


At 3.0 ghz (333x9), and 1:1 divider the ram is running at 667mhz. And he`s got HP 667mhz ram. I don`t see why would ram be the problem apoppin, it would run stock speeds at 1:1.

 

[apoppin]

Originally posted by: kentsfield
...your RAM can't handle it at 1:1 with your CPU @ 3.0 Ghz* - not even 2.66Ghz -not 1:1 on this planet with budget RAM *period*


At 3.0 ghz (333x9), and 1:1 divider the ram is running at 667mhz. And he`s got HP 667mhz ram. I don`t see why would ram be the problem apoppin, it would run stock speeds at 1:1.

oops ... let me rethink this

of course you are right ... i am thinking of something else ... let me look in my OWN bios ...
i thought i had a point about "auto"


edit: *with a multi of 2.0

OK ... he says his MEMORY MULTIPLIER is 2.0

that would make what i said right ... thanks for the correction
- i don't have to look in BIOS

 

[Markfw]

Right, how could he get the ram to test@825 when the memory mult is at 2.0 ??? That means the memory is set to 412 ? Something is off here.

 

[apoppin]

Originally posted by: Markfw900
Right, how could he get the ram to test@825 when the memory mult is at 2.0 ??? That means the memory is set to 412 ? Something is off here.

^^^ this is what i am talking about

 

[nullpointerus]

Look, I know *how* to overclock...

The point here was to find the max. stable FSB of the board in this rig. To that end, CPU and RAM had to be set at the lowest possible speed so that any and all crashes would be solely caused by the high FSB.

All I meant to say about the RAM running fine @ 825 MHz was that the memory had already been tested fine @ 825 MHz.

Originally posted by: Markfw900

What was the all the settings @825 ? The comment below suggests ram:
6 x 360 = 2160 MHz + 720 MHz RAM (appears stable in Windows, may hard-lock on reboot)

My question is, if I told you the RAM had previously tested fine at 275 x 3.0 = 825 MHz, why would you say it's now failing at 360 x 2.0 = 720 MHz? That seems odd. And, even if the RAM is failing at 720 MHz, why didn't additional memory voltage improve the situation?

You multiplier should be set to 2, not auto. At that speed, if it was@825 that would be 412 memory/fsb. But you are saying it won;t do 360 at way below cpu speed. (6x)

Yep, when my RAM tested stable at 825 MHz, I was at 275 x 3.0 = 825.

To be clear, I've never gone past 333 MHz FSB reliably on this PC, and I'm curious as to why. Originally, I selected this board because it was supposed to be able to hit 400 MHz FSB at stock voltage, and much higher w/ a slight bump. i.e. For later quad-core overclocking, IF I ended up upgrading to Q6600 (or a similar model).

What prompted this thread is that I ran into trouble while gathering info. I'm trying to isolate the CPU+MB's max. FSB. I already know the max. CPU clock rate (2.85 GHz) and RAM clock rate (825 MHz) at reasonable voltages, from previous testing. When I get all three pieces of information, I should be able to synthesize the best overall settings.

Besides, I'm just curious as to why:

- My PC can't POST @ 400 MHz FSB even though the RAM is memtest-stable to 275 x 3.0 = 825 MHz @ stock volts!
- My PC can't run higher than 333 MHz FSB without, at least, random crashes.
- Additional voltage, loosened timings, etc. do not help any of the above.

I'd have thought there would be *some* flexibility, yet it won't budge -- not even a little?

It's confusing.

 

[apoppin]

what are your installs like with your CPU and RAM at completely stock speeds?

Where does the instability *start* ?

- My PC can't POST @ 400 MHz FSB even though the RAM is memtest-stable to 275 x 3.0 = 825 MHz @ stock volts!
- My PC can't run higher than 333 MHz FSB without, at least, random crashes.
- Additional voltage, loosened timings, etc. do not help any of the above.

What is your memory divider set to? And can you set your RAM lower than stock - below 200Mhz - while also raising your FSB to rule out the RAM?

edit: you mentioned "auto" ... are you talking SPD or the timings?

 

[nullpointerus]

Originally posted by: apoppin
what are your installs like with your CPU and RAM at completely stock speeds?

I wasn't talking about the WU errors here... per se...

Anyway, I got the WU errors:
- at completely stock speeds and voltages
- at 6 x 200 = 1200 MHz (CPU) and 2 x 200 = 400 MHz (RAM)

So either that WU problem was not related to timings/speeds, or something in my PC was/is permanently damaged.

What I'm asking about this thread is a long-standing feud between my MB, CPU, and 333+ FSBs.

Where does the instability *start* ?

Um... I misidentified the "wierd instability" I asked you about earlier @ 333 MHz because I didn't understand that Prime95's number of tests per FFT size is totally dependent on time (i.e. default: 15 minutes), not on a fixed number of calculations or tests.

Sorry about that!

- My PC can't POST @ 400 MHz FSB even though the RAM is memtest-stable to 275 x 3.0 = 825 MHz @ stock volts!
- My PC can't run higher than 333 MHz FSB without, at least, random crashes.
- Additional voltage, loosened timings, etc. do not help any of the above.

What is your memory divider set to?

2.0 (which is 1:1), for the purpose of finding the max. FSB

And can you set your RAM lower than stock - below 200Mhz - while also raising your FSB to rule out the RAM?

The lowest memory multiplier on the P35 chipset is 2.0, so at 200 FSB my RAM can only go down to 2.0 x 200 = 400 MHz. And, of course, this means that my RAM has to be raised above stock clocks when I go above 333 FSB. No way around it...

But more RAM voltage does not fix my FSB problems, even with a paltry 666 MHz --> 720 MHz memory o/c, which is the odd part.

edit: you mentioned "auto" ... are you talking SPD or the timings?

I mentioned auto timings, which the board calculates relative to (a) the current RAM speed in MHz and (b) the SPD timings.


If I set the board to "auto" multiplier, my RAM would be try to run @ 1000+ MHz, BTW.

:Q

 

[nullpointerus]

Originally posted by: aussiestilgar
On my DS3R board, I had to add more than 0.1v extra (0.3-0.4v actually) to the RAM to get the higher OCs stable. Though this is with DDR2-800 Ballistix, not sure if you could/should do that with your RAM.

Hmm... maybe so...

But mine is value RAM w/o heat spreaders... that'd be asking for trouble!


The question currently in my mind is:

Is a higher FSB more stressful on RAM than a higher divider?

If 275 x 3.0 = 825 MHz succeeds, then shouldn't 360 x 2.0 = 720 MHz also succeed?



Or is the low FSB in 275 x 3.0 preventing memtest86+ from finding errors in the RAM?
Because, at a lower FSB, the CPU is spending more time waiting for the RAM anyway?


That's one possible explanation. Or there could be some other limiting factor in my rig.

I'm no expert with this stuff... any ideas here?

 

[Markfw]

275 * 3 is still testing the ram@275, way below spec. I have mine running@390*9, same motherboard, and I doubt they are that different. Something is wrong somewhere. With that ram@mfg spec (333) and multi=6 ram mult=2.0, and run memtest, it should work with FSB term = +.2 and GMH +.2 (maybe you don't even need that)

 

[apoppin]

ideas ... sure

6 x 333 = 1998 MHz + 666 MHz RAM (appears stable)
6 x 350 = 2100 MHz + 700 MHz RAM (appears stable in Windows, may reset o/c on reboot)
6 x 360 = 2160 MHz + 720 MHz RAM (appears stable in Windows, may hard-lock on reboot)
6 x 370 = 2220 MHz + 740 MHz RAM (hard-locks in seconds w/ Prime95 SmallFFT)
6 x 400 = 2400 MHz + 800 MHz RAM (no POST -- I thought I might pass a FSB hole, but no such luck!!)

Why are you using "six" ... way back when i set up my own gigabyte P35-DS3P MB, i believe i was warned against it ... IF my memory is right ... maybe 'instability'?
--have you tried "9" ?
... it's what i am currently using to get to 3.150Ghz and there is a lot of headroom left.

The lowest memory multiplier on the P35 chipset is 2.0, so at 200 FSB my RAM can only go down to 2.0 x 200 = 400 MHz. And, of course, this means that my RAM has to be raised above stock clocks when I go above 333 FSB. No way around it...

sure .. @ 9 i can set my RAM down to 667Mhz if i want - at 2.0 ... and still keep my OC on the FSB by manipulating the dividers [they go up to 4.0]
-you can too with a simple adjustment.



btw, i got *screwed* by not reading the fine print ... i had to pay NewEgg another $15 for the Bridge Interconnect that Sapphire is too *cheapass* to include ... so i will finish most of our testing with a single GPU'
--till Tuesday

-i shoudda got the XT for $200
:roll:

 

[nullpointerus]

Originally posted by: Markfw900
275 * 3 is still testing the ram@275, way below spec. I have mine running@390*9, same motherboard, and I doubt they are that different. Something is wrong somewhere.

So what you're saying is that, between these two settings:

1. RAM @ 275 x 3.00 = 825 MHz
2. RAM @ 360 x 2.00 = 720 MHz

...option #2 is more stressful, and would require more voltage?

IOW, where RAM stability is concerned, FSB is more important than total clock rate?

You see, that's what would explain everything...
...I just don't (or didn't?) know that it is true.

With that ram@mfg spec (333) and multi=6 ram mult=2.0, and run memtest, it should work with FSB term = +.2 and GMH +.2 (maybe you don't even need that)

It's stable at those settings (6x, 2x, 333 MHz) w/o the extra voltage.

What I'm asking is, why doesn't it go *above* 333 MHz, even with extra voltage, relaxed timings, etc.?

Seems like I should have *some* wiggle room, not a brick wall.

But if what I think you said about the FSB is true, then my 825 MHz-stable value RAM was just a false positive, and the RAM itself really can't go much over 667 MHz even w/ extra voltage.

Originally posted by: apoppin
ideas ... sure

6 x 333 = 1998 MHz + 666 MHz RAM (appears stable)
6 x 350 = 2100 MHz + 700 MHz RAM (appears stable in Windows, may reset o/c on reboot)
6 x 360 = 2160 MHz + 720 MHz RAM (appears stable in Windows, may hard-lock on reboot)
6 x 370 = 2220 MHz + 740 MHz RAM (hard-locks in seconds w/ Prime95 SmallFFT)
6 x 400 = 2400 MHz + 800 MHz RAM (no POST -- I thought I might pass a FSB hole, but no such luck!!)

Why are you using "six" ... way back when i set up my own gigabyte P35-DS3P MB, i believe i was warned against it ... IF my memory is right ... maybe 'instability'?
--have you tried "9" ?
... it's what i am currently using to get to 3.150Ghz and there is a lot of headroom left.

9 x 333 = 3.0 GHz, which, for my chip, takes a *lot* of voltage.

I set the CPU and memory dividers as low as they could go *just* for the purpose of finding the max. FSB. I would be using 1x multipliers for both CPU and RAM if I could. This is *solely* to determine why I can't get my board over 333 MHz, *regardless* of what CPU o/c it would get me.

Knowing the exact limits of my components individually is important. What if Intel and AMD get into another (minor) price war, or new chips come out, and there is a deal on a new CPU @ 333 MHz FSB stock that I want to swap in and overclock? If my other components limit me to 333 MHz FSB regardless of which CPU I put in this board, then that would be a bad purchase, right?

The lowest memory multiplier on the P35 chipset is 2.0, so at 200 FSB my RAM can only go down to 2.0 x 200 = 400 MHz. And, of course, this means that my RAM has to be raised above stock clocks when I go above 333 FSB. No way around it...

sure .. @ 9 i can set my RAM down to 667Mhz if i want - at 2.0 ... and still keep my OC on the FSB by manipulating the dividers [they go up to 4.0]
-you can too with a simple adjustment.

Yes, I *know* how to o/c the CPU, and I *know* what the dividers do.

I'll say it another way...
...IF I were going for the max. CPU clock, I wouldn't be doing it this way... :thumbsup:

I'm simply asking why my FSB results conflict with what I "know" about memory o/c-ing.

btw, i got *screwed* by not reading the fine print ... i had to pay NewEgg another $15 for the Bridge Interconnect that Sapphire is too *cheapass* to include ... so i will finish most of our testing with a single GPU'
--till Tuesday

-i shoudda got the XT for $200
:roll:

Well, it's still available. You could cancel your NewEgg order, right?

 

[Tempered81]

Its difficult to boot with that chip above 375.
the ram should be good for 400 with loosened timings & a voltage increase.
So wall? I vote for yes.

 

[apoppin]

Originally posted by: nullpointerus

Originally posted by: Markfw900
275 * 3 is still testing the ram@275, way below spec. I have mine running@390*9, same motherboard, and I doubt they are that different. Something is wrong somewhere.

So what you're saying is that, between these two settings:

1. RAM @ 275 x 3.00 = 825 MHz
2. RAM @ 360 x 2.00 = 720 MHz

...option #2 is more stressful, and would require more voltage?

IOW, where RAM stability is concerned, FSB is more important than total clock rate?

You see, that's what would explain everything...
...I just don't (or didn't?) know that it is true.

With that ram@mfg spec (333) and multi=6 ram mult=2.0, and run memtest, it should work with FSB term = +.2 and GMH +.2 (maybe you don't even need that)

It's stable at those settings (6x, 2x, 333 MHz) w/o the extra voltage.

What I'm asking is, why doesn't it go *above* 333 MHz, even with extra voltage, relaxed timings, etc.?

Seems like I should have *some* wiggle room, not a brick wall.

But if what I think you said about the FSB is true, then my 825 MHz-stable value RAM was just a false positive, and the RAM itself really can't go much over 667 MHz even w/ extra voltage.

Originally posted by: apoppin
ideas ... sure

6 x 333 = 1998 MHz + 666 MHz RAM (appears stable)
6 x 350 = 2100 MHz + 700 MHz RAM (appears stable in Windows, may reset o/c on reboot)
6 x 360 = 2160 MHz + 720 MHz RAM (appears stable in Windows, may hard-lock on reboot)
6 x 370 = 2220 MHz + 740 MHz RAM (hard-locks in seconds w/ Prime95 SmallFFT)
6 x 400 = 2400 MHz + 800 MHz RAM (no POST -- I thought I might pass a FSB hole, but no such luck!!)

Why are you using "six" ... way back when i set up my own gigabyte P35-DS3P MB, i believe i was warned against it ... IF my memory is right ... maybe 'instability'?
--have you tried "9" ?
... it's what i am currently using to get to 3.150Ghz and there is a lot of headroom left.

9 x 333 = 3.0 GHz, which, for my chip, takes a *lot* of voltage.

I set the CPU and memory dividers as low as they could go *just* for the purpose of finding the max. FSB. I would be using 1x multipliers for both CPU and RAM if I could. This is *solely* to determine why I can't get my board over 333 MHz, *regardless* of what CPU o/c it would get me.

Knowing the exact limits of my components individually is important. What if Intel and AMD get into another (minor) price war, or new chips come out, and there is a deal on a new CPU @ 333 MHz FSB stock that I want to swap in and overclock? If my other components limit me to 333 MHz FSB regardless of which CPU I put in this board, then that would be a bad purchase, right?

The lowest memory multiplier on the P35 chipset is 2.0, so at 200 FSB my RAM can only go down to 2.0 x 200 = 400 MHz. And, of course, this means that my RAM has to be raised above stock clocks when I go above 333 FSB. No way around it...

sure .. @ 9 i can set my RAM down to 667Mhz if i want - at 2.0 ... and still keep my OC on the FSB by manipulating the dividers [they go up to 4.0]
-you can too with a simple adjustment.

Yes, I *know* how to o/c the CPU, and I *know* what the dividers do.

I'll say it another way...
...IF I were going for the max. CPU clock, I wouldn't be doing it this way... :thumbsup:

I'm simply asking why my FSB results conflict with what I "know" about memory o/c-ing.

btw, i got *screwed* by not reading the fine print ... i had to pay NewEgg another $15 for the Bridge Interconnect that Sapphire is too *cheapass* to include ... so i will finish most of our testing with a single GPU'
--till Tuesday

-i shoudda got the XT for $200
:roll:

Well, it's still available. You could cancel your NewEgg order, right?

i have no idea without your rig sitting in front of me ... all i am asking is *why* don't you CHANGE the clock ratio ? ... pick something besides six ... maybe it is just 'unlucky' for you
- you tried *everything else* ... and i know it CAN influence problems you are experiencing


and ... You don't take Visa?

 

[watek]

Probably fsb wall on your cpu chip. You just got one that does not oc well.

Same with my E4500. I can boot 340 x 10, all day long, orthos hours and hours but can't do any FSB above 350.

 

[perdomot]

I also think its the ram. I got my e4400 running at 10 X 300 and am using the ram multi 2.5 with a pair of DDR2 800 sticks. Unless I'm mistaken, you are using VALUE RAM which is rated for a FSB of 333 so going above that is what's messing you up. I'm using this ram: http://www.newegg.com/Product/...x?Item=N82E16820211066
Considering the price with the rebate, I think you should upgrade.

 

[nullpointerus]

jaredpace:

OK... I did find this:
http://forums.anandtech.com/me...=2149025&enterthread=y

It's not the same motherboard, but it's similar, and I do have those BIOS options. I'll give it a quick test just to see if my board has that particular quirk. Personally, I would have thought that *disabling* the board's auto-o/c setting would have been better.


apoppin:

I'll try something other than 6x. :thumbsup:


watek:

I hope it's not the CPU. That'd be awfully hard to confirm w/o buying a new CPU. It's a catch-22: buying a new CPU for this board would be unwise until I confirm that my current CPU is the *only* reason why I can't get above 333 MHz FSB reliably.


perdomot:

Hmm... good deal... but I'm a little tapped out right now. I have a few things to try first...

 

[nullpointerus]

The CIA BIOS option trick suggested here:
http://forums.anandtech.com/me...=2149025&enterthread=y

I went back to try and hit 3.2GHz, making sure my memory was staying at or under DDR2-800 so as not to cause stability problems for my overclock, and no matter what I do, anything above a 333 FSB causes the system not to POST. I have tried different Vcore voltages, different FSB settings, nothing seems to allow me to boot above the 333 FSB.

From what I can tell it's possible that there is an issue with the "USB Mouse support" option and the CIA settings in the BIOS. Apparently the CIA setting needs to be set to Turbo and the USB Mouse Support disabled. I haven't tried this configuration yet on my own, but it has been suggested in a few other forums. I'm going to give it a try tonight and will update here.

Currently I have been priming for 4 hours using the Turbo enabled, USB Mouse disabled setup @ 356FSB/9x multiplier @ 1.275V. I'll update in the morning when I get up to see what the progress is. If I pass 8 hours of Prime, I'll probably start pushing it a bit further to see how things go. Nevertheless, I am feeling good about my results thus far.

...did nothing for my 333+ MHz FSB problem, so this quirk doesn't apply to my MB/CPU.


Also, changing the multipler from 6x to 7x had no effect on stability above 333 MHz FSB.




I'm basically back where I started: the cause of my FSB troulbe is CPU or RAM (or both).

Well, it's not exactly a high priority problem. I was merely curious.

 

[apoppin]

so .. where is it rock solid? what are the last stable settings of your highest OC with no significant errors?

I'm basically back where I started: the cause of my FSB troulbe [sic.] is (1) CPU or (2) RAM (or (3) both)

i'd say (2) - "RAM"