Cold Boot? Memory Died? [email protected] = who you need to contact

[OzSnoal]

Thanks, nice info.

[AndyOCZ]

i have OCZ PC3200 Gold (BH-5) in the San Diego 4000 rig that is listed in the OC Database

 

 

I have OCZ 4000VX in the X2 rig that is listed in the OC Database

 

I have old Corsair CMX512-3200LL rev1.1 (real BH-5) in momma's rig running @ 3.2v with a 90nm FX-55 (one that you guys cannot even buy yet heh)

 

this is why it is so hard for ME to diagnose this problem because I ONLY use high voltage memory in my NF4 boards (TCCD + Angry_Games = big fight cuz it dont like me for some reason)

 

I always use the 4v jumper to give more voltage except in momma's rig (since she has FX55, she dont need overclockin)

 

my OCZ BH-5 = 3.5v

my OCZ VX = 3.4v

 

ive never experienced a cold boot. I've never killed a stick of RAM.

 

personally...until the issue is resolved one way or another, I would go with low voltage TCCD and a high multiplier cpu (like your 3500+, which is the only cpus i will buy now are 11x multi or higher).

 

Let us not forget the 3.3v mod Angry_Grashopper. It can be found here: http://www.dfi-street.com/forum/showthread.php?t=12085

 

It will provide nice voltage up to 3.5v with an adjustable PSU and run your BH5/CH5/UTT/VX as safely as possible without the extra heat created by the 5v (4v) memory jumper. Check it out.

 

That said, lower voltages are always safer and I am a big TCC5/TCCD fan myself.

[Guest burningrave101]

personally...until the issue is resolved one way or another, I would go with low voltage TCCD and a high multiplier cpu (like your 3500+, which is the only cpus i will buy now are 11x multi or higher).

 

Dude, i dont know if you answered me before but i'm still curious why you only recommend CPU's with higher multipliers. There is no reason to run 1:1. People think running 1:1 means your running synchronous with the clock generator but your not. The memory NEVER runs synchronous to the clock generator with an integrated memory controller. There is no link between the clock generator and the memory. The memory clock is obtained from the CPU clock, divided by a factor that depends on the memory specifications.

 

Example = Athlon 64 3500 2.2Ghz 11x multi at 1:1 means the memory is using a divider of 11 to obtain the 200Mhz default speed.

 

So if there are chips like the 3000+ and 3200+ models that have proven to overclock well there isn't alot of point in spending $100-$200 more just for a higher multiplier unless the motherboard your using sucks and it isn't capable of higher HTT speeds. I personally prefer the 3200+ Venice over any of the other venice cores but i'm not real happy with the Venice and San Diego cores because of their crap compatibility with TCCD based modules. Even if you have a 3500+ Venice or 3700+ San Diego with a higher multiplier your probably not going to be running 1:1 with TCCD based sticks anyways cause i know mine and alot of others wont.

[SnooP]

my dodgy speculation. perhaps the 5V system gives out a dirtier voltage with more noise than 3.3V system. but I'm guessing thats already been carefully measured?. I get the impression of this since when running @ 3.2V I had memtest errors on 5V jumper that disappeared when running off 3.3V jumper.

[unreal]

Hopefully I can explain my version of events to you all:

 

I bought the DFI back in January to replace my MSi Neo2 Platinum.

 

The hardware consists of the following:

3200+ Winchester (tested and working for a month in the MSi.)

Tagan 480W (tested and working with MSi K7N2 and MSI Neo2.)

Sapphire X850XT

Twinmos UTT 2*512mb.

 

Now mine shows very wierd behaviour so if DFI would like to test it then I will be more than happy to send it to them, even if it means I need to buy another DFI NF4 to cover me while it is away.

 

The first day I assembled it all and it booted, hardware was all detected, I used the setup guide that the admins posted for all my settings and it was perfectly stable. Xp installed fine but didnt recognise the X850 but thats normal. I installed catalysts (tried 4.12 right up to recently 5.5's) and rebooted into windows to get a black screen! Several reboots and it gets to the login screen and is fine. The black screens only occur after the catalysts.

 

Now here I believe you will think, graphics related... However I sent it off to komplett who tested and confirmed it to be working, as did a friend who tested it also. In these cases it was tested on a board other than a DFI I believe.

 

Now as I said, it boots into windows fine after a couple of restarts so I just did this everyday. I started to overclock and when at 3.3 I killed 2x Twinmos UTT sticks which isnt exactly the highest voltage! The sticks and the mosfets were all actively cooled so to sufficiently avoid overheating. May I also add that both sticks were killed in the same slot. Slot 4 is the slot that would kill the memory.

 

Now this is interested, the 3rd stick I presumed dead was actually not (relates to something I read on Oskar's notes) - after a day of sitting on the shelf it came back to life!

 

Now CPU-Z shows that there is something wrong with my slot 4 because slot 3 recognises memory as the correct chips and correct manufacturing time period however if I insert the same stick into slot 4 then it shows a completely different set of results, it reads the stick as being made in 1999! and a completely different code than to what is actually on the chip.

 

Also I found that when I move the J17 (the 5v jumper) I got the correct voltage however I have since moved the jumper back but smart guardian still reads the same increased voltage on the rail! Now this to me means that the DFI isnt actually switching back to the lower voltage or that it is permently stuck on the higher one - or possibly has damaged the PSU.

 

Also I would like to add that I found that the voltage distribution over one stick would mean that it could boot at 3.1v but to get 2 sticks to boot it would require 3.2v. However this is not what I am discussing right now.

 

Now I have had this board 5 months and I believe I have started to locate the problem. Yes it is regarding cold boots so if all people suffering the problem could confirm.

 

The first time of the day I boot my machine the fans spin up, and they stay at 100%, all of them which are connected to the mobo. Now I find that it will do this everytime you boot until it actually reboots to the time it will load windows. This time the fans will spin up and spin down to the 'normal' level. This would indicate a deffinate problem.

 

Any screenshots and such I will be happy to provide, but with all these fresh installs I will have to recreate the settings which caused the problem and take a screenshot them. So just ask!

 

Anyways if anyone can confirm my idea on the reboots, if not then I will just RMA my board because I know that Slot4 is deffective so shouldnt I RMA because thats a seperate issue in itself?

 

Cheers guys, and keep up the good work.

 

P.s. Phew 40 minutes of typing and thinking there to run through all that I have done, probably missed some important stages so will most more later.

 

Any other results that anyone wants to inform me of, email/msn: [email protected]

[red930]

Playing the odds to get a great overclocker at a lower price isn't worth the hassle for many users like me.

 

I have found that it's easier to hit higher CPU speeds when the default speed is higher to begin with.

Dude, i dont know if you answered me before but i'm still curious why you only recommend CPU's with higher multipliers. There is no reason to run 1:1. People think running 1:1 means your running synchronous with the clock generator but your not. The memory NEVER runs synchronous to the clock generator with an integrated memory controller. There is no link between the clock generator and the memory. The memory clock is obtained from the CPU clock, divided by a factor that depends on the memory specifications.

 

Example = Athlon 64 3500 2.2Ghz 11x multi at 1:1 means the memory is using a divider of 11 to obtain the 200Mhz default speed.

 

So if there are chips like the 3000+ and 3200+ models that have proven to overclock well there isn't alot of point in spending $100-$200 more just for a higher multiplier unless the motherboard your using sucks and it isn't capable of higher HTT speeds. I personally prefer the 3200+ Venice over any of the other venice cores but i'm not real happy with the Venice and San Diego cores because of their crap compatibility with TCCD based modules. Even if you have a 3500+ Venice or 3700+ San Diego with a higher multiplier your probably not going to be running 1:1 with TCCD based sticks anyways cause i know mine and alot of others wont.

[OzSnoal]

Let me start by saying, I don't know what the problem is, I am speculating.

 

The SPD on the RAM - could this be what is causing the problem? I have seen other people posting that their 512MB sticks end up showing 256MB.

 

However if the bios code is implemented according to the Jedec ram standards for SPD, before using any of the first 0-62 bytes (module row/column, ram timings amongst other things), these bytes should be verified with the 63rd byte which contains the checksum for the preceding bytes.

 

Unfortunately the bytes from 64 and upwards are not checksummed. Admittedly they do contain non-critical information such as Part number, Manufacture Date, Serial number and a few more items. But the lack of checksum on bytes 64 and upwards would certainly explain how badly read data cannot be verified and will be used used or displayed 'as is' such as in your case to get the Manufactur Date 1999. This is either because what is in the SPD data has suddenly been changed/corrupted or a 'bad' read of the SPD data occurred.

 

Which brings me back to the BIOS code. Not that I am saying Oskar does not know what he is doing - he most surely does and since I do not have the uncompiled bios code cannot see what he has done. So, this is just written for info:

Since there is no RAM available at power up, all work needs to be done in the CPU registers - I don't believe there are enough CPU registers to store the entire 63 bytes of SPD. Basically what I am saying is that every time a value is read from the lower 64 bytes of SPD, that for it to be trusted a checksum verification is required on the lower 64 bytes. Failing this, there is a risk that bad RAM configuration data will be used to configure the memory controller.

 

I would go one step further saying that using an external SPD reader to come to the conclusion 'the SPD data is correct' is not the same as using the DFI nF4 and verifying that the 'SPD data is correct' on suspect RAM modules. Stating the obvious, completely different hardware is being used.

 

A further possibility is that the serial EEPROMS containing the SPD data are becoming corrupted. Is the 'high Vdimm' voltage being applied to the serial EEPROM chip too? Perhaps the busted modules are getting the serial EEPROMs SPD data corrupted?

 

http://www.jedec.org/download/search/4_01_02_00R9.PDF

http://www.jedec.org/download/search/4_01_02_04R13.PDF

[OzSnoal]

A second thing I have noticed after looking at two seperate datasheets for RAM and taking out my multimeter again...

 

On the dimm slot, pin 184 is used as VddSPD (used to power the serial eeprom that contains the SPD data about timings and manufacturer data), both datasheets rated

VDDSPD = +2.3V to +3.6V.

 

DFI has connected VDDSPD with Vdimm which means that as you adjust your Vdimm, you adjust your VDDSPD (hey now we can overlock the serial EEPROM too!).

 

As there is no reason to adjust VDDSPD, I would prefer if VDDSPD remained at the standard 2.7V no matter what you change your Vdimm to, i.e. Vdimm and VDDSPD should be on their own seperate supplies.

[red930]

In the spirit of speculation, I believe there is a logic problem involved here. Let's take a closer look at your observations.

Let me start by saying, I don't know what the problem is, I am speculating.

 

The SPD on the RAM - could this be what is causing the problem? I have seen other people posting that their 512MB sticks end up showing 256MB.

 

However if the bios code is implemented according to the Jedec ram standards for SPD, before using any of the first 0-62 bytes (module row/column, ram timings amongst other things), these bytes should be verified with the 63rd byte which contains the checksum for the preceding bytes.

 

Unfortunately the bytes from 64 and upwards are not checksummed. Admittedly they do contain non-critical information such as Part number, Manufacture Date, Serial number and a few more items.

If this section is not read by the BIOS it is irrelevant.

Which brings me back to the BIOS code. Not that I am saying Oskar does not know what he is doing - he most surely does and since I do not have the uncompiled bios code cannot see what he has done. So, this is just written for info:

Since there is no RAM available at power up, all work needs to be done in the CPU registers - I don't believe there are enough CPU registers to store the entire 63 bytes of SPD. Basically what I am saying is that every time a value is read from the lower 64 bytes of SPD, that for it to be trusted a checksum verification is required on the lower 64 bytes. Failing this, there is a risk that bad RAM configuration data will be used to configure the memory controller.

Everything after the 64th bit is ignored by the BIOS. The BIOS will verify the checksum and kick the memory out if it fails the checksum. The BIOS never sees anything after the 64th bit so the point is moot. It's a simple pass or fail.

I would go one step further saying that using an external SPD reader to come to the conclusion 'the SPD data is correct' is not the same as using the DFI nF4 and verifying that the 'SPD data is correct' on suspect RAM modules. Stating the obvious, completely different hardware is being used.

 

A further possibility is that the serial EEPROMS containing the SPD data are becoming corrupted. Is the 'high Vdimm' voltage being applied to the serial EEPROM chip too? Perhaps the busted modules are getting the serial EEPROMs SPD data corrupted?

Once again, the data is verified by a checksum? Either it passes or fails regardless of how it is being read. The odds of a corrupt dataset and the checksum agreeing are

 

My logic might be flawed. If it is, please let me know where the problem is located.

[red930]

Good observation. How many other motherboard manufacturers separate these two voltages?

A second thing I have noticed after looking at two seperate datasheets for RAM and taking out my multimeter again...

 

On the dimm slot, pin 184 is used as VddSPD (used to power the serial eeprom that contains the SPD data about timings and manufacturer data), both datasheets rated

VDDSPD = +2.3V to +3.6V.

 

DFI has connected VDDSPD with Vdimm which means that as you adjust your Vdimm, you adjust your VDDSPD (hey now we can overlock the serial EEPROM too!).

 

As there is no reason to adjust VDDSPD, I would prefer if VDDSPD remained at the standard 2.7V no matter what you change your Vdimm to, i.e. Vdimm and VDDSPD should be on their own seperate supplies.

[OzSnoal]

If this section is not read by the BIOS it is irrelevant.

Bytes 63 and up are not relevant to the timings and are for information only.

BUT this does not excuse why the SPD data of user unreal's manufacture date (bytes 93/94) are read differently depending on which slot it is plugged into - which brings the entire SPD reading into question doesn't it? Can you rely on it?

Why is it different from slot to slot?

 

And now onto the checksum, and why I question how the lower 63 bytes of SPD data are read:

Forget about the bytes above 63, they are not important to intialise the memory controller. I don't know how much you know about processors/programming, but you'll have to bear with me. (I sure don't know about how the BIOS code is running )

 

 

Scenario 1 to verify the checksum:

SPDChecksum = 0;

.for (ii =0; ii < 63; ii++) {

...SPDChecksum += Read_SPD_Byte(ii);

.}

.SPDChecksum /=256;

.SPDChecksum ^= Read_SPD_Byte(63);

.if (SPDChecksum == 0) {

...Checksum passed

.}

 

Now I want to use SPD byte 'x' to configure the memory controller. Do I now just Read_SPD_Byte('x') and accept the value given? I wouldn't, i would be fussy and want to know that i got the correct value and want to re-checksum. Why? because it's the safe way to do it and this code is running only once to initialise the memory controller on PC startup. This would be a long winded way, to check it every time...

 

This is how I would do it (if it is allowed to initialise the memory controller in this way and that a certain sequence does not have to be followed):

Scenario 2: (all variables would be a CPU register)

do {

..SPDChecksum = 0;

..for (ii =0; ii < 63; ii++) {

....SPDByte = Read_SPD_Byte(ii);

....SPDChecksum += SPDByte;

....IntialiseMemoryController(SPDByte);

..}

..SPDChecksum /= 256;

..SPDChecksum ^= Read_SPD_Byte(63);

} while (SPDChecksum != 0); // Checksum failed, try again to intialise

[Angry_Games]

Dude, i dont know if you answered me before but i'm still curious why you only recommend CPU's with higher multipliers. There is no reason to run 1:1. People think running 1:1 means your running synchronous with the clock generator but your not. The memory NEVER runs synchronous to the clock generator with an integrated memory controller. There is no link between the clock generator and the memory. The memory clock is obtained from the CPU clock, divided by a factor that depends on the memory specifications.

 

Example = Athlon 64 3500 2.2Ghz 11x multi at 1:1 means the memory is using a divider of 11 to obtain the 200Mhz default speed.

 

So if there are chips like the 3000+ and 3200+ models that have proven to overclock well there isn't alot of point in spending $100-$200 more just for a higher multiplier unless the motherboard your using sucks and it isn't capable of higher HTT speeds. I personally prefer the 3200+ Venice over any of the other venice cores but i'm not real happy with the Venice and San Diego cores because of their crap compatibility with TCCD based modules. Even if you have a 3500+ Venice or 3700+ San Diego with a higher multiplier your probably not going to be running 1:1 with TCCD based sticks anyways cause i know mine and alot of others wont.

 

 

the problem lies in the lower multipliers requiring super high end memory to reach high speeds.

 

I'll give you an example with my Sempron 2800+ 754

 

to be able to run 2400Mhz (default 1600Mhz) you need to be able to do 300HTT...there's not much memory that can do 300HTT STABLE...i see many hitting 300HTT or better...but there's less than 1% that can do it stable (ie pass all the OC Database requirements).

 

to hit 2660Mhz which I did STABLE, it requires 333HTT...

 

this is almost impossible on 2x512MB memory, even the best memory.

 

but, you say, thats when you start popping dividers.

 

this is true...but to a point.

 

Dropping down TOO LOW on a divider gives you such a penalty as to not be able to make up for it no matter how much cpu mhz you can generate. I found this out by dropping down below DDR150 (which starts taking an overall performance hit as it is) to DDR133 and even DDR100.

 

Even cranking the clock up ANOTHER 340Mhz on the cpu, the penalty incurred at such a low divider was serious indeed...what used to be 24,000 3dmark01's @ 2500Mhz ended up being only 22,100 3dmark01's @ 2840Mhz (very stable, but again, the penalty incurred was not worth it, and 3dmark01 was just one of the benchmarks that started losing performance on this divider).

 

I ended up switching up to a DDR150 divider and leaving it @ 2660Mhz, giving me 25,880 3dmark01's. This was the best medium for my clocking. This gave me 333HTT @ DDR166 divider.

 

 

it isn't something that you can just assume you know all about...these guys watched me blast out 1000+ mhz overclocks on this sempron for over a week, and I tried everything I possibly could...I can easily get my sempron to 2840Mhz stable (thats a 1240Mhz overclock!) but again, the penalty incurred at such a low divider just wasn't worth it.

 

I could also crank out 2780Mhz on a DDR150 divider, but my 3dmark01's were only 24,658...thats about 900 points of difference when running 120Mhz slower cpu @ DDR166 vs DDR150...its between that DDr150 and DDR166 divider that you start to see the penalty increase. DDR150 is still a great divider, but I prefer DDR166 or above as at these insane speeds, it really does start to make a big difference.

 

 

but again, this is much different than the AthlonXP/P4 that we are used to from before...on the NF2/AthlonXP, ANY divider incurred such a huge penalty that it was folly to even bother with them (not to mention almost instant corruption of your bios because the NF2 was rather weak about a lot of things).

 

The P4 on the other hand with its 'tricky' FSB of 800Mhz (which was just the ability to do double the data transfer per edge of each clock cycle...a trick really, not a true 800Mhz front side bus) saw less of a penalty, but you could easily tell that there WAS a significant penalty...it could ALMOST be made up in cpu clocks, but you REALLY had to crank them cpu mhz up to try and make up for it. The beauty of the P4 is that with a good cpu, you can easily blast out 1000+ Mhz overclocks.

 

I could do 291FSB @ 1:1 on the P4 2.4C Northwood, giving me 3.5Ghz. I could also do 301FSB @ 5:4, giving me 3.62Ghz. There was a very noticeable difference in all benchmarks when on this divider. Just playing with the machine, you couldn't tell, because lets face it, 3.5Ghz or more is crazy fast (1100+ Mhz overclocks). If you didn't run any benches, you woulnd't have a clue there was any 'penalty' when running the mem divider.

 

But my reasoning then was the same as it is today....if I could do 3.5Ghz / 291FSB 1:1, why would i want to drop down to 5:4 for that extra 100Mhz cpu overclock when the penalty incurred was NOT worth it? I wouldnt.

 

My reasoning is the same on the A64...you have to thoroughly test/bench each setting you can to find your 'best performance' medium...would 1:1 on the A64 be the best? Of course it would, and thats pure and simple to understand logic.

 

but again, who has memory that truly can do all the STABLE testing @ 1:1 when you need 300+ Mhz out it for these crazy clocks that we like to do? less than 1% of users that's who. So you begin dropping the memory divider down, and really depending on your memory (TCCD requires less of a drop on dividers since TCCD can easily do 275Mhz while BH-5 etc needs a lower divider since it usually tops out at about 250Mhz on average).

 

and when you find out your cpu can really throw out some mhz like my Sempron, then you realize that you have to keep pumping up the HTT (FSB), and there isn't any memory that i know of that can do 333Mhz consistently and still be honestly stable...so again you drop to a divider...

 

in my OC Database entry for the Sempron, my happy medium was 333HTT @ DDR166 (actual memory speed of 266Mhz) and it was perfectly stable. This gave me the best speeds/benches/performance out of about 20 different settings/dividers/speeds I tried.

 

 

now, what does this have to do with my recommendation of a high multiplier?

 

higher multipliers are ALWAYS the way to go when you are wanting to max out your entire overclock. 1:1 (200Mhz 1:/01) is ALWAYS the best choice. You might not see but a couple of percentage points in difference between that and DDR180, and again another couple of percentage points from DDR180 down to DDR166...but for some, those percentage points are everything.

 

another thing to realize that Rgone and I have realized since the beginning is that it is much easier to find a good clock with the high multiplier, because YOU DONT HAVE TO TRY AND PUSH YOUR MEMORY AS HARD.

 

take my San Diego 4000+ for example...200x12 is the default.

 

compare that to my 754 2800+ A64 for example, 200x9 is the default.

 

those extra 3x in multipliers can make all the difference in the world. Same as when comparing to the Sempron 2800+ @ 200x8 by default (1600Mhz).

 

to really get the clock speeds up, I have to drop some serious dividers, but the cpu is more than capable of again, doing 2800+ mhz...but to get past 2660Mhz, the penalties incurred by the divider are just not worth it....but if I had a 10x multiplier, I could hit 2660Mhz @ 1:1...since i know my memory does 266Mhz 1:1 easily, I could then begin dropping to a divider like DDR180 and go up to 2840Mhz easily...and more I suspect (i have yet to top out this Sempron in total cpu mhz). If I had a 12x multiplier then i would have even more options, and the ability to run super tight timings with BH-5...240x12 gives me 2880Mhz right off the bat, and I know that I still have 10Mhz more to go on the memory....250x12 = 3Ghz (which i could probably attain if i sat down and really played with it).

 

 

but, you say, those low-end cpu's are more bang for the buck (like a 3000+ able to clock to 2600Mhz for example).

 

that MIGHT be true, but unlike the old AthlonXP and P4, I don't see the ceiling being the same on the A64's (ie you used to buy a 2500+ Barton for $100 cuz it would easily do 3200+ speeds, and the 3200+ would cost $300...and the barton2500 and 3200 would usually clock about the same, up to about 2200-2500Mhz, so you figured heck, im gonna spend $200 less and get a 2500 instead...until they started locking them...then it all changed).

 

i see not only ability to chance for higher cpu clocks easier with a large multiplier cpu (because my 3200+ and 2800+ A64's wont even do 2700Mhz, yet my 3500+ and 4000+ easily surpassed 2800Mhz), but you have a LOT MORE ROOM TO FOOL AROUND WITH.

 

think about it...you got yourself a lowly 3000+ with a 8x or 9x multiplier. You are pretty limited in manuevering room. You get yourself a 3500+ with an 11x multipler, and holy crap you have all sorts of room to fool around and try different things and find the right settings for different memories...

 

is 280x10 @ 2.5-4-8-4 any different than 233x12 @ 2-2-5-2?

 

a bit yes...you WILL see increase in benches with the tight timings...the only one you wont see 2-2-5-2 winning is memory bandwidth, because 233Mhz cannot compete with 280Mhz....but what is the ease of difference in achieving 280 vs 233? For most, it can be an entire universe of difference...because lets face it, almost ANY memory that you buy for overclocking will do 233Mhz (DDR466)...but how many will do 280Mhz (DDR560)?

 

And im not talking 'my memory will do 312Mhz DDR625' and you don't bother testing for stability...I've yet to see ANYONE put up a 100% stable DDR625 set of screenies...lots of users claim theirs are stable, and lots of users (including me) can throw up a benchmark @ DDR625...but not a single one that I have seen has been able to do all the serious stability testing.

 

Any my goal isn't to wow anyone with 'SUICIDE' screens like you see over at xtreme or other places. Suicides are nice to drool over, but they are unrealistic. My goal with the OC Database is to get people to learn stability and the importance of sometimes taking a shot to the ego in order to make sure their rig doesn't BSOD constantly or fail to boot or lock up in games etc.

 

Stable machines don't do this..."SUICIDE" machines are notorious for this.

 

 

This is my goal in teaching the new guys that come here some important lessons...overclocking is cool and can be exciting, but there is some responsibility that you must learn and good habits are hard to learn, bad habits are easy to learn.

 

you may see my overclock of only 400Mhz and scoff at it compared to your 1000Mhz overclock...but if yours is not stable and mine is...then who is scoffing? Not me, I am playing games and working and doing my thing while you are cursing and trying new settings and posting here wondering why you can't get some crazy ungodly overclock and don't understand why you could boot to windows and pass a single 3dmark01 test but 'real' programs lock up, bsod, reboot, etc.

 

(not YOU does not mean any one of you that read this specifically...I am not targeting ANY SINGLE person...it is more of a general YOU meaning 'those of you')

 

 

 

 

then you must keep in mind there will be a few of you who read this and try to argue that I am full of bs and have no clue what I am talking about...and that is cool, you are certainly entitled to your own opinion...but for the rest that are intrigued or curious and want to know how I think I know all of this...keep this in mind...

 

this is what i do all day, every day. As weak as it sounds, I don't have much of a social life because I love my job and run around all day between multiple computers, helping customers by building and rebuilding some, and having some tweaker fun by overclocking the holy crap out of others. I have a lab built just for my job, with 4-8 computers running in it at any given time.

 

All I do all day is test test test, clock clock clock, then come here and post post post. Does this make me a better overclocker than you, the average joe who has a real job and can hit clocks higher than me?

 

No, I am not better than anyone at overclocking.

 

however, I AM more experienced than just about any 10 of you combined, and Rgone and I together probably have more real experience in both troubleshooting and overclocking than any 20+ of you.

 

again, this doesn't make us BETTER than anyone...it makes us more EXPERIENCED than everyone though.

 

Not in Abit, or Asus, or Giga, or MSI...in DFI. All we have are DFI boards. All we eat, sleep and breathe are DFI boards. Day in, day out, daytime, nighttime, anytime, all the time. DFI DFI DFI. I've had dreams of DFI motherboards. I sometimes wake up thinking I was just sitting in front of my computer. When it comes to knowing DFI boards in and out, we are the guys who play with them every day in ways you have never thought of. We have built these different DFI boards at least 100+ times for each model...Rgone and especially me, average about 5 or more builds per day.

 

I build for a customer who has a problem with a PCI-E slot. THen I turn around and drop a different hdd and vid card on one to test for another customer. Then I swap out cpu's and reload windows yet again. Then I test a certain memory, which requires yet another build/reload.

 

then I have customer's who send their rigs to me.

 

 

 

what it all boils down to is not me trying to state I know more than anyone about computers and overclocking. But I do feel like I have much much more experience with DFI motherboards, and I do have more experience in overclocking for the simple fact that it is what I do all day.

 

Again it doesnt make me a BETTER overclocker, as I certainly can easily be shamed by the likes of you guys at times.

 

It does make me more experienced in all the different testings, setups, and scenarios though. Where you might test the DDR180 divider once or twice and run a couple of benches, I have not only tested the DDR180 divider 20+ times on 4 different machines with 2 different cpu's, I have also tested the DDr166 divider and the DDr150 and the DDR133 etc etc and so on and so forth.

 

 

 

and Angry can also be wrong, as Angry still learns all sorts of things every day, not only from personal experience and testing, but from you guys...which is why i built this community and tried to seed it with excellent and intelligent users such as yourself. The instant anyone thinks they know it all (or more than anyone else), is the instant you get shamed and proven to be an idiot...and me, I don't like being an idiot :tooth:

 

 

 

so yea, this is one long explanation of why we recommend higher cpu multipliers. It isn't because of one day of testing. It is because of MANY long months of testing like crackheads that just found a crack truck fully stocked and parked outside their crackhouse

 

hopefully it helps someone. Probably just will irritate most.

 

Also apologize for going off topic and if anyone thinks it is worthy, I will delete this post and put up and entirely new thread so others can read my rantings/ravings hehe.