hehe... sure i know this, but I dont like loose timing

I see. Yeah, different rams will have different OC capabilities, I guess...... what if you put it at 2.2v? How far can it go? Make sure you have active cooling.....

Mine is on red slot..... hmmmm.... never tried on the other slot.....

Edit : Wait a second. You are running your bus speed at 475mhz? My P35-DS3 can't go above 480mhz..... it will reboot straight away...... I'm not too sure whether it is due to the CPU or the mobo's FSB wall..... I've tried vMCH and vFSB up to +0.3v but still not successful....

no active cooling yet, still waiting for OCZ ram cooler from bombman's US bulk, almost 8 month already
the ram cooler made by allngap is powerful but the noise really my dad want a silent rig

mine installed on yellow slot, will try exchange the position & slot later.

Perhaps you can consider Water Cooling if you need it silent

haha...actually this rig is not going to overclock, but my hand inchy

8 months???!?!? Can't cancel meh?

Wished I could sell you my OCZ ram cooler..... it is slightly defective.... lazy to RMA to my seller as it is still usable....

huh? so easy defective ar?

Arrived already defective. But i fixed it myself..... Now using it without any problems.....

So you changed the latency from 5-5-5-14 to 4-4-4-14 ? what vdimm are you using ?

it means that it natively support for FSB1333 bus. Since that E6750 already FSB1333, u already use the default bus. Further oc is possible.


PT was duluxe relabelled while "no PT" is the latest version of Tforce 965. Basically the "no PT" version is better coz had gone through 5 revision if I am not mistaken. The PT version has limited vdimm.

Sorry, I made a typo It's at 14-5-5-5 and I've tried to change it before, doesn't work.

I'll try to push the chip further instead on Friday or so, hopefully hitting 3.5Ghz. The RAM seems to be a limitation because I had problems booting up at 43x FSB 1:1 so some tweaking and voltage adjustments is necessary.

In fact, running at 5:6 and 1:1 is different. It means that even it does DDR1000 (500Mhz) at 5:6, it doesn't mean it'll run 500Mhz 1:1

I think it's more proc than RAM.. if I loosen the timings a lot, I can get 500 Mhz out of the RAM. But the proc won't go past 410-ish, and I've already done 1.6v. How far can you push the RAM if you really loosen the timings ?

---

Added on December 4, 2007, 2:40 pm

Guess what, I finally have conclusive proof that 400 MHz at 1:1 CAS 4 is faster than 500 at 5:6 CAS 5

How to prove it's a processor problem or RAM:
If you can run 450 x 6 @ 1:1, it means your RAM can do 450 stable. In my case, the RAM doesn't even do 43x stable.
I also came to know that L2 stepping has FSB wall but not M0.

Btw, how did you prove it? Maybe you can share on the methods used, etc

But wait.. this seems to defy logic. You can prove that the RAM can run at 500 MHz by the simple fact that it is running at 500. Whether it's 1:1 or not doesn't change the fact that it's running at 500. I mean, how can you say it's not running at 500 when it is ?

Similarly, you can say that the proc can't run above say 430, well, because it doesn't work. I mean, if you can run 1:1, and then you find that the max FSB is say 430, and then you run 400 at 2.5x (which means the proc runs at 400, and the RAM runs at 500) doesn't that mean that the proc is limited to 430, and the RAM to 500 ?



Sure.. getting screenshots and data now

can i conclude this as NB limitation as intel chips have no integrated mem controller?

IMO, when it comes to mem oc, AMD have more fun for the overclockers, i mite be bias thou...

That's yet to be proven. I don't see why it shouldn't run 500 at 1:1 when it can run 500 at 5:6. I'm not really sure how AMD's integrated RAM controller works

wodenus, when memory dividers come in place, it would be different. It means if your RAM can do 500Mhz 5:6, it doesn't necessarily mean it can do 500Mhz 1:1

In order to achieve higher clock speed, I must first determine what's the maximum I can obtain out from my RAM. Hence, I try running the max FSB at the lowest available multi.

If my RAM can do 450FSB max, I just have to change back 8x multi and increase it accordingly. If I cannot go more than 440 x 8 for example, then I know it's the CPU bottleneck because RAM is proven to do 450FSB.

I also would not be able to prove what's the highest I can go with the CPU considering 450FSB is what the RAM can do. Unless I get something that can do 500FSB stable, then I can continue upping the FSB, Vcore, etc until the chip stops working

ah_khoo, not sure about Intel's NB but AMD is indeed fun I hope AMD comes up with super-overclockability chips in future too.

So how do you explain the fact that it is running at 500 Mhz at 5:6 ?

claw, i would agree with wodenus. The best method as everyone knows is to test how high ur rams could go using divider. Only then u'll know ur ram's limitation.

So, i would 95% say, the rams are definitely out of ur equation. It's more on FSB walled already, which is soo much better than any M0s or L2s i've seen

another thing, your NB temps might be limiting ur OC..try upgrading to a better NB cooler or something..replaced the stock tim with paste

Ok, perhaps this is what termed as "FSB wall"
I guess in this case, things are a lot different between AMD and Intels.

---

Added on December 4, 2007, 4:58 pm

This can be determined to the FSB:DRAM ratio in the BIOS?

Er.. IINM the ratio just tells stuff when to wait and when not to wait. 5:6 just means that the processor gets five "slices" of time, while the RAM gets 6. This means that after the 5th slice, the RAM is ready but the proc is not. So the RAM has to wait for the proc to finish doing the fifth slice before it can continue. The fact that the RAM has to wait a bit for the fifth slice results in a slightly higher latency (ie. the RAM is slightly faster than the proc, so it's sitting there doing nothing while the proc is finishing up the previous "slice".) If you run 1:1, no one has to wait, the give and take becomes very smooth.


This guy has a better explanation :



From http://www.overclock.net/faqs/119265-how-c...-save-your.html

Okay think of it this way, you know clocks. They go tick-tock, tick-tock, tick-tock. You always have a pair. What you're doing with 5:6 is setting up the system so that for a set period of time, you have 5 ticks and 6 tocks. So it goes tick-tock, tick-tock, tick-tock, tick-tock, tick-tock, tock. Notice the last one's missing a "tick". That's because the processor hasn't produced it yet. So the last tock has to wait for the last tick before it can continue. It can't go "tock" until the processor goes "tick". The next tick that comes along is paired with the last tock and the system runs smoothly again, until the next fifth tock. That explains why there's a bit of added latency there. I believe this is what they mean by "asynchronous".

The thing is that, clocks can't wait. They run at a set frequency, and whether or not anything gets done it still runs at that frequency. So if the clock runs at 500 Mhz 5:6, it still runs at 500 Mhz, it's just that it's inefficient, because a lot of the time the RAM is doing the PC equivalent of shuffling papers and looking busy.

If you run at 1:1 it's synchronous, so the clock runs like, er, clockwork so there's no wait time, and thus less latency.

I don't know if any of this is true, it's just what I understand

From what I understand, asynchronous means it's not running in parallel and in this case, the FSB:DRAM ratio is not 1:1.

Theoritically it sounds logical but one thing we do not know is how the relationship works between the FSB and RAM. Is it as straight forward as one to one or the RAM would always have to wait for the FSB to pass signals to it even at 1:1? This is very technical but no worries as even there's a lag/delay, it'll be in nano-seconds

I am testing things out further and I booted fine at 6 x 433 now with "SPD" settings on the RAM and Performance Enchance switched to "Standard". It passes Super PI and I'll see how much further it can go.

If it is true that there is an FSB wall and running 1:1 is most optimal, I don't see any point getting good RAM's capable of doing 600Mhz 1:1

Good luck.. the "Auto" setting seem to enable higher oc-ing but it's proven to be somewhat unstable during testing.

Those RAM are for overclockers. People have 6 GHz processors already, why wouldn't they need 600 Mhz RAM ?

If there are 6Ghz processors (stock settings) then it makes sense but there aren't any now

If I can't even max out the full capability of my KVR Hynix RAM and if there's really a FSB wall, I seriously wouldn't invest on any better RAMs. Nevertheless, it's up to own individuals

Oh yes there are... there're just not any being sold



True.. but then there are people with serious cooling who can really turn up the voltage, not every proc has a wall

Yep and as for my case, it's pretty pointless to get a better RAM since it won't budge after that certain FSB I tried doing 6 x 438 and Orthos stable 3 hours, passes Super PI but when I switch to 8 x 430, it isn't even stable at 1.55V

I tried upping to even 1.6V, doesn't help so I guess 3.4Ghz is the best I can do with this chip. I never thought a CPU would carry a wall but it makes a lot of sense if it's the chipset

who is Intel Core 2 Duo E4000 sequence processor owner here ? Can you hit over 400Mhz bus ? I can do max is 365 but then cant boot into windows, so the max is 360.

this is my limit ? 360Mhz bus only ??? damn low wei...

Just exchange the ram to red slot, and able to prime 3.3GHz

increase vCore !

bro kmarc, could you do me some favour?

I need your help to check the vdimm reading of your DS3 rev1.0, from 1.8v to 2.2v

bios set | bios reading | speedfan reading

I'm appreciate if you're willing to help me. Thanks

Eh? How come the yahoo messenger not working? I sent you some message a few times, dunno whether you receive it or not.....

We can discuss on yahoo messenger when we're both online....

Anyway, the readings on my mobo is accurate. If I set it to 1.8v in BIOS, then it will show as 1.8v and in windows, speedfan will show 1.8v under vdimm (dunno why yours show vcore2)....

Now, my settings is 2.1v in BIOS, and it is showing 2.1v in BIOS and speedfan (version 4.32)....

So you're running your 10th anni at 4-4-4 @ 444mhz @ 2.2v? Can go higher?

erm...sounds like your mobo voltage is more accurate than mine... now I set 2.2v in bios, but the bios reading & speedfan is 2.05v only

Even on easytune5?

What about your vcore? I set mine to 1.425v but showing as 1.41v under easytune5.....

easytune dont have the reading for vdimm... abt the vcore, I set 1.4v in bios, but the bios & speedfan reading is 1.36v only

?? How come?? See picture below....

no, this is the option for you to overclock, the reading should be under the "PC Health" there

Oh, I know what you mean..... I was thinking that it was indicating the vdimm as it will show the correct vdimm according to the BIOS setting..... e.g. if I set 2.0v in bios, then it will show as 2.0v in easytune5..... maybe it is just indicating the bios settings.....but maybe it is indicating the true vdimm too?

Anyway, why not try speedfan 4.32 and see any difference or not....

P/S: How come I see you online on Yahoo messenger but no reply?

Edit : Same thing with my vcore. Bios set to 1.425v but showing 1.41v in easytune5..

ya, basically easytune5 does not have the reading for vdimm... I wonder, is it your rev1.0, the vdimm bios reading is same as the one you set? because speedfan indicates the vdimm you set correctly. I'm using 4.32 now as well, but the reading is same as the bios reading.

Errrr.... don't understand.....

For my bios, vdimm set to +0.3v......

You have any experience in running 4x1gb? Now itchy hands to try out 4x1gb as my ballistix is just lying around..... I wonder whether the OC would be lower although all my 4 sticks are almost the same in terms of overclockability....

when you set vdimm to +0.3v, what's your reading for the vdimm? It under the place same as where you check your vcore, 3.3v, 5v & 12v

When out to bios to check again.... hehe, all show "OK".....

Yours actually showing actual vdimm readings in BIOS? Hmmm... maybe I should update my bios, still using F3.... lazy....

ah man...isit set Vcore higher first..then reboot just set higher fsb ?
or Vcore and fsb increase in 1 time then reboot ?

i tried 2.0Vcore leh Idle 79C...after that reset CMOS

LOL not that high la.. unless you have like really good cooling, and even then you have to go step by step Usually you increase Vcore and set a higher fsb at the same time

i see... try update and see, wonder what's the reading for vdimm in bios

can i OC pentium 4 3.0ghz with gigabyte 965p-ds3.?

Try and see ?

last time i tried only to 3.22ghz=215 scared to go more..

Okay this proves it :

(1) Set the thing at 400 FSB 1:1, then run 3dmark2006.

(2) Set it at 400 FSB 5:6, then run 3dmark2006.


You'll see that you get consistently lower scores for (2) than for (1). Not much lower, but still lower. And (1) consumes less power as well

WTA,when I playing COH suddenly my pc crashed and my DS3 set my overclocked E6300 2.68 GHZ to 1.86.....No BSOD screen . The sound shutter for a few seconds and the pc restart.

Is it a Heat Problem ( 60c load ) or RAM problem ?

hmm seems vcore not enough to maintain stability to me
wait for the others to comment