*Official* Alienware M18xR1/R2 CPU Overclocking Thread - Learn How and Share Tips Here

the one that got 5.1 ghz was because the bclk was set to 1.05/49 like the run below which is 1.05/47

 

When giving more flex, you are hitting the TDP limit quicker (more volts=more TDP) and therefore you are not getting the full potential. That is why you have to only add flex when needed, not just max it out right away.

 

Its an unseen temp issue thats related to voltage if you need 900 primary plane (tdc/amps) for 0flex 45x then as you raise the flex voltage you need more amps to maintain the same speeds since the voltage is pushing your speeds down. flex up then raise amps/tdc/primary plane. Just watch temps too high and it will run at those speeds non stop throttle or not.
*********
Just like your talking about Scook except without the amps raised up tdp watts will only get you so far. Its then that for certain 8threaded benches that you will need to raise the amps/tdc/primary plane more.

 

So were they 49/49/49/49 on the multipliers and all cores active from the bios or did you boot in single core mode? Is 5.07ghz the highest you were able to boot at... im wondering the max...
******
John I figured out I was messing my validations up by renaming the name so i left it what is automatically is there which is the name of the computer... now its been working for me.
*************
nevermind i see the 4cores active with HT
John what flex did it take to boot at that 5.07ghz?

 

With the risk of sounding stupid I have to ask, these are turbo freqs right? Not the base freq? I understand these really high numbers are for benching only but what can I expect with an average 2960xm for daily use with gaming? maybe 5% or 10% above stock? 2.9/3.9?

 

yes. BTW, it easy will do 2.9quad to 3.9single no doubt. With 0 flex and baseclock at 101.4 i set 41/42/43/44 and could pass hours on prime95 and many intel burn tests while staying cool especially while gaming. So those ranges are easily reached... for gaming stably. I'm sure higher levels can be reached... origin pc sells an overclocked chip at 4.1ghz quad -4.8ghzsingle core. With enough flex(voltage) you can be at the same levels on a 2960 or 2920xm and enough amperage(tdc/primary plane) and watts(tdp...short and long power. limit) you can run at those same speeds stable in gaming. the higher speeds you set the more voltage needed adds heat and thats the only limiting factor as to how far you or any of us can push these chips. Enhance your cooling well enough and you'll be all set

 

So for those who are using 2960xm's, have you found any higher gains versus 2920xm's?

The last 2920xm I tested, with stock A03 BIOS level 3 settings at max (45,45,45,45) could hit 4.5ghz solid on single and dual threaded apps and 3.9ghz on all 4 cores. All day, everyday Prime95 certified. No Bclk adjustments.

I'm pretty sure in an OC scenario, the 2960xm = 2920xm, but just checking to see if it provides any better results.

 

I plan to try just that. I need to remember to post back here though.

I do not plan to do any record breaking benchmarking. I like to find a 100% stable setup for my gaming and use that as “my” own benchmark. I hadn’t planned on doing extensive CPU overclocking so initially I will just try to the default available BIOS settings.

I would be thrilled with results like you mention, but expect it to be lower than that honestly. I will be happy to be wrong and surprised!

 

Yeah, I just used the stock Max level 3 settings. I didn't go with the modded BIOS. I've tried/tested ~5 2920xm chips in the M18x, and one wouldn't hit past 4.3ghz single threaded and settled in around 3.7ghz on all 4 cores. One fluctuated at 4.1-4.2 single core and settled in around 3.6ghz. An earlier one a few months back BSOD'd on max settings under stock when pushed and I had to dial back the settings. This last one I have hit 4.5 single and dual threaded and 3.9 on all 4 and can run like that 24/7 with the fan only kicking up with 8 threads going. It also broke 7k single threaded in Cinebench R10, so it is easily the best of the bunch I tried (also the newest) so I was wondering what 2960xm owners are seeing.

 

With flex low enough but not too low you get volt related bsod you should be able to run 4.2-4.6 quad stable with enough cooling even without the unlocked bios.

 

For me before unlocked bios... 4.15 At 0 flex was 100% stable in IBT and Prime 95... On small fft large fft and on blend. Onthe 2920xm. 0 Flex gave me the best possible cooling but as you go over the4.15 you will need more flex eventually .

 

My control was to have settings in the normal A03 BIOS standardized across all 2920XM's tested (ambient temp of 72f was always the same due to central air/heating) and the same tube 'o MX-4 to bear out any heat/Vcore/frequency issues which did show many variations across five chips in regards to peak speeds sustainable while running at least 30min Prime95. We all know all chips are not the same, but it was finally with this last one I saw a chip with real potential.

So using the same settings for those with 2960xm's (Level 3, Bclk @ normal, everything to max, no rear elevation or cooling pads), I want to see what speeds they are achieving under similar, everyday usage conditions.

I didn't want to change any of the conditions from the previous data gathered from the first three chips.

I'm pretty sure the 2920xm = 2960xm in regards to OC'ing (like 920xm=940xm), but it never hurts to gather data to confirm the hypothesis.

 

@ Electrosoft,

Same here, Bro. Even after repasting with ICD24 (stock A03 BIOS) I wasn't able to hit above 4.GHz stable on 4 cores.

@Residualvoltage,

Could you elaborate please, what settings would you suggest to try on stock BIOS. I have a 2960 and am currently sitting stock as even the level 3 OC would bring the temps out of whack and slash the core frequencies

 

do you really want help with this Aikimox?

 

Yes, but my main problem is, this machine is my main business laptop which doesn't sit on a cooler pad (ever) and I can't afford keeping it apart to see the full potential of the system under extreme conditions. But would appreciate your help in figuring out the max safe limit for the cpu under stock cooling. (till the prototype HS arrives, anyway)

BTW, I did a 2920XM vs 2960XM (both OEM) Wprime1024 at stock clocks and here's what I got - both will drop to 2.9Ghz (2920) and 3.1Ghz(2960) @ ~20%.

 

i can tell you how that modded heatsink is going to work, since i had a 10 pound copper block sitting on top of it.

 

Yeah, after seeing your benching setup I had my doubts about success, lol. Still, we may actually get a nice cooling improvement over the stock HS. If we can get 45/45/45/45 @ 85C in Prime95 with the new one - it will be a huge success.

 

well...i got my money ready for when it works.

is the heatsink still the same for the 2960xm? dual heat pipe

 

Stock 2960XM HS is a dual piped one

 

Thats a shame - I was looking forward to a nice suprise when I next opened up....not to be huh?

 

who knows, you may still get a surprise (no paste, aluminum radiator, etc )

 

Hahaha - paper stick-it note where the paste should be....... (seen someone post that chestnut before....)

 

Okay no prob me and John can help and plenty others here. To figure out what you will need what program with what settings would you like at the minimum at 4ghz? IBT, prime? if they are at 4... other apps would be able to be higher since they are much harder on the cpu than games and office work like for your buisness. i'll do it on no notebook cooler, flat on a table stock fan trip points etc (so no hwinfo fan control) just let me know which app and its precise settings you had used and i'll see.. since I am repasted the same with ICD24 and it did wonders and I think I have to repaste because I have one single very hot core most of the time. I will do the testing in quad core and at room temp... so i'll plug in your settings and utilize only what was available in stock bios and make sure it was stable for as long as you mention.

 

i think we all have one very hot core. mine was core 3

 

I know for me the criteria (Simulating everyday use):

#1. Room ambient temp = ~72f
#2. Stock A03 BIOS from Dell (Not the modded one)
#3. No cooler
#4. Flat, level surface
#5. A03 Level 3 OC settings then maxed to 45x 1-4.

Using those settings, I can do:

1 Core = 4.5ghz
2 Core = 4.5ghz
4 Core = 4.0-4.3ghz (Fluctuation mamba courtesy of lack 'o juice)
8 Core = 3.9ghz

This was the first 2920XM to break 7k (7357) Cinebench R10 64-bit Single Core for me. The low was 6.2k and a rash of ~6.4-6.6k's for the other 2920XMs.

My primary query is, "Does the 2920XM equal the 2960XM under same overclock conditions?" I'm pretty sure the answer is yes, but you never know.

 

yes and no.
since the 2960 is 200 mhz faster. technically speaking of course. no.
but since the 2920 can be made to match it. then also technically speaking...yes.
so far no one has wanted or tried to push this chip past 5.0 ghz currently held by two of our members.(2920xm) so no one really knows it's limits yet...

i know that's not what you probably want to hear, but it's all we got right now....

 

True, my primary concern is day to day use in stock BIOS overclock conditions, albeit I'm curious to see how it performs when pushed to the wall. It will be interesting to see how your rig in the sig stacks up when you get ahold of it Johnksss.

 

yes...i too am wondering myself...

 

Intel has never used 100% accurate temperate sensors on their CPUs. A lot of "hot cores" are usually just a sign of sensor error.

 

hummmm. dt quad core cpus are monitored using 5 temp sensors. and once you hit 85 on the 5th sensor...the system shuts off or locks up. the problem here is..we dont have the 5th sensor showing...

 

I know in the Sandy Bridge CPUs there is a 5th temperature sensor but I am not sure if this is the one that you are talking about.

IA32_PACKAGE_THERM_STATUS
MSR 0x1B1

You can use the MSR Tool to read this register but I think HWiNFO is already reporting this data.

MSR Tool
MSR.zip

The temperature data is in bits [22:16]

http://img577.imageshack.us/img577/4829/cpupackage.png

These 2 digits will count down to zero just like the other 4 core sensors do as the CPU gets hotter.

0x1B2 is the control register for this sensor and there are a few interesting bits and values hiding in there. If this is the sensor that is causing you grief, you might be able to modify or disable the temperature limit while in Windows by writing to this control register.

Can you post a MSR Tool screen shot of 0x1B1 and 0x1B2
You can run two instances of MSR Tool so you only need to upload a single screen shot.

 

quad core..meaning sandy bridge as well...

i would love to webb, but i dont have my machine at the moment..

 

The Pic is in your email

 

Thanks residualvoltage. Unfortunately there wasn't anything interesting in the 0x1B2 control register.

It doesn't look like this register is being used on your laptop so a different register must control the conservative shutdown temperature.

 

This is going to be a really nooby question.. so sorry.. in advance.
Got the laptop yesterday and specs are in the sig.
I used to have a QX9300 with permanent overclock.
Should I have my 2920XM the same way?
I use the laptop for gaming and CAD mostly.
And the overclocking options are different from M17x R1.
It goes level 1, level 2, level 3.
Where should I set it, if I just want to leave it overclocked and it's for gaming.
I don't benchmark and I don't really care about negligible performance increase.
Kinda struggling to keep up with sandybridge since I'm not a super-tech guy.
I'm just a user.

 

If you set it to anything higher than level 1 the temps may be too high for gaming, unless you have it constantly on a coolerpad/elevated. Even with a ICD24 paste I can only keep it stock for gaming. For example, it would occasionally hit 85C in Crysis 2. With the back elevated, it stays under 75C but that's not a stock cooling any more.

 

I do have a cooling pad, Cryo LX... but I get your point. I'll just leave it at stock. It's still not bottle-necking anywhere yet anyway.

 

Here's an example. multipliers set to 36-36-38-40 (same as level 1). Wprime + temps with room 24.5c, elevated back. That's about as high as I can go without hitting 85c and throttling.

 

Akimox,

im getting a 2960xm stock clocks, So in order for me to do the factory OC i just have to select the level 1 profile?
or do i have to play with the multipliers?

I just want a safe 24-7 oc even if is a small boost. Not benching, just gaming, afew fps here and there wont hurt.

 

The presets Levels 1-3 aren't the best option, they are preconfigured wrong. Better set everything manually. For 24/7 gaming OC' set the multi's to 36-38 on 4 cores and up to 40-45 for single core. Now, as far as I understand, the idea is not to touch Flex, but slightly up the Long and short duration PWR, and find the spot where the temps are still below 85C under full load and no throttling present. I'm sitting @ Long PWR - 65, Short PWR - 73 and 36-36-38-40 for the multi's.

 

do you guys mind explaing what the turbo-boost technology does? my thought on it is that if there is an application that demands more than 2.2ghz, it will kind of do its own "overclock" to like 2.8 ghz?

example: Intel® Core™ i7 2630QM 2.0GHz ( 2.8GHz w/Turbo Boost, 6MB Cache)

 

Thanks Akimox,

I understand most of it, but what exacly is throttling? does that has to do woth the max temps ?
also you said 36-38 on the multies but 40-45 on one core. does it matters what core i set this on? core 1 or 2 or 3 or 4?

 

Turbo Boost technology automatically runs your CPU faster based mostly on how many cores are active. A core can either be actively working on something or if it is in the C3 or C6 or a deeper sleep state, it is said to be inactive. When using the default settings, as more cores go active, the maximum multiplier decreases and the CPU slows down. As cores go back to sleep, the multiplier can then go higher and the entire CPU runs faster. This is happening hundreds of times a second so you can't set an individual core to use a certain multiplier. The appropriate multiplier continuously changes and it is determined by the CPU.

The other most important thing that controls how much Turbo Boost you get is whether the CPU is above or below the TDP Turbo Boost power limit. If the CPU is over this power limit, Turbo Boost gets turned off. Once the CPU power consumption goes back under the TDP power limit, Turbo Boost is turned back on. Once again this is happening very rapidly within the CPU to maintain maximum performance while not exceeding the Turbo Power limit.

Throttling is when the CPU starts to turn off Turbo Boost. You might be able to set a 40 multiplier in the bios but if this causes your CPU to consume too much power, it will throttle back (reduce the CPU multiplier) to keep the CPU from exceeding its TDP power limit.

Give ThrottleStop a try and you can watch this throttling in action. ThrottleStop accurately reports the average multiplier when a CPU is being throttled. In some laptops, initially Turbo Boost can be completely turned off and then the CPU can be throttled further by reducing the multiplier so your CPU is running at less than its rated speed. That's why throttling is like an evil word. No one wants their CPU to be throttled or prevented from running at its Intel designed speed but sometimes this is necessary to control the core temperature your CPU is running at.

 

Ok so i have pretty well stable 3.7 dropping to 3.6 once or twice and kicking back to 3.7, temps were in the 88s-92s. I have my cores at 42, 38, 36,36, and power long is 99 and short is 94, and that watever power is 900. And my time is 247.627

 

Uncleweb said it all
As for the multi's : 36-37-38-40 would mean 3.6Ghz max turbo when all 4 cores/8 threads are active, 3.7Ghz for 3 cores/6 threads...4Ghz for a single core boost.

You are using the unlocked BIOS which is a totally different story. I'm trying to determine max stable OC without flashing the BIOS.

So far, the highest stable for me (system's back is elevated) is 38-38-40-45.
Attached is the screenie of the temps during Wprime 32 and 1024 MB runs.

On a side note, for some reason during a 1024mb run the CPU would default to 2.7Ghz at about 70% of the test and then go back to 3.8 <-> 2.7 in cycles. I'm still trying to figure whether it's a lack of voltage or simply a time limit for the TDP increase?

 

+rep, thanks for the info.

 

What is happening is that is when your long limit actually kicked and that what it was able to maintain... just peaks to 3.7 not a constant 3.7... raise your long limit back up just a little more and it wont drop like that at 70%

 

Yup... sad you can't force the CPU at those frequencies 24/7

 

Here's the new best with the stock BIOS and elevated back (room 24c):

Multi's 40-41-43-45: It would only last a few sec @ 4ghz and then drop to 3.9 for the most part of the 1024 run. Starting from 70% it would drop to 2.7GHz again...

 

Yeah, that sounds nearly identical to my 2920xm. Try a single and dual threaded run at 45-45-41-40 and see if it holds at 4.5ghz for single and dual through the entire run.