Repasting with CooLaboratory Liquid Ultra, any tips before I start?

Is Broadwell supposed to be any better this year ?

Sent from my SM-N900T using Tapatalk

 

okay i will play around and see what i can get out of it what i noticed is, the moment i hit run (i.e. xtu benchmark) the cpu frequency will drop to 3.8GHz and stay there, temps are good, and xtu settings that you proved are applied as well as throttlestop says it's running. =/ i just can't seem to get the cpu to hold it's clocking -.-

 

May I ask you where to find that pressure paper? I'd be interested in trying it on my notebook, but I can't find anything like that.
Thanks.

 

XTU benchmark is not a good test, IMHO. Nothing holds its overclock well with that. Focus on testing with ThrottleStop benchmark, wPrime (a tough one for Haswell), 3DMark11, Vantage, etc. It may hold its overclock with those benchmarks. If you get RivaTuner Statistic Server 6.0.0 it will support native 64-bit so you can use on-screen display to watch what your CPU and GPU are doing with 3DMark11, Fire Strike, BF4, etc. Processor Current Limit and Core Current Limit being too low may also cause the throttling. You might try increasing those setting before raising voltage.

 

okay, so i'll stick to the rest. yeah i will install the riva s.s. today, i didn't need it before but now it's actually a good thing to have since i'm not really sure what's happening.

so i tried some more, when multis amd cpu cache are set to 41 i have no probs, i set the voltage to 1.150 (didn't try lower yet for 41). but when i set them to 42 problems start. i tried up to 1.205V and couldn't get 3dmark11 to finish a run without rebooting the laptop towards the end of the run. should i even try to go higher, or just stay at 41 and accept it? ;-( ;-)


Sent from my iPhone using Tapatalk

 

probably got a handicap cpu from the looks of it. I have had a junk cpu some guy tried to sell me 3920xm and it would only go 43x max and its unstable as hell. and it needed 1.5v for 4.4ghz to run which is nonsense. and I oc it to 4.0 it would drop down to 3.8, with no throttle at all, no temp no clock no modulation throttle shown by throttlestop.

its probably a bad chip no?

 

ah , this would suck so bad -.- because i have no clue how i should get dellienware to replace the cpu.

EDIT: just did another run. it held it's clocking of 4.1 pretty much (min was 4.09). i noticed this before, that there are moments doing benches, where i feel like i forgot to activate throttlestop, but when i check (most of the times... ;-) ) it was running.
is there anything needed to do when running throttlestop on windows7? i remember seeing some post where there was something mentioned like "when running windows 8 in order for throttle stop to be able to do it's work you need to bla bla..." can't remember anymore..
hmmm, once again i'm puzzled by this machine ;-) ;-(

EDIT 2: just saw that in this run of 3dmark11 i just did i got 104xxx for physics score, highest score for me so far.. ;-)

 

now this is really strange: you know how sometimes i run benchmarks when i'm at work, so i will connect using teamviewer and run benches on my aw18 which is at home. now while playing with the settings the aw18 rebooted itself. after oc-induced reboots the bios is always found with the "overlocking" setting set to "disabled" which i then enable again. this time i wasn't able to do this (since i'm nt in front of the machine but using teamviewer). now i just set the oc-profile for 4.1GHz in XTU, rebooted and did 2 runs in 3DMark11. now the cpu score was 104xxx and on the second run 103xxx, both the highest physics scores for me ever reached so far.
puzzled-lvl: 500 ^^
looking at the XTU monitoring the clocking was held pretty much at what it's supposed to...

EDIT: Now I just completed a 3DMark11 run @4.3GHz, 106xx physics score I really don't understand this...and the core voltage is set to 1.190. Didn't try lower yet, will try later on.
I think I will leave the bios the way it is...haha, I don't understand why it will work while overclocking is set to disabled but will not work if it's set to enabled, but I like it... ^^

EDIT2: In the mean time I got the physics score up to 107xxx, but the higher I can get the physics score to be, the lower the graphics score comes out ^^

 

That's what I shared in this photo: http://i.imgur.com/gud2OYi.jpg (see more photos in post #190). I think it has something to do with restriction Dell wanted to have in place for overclocking, possibly due to the problems with 4930MX power consumption and the fan profiles that are messed up and lead to overheating. I am only guessing here, but my assumption is that setting "Extreme Edition Core Ratio Limit" to "Disabled" and using XTU is bypassing whatever was crippled overclocking Level 1-3 profiles. The CPU is "obeying" your settings instead of doing something different.

 

Check your PM about your comments in Edit #2. And, welcome to the world of 330W AC adapter limitations. I'm surprised now that your CPU is holding the overclock that you haven't already tripped the AC adapter breaker with the GPUs overclocked.

 

May I ask what exactly happens when you trip the breaker? Sudden shutdown with no memory dump/reboot/etc? I've yet to observe any crashes that can be positively identified as breaker tripping.. Under too much overclock my screen would just stop displaying or get stuck without actually shutting down.

 

Sure... glad to explain that. It's easy to detect because the symptoms are unmistakable. Basically it's just like unplugging a desktop PSU from the wall socket. Totally dead, LOL.

What you are seeing with the black screen or system freeze is the overclock is probably too high for the GPU voltage and the display driver is crashing with a black screen and/or causing a system freeze. A buggy display driver can cause the same kind of problem. This is really common and it happens to me all the time when I am overclocking the GPUs trying to find the right combination of settings in order to use as little voltage as possible to remain stable. This is one of the symptoms of a TDR error that does not "successfully recovery" as intended.

When your trip the breaker the machine shuts off abruptly if you have an aggressive overclock and are using a lot of voltage for the CPU and GPUs to keep them stable. After the system shuts down, the green light on the AC adapter gets dim and if you wait a moment you will see that it goes out. I have seen a few times when it recovered without intervention, but you almost always have to unplug the brick from AC power, wait a couple of seconds, then plug it back in again to reset the breaker. If you are near the point of exhausting the capacity of the AC adapter and you are right on the verge of tripping the breaker from a slightly less aggressive overclock and over-volt, you may see your display get dimmer, AlienFX may flicker or grow dimmer, then the system may shut down if you don't get a freeze or some other crash from instability.

 

That's what I had thought.. Good to know that the breaker is not just a fuse though

I do wonder why I haven't managed to trip the breaker yet, not that it's anything to aim for. Perhaps 1.1V on the 780Ms is nowhere near "heavy" OC territories by Mr. Fox's definitions..

 

The Alienware 18 also seems to have something on the motherboard that limits its power-handling capacity. If I use the dual AC adapter mod it turns itself off at what seems to be a power load point similar to, if not perhaps just a little higher than where the AC adapter breaker trips.

 

its likely they'll never release a bios for OC at this point. imo windows 8/8.1 is a huge reason to it lol they cant just go against their partner Microsoft although I donno the details I think people should give up on it.

 

Yeah, I think you may be right. I don't know anything for sure about their business relationship, but that's a good educated guess. It's a shame the relationship with Micro$oft is valued more than a relationship with the customers. This is probably true for all of Dell's biggest competitors as well. If they sold systems with an OS being optional and "not Microsoft Certified" they would be more valuable in my eyes because it cuts the ties that are an impediment and restrict ownership freedom. That "partnership" is actually causing harm instead of adding value from my perspective as a customer. I realize I might be part of a very small group that thinks this way. I don't understand why, but there seems to be more people that care about worthless Micro$oft certification more than performance or functionality.

 

This may actually be changing...HP reported that they will offer Windows 7 again and the expected NEW Microsoft CEO currently runs their cloud division. I've read/seen in the news that he actually has linux running as well to try to not be forced to use MS for customers who don't want it...

There is talk that Bill Gates will also step down as chairman...(If you watch his interviews, he really sounds like he's out the door IMO)....

All that said, I think there will probably be changes soon with the CEO at MSFT at least....

 

I don't own an 18, and I never plan on it. What I would personally do is contact Svl7 and open a Paypal donation account. If anyone is going to be able to fix that POS BIOS, it will be Svl7. As for the possible hardware limitations, it the exact cause can be pinpointed, there may be a chance of soldering a new component to the board or manipulating something elsewhere.

And before someone chimes in saying 'we shouldn't have to do things like that', or 'the machine should have worked correctly out of the box', I get that. Unfortunately, you're going to have to, and it does not.

 

actually over on TI| there is a donation account listed for SVL7. but as far as i know through talking to him, atm he's not able to do anything. the ec on those machines (conc. fan tables) is a huge mess. conc. bios i think he can offer to sell unlocked bios chips, which have to be soldered on to the board.
i wonder if some kind of electronic engineer could find the limiter onboard an somehow remove / bypass it..

 

It seems like every few iterations of Alienware laptops has some massive problem that just never gets resolved. The last one I remember is the old M17 and its DPC latency issues. All we can hope is that potential buyers search around and read forums like these. These companies don't care about anything unless it causes their sales numbers to go into the red.

 

Anyway I just ordered myself some of this magical metal. How can I resist!?

Anyone got the stones to try this on their GPU?

Just to be clear.. if the metal touches the red parts I've highlighted in this picture... is that Gameover ?

 

it can mean game-over for sure, unless if you take precautions prior this forbidden-heaven-TIM application(on how - I have posted number of times earlier) you should be more than safe

 

I've actually done my GPU's with Liquid Ultra recently. Sure I gave one of them the code 43 / revision FF (to be replaced by Dell) but on the other it seems to work alright. No dramatic improvement from MX-4 but I think it would be very helpful to apply on both the die and the heatsink or the contact might just not be as seamless as with the CPU.

 

in all honesty, I think GPU's need to stay out of this, GPU's anyway never run too hot, but just about right temps even with heavy OC, like I have tried on my 680m SLI oced to 1.3GHz per gpu couldnt top out more than 84-85c and that was hard beating bench after bench no cooler pad or anything just ICD TIM. CPU's are the ones we have problems with, throttling and powering-off etc etc, now this TIM is a super-solution to CPU regarded problems.
Ofc someone may not be happy with 80-85c on their GPU's and I say nothing but do whatever you feel like you wanna do, maybe it works wonders and maybe not, but I just dont see the point for this paste being used on such high risk application as GPU's where are so many little things that can go wrong if not done right...

Just my 02 cents

 

Yeah it sounds risky. Just lately I've been running a heavily modded Skryim and I need (lol) to OC 120/400 to make it all happen. Temps are pretty good and under 78.. Id just love them to be about 70 since its the kind of game one could play for extended periods..ahem

So brother GTO you think you may have sent your GPU to an early grave ? You may have just put me off trying.

 

Yeah, I left IC Diamond on 780M cards in the M18xR2 and AW18. They run plenty cool with IC Diamond and it didn't want to expose them to unnecessary risk. Just one 780M card costs more than a 3920/3940XM on eBay, LOL.

 

Well the code 43/revision FF issue happened straight after a crash during the maiden 3DMark run following the paste job so I have little doubt that it was due to using the Liquid Ultra. I did examine the card afterwards but couldn't identify any spillage with naked eye. Hopefully the issue goes away with replacement GPU's.

 

That really sucks. Thanks for being brave and trying bro.

 

Was just gonna keep my shame to myself but couldn't just watch as you guys are about to repeat the folly..

 

Repasted my 7970M's with it. No issues and runs cool(910/1300)

 

What sort of temps do you see under full load?

Also for those who have repasted, I'm interested to know if the liquid spreads out much?

 

i'd use thermal pads around the GPU die to make extra cooling and also protect other parts from contacting with cool liquid ultra. but imo, the vram and bunch of other tiny chips on the GPU are the ones that runs hot, the die is usually ok.

 

Gaming: mid to upper 60's
Benchmarks: Upper 60-low 70's

 

code 43 / revision FF is what is normally associated with the back plate grounding out.

[Guide] Curing Shorted Support Plate on NVIDIA MXM Modules (Symptoms: Display "Static" or "Snow" Artifacts - Error Code 43 - Revision FF error)

 

Do you think this could be useful for the cpu too? I mean putting thermal pads around the die.

 

Thanks for the suggestion. I've checked it out but couldn't really see how the same problem could arise from the current-generation backplates. Something's probably grounding out but I had no luck finding it..

 

hey guys
so i finally got rid of my hated aw18 and got back into the R2 game now i'm trying to get some oc in there and since it's the first ivy bridge xm i've ever owned i got some questions. i'm trying to get everything stable @4.5ghz (using liquid ultra for temps), and brother fox gave a hint to mo to use 38 additional voltage in the throttle stop performance profile as well as both package power limits to 120W. what i don't get ist, in the bios performance tab, do i set flex vid override to the same then? and also, is there any other settings i need to play with in order to get the whole thing stable, as for the moment i haven't been able to run 3dmark11 without it crashing on the physics test and i'm already up to flex 43 (just since i saw ther's some insanely low flex around here ;-)

EDIT: so with 43 flex i can get 3dmark11 to finish @4.5ghz (cpu max temp 84 degrees), cpu score is 10610, but the gpu score and combined score is horrible, it took a big (over 2000 points) dive. i'm am bit puzzled =/

 

sounds like to me it throttles, due to either not enough voltage..when under load, or it spiked past the 105 degree point... but probably needed more voltage .

 

It's actually not going to throttle from too much voltage. Not enough voltage won't make the 3920XM throttle, but it will just crash the running program or BSOD. Not having enough Pri Plane (amps/watts) will definitely cause throttling from power starvation.

I can run with 38 flex (same number in ThrottleStop) @ 4.5GHz no problem, but it may vary between CPUs somewhat. You're most likely exhausting the 330W AC adapter if you have a hefty GPU overclock and that's what is tanking your benchmark result. Run 3DMark11 Tests 1 through 4 at 2.9 to 3.0GHz using a ThrottleStop profile and switch to 4.5GHz (using ThrottleStop hotkey manager) only for the Physics and Combined tests and your GPU score will probably back to what is before or even better. You can use ThrottleStop to set Flex, Long/Short Power and Time Window. Run 3DMark11 Physics Test in Window mode and keep the flex as low as you can. If 3DMark11 crashes at 4.5GHz with 38 flex, use ThrottleStop to change it to 40, run it again, etc. Don't go higher than you need to for stability and good Physics score. Also watch for downclocking and if it doesn't hold 4.5GHz on all cores, goose it with a little more Pri Plane power (960 should be enough if auto is not enough... 0=auto, which is the same maximum as setting 896 manually).

 

it should work with CPU yes. i'd put few pads around the die, and on GPU vram and other small voltage regulator to see how much it has cooled. to be honest theres almost no way to monitor the temperature other than GPU die and other parts of the laptop unless you use external tools. Hwinfo64 only shows graphics Vram temp and shader temp on 6990m, after that none of the graphics card showed anything.

primary GPU will have a hotter shader and vram due to heatsink design (left when facing monitor), 2ndary GPU will have hotter core again due to how heatsink is designed. if you compare R1 and R2 heatsink you'll realize they did some adjustment to this and R2 GPU heatsink is a bit better modified for this issue.


@everyone, I like my CPU (after trials of many chips) at low flex

@mr. fox, is there a reason not setting pri plane to max? I just leave mine at max, at 8100 ish. 2nd plane is for iGPU so I have that at 0

 

so it's not rock solid yet, i got 38 flex at 4.5ghz and once the intel processor diagnostics tool still crashed the laptop making it reboot. i will try upping flex just a bit to see what will happen. as far as i can see the cpu not holding its clocking on all cores is not an issue, will check again with riva ss later on.

EDIT: what is the max flex that is reasonable for 4.5ghz before considering that it would be better to focus on getting 4.4ghz stable instead of 4.5ghz?


Sent from my iPhone using Tapatalk

 

That's strange. I read on a few websites that tx 4 is one of the best ..?

Sent from my SM-N900T using Tapatalk

 

I tried using TX-4 with the M18xR1 and 2920XM and it really sucked. It lasted only about 4 or 5 days of heavy overclocking. It is one of the worst I have ever used for CPU overclocking.

 

Still going like gangbusters... Alienware should consider using Liquid Ultra as their "factory" TIM. Here's a FFXIV bench with the 4930MX @ 4.5GHz.

 

Nice job brother Fox!

I got my Liquid Ultra in the post yesterday. I'm procrastinating because I'm trying to make a decision to do the GPU's or not. Doing the GPU's should result in better efficiency no ? I'm really bummed out being shackled by this 330w PSU so want any boost I can get now...of course cool temps is something I want too.

Just to be sure.. these little metal lugs around the core. Are they Aluminum ? If so I might not do it until I'm sure of how the TIM spreads after pressure and stress... Really not sure.. I can't decide

View attachment 108683

 

Those are surface mount components and shorting them will kill your card.

 

Thanks Meaker. I think ill leave pasting the GPU with Liquid Ultra alone for now until I know exactly how it spreads.

 

Please don't risk it on your video cards. Use IC Diamond or some of that Gelid GC Extreme paste on the GPUs. I'm not going to use Liquid Ultra on my video cards. They don't run hot enough to need it and it's just not worth the risk. It's totally worth it on the CPU and I'm sold on the Liquid Ultra for the CPU.

I am surprised your video cards don't have the black plastic cover on those surface mounted components. My 680M and 780M cards have all had a plastic shield over them.

 

I'll take your advice. That pic was just a random google image i used.

 

Man I dont know about you guys but I wouldnt want anything like that near my system Gallium Induced Failure of a Heat Sink - YouTube