How to Overclock the Alienware 18 and Haswell CPU (or actually have it run full stock Turbo Speed)

thanks for your replys guys, my wife seems to have some stomach problems, so i will post the links to the benches, then look after her, and then come back and read your answer and respond

here's the link of an 11 run on stock clocks NVIDIA GeForce GTX 780M video card benchmark result - Intel Core i7-4930MX,Alienware 01W2J2

and this is one with an oc on the gpu (+130/0...just to test): NVIDIA GeForce GTX 780M video card benchmark result - Intel Core i7-4930MX,Alienware 01W2J2

talk to you later

 

HWBOT uses one authorized version of numerous benchmarks to avoid apples-to-oranges comparisons. If everyone is using the same version then the results are meaningful. As an example, you will notice similar differences with Unigine Heaven 1.0, 2.0, 3.0, 4.0. You have to pay close attention to the software version before you get overly excited or disappointed when comparing your score to someone else's because the difference in scores can be fairly great running a different software version than the person you are comparing scores with.

As far as why your 4.1GHz wPrime is better, that can vary even on the same machine from one day to the next. What processes might be running, what the temps are, etc. If you run it again you might get an even better score or a far worse one. I have seen it vary like this to some extent with all processors, but with Haswell the results are even more consistent. I just ran the same wPrime 32M benchmark with identical CPU settings as yesterday and got a 6.199 sec score. I rebooted the machine and got a 5.891, logged off and back on and got a 5.891 without changing anything.

Looks like you need to spend some time testing different CPU power settings to find something that works better. If you are using default BIOS settings your CPU performance will be poor and it will affect your graphics score. With that moderate GPU overclock your physics and graphics scores both took a nose dive. Notice your CPU clock speed on one test compared to the other.

 

hmm, but the cpu clock speed was the same on blth tests, it wasn't oc'd. i'm not too much of an xtu expert, i'm not sure which settings i should play with except for the multis ;-)

EDIT: sorry, I know now what you mean..didn't really get it at first.

 

Glad you figured it out... in case anyone else was trying to figure out what I was referring to...

Compare: Maximum turbo core clock: 3,791 MHz vs 3,396 MHz (400MHz difference in turbo speed between runs and 2.1GHz is also not the correct non-turbo clock speed... default minimum should be 3.0GHz)

Check your BIOS and confirm Non-Turbo Flex Support is "Disabled" and the Flex Ratio Override is set to 30 or 31. If it is less than 30 or 31, temporarily enable it, change the value, then disable it again.



And, this is only the information captured in the "countdown" scan before the benchmark starts. The CPU could be clocking down far worse once the system comes under load.

Use the 32-bit version of 3DMark11 and on-screen display to watch what is happening, or look in the HWiNFO64 sensors to see what your minimum CPU core clock speeds were. Using on-screen display works best because you can see in real time what is happening and how long the problem is lasting. (The on-screen display does not work with 64-bit games and benchmarks, that's why you need to use the 32-bit for this diagnostic testing.)

If you're not sure what I am referring to regarding on-screen display, see the opening post in this thread:

http://forum.notebookreview.com/ali...mperatures-game-benchmark-screen-display.html

 

Is it sad that I can tell by eye without even looking at the FPS if my 3dmark run is going well or badly?

 

Yes, if you are used to running it frequently like we are that is easy to tell. You will know within 3 seconds or less after test #1 begins that it's not going well and you might as well hit the escape button and stop the benchmark, LOL. That is kind of sad in a way... you know you've been benching a lot when you can do that. Using the OSD should be helpful to pinpoint what behavior (CPU/GPU) is causing the system to run poorly.

 

Yes, or even having something like MSI afterburner/XTU logging in the background and looking at the usage trace for the CPU/GPU.

I've been scene watching on 3dmark since 3dmark 2000 *sigh* I am a bit sad on that front but oh well

 

I get 9800's in 3dmark p scores... but my gpu's don't hit their performance mode, they stray at the 384mhz speed. soo annoying. In games and everything else they work fine, but just not in benching. As they hit nearly 100fps already. so annoying. they stay at 50 degree's I have even tried with stock driver and clocks. but power never passes 52-53 on them.

Any idea's ? is my cpu's power stopping them somehow ? Its weird only getting graphics scores of 16000 and then 5000 for physics and whatever else. shouldn't they be much higher, at least for graphics ?

Im wondering is the unlocked vbios is the cause..

 

yeah, non-turbo flex support is disabled. flex ratio override is set to 32 as in the screenshot you posted in the first post of this thread.

i did one more run with an oc on the cpu, all set like you mentioned in the first post and the multis set at 43, this was the outcome: NVIDIA GeForce GTX 780M video card benchmark result - Intel Core i7-4930MX,Alienware 01W2J2 NVIDIA GeForce GTX 780M video card benchmark result - Intel Core i7-4930MX,Alienware 01W2J2
cpu hit 99 degrees celsius.

now i'm about to install 32bit version of 3DMark11 and osd.

 

I am running the same unlocked 780M vBIOS from svl7/johnksss so that should not be causing problems. Same vBIOS is being used for Alienware and Clevo.

The GPU core clock is never reported accurately in the 3DMark11 results. The memory speed is, but the core clock will always show 324MHz regardless of the GPU core speed.

The most reliable way to watch what is happening is using OSD and 32-bit 3DMark11 and observe in real time what the CPU and GPU are doing. What you see may be an eye-opener. I have seen the 4930MX drop below 100MHz under heavy load with the wrong settings. Of course, this totally trashes both graphics and physics test results because the machine can't run well at all with such low CPU clock speeds.

You should be quite a bit higher on graphics and physics score. If you are not already doing so, use ThrottleStop to help keep your CPU clock speeds from dropping under load.

This was a pretty healthy overclock (4.6GHz, which power throttled down to a humble 4.2GHz during the Physics test) but most should be able to get fairly close to this... at least in the mid P14k range, at least 9k CPU score without an insane CPU overclock and around 16K on GPU score with stock GPU voltage... about 975/1500 GPU overclock.

3DMark11 Search Results for 4930MX and 780M SLI



Here are the settings used for this run... I believe 1.025 GPU voltage (much more GPU voltage than that shuts down the Alienware 18 due to some unidentified system power limitation)


(click to enlarge in new window)

Well, that is an improvement. Your CPU did better on that run that it seems to have on the previous runs... 700 3DMarks higher CPU score, so you're moving in the right direction. Don't worry about the 99°C CPU temp. We already know Haswell is a hot sucker and the thermal trip point is about 105°C. If it gets too hot it will throttle or shut down, but it won't harm anything in a short benchmark like this. It's made to do that. If you have sustained core temps like 99°C it's not a healthy situation, but Intel CPUs are self-protecting in the situation of a spike.

 

damn, fell asleep last night..
in the end i was so tired i couldn't get the hwinfo64 to work with osd ;-) i will do that today. what i did try was setting the multis to 44 instead of 43, but i guess the cpu probably reached 105 degrees, as it suddenly shut down when doing the 3DMark11 cpu test. will finish hwinfo and osd later on and then check. i so hope dell with fix or release those fantables, its such an annoyance to not be able to hsve the cooling done right.

BTW: and i'm - sadly - running on windows 8... i think i might bump a hyperx ssd in the system within the next one or two months and then get rid of the 8 and install my windows 7 ultimate.. just don't feel like installing windows 7 right now since i will have to do it again anyways as soon as i get the ssd.

 

Switching to Windows 7 will give you a better OS if you're not a fan of the cartoonish Windows 8 UI mess. Windows 7 has a minor advantage where the CPU performance is concerned, but it's not going to fix what you're seeing in benchmarks unless something is seriously wrong with your current Windows 8 installation.

 

yeah, it was MSI after burner that was reporting that the entire time, my gpu's stayed underclocked.
I basically detach the monitor and make it wide, and long enough so I can view at least 10-15 minutes of data

 

The results you are seeing definitely indicate something is malfunctioning. Not sure what is causing it. The behavior sounds similar to what happens trying to game or bench on battery or using the "Power saver" profile. I know you are not doing that, but it behaves in a similar way.

 

yeah, but oddly in all games, Im fine. though this si 64bit windows 8 3dmark so who knows

 

Don't know that this will work, but I saw something similar posted in another thread where 780M was not clocking above 324MHz... try a power drain. Disconnect system battery (not CMOS) and AC adapter, hold down power button 30 seconds and then reassemble. Easy enough to try and nothing to lose. As soon as I read that I thought of your situation. Hope it works. *crosses fingers*

 

A problem that I frequently run into, is that the machine shuts down as it exceeds the power draw limit. To prevent that the GPU cannot be clocked as high as they could be, because I couldn't really raise the GPU core voltage much higher.

 

Well, woodzstack is running into problems with the GPU clocks staying in a low P-state and not even running at stock 3D clocks in 3DMark11. If you are pushing the CPU and GPU overclock, it is imperative to use ThrottleStop to control the CPU or it will sap all of the power away from the video cards and result in terrible benchmark scores. This is not something new on the 18... M18xR1/R2 had the same issue, but it is not difficult to manage.

There is definitely a power handling limitation on the 18, but it doesn't seem to have a remarkable effect until you start pushing it too hard, then it shuts down every time.

@woodzstack - in addition to what I have already suggested, try disabling SLI and see if a single 780M exhibits the same behavior. That might shed some light on whether the GPUs not ramping up in 3DMark11 is a power depletion issue. If a single 780M behaves in the same inappropriate way, then it's not a lack of power.

 

Some months ago Xbitlabs ran some stress test on the 4770K, clocked at 4.4GHz running @ 1.2V. It was running using NZXT Havik 140 cooler, which is not the best CPU cooler, but with 6 heat pipes and 2 140mm fans, you would expect it to be not bad.

Nonetheless one core was still hitting 96C, which shows how hot the CPU is: http://www.xbitlabs.com/picture/?src=/images/cpu/core-i7-4770k/17.png

Even without overclock, the 4770K was shown to be almost 20C hotter than the 3770K at full load : Intel Core i7-4770K CPU Review. Intel Haswell for Desktops: Ruin of Our Hopes?. Page 12 - X-bit labs

 

That is a really good article. Thanks for sharing it.

Here is two of the very meaningful comments:

All of the bogus marketing hype around Haswell reminds me of the " The Emperor's New Clothes" folktale that many will remember from their childhood. It's nice to see when someone writing a review steps us and boldly declares "Hey, the dude is nekked" LOL.

Improvement of the still very lame Intel integrated graphics is a farce and a huge distraction. It's pretty sad if they thought we would actually see any value in that. I rather see them use that processor real estate from something meaningful, like more CPU cores.

 

moderators should sticky this thread. Its large enough.

Anyways, I wonder if any aftermarket cooling options will come out for the new 18's.

Personally I'd have combined the heatsink fins, into a single piece, to make more use of the fans.

 

I have mentioned that several times, they need to learn from MSI and link the heatsinks with copper pipes. But do you think it will be picked up at all? Nope, a lot of times innovation is looked down upon because of the cost of redesigning and manufacturing. Any tangible improvements to the chassis will only come next gen.

Frankly I don't see why they have to remove SPDIF out and e-SATA (which also works as a USB port) port that used to be on the R2.

4x 2.5" drives on the R2 vs 3x 2.5" drives on the 18, and 5 USB ports on the R2 vs 4x USB ports on the 18.

 

Kind of odd that they did that right, especially considering there is so much more room too in this new chassis.

 

so i got a little further ;-) i reflashed the cards with the vbios of my compatriot svl7, now things look a little different: NVIDIA GeForce GTX 780M video card benchmark result - Intel Core i7-4930MX,Alienware 01W2J2
this was on standard voltage..didn't care to raise it yet. cpu was held at 4.2ghz and throttlestop kept it from throttling when those temp spikes in the combined test arose. +180/+320 on the gpu for that.

 

Yay! That looks a whole lot better... +1766 3DMarks. Nice!

 

yeah that makes it interesting again
thanks a lot for your help i'm wondering about some xtu settings which are ok to run for 24/7. i don't really want to have the cpu held at 4.2 ghz all the time...at least not until there's some decent fan tables ;-)

 

hmm, so i can't get this to work... can anyone explain to me, how to set xtu in order for the cpu to have all cores @42 multi, but, when not under load, clock down (and not hold those clocks the whole time). i'm reading through everything concerning xtu i can find, but somehow don't get it to do what i want xD

 

At the very least you will need Core Current Limit set to 112.000A and Processor Current Limit set to 123.000A. Under load it is always going to clock down some because it is a Haswell and that's how they behave. You can probably get away with Core Voltage being lower than default... try 1.175V, 1.165V, etc. Look at the opening post. You may need to tweak the settings some to find the sweet spot for your CPU.

 

ya I never used throttlestop until last night, wow, it keeps things from throttling.

With a XTU profile that usually gives me 910-915, I now get 950-960. Im considering going for the 1,000 mark.

Throttlestop seems to stop the graphics cards from throttling too.

 

Cool. Yes, ThrottleStop is amazing. I have been using it for a very long time. I am glad it is working well for you also.

It can sometimes have conflicts with XTU and HWiNFO64, since all three of them access BIOS sensors. If you launch XTU first, choose the settings you want, apply them, then launch HWiNFO64 and ThrottleStop, it will avoid the conflicts. For example, you can change BCLK (reference clock) in XTU without rebooting. However, if HWiNFO64 and/or ThrottleStop have already launched, you may be forced to reboot. The CPU power settings you have applied in XTU should also be visible in ThrottleStop.

If you haven't already tried it, the Unpark CPU App seems to help a lot with Haswell. It helped Ivy Bridge some as well, but a lot more with Haswell. As the name implies, it help circumvent Windows from being stupid and overzealous about parking CPU cores to conserve energy. I don't know if it works with Windows 8/8.1... will have to try it to find out. Using the app is easier than using regedit, but accomplishes the same purpose.

 

Okay, well getting a stable system with throttlestop and XTU running seems insane if benching. I always crash, have not completed a single bench... even though I kept temps below 60 degree's Celsius.

I can complete a bench without throttlestop, and it of course, throttles, but with throttlestop - no way. Always crashes, and its annoying. Any advice ?

I was using anywhere from 1.15v to 1.40v, current from 130A- 175A, tried over 100 configurations. 75% of them stable without throttlestop running, temps fine, all below 80 degree's, save anything with voltage above 1.3+ then it stayed below 95 degree's..

I always unpark CPU's, in XTU you can sometimes see it saying core count 1 or such, and go figure, the cores parked again.

However, as much fun as it is, wishing to rip my hair out, I got a few benches but at 4Ghz giving me 960-968. then of course, anything past a certain power take from the cpu does the whole ZEEEWWWWP! *black screen*

This CPU no matter what you do, is never going to beat those scores we have, not until something drastic changes about the power consumption or limiting factors somewhere. I suspect the power bar too, is limiting it, because a few times, it shut off, and wouldn't work for a few moments, and was incrediably hot.

This was when running around 1.56v and 180watts for the cpu at 5.0Ghz.; I can complete the bench, but not with throttlestop. and it throttles around 3.9-4.1ghz the whole time. spare occasion it jumps to 4.5Ghz and 4.3 but only for a split second once. and then bzzzzwwwp! noise and some warning message about having no AC adapter when you turn the pc on, and some new windows 8.1 screen asking if I want to run diagnostics, and AW diagnostic options coming on. After 5 minutes, my AC adapter turned back on.

I get better results when I move the AC adapter outside the house ( 6-7 degree's outside) and run XTU. Things are much easier to be stable. It makes more difference for me then trying to keep the laptop cool. So, the 300watt AC adapter we have, is a very large limiting factor, might even be the single most handicap on the system, IMO, but I suspect theres the same limiting circuitry inside the laptop too, as dual PSU mods make no difference at all, save they won't overheat and shutdown like mine did, but add no other benefit.

an ADC says we pull almost 560 watts from the wall, when I use 2 of these. and 450 watts when im using a single Ac adapter, and somewhere near 150ish when not in load, and idle, though unparked, and throttlestop running. I am starting to hate this laptop for all its bogus mysteries not adding up.

 

I might be able to share something helpful, but can you be a little bit more specific? 3DMark11? What kind of crash, and at what point? BSOD with a bug check code, hard freeze in Test #1, lockup in the Physics Test?

It is terribly frustrating. That's why I haven't burned many calories on it lately. I have spent countless hours tweaking, just as you have and kh90123 has, without achieving anything particularly phenomenal. To be completely clear, the 18 does an acceptable job of gaming and it's a nice system, but it's not living up to the legendary reputation for extreme performance.

Really great XTU benchmark scores are kind of meaningless when the rest of the benchmarks, especially the Futuremark benchmarks, are not superior. I have some better scores at some things, but they are far below my M18xR2. Something is interfering with you getting similar results with the 18 and maybe we can isolate what that is.

 

I should, as we all should, keep sharing what little things we do find. I have found, that after installing the unlocked vbios thanks to T|I guys, you need to triple check your graphics drivers. I would recommend some simple 3d tests and such, see if they score better, run smoother. because all in all, they absolutely should no matter what, run better, no excuses here. stock for stock. If they do not - maybe try installing Dell's stock drivers, then installing whatever driver you actually wish, next. for me, and from what Ive seen and read, Ive notice this worked for a few people.

Now, CPU - in bios, first two options - enable, second two option disabled. easy to remember.
Hidden things we still can not see - if you have OC option 1,2,3, the higher the number the better. select the OC 3 state, thewn disable that option anyways, however, you might notice the clocks stay, go and change those.
Set your clock speeds to something reasonable, x39-x42 ish on the cores. think of it as its going to take you x hours squared to get it running smoothly, where x = (chosen multiplier-39).


If Flex is still selectable, you can set it between 1.165v - 1.22v (1165 -1220mv) if not, no worries.
set the time to 128 window

set your power maxes here, to inbetween 75- 115,( short term is higher, long term is the sustainable long term run..)

you shouldn't be needing to come to bios after this, again.
next XTU will do all the higher clocks, changes, but this is enough to get things started/rolling and make XTU work. (Usually, every system different)

when in windows, UNPARK your cpu's
then close the appt.
check with cupid
run your hwmonitor to check temps etc. close them both.
run XTU

play around with your settings and find something that can actually finish a bench run in the XTU and then save the profile and make it default when windows starts.
I find, if you notice things reversing positons in XTU, like core cache and cpu limits.. then make them both the same. who cares then it they switch.
It seems, more often then not, having the voltage 20-30mV less then we want, and using a dynamic offset for the core, of that missing value, helps.
don't add more then 1mV to anything else, and only if your system is really cold, if ambience is really warm, you might wants to try as much as -1mV but this could cause severe crashing, as the I/O bus and such use very very little voltage, something like 0.02mv to operate or around there, so a +1mv is huge.Just needs enough to latch and relatch etc..

I believe everyone should be able to find a 4.1Ghz stable x 4 core clock at 30 degree's ambience. you can also get 95% the same performance out of 3.9Ghz and overvolting like mad. extra heat, and about a year less of cpu life for sure, but it can work (for those with 4900's, 4800,4700)

These cpu's will ALWAYS throttle if they get too hot. Mechanically speaking, throttlestop will NOT change that. But throttle stop can and will stop it from throttling from too low voltage, when it has a load, and it trying to throw that load around at ludicrous speeds. you will usually get a insta bzzzzp power off. that means you didn't have enough juice. when things are running too hot, you get errors and corruption, artifacts, and things happening that shouldn't happen, waaay to fast and waay to slow timings for things.. when you get insta-shut offs - that's power. Remember that for the future. its been this way for 15 years so far, and been the number one tip for overclocking, something most people don't seem to pass along. Its why people sometimes buy a extreme core and get some crashes and then never end up OCing it enough and think other people just have better cpu's or better luck - ADD THE JUICE increase the power. CPU's don't just fry. motherboards, sometimes do, usually if you start overclocking the FSB and southbridge, and QPI lanes or stuff. but never the CPU. Trust me.The CPU's will always take care of themselves, at least since p4's existed. p3/p2/p1 and before, ive seen some fry - but 2700Mgz on a p2 is waay to high.

Throttlestop, seems to prevent the GPU's from throttling low too. Be careful there, if you OC them too high, or don't feed them enough juice, or if they run for five minutes going slowly from 92-100+ degree's.. as I've seen. Crash imminent.
stock seems fine, usually. with unlocked vbios, stock is great, infact I lower the max power, and core. Anything else, and I can not sustain my CPU OC. But I've notice, I still get better graphics performance, the better my CPU runs, no matter what.

Now, whats causing me to not get higher then 13500 ish on 3dmark idk. but theres apparently still 10% of people faster then me. which is unacceptable, lol.
before throttlestopping, I used to score 5k less.

I have force 16 frames to be prerendered. I usually force highest quality and all options to max, or at least what NVidia inspector lets me use. I use either a 4k monitor, or a old hp 3065 30'' that does 2560x1600@60hz.
Never use v-sync .
I use 331.65 drivers,.

 

Flex VID override shows up in BIOS when you set the voltage mode to Static in XTU. If you disable Static and go back to Adaptive mode the option will disappear. Adaptive mode doesn't work for core voltage below 1.167V, so if you're running it lower than that voltage you will have to use static voltage. AFAIK, in static voltage mode, setting a negative offset to get the voltage you want, is the same as setting Flex VID directly. 1mV is more than 1mV, you can see that in ThrottleStop. The min step is closer to 5mV. From what I have seen, Flex VID = core voltage.

Trying to run the CPU at 4.1GHz with the overclock and overvolt on GPUs have resulted in hard crash in combined test for me. Most likely due to the system shutting itself off as it pulls too much power. It's either that, or hard crash from CPU overheating and the fan still idling. I have since gone back to 4GHz as it's a bit more stable.

I noticed, using ThrottleStop to disable Turbo in 3Dmark11, that, at the end of test 1 sometimes the CPU clock speed can go down to 1.5GHz. Maybe I am not giving it enough power? It doesn't happen in other tests, only in test 1.

 

No, you are not doing anything wrong. Giving it more power doesn't help, but it does seem to be power related... perhaps more heat than power. It's hard to tell and it's not consistent. I have seen it go below 800MHz now and then and it ruins the benchmark result when that happens. I don't know if the FIVR is messing up or what. Haswell does some really funky stuff.

 

If I just set the CPU to max clock using ThrottleStop, then the drop in clock speed doesn't happen. I agree with what you said. It's probably heat related. It throttles in some ways that I don't know. In hindsight, this might also be an oversight of Dell/Alienware. They have probably designed the chassis before Haswell was released, and then only they found out how hot Haswell is, and how the current heatsink can no longer dissipate the heat from the CPU. Redesigning the chassis isn't hard, but re-manufacturing everything is costly.

Finally I received the J77H4 11.7 cfm fan. So it fits fine in the chassis, but the catch is the cable is 2cm short to reach the connector on the mainboard.

I plugged it in (without screwing it on, just holding it) and set max RPM, and tested with both stock 5.5 cfm fan vs the 11.7 cfm fan. The stock fan actually has more motor noise. The amount of air that's moved seems to be about the same, but the 11.7 cfm fan seems to blow stronger. I will compare the blade pitch and CPU temps once I get the cable lengthened. Blade shape seems the same between the stock fan and the J77H4 fan, it makes me wonder if it's just the manufacturer quoting the CFM at different RPM. Technically, CFM or static pressure alone is useless. What they need to provide is ACFM (actual CFM) at certain static pressure value. Or certain static pressure value at certain RPM. It needs to be relative to makes sense.

I will have to go out to the local electronics shop and ask them to cut and solder another 2 inches of wire on it.

I disassembled the fan completely at first, trying to reach the motor, to see if there is any easy way to remove the fan from the shroud. Nope, no easy way to get it off. The C shape lock that's made of spring steel is a pain to deal with. One slip and it leaps 100m away.

 

View attachment 104836

View attachment 104837
The person did a pretty good job of soldering and covering the joints.

Hmm I don't really notice any difference in temps, at most a few Celsius. There's perhaps a bit more air coming out from the back, but I can't tell for sure unless it's measured. The max CPU fan RPM shows up as 4200 RPM. Mr. Fox, can you confirm that this fan spins even faster in the R2?

 

thx man, playing with all the settings now ;-)

 

The heat coming out of this monster CPU is incredibly. I could feel the air coming out of the vent, and it's scorching.

I have since clocked down to 4 GHz to maintain stability. I keep transient max power at 85W, and steady state (long duration) power at 70W. Any higher and it's not stable when I run wPrime 1024M. With 4GHz, it took about 183s to complete, which is slower than some of the Core i7 870. Then again, the 4930MX beats them in Cinebench 11.5, for some reason.

I think after changing to the new fan the idle temp drops by about 3 Celsius. But under full load I have one core that's almost 10C higher than the one other core with the lowest temp. The core with the lowest temp must be the one beside the GPU, with all those dark transistors acting as heatsink. The one with the highest temp is probably the core that's sandwiched.

Look at this image from Wccftech and you will get what I mean: http://cdn2.wccftech.com/wp-content/uploads/2013/05/Intel-Haswell-Core.jpg

Going from right to left, it's probably core 1 to core 4. My core 1 has third highest temp, core 2 has highest temp, core 3 has 2nd highest temp, and core 4 has the lowest temp. It makes sense all the heat from the last core could travel to the GPU that's off when dGPU is active.

If there's one thing I learnt about laying out transistors, is that you want to do things symmetrically, to even out variables. Maybe Intel did it this way to reduce the signal path for the L3 cache and the intra core interconnects. Sadly this configuration isn't the best at spreading out heat. Having a large GPU on die probably makes it harder to lay the cores out symmetrically in a square.

Either the CPU has a hotspot, or I need to remount my heatsink.

 

ha, i set c.c.l. to 112.000A, p.c.l. to 123.000A and had the multis at 40, cpu was hovering @~4Ghz, startet prime95, temp shot up to ~97 degrees, throtteling of 100% (shown in xtu) down to 800mhz. even though already said a hundred times, i'm "impressed" every single time about 1. how hot this asswell cpu gets (i know, fantables aren't really playing the game either), and 2. the funky happenings of it you already mentioned...

speaking of throttling, i wasn't using throttlestop because i just wanted to see how the temperature is behaving with those settings. i came to the conclusion, not at all ;-) will have to check with games later on, as that's what i'm looking for, a 24/7 setting for gaming until someday the may "enhance" the fantables some "more", not for benching or prime95 usage etc.

 

TBH I love this 4th gen. Its still pound for pound more powerful then other cpu's. Its the difficulty in using it, that has us concerned. But you know what, imagine if it was soo easy, that really, all we had to do was slide a slider to the left or right, and everything else was done for us.

We wouldn't want it then, unless there was people modding it and tinkering with it in unseen ways. But, essentially, here we are with this oddly incredible cpu that takes a genius to figure out how to clock it. Some will succeed, others will hate, and move on. Then theres those of us, on this thread who are pushing it beyond what was expected of it, and blinking our eyes and shaking our heads at the things we keep finding out.

This processor goes from silly to frustrating in a blink of an eye. I do not blame the CPU, or INTEL, but the fact we're using a system designed for at MAX power 330 watts.
The cpu should be taking 100 watts
The gpu's both should be taking 100 watts
that leaves 30 watts for... 4 harddrives, soundcard, 4 usb 3.0's DHMI, mini-D port, powering the 18.4'' monitor , Lighting the lights all around the laptop, speakers, subwoofer....NIC, RAM, FANS/cooling....

LOL, it just makes no sense.

4.3Ghz this cpu should use somewhere near 155watts at full load
2 780m.. well , they should use 110watts full load and overvolted, realistically, 85watts, is the highest we should see if we go more insane then anyone ever has, and max it as much as we can. most of us Im sure don't pass 80 watts/80% power on these.
Right there, theres not enough power.
already hit 300+

how this laptop is supposed tp run on a 5$ LiPo battery like we have, no idea. But, 87w/h is garbage. I run 40C type 1, 17.5v, 20A , RC car batteries, for my RCF hobby, which cost me 60-70$, can do about 150watts for a constant 3 hours, and are just as small as the one in this laptop. But they can also dish out 300watts for a straight 30 minutes. It would melt whatever is near it, but it can do it.

So, this is basically Dell's fault, for re-using an old laptop AC charger/adapter, and the current mix of its circuitry onboard and crap.

 

you pretty much said what i think. except for the rc hobby, i don't have a clue about rc ;-)
one more thing i blame is dell for using those ips pannels on those machines. i absolutely love the screen of the aw14 and also absolutely hate the aw18 for it's p.o.s. ips pannel with all the bleeding and glowing. it makes this thing look like some 200$ best buy laptop screen. then again, they did a louse job on the fan tables and even though there was supposed to be some kind of fix applied with the a03 bios, i can't seem to have found it yet. and the ac charger / adapter as well as the battery just go right into the fail-section as well... i wish dell never bought alienware... but concerning the cpu, i love it as well, i love the challanges of attempting something difficult and step by step achieving something by working on it. but then again, temperatures seem to hinder the progress just a tiiiiiny bit as well ;-)

 

We are probably not going to beat TBonesan's CPU physics score of 11k.

http://forum.notebookreview.com/ali...benchmark-thread-part-3-a-44.html#post9450430

NVIDIA GeForce GTX 780M video card benchmark result - Intel Core i7-3920XM Processor Extreme Edition,Alienware M18xR2

 

I like a challenge too, but I don't like wrestling with futility or wasting time with something that needs an excessive amount of hand-holding to be "good" rather than awesome. The issues are not limited to the Alienware 18... far from it. Haswell processors are not good for overclockers. Lots of overclockers with more experience than I have jumped on the Haswell bandwagon to only be disappointed with the outcome. It is difficult for some, and impossible for others, to get their Haswell CPU to perform as well as their Sandy/Ivy CPU that their shiny new Haswell replaced. It performs better at lower clock speeds, but what overclocker really cares about that? What matters is how hard you can push it and how far you can go with the overclock before it lets you down. There are a few silicon lottery winners with a Haswell X or K CPU, but not very many. Haswell comes out of the chute bucking hard, then wimps out after a few hard kicks.

 

yeah, well...true foxy

as for T-bones score, try not using SLI and dedicate the other GPU to physics and OC it.

 

I hope this isn't considered off topic as it is related - if someone can recommend another thread, please do.

I'm just curious if anyone has seen talk of a law suit over this issue? Dell continues to not only sell but claim "Dell Recommended" for the 4930mx version at a $500 premium over other options with claims of clock rates up to 4.3 GHz, clearly well beyond what these systems are capable of as shipped. Every attempt that I have made to contact Dell through normal and other channels since purchasing my system have either gone unanswered or passed on to some other contact ad infinitum. True, that is _the definition_ of customer service at Dell these days, but they seem to be going much further in this case.

Too me it seems that Dell and possibly Intel have really exposed themselves here so I'm surprised there aren't lawyers clambering all over this one. As a victim to this fiasco, it would be nice to see a resolution one way or another, especially given the fact that this system is now well into it's life cycle.

Cheers

 

It only has to hit 4.3ghz under a single condition and it would pass that sales pitch.

 

So any news or anything new that was found, that can make the CPU run cooler?

 

Maybe sometime in the near future I will pull the palm rest, plug in a USB keyboard, and see if Haswell runs OK with a block of dry ice sitting on top of the CPU heat sink. I expect to find that it still throttles under heavy load, but it will be interesting to see. Not sure when I will get around to doing that, but I will post my findings if and when I do.

 

You're a madman Mr Fox...an absolute madman