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

No, but it can be disabled in the BIOS... if you have a BIOS with unlocked menus. The driver still installs as part of the chipset, but it does not interfere with performance when disabled in the BIOS. I have it disabled on the M18xR2.



Care to elaborate more (maybe share some photos) about your TEC and fan?

 

@ Rafix

Rafix, I did the CPU stress test and it does not complete. The system shuts down after 30-40 seconds.

After that all values are still there, the only thing that has dropped in XTU is the "Processor Current Limit", which always goes down to 55.000 and the processor running around 3,8.

When I raise it back to 135.000 the processor goes up to 4,1. It completes the benchmark at 907 but not the stress test.

Rafix or Mr. Fox, any suggestions ? And is the behavior during the benchmark test ok ?

 

Is it overheating and shutting down, or simply shutting down? What are the temps when it does that?

 

I also believe that it might be induced by too higher temps, because it shuts down after 30/40 seconds and not right away as soon as he starts the stress test, as it happened to me.

@Omadon. You didn't confirm if you turned off the VR settings?

 

@ Mr. Fox

Temperatures go up to over 90, once they reach 99 or 100 and stay there for a couple of seconds it shuts down

The same was happening before during the benchmark test, but not any more.

@ Rafix

VR settings are both disabeld

 

Your system is behaving properly. The Haswell CPU triggers a thermal shutdown at 100°C. Without an unlocked BIOS with access to the thermal configuration menu, you can't change this behavior other than taking extra steps to help your machine run cooler. Using HWiNFO64 for running the fans full blast would be the ideal way to mitigate this problem except for the fact the CANNOT because of the messed up fan tables... doing this kills power to the GPU fans and causes them to overheat instead of (or in addition to) the CPU.

If you watch the temps, you will most likely see them go above 95°C before any evidence becomes apparent that the fans are working. They probably kick in a couple of seconds before shutdown, or not at all. They are too slow to react. The fans should be audible at 75°C and running at 100% no later than 85°C. They should react immediately to temperature changes. Unfortunately, they do not... there is a delayed reaction and it is a huge problem. Heat-induced throttling begins at 90°C unless you modify that behavior with an unlocked BIOS.

Try using external cooling techniques... raising the back of the laptop or setting it on a cooler pad might make a minor difference, but probably not enough to be meaningful. Until the fan tables get fixed you're going to have to run a lower overclock or find a cooler place to run the machine. I use a 10,000 BTU portable AC unit for benching, but that's not something most people are going to be willing or able to do.

 

If you don't force your GPU at 3D state all the time, and allow them to downclock, you could safely use the HWiNFO64 way to max out the CPU fan RPM. That will prime the system for a benchmark by making the CPU cool. Since the GPU fans stop working you will see the GPU temp go up a bit. After you finish the benchmark you could reset the fan back to Auto in HWiNFO64, and I think the last time I did that the GPU fan started to spin again.

I am not sure but I presume Mr. Fox locks his GPU clock at 850MHz. Am I about right or wrong?

Shutting down means it's a black screen crash, due to overheating. You will have to do what Mr. Fox suggests, lower overclock or, spending more time to fine tune the OC. I suspect my 4930MX won't complete the 1024M wPrime benchmark too, although I haven't run that test for a while now.

With Haswell running this hot, unless you have AC running under the laptop, I think the key is not about achieving highest overclock, rather it's about finding the sweetspot in terms of power, current, voltage and temperature, in order to minimize thermal throttling, so that the CPU can run at whatever maximum multiplier you set. In short, you want to lower stuff so that you won't overheat the CPU, but still make sure that it can run as fast as possible.

 

Um, that's not true for me. My 780M cards will hit 90°C doing nothing but wPrime, ThrottleStop and XTU benchmarks. If I run Goggle Chrome, they will hit 100°C and cause a thermal shutdown because the fans are turned off. I recommend that nobody use HWiNFO64 for manual fan controls on their Alieware 18 unless you plan to watch the GPU temps like a hawk. Video cards are more fragile than Intel processors, so be careful.

No, I don't lock it at 850MHz all the time. I sometimes lock in P0 with K-Boost when gaming and benching, but just having a browser window open causes the GPUs to enter the P0 (performance) state and stay there. Use the Open Hardware Monitor gadget and you can watch the temps climb and see the GPU clocks move to the P0 state when you open a browser window. There are many other applications that invoke 3D clock speeds, including something as simple as the Windows Snipping Tool. See below...

 

I will test that out when I get mine repaired. Last I did that, it rose up to 70C doing one pass of XTU benchmark, but with nothing else running. I changed the fan mode back to auto immediately in HWiNFO64 after that.

Also, in the Nvidia CP do you have the global preference for 'power management mode' set to 'prefer maximum performance'? I don't notice my GPU ramping up that much. I set it manually in the Nvid CP for each games.

 

I always set everything to maximum performance on a global basis. I typically do not use power management modes on any systems, Alienware or otherwise. Even my crappy Lenovo ThinkPad for work is set for max performance, LOL.

 

:d

 

That would explain why the GPUs ramp up to 850MHz, even with Chrome

And the high temps when the GPU fans are disabled through HWiNFO64. I only set maximum performance for games, and benchmark, or when I want it. Seems like setting it to maximum performance bypasses the 2D mode, so it skips the 2D clock speed.

 

Probably so... I conditioned myself to be that way over the years and like it better. On a system that doesn't have messed up fans it works fantastic running in performance mode by default.

 

I typically rest on balance because the impact to game performance is negligible and it tends to keep things quieter. Then again I tend to adjust my fan profile.

 

Oh in case you didn't get my point. What I was trying to say was that turning off the GPU fans (by changing the CPU fan to manual in HWiNFO64) to run benchmark or test, won't make the GPUs overheat if you don't have them set at max performance in Nvidia CP. I am suggesting that, in case you want to run a CPU benchmark on the 4930MX, and you don't want the GPUs to overheat, you could set the power setting to 'adaptive' in Nvid CP for the duration of the test.

I noticed the 2D clock skipping when I set stuff to max performance (and it's the same on desktop GPU) a long time ago.

Then again, having manual fan table is the best. Also, having a shortcut on the keyboard to set the fan RPM to max is also a very convenient idea. I believe it's Fn + 1 on some Clevos. Although Clevo don't have the best build quality, esp the chassis, the lack of throttling due to messed up BIOS, and the ability to control fan profile manually suddenly seem enticing.

 

Yes, I understood. It's really sad when something inferior looks better. All they need to do it fix it so things work perfectly just like the M18xR1/R2 and then everything will be peachy like it used to be.

 

So I'm messing with the HWiNFO64 Program's sensors to detect change in my graphics card's temperatures with fan speed. When i go to set the fan speed to my preference, the fans speed up and the first GPU rises in temperature heading for 100C. I bought this mini 4 inch fan yesterday for my room, and i had an idea for fitting it right behind my new Alienware 18 on the LEFT side. I went to speed the laptop fans up and again the GPU's temperature started rising. But then, i turned on the fan that i had on the left side of the laptop, and aimed it down toward that one grill. The next thing i notice, the gpu's temperature started dropping till it hit about 57C. If i'm not mistaken, that's where the CPU exhausts from while the 2 GPUs exhaust from the right side.

Problem solved...?

 

I wouldn't say solved, since the cause is a screwed up fan table, but patched for sure if you can keep GPU temps in check while gaming. Using this method what temps do you see under heavy load? let's say, when playing Crysis 3?

 

Hi Mr Fox,

How you keep the CPU at 4300MHz all time ? The Vcore is static ? I don't see Vcore monitoring on your screens.
Mine go down when i'm IDLE, if i put 100% on minimal state it stay around 2900MHz...

I can't let it stay at 3600 or 3800MHz, on turbo state, and the Vcore moves too witch cause BS in IDLE mode, to low.

Thanks

 

What you want to use is Throttlestop. It's what a compass is to a good boy scout.

 

Correction... the FAN (singular, not plural,) speeds up. Only the CPU fan does. If you attempt to manually control the fans with HWiNFO64 (or Almico SpeedFan) the GPU fans turn off. That's why the GPU temps start climbing.

 

Hi TBoneSan

With Throttlestop the max i can set is 2800, when i set the multi to 30 or more nothing change...

 

Maybe this will help. Change the video resolution to 1080p and go full screen, then you should be able to see better.

<iframe width='1138' height="640" src="//www.youtube.com/embed/xb2HPbk_O1I?rel=0" frameborder='0' allowfullscreen=""></iframe>

 

What i don't understand is why the GPU temperature decreases when im cooling the left side of the laptop. Do the GPU fans just re-activate?

 

The GPU fans will react to the die temperature of the GPU itself.

 

Watch this video.

Alienware 18 teardown - YouTube

 

Would it be possible to physically disconnect the fan(s) wiring harness from their present connectors and then using a new (to be invented/designed) modified interconnect harness ( I'm thinking of something that has the proper plugs/jacks/connectors)..plug them into a different source of power....say the source for the tron lighting? I'm not thinking of anything that would change speeds.....just a constant source of power for the fans to run flat out...that would start when the laptop boots up and shut down with the lappy. Does this make any sense?

 

It's interesting, but it's not a really useful tear down guide. The video fast forward many times and skips all the parts where he takes off the screws, and unplug ribbons and cables, which I think are critical to see. Thanks for sharing.

 

Yes, that make sense. What I have heard one person considering is connecting both of the GPU fans to the CPU fan header so they could all three be controlled by HWiNFO64 in unison. If a person could find terminals to make a 3-in-1 pigtail, it would not require any mods... simply plug them in and you're done.

 

I posted this in the main thread but I guess it's interesting to let people know about my experience. I will run some benchmark and see if I could still reach 4.1GHz without bricking it.

"I finally got my machine back up and running, after 2 mobo replacement. I spent the better part of today trying to update to 8.1. In Win 8.1, if you don't have Secure Boot enabled, there's a permanent watermark on the bottom right of desktop saying "SecureBoot isn't configured correctly".

Also I found from all the pain I went through, that, disabling fast boot is a better idea. I don't know what hanky panky stuff Win 8 does with fast boot, but it skips certain hardware checks, and if you have any problem it won't POST. Without fast boot, when you boot up, and reach the Alienware character in the middle of the screen, you will see the red bar at the bottom. With fast boot it's some spinning circles at the bottom.

And, it seems that this third mainboard could be a little bit different (ie, revised). Instead of running into no POST, it will POST and reset the BIOS to default settings after a few times of turning the power button on and off. Now I have not confirmation about this behavior but it appears to work this way.

It also seems that the certain chipset have been changed, but I can't confirmed this. Immediately after replacing the mainboard, unlike the 2nd replacement I had BSOD saying Memory_management, and on top of that the system wouldn't detect the GPU anymore. So I tried installing Nvidia latest 331.58 driver, but it didn't work either. Sometimes it detects one GPU, sometimes it doesn't. I thought that my 2nd GPU was dead, or that the SLI bridge is broken.

In the end, I updated to 8.1 and installed all the drivers. Several drivers have been updated, but Dell doesn't list all of them on the driver page. Some of the drivers doesn't install nicely in 8.1, so a bit more effort was needed than usual. For a long time I had one unknown device in Device Manager, until I installed the new Intel Management Engine Interface. DPTF installation didn't work for me in Win 8, but now it works in 8.1. Somehow after many reboots I got the SLI GPU back, and everything is working properly now. Even the Dolby Profile Selector is working (well I actually had to make a new task in Task Scheduler, but that's just the quirk of Win 8).

I am not optimistic about Dell releasing any further firmware to address the throttling on 4930MX. It might be an intended design. Running the CPU at 3GHz, it takes 35W and the CPU temp never goes higher than 80C. Seems like it's trying to keep it at 77C, about the same range as the stock GPU. At this temp the system is whisper quiet. Haswell CPU runs extremely hot.

If anyone is interested to get Win 8.1 from 8, you can get the ISO by reading the websites below.

How to download the Windows 8.1 ISO using your Windows 8 retail key | WinBeta
Here is how to get the Windows 8.1 ISO and create a USB install stick - Neowin

And when you install 8.1 using the DVD/USB drive, you will have to key in product key. Use the placeholder key here:
Windows 8.1 fresh install with Windows 8 licence - Super User

Win 8 key doesn't work for that. Sucks isn't it. Well it's just another day of Microsoft. You will be able to activate 8.1 once you finish installing it, with the Win 8 key. Installation takes a long long time, even on a Samsung 840 Pro."

 

Just a quick question, would there be any problem overclocking the 4930MX on the P157SM, since its not an 18 inch nor 17inch laptop? Regards to heating temps and all that..

 

The short answer is "no problem" because the Extreme CPU is built from the ground up with overclocking in mind. It is the main reason to own one and no mobile CPU can be overclocked anywhere close to as well as an Extreme processor. The difference is like night and day. If you're not going to overclock, buying an Extreme CPU is of little or no benefit.

How much of an overclock you can implement will depend on what your BIOS and XTU allow you to adjust and what your P157SM cooling system is able to handle effectively.

 

The P150SM is going to do ok up to around 4.1ghz most likely, more than that and you need the copper heatsink from the 17 inch models and need to think about modding a higher PSU.

 

Cross post from another thread, but I figure people might want to know about this.

I noticed going from Win 8 to 8.1, the XTU benchmark scores dropped by 200 points, from 800+ to 600+. For my 4930MX, the clock speed during the benchmark is the same in 8 and 8.1, they shift around a bit from 3.6GHz to 4Ghz (I have it at 4GHz right now) as the temps hit 100C.

wPrime 2.09 shows 6.8s in 32M test. So this is about where we expect the performance to be, but the drop in XTU is inexplicable. Any further drop in voltage will result in better time.

Weird age of computing has come, I think. As an engineer that will most likely be working in the semicon industry, I am all for efficiency and die shrink is one way of achieving it, but transistor leakage gets really bad for small transistors, and they have to fix it, if not overclocking room is just going to get smaller and smaller.

And to Mr. Fox, the dynamic CPU offset is a offset for the voltage of the CPU. Imagine if you have a voltage vs clock speed curve, the offset just pulls the whole curve up and down, so the offset will affect the idle voltage too.

 

For those who use XTU to overclock the CPU, by default the XTU service is set to start automatically after bootup, but it's a delayed start. There's a certain delay before the service starts, does anyone know what's the delay for it, and how do I edit it in the registry?

I tried Googling but couldn't find much about it. If I set the startup type to "automatic" instead of "automatic (delayed start)", it makes the startup time too long for the other apps and services at Window startup. The default delayed start delays it too long, possibly 2 mins but I didn't measure it.

 

Intel may have done that to prevent any conflicts as windows starts.

 

Intel Rapid Storage Technology is also delayed startup. I have always changed it to normal startup in the services properties without any problems.

I think Meaker is right on XTU. I would not change it from delayed startup. If you have ThrottleStop or HWiNFO64 launching at startup, this can cause conflicts if XTU tries to launch simultaneously with those utilities because all three have hooks in the BIOS.

 

Another quirk I found out about XTU and the BIOS CPU settings.

The rest of the settings sticks (core current limit, core voltage, etc) but sometimes after going into BIOS and saving it, and sometimes for unknown reasons, the processor current limit always get reset back to 55A.

 

Yes, I am not sure why that does not stick. Sometimes it does and sometimes it does not. What is not entirely clear is the relationship between Core Current Limit (Pri Plane) and Processor Current Limit. They both need to be set with XTU, but they seem to be somehow related and they affect each other. The Processor Current Limit is something new with Haswell. It does not exist with Sandy Bridge or Ivy Bridge.

 

Still to this very day I have trouble getting BLCK to stick using XTU. Its so annoying. I always think I've finally made it stick it and then it changes back.

 

Have you tried without HWiNFO64 and ThrottleStop having already been launched at boot? Once they have launched one time, BCLK adjustments with XTU may not stick. I have both running at startup. I temporarily disable them, reboot and then set BCLK with XTU. I also add the check mark to "Force Boot" in XTU, reboot and it usually sticks. I then re-enable HWiNFO64 and ThrottleStop to run at Windows Startup.

If a hard crash event takes place, this will sometime cause BCLK and other XTU settings to be lost, but not very often.

 

I have tried after I disabled them from Task Scheduler boot, then enabled them after I made the changes and rebooted in XTU...

 

Well, I'm not sure then. Something about your system behaves a little different. The new 18 also has trouble holding XTU settings... especially Processor Current Limit.

 

Overclocking the 4930MX on the 18 is like dealing with a hooligan. I spent another night trying to get stable settings. At 4GHz, the CPU eats twice the power compared to when it's at 3GHz.

I have repasted time and time again, maybe the die contact area is just too small for the CPU to transfer the heat to the heatpipes efficiently. The CPU uses less than 100W at max load, which is much less than the GPU, so it shouldn't overheat. Maybe like someone has suggested (was it Mr. Meaker?), we need direct contact heatpipes.

The processor current limit adjustment behaves weirdly in XTU. The moment I adjusted it from its original 55A value (even to 56A), the system would crash in Cinebench 11.5 (probably due to overheating). Even at 3.2GHz, the CPU reaches 70C with just gaming load.

Also, why does adjusting other values stick in XTU (core current limit, cache voltage, core voltage, offsets, etc etc), but the value for the processor current limit do not stick. This is some black magic going on under the hood.

 

Crazy thing was it didn't always behave like that. Previously I could actually uninstall XTU and the BLCK remain set.

 

I set my settings exactly like the ones in this thread adn when i run the stress test i immedately get BSOD. I dont really know about overclocking this thing so i guess im screwed. just wondering why mine immedately gives me the BSOD

 

I had the same problem. Load BIOS default. Set your BIOS as you see it on Brother Fox's screen, but lower your CPU OC to 41/41/41/41. Set BLCK at 100.24. Reboot. Set XTU as you see in the screenshot, but lower the CPU clock value all to 41. Make sure you click on Other and set Processor Integrated VR Faults and Processor Intergrated VR Efficiency Mode to disable. Try if you can finish XTU stress test this way.

 

whats worked for me, since day 1, without issues... go into bios, select oc profile 3, then change the values to all 41. keep all other settings, and enjoy. Always at 4.1Ghz

 

Run a Cinebench 11.5 CPU test and see if it downclocks. I think it will downclock to 3.2GHz or lower.

 

I have some new findings about resetting the BIOS. The key is the CMOS battery and not the CPU. It's here:

http://forum.notebookreview.com/ali...-into-no-post-messed-up-bios.html#post9422971

I added a few washers to the CPU backplate to increase the pressure on the CPU. Although it was annoying as I ran into another no POST, but in the end it turns out that what I did was worth it, the temperature went from 100C to 70C in Cinebench 11.5.

I suspect it's because a few months ago I added a thin Fujipoly thermal pad to the mosfet above the CPU. Since the heatsink is only pushed down by a few springs on the screw, the pressure isn't that strong anyway. That thermal pad must have pushed the heatsink further from the CPU die, resulting in high temperature.