I Cannot Overclock My Alienware 18 to 4.3Ghz?

I just found out my CPU wasn't overclocked to 4.3Ghz. I called Alienware and they said it's impossible to overclock this CPU and has been a mistake on their site for MONTHS. Is this right? The CPU is impossible to overlock? Why would they leave a mistake on their site for months?

What the hell?

 

Hold on, bud. It is not a lie. At best, it's a simple mistake. I changed the thread title to correct the misinformation.

No... It is not impossible, but they are working on a BIOS update so it will be possible. I take it you have not noticed the previous posts where I have been assisting others with this issue. I have had the 4930MX overclocked to 4.8GHz, but not without some special settings. Hang tight and we should see a fix coming before long.

If you want to get more out of it while waiting, you can download and install XTU 4.1. Set the multipliers at 45 on all 4 cores. (You will need to set them at 45x4 in order to hold 4.1 to 4.3GHz due to an incorrect power setting.) You will also want to use the CPU Unpark Utility to make sure none of the cores are being parked by Windows 8.

Lower the Core Voltage from the default 1.2V to around 1.167V and change the Processor Current Limit to 120.000A. This will allow the CPU to hold between 4.1 and 4.3GHz under load for approximately 30 seconds, then it will begin to drop. Set the Turbo Boost Power Max and Turbo Boost Short Power Max to at least 100W (you can go to 150W, no problem) and the Turbo Boost Time Window to at least 128 seconds.

The reason for the misinformation is that the Core Current Limit is set at 32.000A and it needs to be at least 112.000A for 4.3GHz. The MSR Lock is enabled and needs to be disabled. XTU will not allow us to change the Core Current Limit to 112.000A. If you try, you will be prompted to reboot and it will revert to 32.000A after rebooting.

The above settings are not a permanent fix, but they will allow you to play with the machine and have some fun with short benchmarks like wPrime 32M while we are waiting for an updated BIOS. The clock speeds begin to drop after roughly 30 seconds due to a combination of heat and power depletion. Once we have the Core Current Limit set to 112.000A or high, the Processor Current Limit can likely be set much lower and the clock speeds will hold without the introduction of too much heat.

[ Download | Intel XTU 4.1 ]

[ Download | Unpark CPU App (Select File > Download from top menu) ]

 

I figured it can be overclocked, I'm not mad about that. I'm mad about the fact it's been like that on their site for months without them changing it and Alienware tech telling me "that CPU is impossible to overclock". And yeah I missed have missed the other threads, sorry.

 

No worries. Have fun experimenting while we are waiting for an update. From the perspective of the rep you spoke to, it might be impossible. Using the settings available in the BIOS, that would seem to be accurate. So, in all fairness I would not categorize it as being a lie... It's only my personal opinion, but that is a pretty harsh comment. However, I do understand your frustration. You dropped some big bucks on that bad boy.

 

What about with this new bios that came out yesterday? What did that even fix? I don't see any notes anywhere.

 

Did you read the release notes posted on the download page?

Alienware 18 BIOS A02 | Driver Detail - Dell US

It fixed a ton of stuff. Not the CPU power settings yet. But, I know Dell/Alienware is working on that.

Below is what is posted on the download page.

 

I missed the drop down menus, thanks.

 

It improve performance. My 3dmark11 on dell stock driver when up 500 point!


Sent from my iPhone using Tapatalk

 

Cool! Once we get the BIOS update with the ideal settings, it will be even better. Thanks for sharing.

 

Tangential:

Mr. Fox,

Have you overclocked the CPU in the 17"? I'm hoping that it has good CPU cooling this year.

 

Yes. Similar XTU settings (slightly less aggressive) will work, but I don't have an Extreme CPU or GTX 780M in the 17, so it is not as much fun to tinker with. The 4800MQ has similar behavior and the performance tweak lasts about 30 seconds. It is not overheating, just running out of power to hold the overclock. I love the 17, but I am an Extreme CPU/dual GPU overclocking nut. It is more suited for a great gaming experience rather than extreme performance, and it is fantastic for business use.

 

Anyone who's not happy about this should PM Luis or Pardo or the Alienware reps. I have explained to him and hopefully he'll forward it to the engineers soon. Once there are enough people complaining, hopefully it will be prioritized and fixed soon.

In the meantime, Saint (nice name isn't it), I think you can go play with overclocking the heck out of the GPU. It's basically a GTX 680, and can take quite a lot of voltage, and thus overclock fairly well. It all depends on how fast the heat can be extracted from inside the machine. I made like 4 OC levels, one for low requirement games, the level 2 and 3 are for higher requirement games (depending on whether it's CPU or GPU intensive), and the last level is to explore unknown alien territory.

 

?


If the processor is not overclocked up to 4.3 as stated and confirmed on my order form and paid for.....what is the processor speed going to be when my M18 arrives?

 

Stock 4930MX speed? ARK | Intel® Core

Perhaps Saint Satan could chime in. It's probably 3-3.9GHz.

 

My understanding was that the M18 chassis allowed a better cooling platform to allow the overclocking...please correct me if I have misunderstood.

If there is a limitation preventing overclocking in the M18 chassis, or more accurately a limitation preventing overclocking to 4.3, what advantage does the m18 have?

Sli is available in smaller chassis with the normal clocked 4930mx.

 

It simply comes down to a BIOS update that is required. Once that comes you'll hopefully be able fully exploit the XM chip beyond 4.3 Ghz like can with previous gen XM's. If it doesnt come from Dell its probably going to come from SLV7 (we shouldn't expect/demand this either) if we ask very nicely.
I'm guessing now the BIOS mulitpliers only allow you to do (around) 43,41,40,39 with a level 3 OC. Technically 4.3 Ghz but only on 1 core when the others are off.
Once unlocked you'll be about to do stupid crazy clocks like 50,48,46,45 and beyond and actually hit and hold those speeds.
I wouldn't worry about things for now.
If you really can't wait and want an unlocked XM doing 4.5Ghz ill swap you


SLI isn't available in a smaller chassis unless you want to grab a m17x r2 from a few years back.

 

I have visions of a sterile laboratory with racks and racks of spectrum analyzers, winchester drives, tape drives, blinking multi coloured lights, probes with wires snaking everywhere....huge wall mounted panels with blade switches, various and sundry warning placards shrieking "DANGER", "HIGH VOLTAGE", ......technicians in white lab coats behind glass shields....clipboards in hand...flashing red warning lights...waiting for the smoke to come out of the box....

For the techno-peasant like myself......may I be so forward as to ask...." could we just have a knob or slider.....that we just turn to 11 ?"

 

My M1730 XPS with the x9000 processor has the ability to be overclocked to 3.4 When I received the laptop, I followed the instructions, delved into the bios, moved the slider all the way to the right....did the restart....fans came on, processor speed was up to 3.4.......has been there ever since.

I guess that I was under the impression....that the bios adjustments on the new laptop would be the same....slider to right, fans come on, speed of 4.3,...... we're cooking with electricity.

 

False advertising needs to be fixed too. It's crazy that they'd leave it up for MONTHS knowing damn well it isn't overclocked. How long does it take to change the text on the site? 30 seconds? Absolutely disgusting for them to leave it like that.

This is correct. Max is 3.9Ghz.

 

So if you change to a level 3 OC in the bios the max multiplier is 3.9 ?!?!?!

Can't you key it in yourself ?

 

I've actually had the 4930MX overclocked to 4.8GHz. Alienware is working on a BIOS update and everything will be as it should be in due time. Core Current Limit is locked at 32.000A and it should be 112.000A. It should be adjustable in XTU, but it is not at the moment. 32.000A is way too low. So, yes... good things will come with some minor setting changes.

Here is a link to my 4930MX XTU score of 968 clocked at 4.5GHz. This is higher than my XTU score with 3920XM.

You guys with a 4930MX can download and apply the XTU Profile in the link above if you want to play around with it. If you are in a hot environment it might not work quite as well for you. I keep my house nice and cool... 70°F or cooler all the time.

 

Do you know what's the difference between processor current limit and core current limit?

 

Yes, they are totally different settings. Processor Current Limit is the CPU voltage for the IA cores. It is called Flex VID in the InsydeH20 BIOS. Core Current Limit is the Turbo Boost power, (expressed in amps in XTU and watts in the BIOS,) which is identified as Primary Plane (Pri Plane) in the BIOS. However, that setting is hidden unless you have an unlocked menu BIOS mod. Everything that needs to be adjustable for overclocking is available right now with XTU except for Core Current Limit. That register is locked and set at 32.000A maximum in error. If the BIOS update that fixes this unlocks that register, it won't make a lot of difference what the default value is at if we can change it with XTU.

The Core Current Limit being set too low is why all of the Haswell CPUs start to downclock under load. They don't receive enough power with 32.000A, so they begin to fall. Cranking up the voltage as a workaround is not good solution. It induces heat and the performance boost lasts about 30 seconds.

There is also another setting for Core Voltage for Haswell CPUs. It is different also. And, there is even a separate voltage adjustment for the uncore. Most of this stuff was not adjustment with mobile processors before Haswell, so it will take experimentation, trial and error to determine what is ideal for each overclock level. Even when we have a BIOS update, customers are going to need to explore different settings to find their groove. It's not going to magically run like a banshee automatically. What I have learned by trial and error as a temporary workaround will need to be re-learned with appropriate Core Current Limit adjustments.

 

I'm kind of curious now.

Where did the original spec of 4.3 ghz come from?

I guess that I am labouring under the impression that the chip/board/chassis combo had already been tested and validated to run at that speed prior to the M18 roll out.

I am also labouring under the impression that 4.3 ghz is a sustained speed.....

If I have misunderstood ....please correct me...

 

I'm guessing 4.3 Ghz came from Dell just applying a standard level 3 bios overclock. which would be something like 43,4x,3x,3x and the 4.3Ghz was only symbolic of the CPU hitting that speed on 1 core anyway, not the 4 cores (although its up to the task).

 

I would more expect a 43,41,39,39.

 

I believe it was something that someone arbitrarily decided. While it might be arbitrary, it makes good sense because 4.3GHz (on Ivy Bridge and Haswell Extreme CPUs) does not require elevating power settings and generally runs stable with BIOS defaults using all 4 cores.

In the grand scheme of things it doesn't matter with an Extreme CPU what the default multipliers are. The limitations are power and temps, not a symbolic BIOS preset. Otherwise, you set what you want to go.

I generally set all 4 multipliers the same. Personally, I don't find a lot of value in staged or descending core ratios. I don't use enough single-threaded apps to care what the speed is on one or two cores.

 

Now, I need to explain that at this point, Mr. Fox, you are educating me. If, as you indicate, all 4 multipliers are set the same to arrive at say, 4.3 ghz; does the cpu constantly run at that fixed speed, or only when the cpu decides to go into "turbo mode"?......I guess I'm not really certain what exactly causes the cpu to go into "turbo mode". Or is "turbo mode" determined in a varying manner proportional to a load that the cpu faces at any given time?

Thanks for your patience and understanding.

 

I think that's your answer there, in bold. Refer this:

Intel® Turbo Boost Technology—On-Demand Processor Performance

Intel® Turbo Boost Technology Improves Application Performance

 

My bad....I didn't do any research before I asked the question.....again, thanks for your patience....

I have some reading and research to do.....

 

Asking questions is good. That's part of how we learn.

Unless you change settings, the CPU will not run at Turbo Boost speeds all of the time. Mine does because I have set it to do so by overriding some things. Under normal conditions the core clocks will lower when demand is low and increase when demand increases. Extreme CPUs offer a serious performance advantage over the QM/MQ garden variety in that all 4 cores can run at maximum clock ratio (stock or overclocked) simultaneously.

 

Isn't the 4930MX is the highly binned version of the 4900MQ, and the 4800MQ is the higher binned version of 4700MQ? The extreme version CPU has unlocked multipliers to its advantage. The 4800MQ can do +400Mhz on 4 cores, making it 3.9GHz. The 4900MQ should be able to do +600Mhz, making it 4.4GHz.

Assuming that Dell releases a BIOS update fixing the issue, the PSU power limit will be the next limit, but we should definitely see some improvements. The question is, how much? Mr. Fox how much can you go on 3Dmark11 using single PSU before it starts tripping out?

 

We don't know the exact binning process intel uses, though power consumption at specific frequencies is likely targetted.

The 4900MQ can go up to 4.2ghz on all cores and 4.4ghz on a single core when using the partial unlock.

 

Not at all. An Extreme CPU is different architecture. Power limits are much higher and the non-Extreme processors have fixed TDP thresholds. They will never perform on par with an Extreme CPU simply because they cannot. In addition to power limits, core ratio limits are capped and the non-Extreme CPUs cannot run their max multiplier ratio on all cores simultaneously. Intel charges a lot more for the Extreme CPU because it is capable of a lot more. The non-Extreme CPUs are still excellent processors that serve the needs and wants of the majority very well.

Binning is a different process, theoretically, where non-Extreme CPUs are designated to a lower model number based on their performance capabilities. As Meaker pointed out, we don't know the finer details of Intel's process on binning today. It may not be what we assume, and it may be a different process than what was used in the pre-Core i series days. If Ivy Bridge set an example that Haswell followed, +400MHz OC capacity is the limit for the partially unlocked mainstream processors. a 38XXQM and 37XXQM both had a +400MHz limit on their maximum core ratio. It is not a reflection of their capacity, it is a limitation or threshold on what settings are allowed. Stability and TDP are probably a basis for the +400MHz OC limitation. Rather than binning the chips, they would be able cull and discard the rare example of processors that are incapable of achieving the modest expected +400MHz OC limit as defective chips.

Dell is working on a BIOS update to provide optimal CPU power settings. The PSU will not be a limitation for the majority of users. For a very small segment it will be. Those affected will be the OC enthusiasts with Extreme CPUs running much higher overclocks with much higher TDP limits than what the non-Extreme CPUs are capable of achieving. 3DMark11 is an exception even for the gaming enthusiast group. There are few games and benchmarks that are able of maxing out the power draw of the CPU and GPUs simultaneously. In most games and benchmarks the CPU and GPUs are trading off workload and the 330W AC adapter does not get pushed beyond what it can handle. Remember, the 330W capacity is not peak output. It can actually output much more than 330W for a short period without turning itself off. I suspect it will be unusual for a system with a non-Extreme CPU to have issues with "tripping" the AC adapter. If it happens, it will likely be an infrequent anomaly. I have not seen an example of that (yet). I have seen multiple examples of the Haswell MQ processors turning off the system when they were pushed beyond their power limits, but the AC adapter was not affected.

 

The extreme CPUs are not a different architecture, they are the same electrical layout as all the other CPUs, they are just picked under slightly different or stricter limits.

The power limits in the right system can now be raised on all the other CPUs, so a 4.2ghz 4900MQ will perform like a 4.2ghz 4930MX.

 

I agree with Meaker. Although technically, one could say that structurally, the 4900MQ/4930MX is different from the 4700MQ/4800MQ since the latter have 2MB less L3 cache (SDRAM). I am guessing that they all have 8MB when first laid out, but then are laser cut disabled.

 

Mr. Fox, I read with considerable interest, your "how to" for the dual psu setup. My question is, why the rectifier circuit?....Are you not essentially feeding the laptop with a parallel power circuit using 2 psu units? Like 2 batteries in parallel?

 

But it sure as heck wont perform like a 4.8 ghz one.

 

I doubt they will accept the same power limits because it seems as if their TDP is capped. You may be able to apply a setting with software, but I am not certain that it is actually utilized. You might be able to tell using ThrottleStop benchmark and watching the voltage and watts. Maybe architecture isn't the best word choice, but there is no question they are limited and will never keep up with an Extreme CPU. Intel is too smart to sell an inexpensive CPU that has the capacity to perform identical to their flagship. There is nothing wrong with them... they are just limited, and that is by design. There are no shortcuts. To say they will perform the same as the 4930MX if the 4930MX is slowed down to match the lesser CPU may possibly be accurate, but the comparison doesn't make good sense to me. It might be closer to accurate with "K" and "X" desktop CPUs than it is with the mobile processors. And, that still may not be completely accurate because lower efficiency of the QM CPU might result in lower performance, throttling or some other unexpected inefficiencies. One would need to do some scientific testing to determine if that is completely accurate. To me it isn't a matter of what the MQ CPU can do versus the MX... it's really what the user wants to achieve and what they are able or willing to spend to achieve that goal. If they are willing to settle for less, there is nothing wrong with that decision.

Exactly... we can call it whatever we want to and the end result is the same. It's not an Extreme and it lacks the capacity to perform to the same extent that an Extreme CPU does. It's a difference between good and great.

This discussion reminds me of the thread a while back with the 3720QM ES that was supposed to be a 3920XM in disguise. Was it better than 3720QM? Most definitely. Was it the functional equivalent to the XM? Nope.

It is an added safety feature. It helps to keep the AC adapters working in sync, smooths the output and keeps the current flowing in the right direction. You can try doing the mod without one, but there might be unfortunate consequences. For less than $5.00 for the cost of a bridge rectifier, it's not worth the risk. I am not an electrical expert, but I did take the recommendation of an electrical engineer and incorporated the bridge rectifier based on that advice.

 

Yeah, I just listen to svl7 on the rectifier and then waited for Brother Fox to be the beta tester. And after all was well...put mine together...hahaha

 

This may be of interest to some people: Process corners - Wikipedia, the free encyclopedia. Each process will have different process corners. A good design might work in some corners, a great design will work in all 4 corners.

To elaborate on what has been said, the line up of CPUs is a convenient result of process variation (across die, across wafer, etc etc - Process variation (semiconductor) - Wikipedia, the free encyclopedia). It's a statistical variation that's due to natural physics. And as the process node is scaled down, the variation becomes a bigger factor. A simple example would be, think +/- 1nm of gate oxide thickness of a 22nm transistor vs a 45nm transistor. Because of that the gate turn on voltage variation is a bigger issue on the smaller process node. And there are many more variations, each of which is typically plotted by some sort of Gaussian plot.

And because of that, there are variations even thought the chip design is fundamentally the same. I agree with what Mr. Fox says, but where I will diverge is, it is not by design that there's the higher end or lower end chips, it's by process variation. What Intel could do is characterize the variations, and come up with some sort of test program (think Schmoo plot) to differentiate the chips. Watch this:



So, process variation -> different class of chips -> different lineups/SKU. The extreme edition CPU is so expensive because the yield of it is so low. On a new less mature process, the yield could be in the 70% range or 80%. And based on my experience, I dare to say that less than 25% of that will make it into the highest bin. To think about it, because the 6MB L3 cache CPU will have to be laser cut before sealing with the IHS (integrated heat spreader), chances are, 4700/4800MQ are binned different. The 4930MX is binned from 4900MQ, most likely.

And indeed the extreme edition being cherry picked will outperform the 4900MQ clock by clock, probably by efficiency, and from that thoughput too since the 4930MX will probably has less leakage contributing to less heat, so it will hit the thermal ceiling later. But then again, because the CPU is being pushed beyond the Gaussian curve of efficiency, eventually the extreme edition will probably be less efficient than the 4900MQ. But then again all these data is hidden in Intel's fab.

I work with IBM process, and do mostly analog design, so I have less experience with VLSI. But when you have many repetitive blocks, it becomes sort of VLSI too. Historically it's only until recent years where Intel started to bin aggressively.

TLDR: 4930MX is like ice wine to the normal wine, which are the rest of the lineup.

 

Also, GentechPC sells the 4930MX for 1145 a piece. Anyone interested should get from Ken. He's a nice guy. I am waiting for a BIOS update before I act.

 

Seriously...What does this have to do with a non extreme vs a extreme? Nothing.
Since non of that has to do with why he can't do 4.3.

 

Until the BIOS update is released that corrects the Core Current Limit, none of the Haswell CPUs are capable of performing as well as they should. Dell is working on it, and yeah... this is all off-topic, LOL.

 

It's all for background knowledge. I am first an engineer, second an IC designers third an overclocker. Engineers are fastidious about stuff being right. Some of the stuff said are not technically correct, but I'll leave it there. Anymore said I could land myself and the whole department in trouble. Happened to Stanford once and that's embarrassing to the faculty and the people in the industry.

I really need to know how much more can I get from the 4930MX (assuming there's BIOS update and fixes) before I shell out the money to get one, running on a single PSU.

 

Good thing this is not engineering class.

Sure, some of that stuff is incorrect, but the bottom line is...
I's the bios that is messed up and it's messed up by one setting only at the moment which seems to be taking forever to get authorized and fixed.

 

Could you, based on your experience, give me an estimate of how much more performance (for eg 3dmark11) I could get from the 4930MX running on a single PSU?

 

It's not a 1 test situation. It's an over all situation.
So if we take 20 test, the 4930xm has the potential to come out on top in about 80 percent of them. That is the best way to put it. My 4930 will not run like yours because I would do things a bit different than you. I may be able to operate daily at 5.0 Ghz while you can't get past 4.5 Ghz. Lots of variables in there.

If you "can" afford it, buy it. You wont go wrong. If you "can't" afford it then let it go and get something lessor. EVerything I own has an extreme in it. Why? Because I want to be able to run at what speeds I feel I can get away with. (To a degree)...But running locked multies for me...not really an option.

 

Availability is a bit of a problem right now, which drives the price up. As soon as Dell release a BIOS update, I will feel more encouraged to get one. I am also somewhat worried about the idle noise of the fans. As of now, if I leave the 18 on flat table with the 4800MQ at 3.9GHz, the fan ramps up quite a bit and it's rather noisy for me.

On top of what, I will also be moving to somewhere which is hotter than where you are right now. I am worried that heat will be an issue, but as what Mr. Fox said, the better efficiency of the 4930MX at high 3GHz range vs the 4800/4900MQ, which might make the difference of a few celcius, which might be all that I need for it to be quiet during idle.

If you know someone selling the 4930MX, please let me know. And thanks for the advice.

 

Mmmm...I'm all for the safety aspect....I really, really don't want to let the smoke out of the box........I've done that to other electronic devices through the years....I end up keeping the dead devices for a source of itty-bitty screws....