AW M15x Throttling Issue Investigation - Stock clocks and overclocked.

Yeah, if the performance correlated to the severe dips, you would see your fps dive very drastically.

But Prime95/Furmark put a hurting on the 920/260M on a different level than games and most apps.

I mean, will you encounter "real world," scenarios that stress your system on the level of running Prime95/OCCT/Furmark in some type of simultaneous tandem out there? Probably not.

But there are some variations in clock speeds even running WoW unless GPU-Z is suspect which I'm wondering.

 

That's what I'm starting to wonder...its "display" doesn't update like it should, and, during the actual graphics test, GPU-z didn't log, the readings I'm seeing are post test completion. Strange.

I am curious though. According to the Inspiron forum, the ones also experiencing this issue, supposedly they found a fix and it is the power supply. Seems there system comes with a 90W brick and they are testing with a 130w brick and no longer have the throttling problem. I wonder, If my M1730 brick would work. It is 230w vice the 150w provided. I'm a little weary to try it though.

 

Well, luckily Joker has one on the way so we'll know fairly soon. If the PSU resolves the issue, you can bet I will be howling to Dell about it.

 

If you read that Inspiron thread a little more carefully, the 130w brick doesn't solve it completely. A few of the more power hungry systems there still experience throttling.

 

No, what you're describing is Powermizer doing it's job properly. You don't need max GPU acceleration to watch an CGI movie from the game.

Throttling is different. During a 3D render, the card should remain at max clocks at all times - period. Throttling is when you are playing Crysis and the clocks go down for some reason.

 

Got similar results on mine, I'm still using stock drivers. I tried exiting GPUz and relaunching it but on restart it had fluctuations from the word go.

Used P95 and Furmark with the settings as you described:

 

not sure if ya'll saw this but maybe this is the update you are looking for?

http://www.engadget.com/2009/12/03/dell-releases-bevy-of-firmware-updates-to-fix-most-of-what-ails/

It says that this should fix Alienwares throttling too. I am not a tech guy so don't flame me if i'm wrong. However, if the power brick being upgraded fixes it then all hell is going to Dell.

 

okay sorry! its not for M15x...they just state that the hear the problem now.

 

ya there was a similar post on the Dell blog about throttling. It relates to the business laptops, but it might be extended later on to the M15x, since this issue seems to be gaining momentum. Here's the link:

http://en.community.dell.com/blogs/...009/12/02/dell-on-laptops-and-throttling.aspx

 

By the way Joker, what is the best route to contact Dell/Alienware customer service?

My computer is still in production, and I don't want it to ship until this issue gets resolved. If it is indeed a BIOS problem, I'm probably canceling my order as this is a pretty major defect.

 

I e-mailed engadget about the m15x as well. I'm glad they mentioned it.

 

Well I guess this does prove that it's not just my system. An interesting tidbit for you guys: last night before going to bed, I downloaded RivaTuner and forced 3d clocks of 550/950/1375 and ran furmark + p95 + gpuz (for monitoring) at the same time. The clocks couldn't throttle because rivatuner wouldn't allow it and about 15 seconds into it, the furmark scene began flashing and the display driver crashed. This tells me that there is definitely a power issue involved because the GPU became unstable.

 

I usually call Dell financial services to cancel or modify orders. But I have good news, the adapter SHIPPED OUT! It should be here tomorrow and is already en route via fedex. So we won't have to wait until Monday for a definitive answer and you will still have time to cancel your order.

 

Hellz YEA! I expect your report on my desk by 5pm tomorrow.

 

It will be attached to my copy.

 

Place your bets everyone. $5 says the 210W PSU doesn't fix the problem.

 

SS, OTC, what PSU does your Clevo come with? Considering the similarity in specs, it could be useful information.

 

I use a Li Shin PSU rated for 120W output (20V, 6A). However, it has a nominal output of around 150-160W based on my KAW. This was tested back on my M570TU though; I have not yet tested the one on my W870CU. All of the performance Clevo's, aside from the D900x and the M980NU, use the same PSU.

 

I think there will be some improvement, at least at stock clocks. Whether it solves the problem or not, I doubt it as well. Not until Dell fesses up and tells us if it's a motherboard problem or something that requires a bios update.

 

Dell usually isn't the company to admit something wrong. Historically, they would just silently launch a BIOS revision or silently "fix" the problem in the next manufacturing batch or succeeding refresh, and that's already being optomistic. Dell has already recovered itself as a premier notebook company with the acquisition of Alienware and the launch of the M17x. I don't think they would want to bring up the power issue as it would break their streak. But if Dell does address the issue in public, they may just pull a textbook move and say that no user would tax the CPU and GPU like that, or something similar of that sort.

I am hopeful as well that the 210W PSU will help a little bit, but it won't help enough that it would open up that much leeway for overclocking as users wish.

 

At this point the issue is publicized at engadget and I'm sure other websites will follow suite. Dell has acknowledged an issue with the 1645 and their other notebooks so hopefully they do the same for the M15X and issue a fix for it as well. If it takes a motherboard revision, I hope they are willing to send out new ones to current owners via overnight like they usually do for part repairs--though that's being VERY optimistic.

 

Do you have the link to the article at Engadget? The computers here in the lab have more trojans in them than the streets of New York City; so slow right now.

 

I have not done any significant testing yet, but having played Modern Warfare 2, Crysis, and Aion for more than 8 hours straight without any performance decrease, clearly indicates I do not have a problem.

Our old friend, logic, now takes the stage...

If each and every owner of the AllPowerful M15x does not have this issue, then it is not an issue with the design of the notebook. Ergo, the select few experiencing issues should simply have their defective machines repaired. It does not take rocket science or a multi-page thread to figure out you just need to get AW to replace your system.

 

your base case is flawed. the throttling occurs when pushing cpu and gpu to the max, not during gameplay.

 

Just wondering how this will effect regular user, just playing games, photoshoping and multi-tasking? Sorry I'm an idiot and don't know much about the technical stuff.

 

Out of everything I've tested, I've noticed no throttling to point of affecting gameplay, watching Blu-Ray, or encoding video with Nero. At this point, I think it's an extreme test thats being done, and that's where the fault is recognized. My FPS are on par with Dragon Age, WoW, Far Cry2, DoW2 and my 3dMark06 scores are where they should be, just slightly behind my M1730 ( i know vantage is better, but can't get it to run on the M15x due to the resolution not supported crap)

I think end user will be fine with their M15X as it is currently. Those that Overclock the out of it and benchmark probably won't have the desired results due to the throttling.

Now, time to quit pissing around and enjoy some gaming. Cheers!

 

This doesn't prove anything. Like I said previously, WoW plays fine in heavy taxing areas with no large variances in frame rates (+/- 3fps which it does naturally on all my other systems too), while the GPU graph is all over the place with such drastic "reported" decreases in GPU clocks that I then said makes GPU-Z perhaps suspect. I am not invalidating the power issue, but I am, at the moment, invalidating GPU-Z as a totally reliable means of reading GPU clocks with the M15X.

Go ahead and load up Prime95 with GPU-Z running and see what happens to your sensor readings. Go ahead and run ANY of the tests outline in the initial posts and report your results.

If I just used a game or games as a indicator of system stability, it would pass with flying colors. Then again, many systems over the years ran games just fine, but when put to a real stress test showed cracks.

Which then leads to the argument, "If your daily computing will never push your system to the level of producing potential stress faults, what does it matter and why are you beating your head against the wall?"

 

Totally agree!

 

It's not even close to pushing it to the max since it occurs when the GPU is underclocked. All it takes is to push the CPU to load which plenty of games can and will do.

 

All it proves is that you haven't stress tested your system with the gameplay you've done. These tests specifically highlight a design flaw and/or bios issue with the system when all 4 CPU cores are loaded. During that event, the GPU throttles itself even when below stock speeds. The proof has been outlined in the first post. Oh by the way this post also shows another M15X owner has the exact problem when the system stress tests are performed so clearly it's not isolated to my system. Run the tests on your system at stock GPU clocks AND underclocked, I guarantee you that if you do it right, you will see throttling. For a botique top end notebook, this is unacceptable because an upcoming or even current game that pushes the CPU will cause the performance to degrade significantly.

 

Give it a shot with RivaTuner, it shows the same throttling unfortunately.

 

I'm in total agreement there is a potential power issue. I will give Riva a run tonight. But I am sure Riva reads the sensors the same way GPU-Z does too. We can't totally rule out a problem with the sensor, bios issue or both. Perhaps I am showing more drastic drops because of the 920XM with its +10w consumption level.

I wonder if we can get an owner of a M15x and a 240M (I know there were a few in the Owners thread) to run some of these tests. With its drastically decreased consumption level (23W vs. 75W), it would really add to the evidence of a major power issue if their system doesn't have reported GPU downclocks or it still downclocks and there is a bug/bios issue elsewhere.

 

One reason I tend to believe the sensor readings is because I forced the GPU to stay at stock clocks using RivaTuner and when I did that, the furmark screen started flickering (as if the GPU was choking without enough power) and then it crashed. So that strongly suggests the sensors are accurate and there is indeed not enough power to supply the GPU when the CPU is near 100% load on all 4 cores. To those wondering, "how does this apply to me on a daily basis if I surf the web, play typical fps/mmos and watch blu ray?" it doesn't. But then again, if you are that type of person, why buy an AW in the first place? A dell studio xps 16 would more than sufficiently do the job (but then again it is also plagued with throttling). As a premier gaming notebook, this thing should be able to easily handle 100% CPU stress without having to shunt power from the GPU by throttling it--that is a design flaw. The competing products from Sager and even the budget Asus G51 systems do not have this problem, why should we say it's okay for AW?

BTW electrosoft, I think your 920XM is probably causing instability for your system faster than it does mine because of the additional power consumption. Having someone with a 240M run the stress tests is a good idea. I also have the 210W power brick showing up tomorrow, I'm guessing it will help stabilize the system at stock clocks even when the CPU is pushed to 100% load.

 

I missed the furmark, forced speeds scenario. Then that definitely lends itself even moreso to the power issue.

Yeah, that's why I noted the +10w difference and the more drastic drops in clock speeds. Just as you said even the brightness levels affected performance, this system is walking a fine line under heavy load.

Definitely looking forward to seeing if the 210W PSU will remedy some or all of the problems.

If a M15X owner with a 240M would run even a few of the tests outlined in the initial post by Joker, it would be greatly appreciated.

On a side note, I watched the season premire of Scrubs this week (both episodes), and when, "The Todd," made his appearance I immediately thought of the notebook forums and you 5150. I tried to explain it to the ball N chain why I was laughing, but she didn't get it, heh.

 

I agress with you joker, with a system this price, this shouldn't even be an issue. I was looking at a different post by tangtang when he was playing gta4, he was having issues with it, could this be the cause of his problems.

http://forum.notebookreview.com/showthread.php?t=430598

 

Considering GTA4 pushes all the cores and the GPU (and is one of the few games to do it) coupled with the fact on the desktop GTA4 LOVES i7, this is a plausible scenario.

 

So I could be rite. I knew it, I'm a genius Now can I gave myself some rep? LOL

 

I can't believe some consumers would take defects for their $2k + and let Dell walk all over their faces.

 

Easy Panda!
Nobody is walking on anyone's faces. All I am saying is that this system was purchased for gaming, and it does perform exactly as it was intended to do. Hell, I'm not trying to get Toyota to give me a new truck because it has a 4.0L V6 and can't run the quarter mile in 12 secs. It's a truck, it drives over and through stuff, as it was designed to do.

 

Try GTA4 on your new system and I'm willing to bet my $1k Asus will walk all over your FPS.

The underlying problem undermines all high end usage of the system - including gaming.

 

Well, I'd qualify that G51 statement. The 51j doesn't throttle the GPU under the same circumstances, but it immediately throttles the i7 CPU from 1.7GHz to its idle default of 933MHz. Like the M15x, I haven't noticed this happening under today's normal stress apps but todays games aren't tomorrows games so...

Its a problem. Even on the G51 systems, albeit with reversed components.

 

I just received my AW back home and now i'm feeling a wee bit nervous about using it with this issue. Seeing that I am doing CAD, CS4, and 3d's max...they are intensive at times.

I really hope it is the power supply. I really hope it is not the motherboard. BIOS can be updated. Are these tests that people have produced from earlier versions of the new M15x? I just got mine so I am not sure if they maybe "sneaked" in a fix.

I will test it with 3Dmarks06 soon.

 

Your best test scenario, tedpud, is to run those very apps you hold as mission critical and observe their results under heavy load.

 

NBR trying to reach a broader audience? Just noticed this:
http://forum.notebookreview.com/showthread.php?p=5584424#post5584424

 

Does anyone know the return policy? Is it 30 days from purchase? or 30 days from shipment? I'd hope to have some clarification on this issue before that period is up. For over 3K this is unacceptable, IMHO of course.

 

seems like Sager 8690 is better than Alienware M15x now.

 

Try getting one of them repaired and let us know the turn around time. Dell, assuming the parts are in stock will have a tech on your doorstop next business day with parts inhand. Heck if you ask them to we will go to your workplace to do the repair work. Quality goes beyond just specs ya'know.

 

Lets keep brand bickering out of this.

 

download intel turbo boost technology monitor:
http://downloadcenter.intel.com/Detail_Desc.aspx?agr=Y&DwnldID=18353&lang=eng
I haven't seen my core i7 drop to 933mhz, it drop to 1.6ghz but went to again to 1.73 or higher within a second.

 

fuyuki: Its actually a Good thing to have the i7 throttle down to 933MHz during low CPU usage situations such as web browsing, office work, etc. I'm thankful for the feature heat and battery wise.

The problem is it doing it under extreme CPU+GPU loads, not in today's 'normal' situations. Follow the test on the first page of this thread or simply run prime95 stress test then add in furmark with at least 4xaa and 1600x1200 in windowed mode while monitoring with something like CoreTemp (or possibly even the Intel monitor - I'll try that in a bit). I'd almost guarantee you'll see the i7 drop down to 933MHz.

edit: actually just installed the Turbo Boost monitor and it doesn't show CPU speed at all if not in turbo boost modes. So, its not a very accurate tool to say the least. What I am surprised at is CoreTemp doesn't show the Turbo Boost modes at all, its maximum is 1733MHz when Intel's tool shows 2.8GHz etc depending on load. CPUID's TMonitor does better but is so damn fast to watch. hmmm.

Are there any other good monitoring tools anyone knows about?