FurMark GPU and Performance Degradation on AW Systems

FurMark GPU and Performance Degradation

Problem: Owners of Alienware systems may experience performance degradation issues while running the FurMark GPU benchmarking application.

Discussion: The Furmark application is a very intensive GPU benchmark that uses fur rendering algorithms to stress test the graphics card in your computer. The program is purposely designed to overheat the GPU by creating power draw above and beyond the design specifications. Using this program is likely to result in noticeable GPU performance degradation on your computer. If the GPU is pushed to the extreme, using this program could even result in thermal shutdown of your computer.

Solution: Alienware and our GPU partners do not recommend using the Furmark application as a GPU burn in tool. Although Alienware systems have thermal protection mechanisms built in, running programs such as Furmark could lead to potential system damage. Alienware owners should use extreme caution when using benchmarking tools such as Furmark.

Official statement from NVIDIA:
“FurMark is an application designed to maximize power utilization from the graphics card. Running FurMark on your GeForce-based PC can create power draw well above the design specifications of the graphics card and thermal solution. In some cases, this could lead to slowdown of the graphics card or even shutdown of the PC. While the slowdown and shutdown features are designed on the GeForce graphics cards as protection mechanisms for the hardware, running Stress Tests such as FurMark may still have the potential to damage the graphics card and should thus be used with great caution.”.

 

Thank you for the post. I really appreciate you coming over and posting this.

However, what about performance degradation when playing games or running 3dMark06. In both these situations my set up has "Performance Degradation". The issue gets even worse when I start plugging things into usb ports (e.g. wireless mouse). Then I plug in and external dvd drive and it gets horrible.

I am sorry to sound kind of harsh, but is the next solution going to be to unplug every usb device from the alienware when gaming? Am I supposed to play with the touchpad ? Does Dell design systems which have no head room for power spikes? I would have assumed that you should at least give the system sme 30-40W headroom for power spikes (e.g. harddrive starts revving up along with the optical drive). Additinally, one can at least plug in 3 usb devices, each taking up to 500mA, which results in 1.5A current draw if all of them draw the maximum of current.

 

This is a very important and informative post that I'm sure many are unaware of. I have been informed of this potential danger from a tech agent at Alienware and have since halted it's use in fear of damaging my system. Also, is it a possibility that after using third party apps such as Furmark and damage does in-fact occur, that Alienware may or may-not fix hardware issues as a result? Thanks for the info DELLChrisM!!!

 

I ask you how about you bought new computer just one month old then running game and face throttling..you still don't use furmark or do any intensive stress on it..just by playing games and benchmrking... because of furmark is the main evidence here..thats why the denied it..they try to sneak out from be responsible for throttling problem.. thumbs up for wattos on giving other solid evidence about power issuse when do benchmarking and games..

 

@ hackcremo

I am aware of the throttling issue witht he 920XM and 5870's but as DELL ChrisM mentioned that Furmark stresses the gpu's well above the design specifications. This is not saying that Dell/Alienware may be using Furmark as a scapegoat, however the fact remains that Furmark can be potentially dangerous for those that are new to stress testing hardware on a new system. I mainly game on my system, aside from doing school work (CS4, sketchbook pro, 3ds Max, and Maya) and the occasional benchmark here and there. I have not truly run into any real problems as of yet.

 

agree 110%
ey wattos, havn't you ever played SC2 with a trackpad? how about one that works at 20hz? i can unfortunately say yes to both questions >.>

hehe, imagine playing LoL, HoN or Dota with the trackpad.

truth be told Chris, the system itself is so close to the 150watt limit when coupled with the i7 and 260m, it should not have been in the design spec to begin with. undervolting the vbios was a bandaid fix at best. When everyday apps/games throttle due to being "well above design specifications" there is a problem

 

Folks, just a note -

I have mentioned this before in each of the benchmarking threads.

"While benchmarking / number chasing can be quite a bit of fun, please be aware that benchmarking your system is pushing your system to its limits. It can and will stress your components. Do so at your own risk - nobody is responsible for this but you."

This includes Furmark and all of the other stress apps.

What I see in ChrisM's post is a warning from NVIDIA. This is the reason why I chose to sticky this for the time being. I do not see anything in post#1 which states that Furmark is responsible for Throttling on the M15x.

With that, any comments on M15x throttling need to be posted here:
http://forum.notebookreview.com/ali...e-investigation-stock-clocks-overclocked.html

M15x throttling posts belong in that thread - this thread's topic is on Furmark and Dell/NV's warning that using it can lead to damage. Please do not post off topic.

Thanks.

 

Hi Batboy,

I actually have to disagree,

The topic clearly says

and the first line is

Unless performance degradation means permanent damage to a piece of hardware, then I cannot agree with you. In my humble opinion "performance degradation" actually means throttling. As far as I know the M17x does not have any throttling issue when running furmark, so I dont see why discussion about the M15x would be off topic.

If indeed the message was a warning, then I would assume that this does not only apply to alienware systems but to all Dell systems. Furthermore, it would most likely have the warning at the very beginning of the post, not at the end.

 

Good point Wattos and i see where you are coming from.

I would imagine this would apply to all systems - not just Dell, AW or a specific make. Furmark is a stress tool. It will stress and can damage (IMO of course) a GPU if used improperly.

As for the 15x specific throttling, ultimately I do not see the need for another thread to discuss this. The 15x throttling thread has a very high post count which helps put visibility on the issue. We'll play it by ear for now.

Thanks for adding in your thoughts...

 

^^ I'd totally agree with this statement if we were talking about overclocking. But weren't these systems designed with gaming in mind? As far as gaming goes, you could say the same about quite a few games. Some, like Crysis would stress the GPU close to the level of Furmark' and we play those for many hours! I have/had 12+ hour long sessions and can tell you that the GPU can get quite hot, maybe not as hot as Furmark pushes it but comparable, IMHO. In fact, every respectable reviewer does run Prime and Furmark for many hours to see how well do systems cope with high load. That's normal, since many gamers can and will load their machines in similar way.

Now, to the point. I see a serious problem with the current cooling approach on AW machines. While it works perfectly for GDDR3 GPU's, it fails with GDDR5. Just by replacing the stock pads on memory chips with a high quality thick thermal paste, Dell could solve the issue. The memory chips wouldn't get as hot anymore. That's how it's done in business grade Lenovo and HP systems.

Throttling is a different story of course and belongs in its own thread.

 

Are you m15x user?? Based on your signature it said that your are m17x user..m17x don't have any throttling problem because the power require for the laptop to run on full performance are fully supported by bios and power supply...

This throttling issue not occur only on 920XM and 5870's..it occur on cpu and gpu combination that's draw more power than 150 watt...for example 740qm and 260m, 920xm and ati 5850...

You said it potentially dangerous, how about you doing stress on stock clock...only on stock clock, no overclocking or modded whatsoever..then you use the furmark as stressing programs..then it failed..what are the point you cannot run 100% performance only at stock clock..?? We were running furmark on stock clock never reach so high temperature until give performance degradation...and with help from cooling fan i don't think it will cause any problem on stock clock..

 

So back on topic to make BatBoy happy

Well, to be honest, if I were to design a high end top class (and expensive) laptop, the first thing I would do with the prototype is to run prime + furmark for at least couple of hours and see how it copes with maximum load. This would actually benefit dell, as if the laptop passes that test, you can be sure it will cope with anything you throw at it and you wont see that many hardware failures due to overheating, etc...

Just remember, Alienware users arent your average Joe's. We want maximum performance, overclockabily etc. If a system fails at such a stress test, it can hardly be called a premium system, but rather a medium line system created for the average user

 

Very well written posts Wattos.

 

Agreed... good posts Wattos. I would like a little clarification myself. There is a huge jump between performance degredation and total hardware failure. And writing a blank warning about Furmark is a bit unfair.
There are multiple settings within the program. I have never tried out the Xtreme Burning mode. I have used the benchmark and the stability test. With that said I have monitored my temps like a hawk. Eye watering blank stare type of monitoring. <-- Kinda like that
Thus far the only system that had trouble with Furmark was the one that failed all other benchmarking software testing and failed to play games like Crysis without a meltdown. That system would reach 106C in furmark and 104C in Crysis. The other two alienwares did get warm but never, I repeat never came close to such stressing temps. (Nothing above 80's C)
It's good to warn people of the possible dangers especially if they have no experience and don't know that such tests can potentially harm a system. Yet I fear they (AW) will use this Benchmarking as an excuse not to honor the warranty. Maybe I'm just paranoid.

 

@hackcremo,

Obviously this thread is dedicated to Furmark users and the potential dangers from its use, regardless of system, system specs!!! Yes, I have a M17X R2, and actually the M17X R2 does have some throttling issues. Since you love NBR soo much take a look at the M17X R2 4870/5870 Throttling issues thread!!!! Furthermore, my system is stock, running stock clocks, with no mods or OC. I have used both Furmark and Furmark Multi-GPU apps to stress test my GPU's. Currently, the temps on my GPU's hit 100C after 15 minutes of running the Furmark multi-gpu app. So yes, Furmark can potentially damage your hardware!!!

 

My two cents about furmark: I also believe it does stress a video system more than most games will do so. However, as others have pointed out, there are some games out there that can approach that level of GPU use. Furmark's goal is to use fur rendering to push the GPU load to 100% and keep it there to see if the system being stressed is stable. It's not a performance benchmark like 3dmark.

Did anyone notice ATi's opinion on the subject was noticeably absent? That's because their video cards consume less power than their nVidia counterparts and it also happens to be nVidia's 260M that is one of the main culprits in the M15x throttling. Coincidence? While the 5850/5870 memory controller temperatures can get high, even with furmark on a properly built Alienware they do not exceed 90-95C which is well below thermal shutdown specifications. The most important point is that the 5850 used in the M15x consumes far less power than the 260M thus does not induce throttling/"performance degradation" so readily. Lastly, Asus (another major ODM) routinely uses Furmark to stress test RMA systems for their customers. Someone should send a memo to Asus that they don't know what they're doing.

 

well, I guess it is just coincidence that this was posted just before the release of A08 which is supposed to fix throttling on the M15x.

I just hope the next post wont look like that:

Problem: Performance degradation when playing demanding games like Just Cause 2, Mafia 2, etc...

Solution: Do not play games on the laptop.

 

Initially it will be something like this:

Problem: Performance degradation while playing WoW, SC2, Mafia 2 etc.
Solution: Disable turbo on the i7. Together with nVidia we have discovered that Intel's turbo mode can cause power spikes and degradation thus it is unsafe to have active.


The sad thing is they really are telling their customers to disable turbo mode in the bios so I guess my hypothetical is true. Your post is the next step they'll recommend after the Intel turbo disabling doesn't work out.

 

It sounds like we need another Brian for the M15x users.

"Dont run performance tests on your laptop, because your laptop wont handle it"

Not good enough.

Instead, get someone like Brian on the issue to diagnose the root cause, and work with engineers to fix.

Supply new power bricks if need be - just like with the studio xps 16 issue.

Considering Alienware is supposed to be the best of the best and all powerful, its kind of a joke and an insult to tell users they cant run certain apps because the laptop wont handle.

NOTE TO DELL: ALIENWARE IS NOT ANOTHER NAME FOR DELL. IT IS YOUR PREMIUM GAMING BRAND. TREAT IT AS SUCH.

-Ash

 

This is an absolute joke. The whole point in benchmarking is to see how your system copes at 100% load. If a system at stock speeds cant cope under these conditions, then its a design fault.

Every single manufacturer should be testing their products to ensure that it can run at 100%. How do you think that car manufacturers calculate top speeds? They 'benchmark' or stress test the engines, which is no different to what every single computer manufacturer does. Last time i checked, lambo werent telling people to drive slowly in their cars incase the engine overheats.... The car is designed to run at high speeds and high heat.

Yeah, sure, benchmarking 24/7 on your system isnt healthy due to the heat that can damage components, but you should be able easily bench for up to 6 hours without any issues. As i have already said, if a system cant run at 100% then the manufacturer is at fault due to a bad design.

 

I'm keeping an eye on the M15x stuff, but I don't have the bandwidth to work in depth. I have a new peer who is coming up to speed on the NB Forums. He should be out here more as we go forward...

 

I think you guys are reading WAY too much into this statement.... The statement is in reference to Furmark specifically, due to the way it overdrives the GPU(s) - think about redlining that Lambo and then holding it there - I would bet that the engine is going to have problems at some point... If I had one, I certainly wouldn't do that to it. The engineering team is readily working the throttling issues on M15x and this is not an excuse or justification for problems in games or other benchmarking applications. These issues will be worked and resolved by the engineering teams...

Thx.

 

Mind telling us what are your engineers REALLY do since this problem emerges form last year???

By releasing new bios until A07 version that's are useless and can't solve throttling issue..In the mean time, you aware of throttling but why still sell m15x with gtx 260m?? Because you can gain more profit compare to ati right?? Comparing price spec on ati 5850 and 260m..for sure 260m will be more profitable but when user found that it not perform well..you give an excuse..

 

It sounds like an excuse for throttling issues to me. I dont see many other manufacturers complaining about benchmarking on their systems?

Alienware and Dell have been working on a fix for the M15x for nearly a year now, and to be honest, i dont think that they want to admit that there is a fault with every single one of their M15x systems manufacturered in the past 12 months, it would cost them too much to correct. Had they of acted and resolved the issue when they originally heard about it, it would have cost them a lot less to fix.

I only used the lambo explanation because its simple to understand, and my dad has one. Trust me, high performance cars like lamborghini's are designed to run under stressful conditions without overheating. Its the same with any high permance product though. The point that i was trying to get across though, was that Alienware shouldnt be selling high performance laptops if they cant perform under stressful conditions for a few hours. The gaming industry is already at the stage now where games run modern GPU's at 100% load.

 

Alienware are running dangerously close to someone filing a lawsuit against them. Selling products with a known fault and then lying about it is illegal in most countries.

 

what a joke!

 

This here seems to indicate to me that Dell/Alienware are starting to wise up a little.

Before the last few months or so customer support for various known issues was just not done by ANY manufacturers.

Then Brian starts working on various M17x issues, and now teaching someone his ways to help him out - looks like Dell are beginning to see the value of their customers.

As I have said before, try not to rage at Brian, as he is your first and last line of support when it comes to real technical issues, and -while it should be- trawling the NBR forums and looking for problems to fix is not his job.

You catch more flies with honey than vinegar.


Yes, Dell support previously has been utterly a joke, but its starting to change now and we need to realise that and rage less.

-Ash

 

Brian is awesome, he does what he can for us. The dell executives and engineers that designed these throttling algorithms and other shortcomings into the M15x need to be demoted or fired. If I was in charge of AW's design group, I'd make sure the systems were designed in a forward looking fashion that could anticipate greater thermal tolerances than currently on the market as well as more than ample enough leftover power for future upgrades and/or overclocking. The M15x was crippled from the start, they always mentioned the 65W limit. Whoever made that decision should've been fired.

 

I am really grateful to what you have done in the past weeks. But you also have to consider the frustration of every M15x user. The throttling problem has been known since November 2009 (thanks to all the work Joker has done) and the first response to the problem was that this was not worth fixing since furmark is not a real world scenario. When we provided examples real world examples, Dell suddenly stopped responding in January.

I would really appreciate if you could actually have a quick look on thaqt issue if you get the time.

 

The reason why Nvidia had to put that money aside for warranty repairs in 2008 was because almost every 8400m/8600m manufacturered was faulty... Its not actually related to people benchmarking on their GPU's.

Source of information? Pick a website... Google Search


I dont think that anyone is personally blaming Chris for any issues, from the feedback that i have seen he is doing a great job, however people are concerned that alienware and dell will use benchmarking as a scapegoat for design faults.

 

I want to preface this with: Brian you are totally awesome and you go above and beyond your duties at work to the point you have restored my faith in Alienware support.
Now holding to your example with the Lambo. I agree that I wouldn't redline a brand new Lambo. BUT if the company says I can go a stable 210mph in their car then I'm going on the Autobahn and give it a go. Now I won't stay at that speed for an hour but if they advertise a stable speed of 210 that Lambo better be capable of it. It may redline when it tops out, but they would have advertised a top sustainable speed of 200 and redline at 210. Done with analogies.
If a laptop is advertised as stable running for hours of gaming then it best be able to handle it. No throttling, no overheating. etc. As was said most new games are giving the gpu's a high load already. And as joker said; this is a design issue and in some instances an easy fix implemented at manufacture. (ie: 5870 memory/gpu cooling)<--thx Ash
If I am advised against benchmarking a new system out of the box... I will be wary of making the purchase. Maybe it should be a more specific disclaimer advising the user to take caution and monitor temps if using benchmarking software. This kind of blanket statement makes me worry. It opens up all kinds of loopholes for excusing design faults.

 

furmark can do damage to your laptop because the quality checks are so low or perhaps not even there! after all they don't recommend it, so perhaps they don't even bother themselves!

mobile graphics platforms not meant to be tested, perhaps... but this is a gaming mobile graphics platform, different beast alltogether my friend. also, i beg to differ about the cooling statement, in fact the m15x has a GREAT cooling system in comparison to many other laptops i've had/seen. the issue here is the type of paste (pad) they use... yes it's actaully pre-applied to the heatsink as some sort of pad... pretty sad actually. and the somewhat apparent lack of testing prior to being shipped out. now that is really sad because the tools for testing are free... one is called "furmark".

also, i've ran furmark for a constant 1/2 hr with a 5870 at full clocks in the m15x and it never reached 80, thats with no cooler or mods. i'd say a little care in application and good quality thermal paste makes a difference, a HUGE DIFFERENCE. ask the people who did the gfx card mod in the m17x forum what a difference a little mod can do, heck they probably saved dell more returns and refunds than anyone!

in our case the performance degration is purely due to lack of power, not temps. i had the 920xm and 260 throttle with the temps not even hitting above 70C. thats both in-game and in furmark.

i'm not saying it's not good information, in fact i think it's important information. people just running furmark who have had no prior experience can be dangerous to be sure and perhaps many here have tried it without knowing when to stop it. however, the laptop in working condition should be able to handle such a test and pass, if it fails it means something was wrong to begin with.

its good information at the wrong time to the wrong crowd.

wow, people abusing their systems would see it first, so that means... its ok to have a problem in the hardware?? am i missing something here?

 

Thats because about 80-90% of laptop consumers dont play games on their machines, they use them for web surfing and youtube, so its going to be pretty obvious that the remaining 10-20% will notice the faullt there first because they are using their graphics card for its designed purpose, to process detailed graphics. You cannot assume that the 10-20% of buyers which escalate the fault first are all benchmarking. Very few people as a % benchmark, i bet that its not even 1% of the consumer notebook market.

Running any product at 100% will reduce its lifespan, but it should be able to run at 100% for a set standard of time before shutting down or throttling back, and im not talking a matter of seconds or minutes here.

Finally, not wanting to start a debate here, but Nvidia have really lost it in the past 3-4 years, ATI are starting to gain market share pretty quickly

 

Do you work for Dell or Alienware by any chance?

A laptop should never get to that stage at stock clocks unless there is a design fault with either the electronics, or the cooling. Effective and efficient cooling systems should be able to dissipate heat away from the components and out of the laptop. One of the things that Alienware are renowned for is their cooling design in notebooks over the likes of HP, Acer, FTS etc.

Also, manufacturers implement security features to stop hardware from failing due to overvolting, heat etc, so the component should throttle back or shut down as a last resort to avoid damage

 

These machines were designed for enthusiast to play games, not fry eggs-
My last Area51 m17X and its sli 9800GT's could fry eggs/ boil water and grill steaks. Couldn't figure out how to get it to brew coffee properly...
In all seriousness people who don't know what they are doing, don't care to monitor their temps, probably should not be benchmarking in the first place. I say probably because you have to start somewhere in the learning process and those new to it just need to take caution knowing that it is a serious endeavor.
I agree wholeheartedly with your entire statement about computer components not handling excessive heat for long periods. And it saddens me to think that many of the cooling issues can be solved through more careful consideration during manufacture. (gpu memory thermal pad thickness and quality/gpu die thermal paste quality/ seating of heatsink on die) <-- all of this info thx to Ash in his cooling mod. Then the throttling band aid fix for the fact that it's possibly cheaper to let it throttle than to change manufacturing procedure? I don't know and won't pretend to understand corporate culture or decisions. BUT knowing all this I did buy a 3yr complete warranty for a reason. If furmark is tough on a computer (due to intense nature)I hate to think what my longterm constant use of Autodesk software suite is going to do. Building, compiling and lots and lots of rendering hi res.

 

My two pennies worth,

Any properly designed gpu with the recommended cooling and running at stock speeds should never overheat or be capable of being damaged, (that's what the stock speeds are for!!). Saying that a computer program that runs a 3d image will damage a board does not say a lot for the manufacturer.

I agree that this is an attempt to deflect criticism, and hence ignore, the issues surrounding throttling - it may work over on the Dell forums, but we are a little smarter than that over here!!

 

yes there are many who have asked that question. Just as many have stated what an improvement using aftermarket paste has had. what am i to conclude then?
the only logical conclusion is that if you want something done right, you have to do it yourself, because quality control low... so low infact that now we're being warned that testing apps may cause hardware failure! i find this insulting to say the very least.


Right heat has to disappate, or rather energy has to be transfered from the chip to the heat sinks, then the heatsinks are cooled using air. the problem is poor contact between the chip and the heatsink resulting in poor transfer of energy. if replacing the paste made such a huge difference as many of us have found, what then is the problem?
cheap paste?
poor application?
all of the above?

oh and thermal pastes operating temps range much higher than any cpu/gpu can handle, your hardware will burn out prior to the paste crumbling (enthusiast pastes). last i checked AS5 was operational to 130C continuous!
180C for spikes. if you ran the gpu at 130 i'm sure it will fry... before the paste even began to crumble. oh and AS5 also does not dry out. it remains just as sticky and thick as they day you applied it... even after 5 years.

 

Nvidia/Dell can not have it both ways - on one hand the stock clock speeds are set to position the gpu at a certain performance level to complete with ATI. To then tell us not to run a program as it will over stress the components is insulting to our intelligence, (especially as the program is a key player in a long running issue).

They know the easiest and probably best way to stop the throttling issue is to down clock the gpu - however, this moves the GTX260 further down the performance/price ladder and places ATI's offerings at a much better point so much so that no one would buy another mobile graphic cards from Nvidia if given a choice.

 

The M15x and M17x R2 are not the only laptops with stutter/throttling issues. The M17x R1 has been shown by me and others to also suffer HORRIBLY from throttling that Dell actually made worse in the later BIOS versions (A02 allows MUCH higher power use than A03). That being said, a stock QX9300 and GTX 280m SLI can easily exceed this throttling power threshold and induce throttling of the CPU thus crippling performance horribly.

Heat is handled well enough by the stock system but the Power Supply is barely sufficient at stock though, and we never get to max it out as the CPU throttles long before that point is reached.

If dell was not so cheap with the Power Supply they could have allowed proper performance in the BIOS for the CPU under a high system load and then made MANY customers happier.

 

I can concur with Scook's comment! Great point, as I too experienced this with my late R1. The R1's mobo was a disaster, not to mention the dpc/stuttering/throttling issues that the system was plagued with. The only fix in my case "R2 Replacement".

 

great points, this is not the first time this has happened. i hope things will change for the next lineup of laptops they release! doing the same mistake the 3rd time round would be shooting themselves in the foot.

the sad thing is the bios limitation means larger psu's will not work without errors, but it seems that's the only solution to the problem... oh the irony!

when i go out and buy a new system, the psu is something i do NOT go cheap on! you can go value ram, cheaper motherboard without extras... but psu??? actually i don't think they went cheap, i just think they did not expect that amount of draw from the m15x. the fact that electosoft managed to have the m15x draw 195 watts without his psu giving out is amazing. mine crapped out a little over 160

 

Can the M15x use the 240W PSU of the M17x? That would be something worth testing.

 

I used the 240W but I dont know if I gained anything with that

 

The major gain is the lifespan of your PSU and removing the PSU as a potential issue throttle wise for benchmarking and stress testing (not to say the system won't throttle by design ).

Instead of cutting corners (and still charging a premium), Alienware and similar gaming laptop vendors needs to start using real stress testing tools and design their systems accordingly.