S-XPS 1645 Throttling Info. and Updates

well this sucks, where is A07? waaaaaaaaaaaahhh..... *sniffle*

 

Most people "new" to computers fall quickly into the pit of benchmarks and synthetic tests. Soon people care more about 3DMark '06 scores versus FPS. Shameful.

I think the system is "throttling" when the either the GPU or the CPU underclock themselves under real usage. That's throttling. This is far bigger than what "XPS" stands for and whether "gaming" was mentioned in the details pane on Dell's website; this is an issue of knowing what's real and what's fake.

Benchmarks are a means to an end, not the end themselves.

Fine, you run a stressful program in your daily use. Run it and check if it throttles with A07+130W via ThrottleStop & GPU-Z logs. But that program better not be "3DMark06.exe" or "Furmark.exe". Or else your life has no meaning, lol.

~Ibrahim~

 

Couldn't agree with you more. If I can just play some intensive games without it throttling on me, I'm fine.

 

To me a synthetic benchmark is a benchmark designed to eliminate any possible limiting factor to test one component solely, and in a way that would other wise not be useful for real world work.

On the other hand, A benchmark is something to that tries to mimic how a real world application will run on given hardware. Unlike furmark.

I don't know what your definition is, or the real definition is. But to me 3dmark 2006 is not a synthetic benchmark, but rather. Not a very good real world benchmark.

I think if it throttles in 3dmark 2006. It will probably do so in real world apps.

 

Wait for it.. i have a feeling tommorrow morning will be like christmas.

 

It's still be4 Chinese new year
*eagerly awaits*

 

Interesting point. I'd like to see if it throttles in 3DMark 06 without anything else running and the screen brightness all the way up.

 

@SlyNine

Fine. But why do we have to have a middle-man, per se, here? Why can't we just test the programs we will be using in real-world usage? Why do we have to add another slant to the problem? Why, why, why?!

@all

If someone DARES to say "well, I don't own a game to test with so I have to use 3DMark '06", just...stop. It's called a demo. Find one. It's also called Google.

Why can't everyone be happy for what they paid for? A laptop that performs at the speeds it is supposed to on the programs you will use!

~Ibrahim~

 

Demo is fair. But I was doing some testing on some games. And found Dirt 2 and Empire earth to be very demanding games, CPU wise. Dirt 2 consumed 80% of a Q6600 @ 3.2ghz.

Empire earth under a 8 civ skirmish with CPU taxing settings on used up to 80% but wasn't near as consistent as Dirt 2. I think GTA 4 will beat them both. Getting ready to test that out.

I still think running prime+CPU intensive games is the best way to go to get a feel for what future games have in store. The GPU will turn FPS a fast as it can, but the CPU can only run multi threaded games today, if they exist. However games that push more threads, but are not necessary more graphically intensive may exist in the years to come.

DX11 ( keep in mind, the DX11 API will still run on the DX10 based 4670) is a lot about multi threading, the Xbox 360 can run 6 threads. I think its safe to assume games that are more threaded are coming.

 

I disagree, if a component can't perform at its full potential under full load, then it's not working satisfactorily. To me, it's like buying a dozen eggs where one is rotten, and being told not to care since I shouldn't be eating that many eggs anyway. Who exactly are you to say what is satisfactory to me?

 

But with that said, keep in mind Furmark is considered a power virus by AMD. It has burned out cards. So if you must run it, don't keep it on for to long.

 

It's wrong to dismiss synthetic benchmarks like that. Sure, it's also wrong when people obsess over them and care about nothing else, but there's good reason to use benchmarking tools like 3DMark, Furmark and such: they provide good control factors for actual testing.

Real world testing is great and all, but it's not easily possible to reproduce a set of results over and over again. Some games come with timedemos and loops you can run, but how is this really much different than using something like 3DMark? A tool like that can give you accurate results that you can reproduce over and over again if desired. Change the clock rate on your vid card? Run the same benchmark a few times to see how much this affected the performance.

Eye-balling games is just not the same, nor as meaningful.

It's also incredibly asinine to say that this laptop shouldn't be used to run those benchmarks; I'm not saying that you yourself said that, just a general notion I've seen on this thread a few times (and from Bill). I'm not expecting my laptop to be capable of running Prime95+Furmark for 30 hours in a stretch, but I DO expect it to be able to run Prime95+Furmark for a few minutes without throttling. If heat is a concern, throttle the system when it reaches a heat amount. Not long before this actually happens (which is what is happening right now). Or just trust Intel/ATI on their already-included throttling mechanisms for overheating. They work just fine.

 

No. It's like saying there is freckle on an eggshell when you're going to be making scrambled eggs. It.does.not.matter.

If you have a legitimate complaint where your real-usage program causes throttling, fine, rant, yell at Dell, file a suit, I'll be right there with you.

Then your freckled egg actually was rotten on the inside, to continue this analogy.

That's completely right. I have no special privilege to tell you what you think, none at all. I am just trying to explain why I think some people might be expecting too much.

@asherek

In-game benchmarks and time-demos are fine. That's real-world usage. A benchmark is different in that it attempts to recreate real-world usage. And NO ONE is going to be eye-balling games in an attempt of submitting real evidence. That's incredibly asinine, too. 3DMark '06: truthfully, I don't mind that much. A computer should be able to run that without issue as it was developed specifically to recreate real-world usage. It's all this P95+FurMark that gets my blood boiling.

And, why do people think it's throttling "long before" heat would be a concern? Without thermal paste, my SXPS 1645 easily broke 80C on P95 (alone!) after 10 minutes. That's pretty freaking hot. How high should it be allowed to go, then?

~Ibrahim~

 

+1.

Anyone 'testing' A07 under NDA (bill has hinted earlier that NDA would be applied to those people) and advising Dell that though it's still throttling, it's ok because it's only 'under certain circumstances' (but not related to heat), is raising a big middle finger against every end-user.

Bill could end all this by saying exactly what Dell are doing: latest comment from Bill implies throttling will not be eliminated because 'the system is not meant to run at 100%', effectively.

Those talking about fan noise, this won't be a problem for those who don't want to use 100% of the system. Those who do know it requires cooling and therefore don't complain. These people may even take it a step further and get a laptop cooler or get right in there and apply better cooling paste.

 

80C isn't all that hot for a Core I7, I know its not what you are used to when thinking about CPU's. I was always used to 60C being to hot. But now the Core I7's can stand 100+ C. If it doesn't reach that on Prime 95, It wont reach it under most, Prime pushes it about as hard as it can be with normal apps. Not that I consider prime to really be normal.

In all reality, a normal CPU app with prime will cause less intense calculations to run. In my benchmarks (I have over clocked and built many systems) this normally lowers the temps some.

However with the GPU and CPU being on the same heat spreader/heat pipes, Temps could go up a bit, it will be something I'll be watching.'

Do you mind if I ask you what monitoring software you used to check your temps?

 

100C has been the TJ-max of Intel's for quite some time, I think. My E6600 had a TJ-max of 100C, too.

100+C is not going to break a CPU AFAIK, but it sure can't be safe when that's Intel's own limit.

I have never seen an application that puts a CPU to 100% usage constantly in my life. Ever. And 100% on one application can cause a different temp than 100% on another app: LinX and P95 are perfect examples.

I used RealTemp.

I didn't know app+P95 would lower the temps...interesting. Could it be a priority issue?

~Ibrahim~

 

Why again does it not matter? If I pay for X performance, I do not want X-Y performance, I want X performance. If I wanted X-Y performance, I would have bought an X-Y component.

No matter if X-Y is sufficient for what I do *now*. Who's to say what I want to do in the future? Part of the reason for buying high end components is to have the tiniest bit of assurance that if for any reason my computational demands rise in the future, I won't have to go run out and buy entirely new parts.

As far as I'm concerned any piece of hardware should be able to full at full load, at full performance, indefinitely. If it can't then it should have been spec'd lower in the first place.

 

I think Bill is doing everything he can to help us, honestly. He's done this on his own time it seems. He cannot change what Dell engineers do. Nor can he control 100% what he is able to say, on a personal level.

I didn't like this FurMark thing from the start. But to be honest I'm not sure if there is a better sure fire way to test for throttling, I can say if it does not throttle under Prime95+ Furmark+Max Brightness right off the batt. The only reason it will throttle is if the heat starts mounting and hits thermal limits. In a laptop you must understand there is some give and take.

Also, if you run Prime+Furmark+max everything, I can promise you your system will not last you very long. Your car might redline at 6grand, but hold it there for an hour under load and see what happens. I bet your car makes it about 7-15 min before really bad things start to happen.

Please guys, lets just wait and see what happens.

 

You say this now, but it severely concerns me when I see Dell reps (Bill included), who inform us that it's acceptable if Crysis runs at medium settings at an acceptable framerate. What concerns me is that this honestly feels quite like eye-balling.

Maybe I am reading too much into it, but I like to see more concrete evidence than that, you know?

I don't see why it's not a legitimate concern. My Envy is able to run P95+Furmark without throttling. My Asus is able to run P95+Furmark without throttling (Yes, I own too many laptops, but that's a separate issue altogether ). I don't see why it's so unreasonable that I expect a newer laptop to be able to do the same things as my other ones. As I said, I'm not expecting my laptop to perform the same as my desktop computer or run P95+Furmark till it explodes into a giant ball of flame.

As high as Intel has deemed to be safe running on their CPU's, which is 100C. Yes, that's super hot, but frankly, it should be upto me to determine that, not Dell. The chip is rated to run upto that temperature safely, as far as I know. At this point, the chip itself throttles back. Perhaps Dell could at least give us more control over this. Keep it in the BIOS or something, so most users won't even go in there in the first place to muck around with it.

 

Depending on the process, certain threads can use more functions of the CPU, or take more time on wasted cycles or even cause branch misprediction. Hyper threading might change this dynamic some.

 

@zengei

From what trust I have in Bill, this (A07+130W) won't be an X-Y difference from a perfect non-throttling laptop. It'll be like an X-x. Bad example, but this update should get us very close, if not completely there, to fixing the throttling.

But, I will see I get what you mean now clearer than before. An i7-720QM should perform identically anywhere it is put or else that really screws up the whole system.

@asherek

I'm hoping that Bill will provide us with some concrete numbers comparing A06 with 90W to A07 with 130W before the engineers/he makes the call on whether the BIOS is "finished".

A BIOS thermal limitation would be handy. There are some programs, IIRC, that can do that software-wise, too.

~Ibrahim~

 

I was typing up something but you basically said what I wanted to. If its going to throttle it down when other laptops are not then I want that to be advertised in the technical specs with some degree of testing so that I know exactly what I'm buying. I'm buying with an expectation of performance whether I'm using that or not right now is irrelevant. I can't say that I wont need it 6 months from now.

 

The Core I7s run hotter then the Core 2 processors. However, if you think that's scary, I had a 8800GT that ran at 115C.

I have to agree with what you say to an extent, at 115C that card would crash, even though it was supposed to handle 120C before throttling.

However I had a Core I7 920 at 105C for a while with no performance problems at all. Don't ask. I don't really take pride in to punishing something like that. But that Core I7 works flawlessly to this day overclocked to 3.5ghz.

I love the Core I7 CPU's, they are beasts. That's why when I seen a Core I7 paired with the RGB screen ( alittle disappointed in that one) and a good videocard. I pulled the trigger.

If this throttling problem is solved, I then have to battle with the RGB backlight problem ( half moon backlite bleed and flickering, sigh.) But the 4670 is right where I thought it'd be at least.

 

Running P95+Furmark is like testing your home's wiring and circuit breaker panel by plugging a 1500w heater into every receptacle and expecting it not to trip any breakers.

Modern computers have a lot of functionality and massively parallel systems that offer great performance when applied to their intended uses. I don't think it is realistic to artificially saturate the CPU and video card and expect no throttling. I'm confident we will see 100% performance in every real world application.

 

I respectfully disagree. Your example would be valid if you replace P95+Furmark with "overclock the CPU/GPU far past rated specs". At the default clock speeds, this laptop can and should be able to handle P95+Furmark for a reasonable amount of time without throttling. It's artificial limitation that's quite unnecessary on Dell's part.

 

While its true, that furmark + prime95 isnt realistic. Though the new intel cpu's tend to max out its power envelope. Something to think about, we measured around 110-115w peaks at the wall with a kill-a-watt in real world use. With furmark + prime95 going, the max it ever got was 130watts through the wall, not really all that far away is it.

Theres plenty of real world results based of games etc that we have. This has been debated to death, although it is true a lot of other laptops can max out. I believe the laptop should be sufficiently powered for anything, as its a safety/just in case scenario.

the intel cpu's may be able to get up to 100C and what not, but i will have to disagree with webb if thats safe or not. As it also makes everything around it super hot, The heat pipes etc are going to be near that hot, blowing even hotter air on the screen (for whatever reason they decided to do that i do not know). Without ventilation inside the laptop itself itll just cook, some laptops have that ventilation though, the envy seems to as a few others i think.

 

I think that's where you and I completely differ. I take large pride in doing that.

 

I couldn't agree more. After all this, if this machine still throttles, with A07 and my 150watt, it will be going back to Dell and I would never trust Dell again with my money. After all this and all the time and effort everyone has put into begging Dell to fix this, any throttling would be unacceptable. There is no reason an i7 in an XPS should perform less than an i7 in any other laptop.

 

I'm not trying to unnecessarily stir the pot at this point, so my apologies, but I would say that it's Dell fault for creating an idiotic venting system then. I still don't know why on earth it vents the air over the screen. There is no way on earth I'd design a laptop to do that, regardless of the throttling issues or not.

 

While true, that isnt great design on dell's part if that is the limiting factor. I didnt like it to much at all either. I dont think the 1645 has heat issues what so ever though, its cooling system was quiet good when i had mine. but the 1640's may. at this point it is what it is :confused2:

 

Running Furmark+prime95 should be ok till either the CPU or the GPU hit the thermal threshold or near the threshold by which time it must throttle. This is exactly what happens in other powerful notebooks.

However, throttling based on power consumption is not supposed to occur.

So, throttling is acceptable as long as it is not conservative.

 

Just as long as the system was designed to disperse the rated TDP from the I7and GPU. I'm ok with that. I imagen FurMark pushes GPU's beyond the rated TDP.

But we just have to sit tight a little longer. I bet that BIOS will be out before we know it.

 

The game is up Dell.

Its very simple.

Dell is artificially limiting the power to their laptops which has the side effect of the machine throttling.


Its time to contact the media and for this story to come out.

We have had enough of your silly games.

 

OMGoodness quietcat, your posts always make me crack up and are a real treat. You're most definitely not quiet. Personally, I do agree with your bolded statement but let's at the very least give them until this BIOS. They're trying something at least. If that doesn't work then we can do what should have probably been done months back (publicize!).

Let's all just wait for the BIOS fix...

 

I know at least one site that wouldn't support our claims if we used FurMark as proof, Anandtech. We need something more substantial. But thats IF the BIOS doesn't fix it, I'm still banking on the Bios fixing it.

 

A07 is live.

 

ITS OUTTTTTTTTTTTTTT A07. Let the testing begins

Enhances CPU Power management...

 

I don't know if you guys noticed it or not, the A07 Bios is on the download page for the 1645, i don't own a 1645 but am planning to get one soon, so been following this thread for a while now, and have no issues in waiting for the throttling and the RGBLED issue to be solve.

I'm from India btw.

 

A07 is here, run the tests!! lets put this to rest

 

Finally. I'll test tomorrow. I'm too lazy to reboot and besides people far more excited than me will have probably posted results already.

 

Guys, I'm sorry to disappoint everyone, but it still throttles. A07+130-watt+Prime95+Furmark = 9-10x multiplier, however modulation isn't dropping below 100% meaning it is somewhat improved.

 

Whether 100C is safe or not isn't my decision. Intel designs and tests their CPUs and has decided that a peak core temperature of 100C is perfectly safe. That's their specification and if you are under that temperature then you are running your CPU within spec.

Keep in mind that Intel locates their thermal sensors at the hottest spots on the core so it can report a peak core temperature. Testing on a Desktop Core 2 CPU showed that when running a stress testing program like Prime95, the heat dissipates very rapidly over a very short distance. From the hottest spot on the core to the center of the core which is less than the width of your pinky, you will see a temperature drop of approximately 25C at full load. When you move to the outside edges of the cores you will see even greater temperature gradients.

When you purchase a laptop with an Intel i7-720QM you should have the expectation that it will be able to run reliably at the specifications that Intel has set for it. This CPU should be able to maintain its full default multiplier even when fully loaded and it should be able to do that right up to the thermal throttling point that Intel has set for it regardless of how high that temperature seems to you or I.

If a laptop manufacturer builds a laptop and for whatever reason it can't run at the performance level that Intel has set then they need to be upfront about the product that they are selling and warn consumers of its limitations. Dell does not do this.

Same thing with the HP Envy 15. There's nothing wrong with it throttling the multiplier down to 7.0 when running on battery power but since this results in the i7-720 running significantly below the Intel specification, they should warn consumers up front of any significant limitations like that. Same thing if your laptops are deliberately throttled so you can't fully load the CPU and GPU simultaneously.

 

The results are in- still thottles:

A07 + 130-watt adapter + Furmark + Prime95 =

 

gaah: Throttling during that test was expected. Does ThrottleStop help out any?

The important thing to test is during gaming and any other real world application you can think of. That screen shot shows your CPU operating at about 75% of its true potential when 4 cores are fully loaded.

 

A07 + 130-watt + Prime95 alone

 

Why is it expected when other laptops in the same class can run it without throttling at all?

 

Thats a bit dissapointing, although tons better then what it used to be. Seems the full use of the 130w isnt being utilized? what are the temps, is it super sensitive heat throttling perhaps?

webb, i was saying 100C is 212F way to hot for a home use item, as if its that hot, the heat pipes are near that hot and it will conduct through the body as there is no airflow to stop it. not arguing that if it does get that hot, there should be an adequate cooling system keeping it below 80C in the design on dell's part. Basiclly was saying to allow it to get to 100C constantly isnt realistic even if intel says it is, but a cooling system should be good enough to never let it get that high, IE throttling isnt a solution. I think this topic is out of the scope of this thread, unless the 1645 all the sudden has heat throttling. As far as i know it was never an issue

 

It's expected because this laptop was never designed with that load in mind. Even with a 130 watt adapter, bios A07 is not going to let you pull the full 130 watts. How many watts they have decided on is anyone's guess. Testing with a Kill-a-Watt will give us some idea of how many extra watts they have allowed and if there is still room for further improvement.

The design goal of A07 is not zero throttling which is going to be disappointing to many users. Dell is hoping for less throttling during normal use and being able to play a game or two without them becoming slide shows.

atlstang: I just wanted to clarify that a peak core temperature of 100C won't result in the heat pipes or your lap also being at 100C.

 

ThrottleStop still helps, and results are the same on battery as 130-watt. Actually, the multiplier is higher when on 130-watt AC. When I enabled ThrottleStop on battery, the system immediately shutdown without warning. On AC, it looked like this (when only the multiplier lock is enabled, extreme clock modulation reduction occurs):

 

After these tests, my lappy is super hot! I think whatever remaining throttling might be for the better, I have never felt a laptop get hotter than this before. The whole thing is burning up. lol

And I'm cooling with Arctic Silver 5..