S-XPS 1645 AC Power Throttle Issue Investigation

Thank you. I was looking for stuff like this. Thats great. The i7 studio 15 pulling 90w max. I wonder if thats the max they could measure, if the ps can only do 90w? With the upgraded graphics card in the 1645 i would assume more power draw like you mention. Which would elude to the problem we have in load conditions

Good work, great couple of posts.

 

Yeah, that happens if you disable EIST in the BIOS. BUT, if you disable it with Real Temp, you can get the 13x multiplier.

But, do you guys want to hear something that is preposterous? With EIST disabled, you can only get the 13x multiplier if the brightness is at 100%. Drop it down and you will go back to the 12x multiplier.

13x full brightness:


12x low brightness:


On battery, no matter the brightness, it remains on 12x. I found some PRETTY freaky stuff running Prime95 on the battery. If I went to the maximum brightness and went up another step (hit Fn+Up again), the screen would BLACK OUT! Dropping it a notch down brought back the screen. And sometimes switching the brightness on battery caused a green flash.

Scary stuff.

~Ibrahim~

Quick performance break-down @ 100% brightness:

EIST enabled, adapter = 5000 3DMark '06
EIST disabled, adapter = 6300 3DMark '06
EIST enabled, battery = 7100 3DMark '06
EIST disabled, battery = 7100 3DMark '06

 

From the pictures, your getting the 13x because it is not working at full load 54%, when it works on full load 100% the temperature increases and hence the switch to the nominal 12x or even less.
Also you are not stressing the GPU as well, you will be shocked to see the CPU working at 7x on charger vs 10X on battery.


I also replicated your Prime95 and RealTemp with EIST disabled.
1. It didn't lock the processor at any speed.
2. As load from prime95 reached 100% all the threads worked at 12X no matter charging or on battery, maximum or minimum brightness.
I have the WLED not the RGBLED.

Now consider the case that your power adapter is faulty and provide a power that is even worse than the underpowered nominal 90W.
This explains the 5000 under adapter vs the 7100 for working under battery.

No idea about the difference between the 6300 vs the 5000 when EIST is enabled or disabled. Can't really judge it until solving the problem of the reduced performance when the charger is plugged. It may turn out not to be a valid problem as the design of such algorithm is based on the assumption that the system has enough power.

Also, I don't have the mentioned problems with monitor. Try first to confirm the problem with another RGBLED owner, it may turn out to be a defective connector or so as it is obviously not related to the power problem.

 

if you disable speedstep, doesn't the cpu run in the lowest config? that would explain why it fixes it if your cpu never goes past 1.6 (or 1.73 in my case, 820qm)

i tried the 130W cord briefly, and there is no stuttering in CoD MW2, and im installing crysis right now, to test that. what sort of parameters would y'all like to see in my benches with 90 vs 130?

are you saying above, that if i disble speed step w/realtemp, it will lock my multiplier at the highest? in teh main realtemp app, the EIST box with teh check is grayed out. can you give me details on how you disabled it?

 

Please run crysis benchmark(or any other GPU intensive Benchmark) and prime95 at the same time so that you gt a 100% load for the GPU and the CPU.
Run it under the max performance powerplan once on adaptor and another time on the battery.
Max brightness in both cases.
This configuration sucks up the as much energy as possible.

Report the result in both cases. If the performance under adapter is better than or at least equal to battery then you solved the power problem.

 

alright, will do, I'll have those up ~midnight, it.s currently 10:42PST.

 

Can you please run 3dmark06 with full brightness with 130W AC adapter plugged in to compare with your previous result with the 90W AC adapter?

Thanks

 

I'm coming to accept the fact that it does seem like a power issue with the ac adapter, but that doesn't change the fact that this post made me laugh.

I'm done trying to explain it though. So heres to waiting for the results of the 130 watt adapter tests.

 

I would say for the Crysis bench, check out the config/settings that Anandtech uses. I think the bench loop they use is custom so we wont be able to use that, but they use what i would consider the "preferred" high settings: a blend of AA, AF, higher res, and high detailss

 

Never mind i got it to work (lol, im talking to myself)

 

Looks like I won't be home until about 3 AM so I'll do the tests then.

took me like 6 min to get this post up using the little tiny buttons on my BB

 

My call was passed to research team.
They will call me on Monday and let me know the results.

 

I'm looking forward to those results. The teaser of no slowdowns really has me curious to the outcome.

 

Same here. Be good to put this issue to rest if it is.

 

Same here. I really wonder what max's test results will uncover. If this solves the issue I'll be calling my Dell rep right away to make sure I receive an 130W together with my 1645 at the same time

 

Someone must have been tired after getting home at 3AM. The entire world hangs at the mercy of your fingertips, max420! Enlighten us and release us from this otherworldly suspense!!

 

Haha this is starting to look like a soap series

 

The Soap would be called something like

"Living with 90 Watts"

 

lol I should really stop refreshing this page all the time I'll check back later

 

Haha, that's what I'm doing.

 

I've just received my 1645 today and I'm just having a play around with it now. I also happen to have a power meter which I've just plugged the laptop into.

This only measures the total amount of watts used by the charger and system and the charger won't have 100% efficiency so I'd guess that for the charger to put out 90 watts it'll have to take in quite a bit more, at a guess around 120 watts. I hope that makes sense.

The spec of the machine is a

i7-720QM
wled screen
128gb ssd
no bluetooth
Win 7 64 bit Pro

With the machine sitting at the desktop doing nothing it pulls 37w with the screen on the lowest brightness setting and on the highest it's 43w.

Incidentally I performed this test with the battery at 100% and also with it removed and the figures were the same.

I'm just downloading 3dmark06 now and I'll run a test through and see what it does.

 

Incidentally I've just done the same test but with a with an external monitor attached too, so at idle on the desktop with a laptop set to max brightness.

laptop only - 43w
laptop and 1600 x 1200 external monitor - 46w
external monitor only - 33w
two external monitors (no laptop display) - 36w

As a side note if anyone wants to hook up one or two external monitors either the machine itself or Windows 7 handles it beautifully as all of my monitors are different sizes and it just chooses the correct size for each as the are connected/disconnected.

 

Dont even bother having THAT argument with dell tech dept.
They gave me every excuse from
'you dont have yours so how do you even know its affected'
reply 'because it doesnt take a genius to work out it's underpowered'

to 'this is an issue in the US it might not happen in the UK'
reply 'so the two major power consumers, the CPU and the GPU, might be different components to the US?'
'yes'
'but i'm paying for the same (and more)...grrr grrr grrr'

basically the long and short of it:
they wont accept any problem exists until you either recieve your system AND have the tag number AND investigate other causes, OR their tech support system, which apparantly takes 15 days to update :S:S, shows a known issue (specific to your country) AND an action plan has been put in place.

 

yeah your best waiting til ya get your laptop before asking to a replacement adapter. and looks like best strategy is to play somewhat stupid at first? pending results from max ill be ordering a 2nd adapter for backup/put at the gf's house, and hoping to get the 90w replaced by dell as well.

thats interesting stuff ollie. do tell anything else you find.

 

Sweetness. Well, that's 10W for the WLED. Wonder how much for the RBGLED?

Do run 3DMark '06 and/or the Prime95/Crysis combo.

How to disable EIST: If it's grayed out, I think that means it is disabled in the BIOS. Enable it there and then uncheck it in Real Temp.

Yeah, I'd love someone to explain the 1300 3DMark '06 score increase with it disabled.

~Ibrahim~

 

Right, I've not used 3d mark for years but I take it I'm right in thinking that all I need to do is a standard install of the free version and then click run test.

If that's the case then this is quite interesting as I've just got the following results.

AC adapter - No battery in laptop - Max brightness - Wifi off
3D mark – 6970
SM 2.0 – 2465
SM 3.0 – 3050
CPU score – 2975

AC adapter - Battery in laptop 100% charged - Max brightness - Wifi off
3D mark – 6931
SM 2.0 – 2462
SM 3.0 – 3023
CPU score – 2955

AC adapter - Battery in laptop 100% charged - Max brightness - Wifi on
3D mark – 6947
SM 2.0 – 2446
SM 3.0 – 3066
CPU score – 2918

I've not altered any power settings with the laptop, I've just used it as it is.

Also of interest is my power meter never went above 86w and generally stayed around the mid to high 70's and do remember this is the power the ac adapter is consuming not what it's putting out which is suppose to be 90w.

Also bear in mind that I'm based in the uk so our power is 240v so that may alter the efficiency of the adapter.

 

So what other tests do I need to do to really stress this machine out?

 

It could be it downlocks as safety because of power issues, when disabled it doesnt do it. even though physical limitation of the processor could run at higher multi, but it might not know that, it might even bring instability. Possibly like undervolting the processor, Its not supposed to run there and milage does vary but it still does run anyway.

Kind of my guess on it

 

Hah a new power related error.
There is a post above that mentions the battery "plugged in, not charging" and the "do not recognize the battery error". This came up for me when I was playing HL2:EP2 a few minutes ago, and the white rings on the monitor pivot mount started flashin orange, which according to Dell's documentation, means that the battery is "critcally low on power". Strangely enough, the battery said it was at ~60% power. I restarted, because plugging in the charger gave the "not charging" comment, and once restarting I got the "battery not recognized" dealie. A hard boot solved both these glitches.

 

Please do the test with the use of two software one to work at 100% GPU and another one to use 100% of the CPU using the adapter and maximum brightness. This combination will consume the maximum power.
The processor is so powerful to the extend that when you do 3dmark alone or crysis, the load will not reach 100%.

 

I don't know if this is related to what you guys are seeing, but PC PRO (UK magazine) tested the Studio 1557 and got slow downs that Dell told them were related to the throttling temperatures being set far too low.

See http://www.pcpro.co.uk/reviews/laptops/353653/dell-studio-1557/2

A beta BIOS update improved the problem but did not eliminate it.

 

Ok, that makes sense. So which bits of software do you recommend, sorry if it's been mentioned lots already but I've seen lots of things mentioned.

 

So with EIST disabled does the CPU still dynamic adjust its clock speeds between 1.6 and 2.8 or no?

I've yet to disable it because WoW runs great (no slowdowns ever, weird). L4D2 stutters, but if i run it on battery until the hinge lights start blinking yellow and THEN plug AC in it doesn't seem to stutter until the battery is charged (weird again).

 

This testing is all done with the 90watt adapter that dell sends with the laptop right? What is your adapter's model number?

The power draw suggests we might be on the wrong track and that it very well could be a software glitch.

 

He never went on battery power.

 

Yes I think it's just the standard adapter, it's very flat and puts out 19.5V 4.62A. The PN is CM889 and the model is FA90PE1-00

 

I'm just trying to do the battery power tests now but they're really low at around 3000 for the 3D Mark result.

I've been into the power options and selected high performance mode but I'm obviously not setting something else. Can someone point me in the direction of things to check please to make it work properly in battery mode?

 

lol, my bad, im up now though. i'm going to start the tests after this post. =D

ya, doing it at 3:30am didn't work out so well. anyways, in case anyone is ordering, the pa-13 is a huge . can anyoen confirm there is a slim 130W? dell says there isn't. im just waitin for my battery to hit 100%, then ill run the battery bench, and then the 130W bench. ill do th e90W bench last. so youre saying run prime95 to max the cpu, and then make note of my crysis benchmark?

 

ATI has a Power Play option that adjusts your GPU rate. might check that

 

This is correct, check the High performance power setting advanced properties. By default power play is set to maximize battery life.

 

That looks like the fella, with it set to maximum it looks to have greatly improved the fps. I'll post the test results when they're finished in a few minutes.

 

Don't take this the wrong way, but I really hope you have the same problem

Not because I'd be jealous if you don't, but because if you don't have an issue then we're back at square one.

Just to confirm your tests you've done/are doing:

1) Max brightness/Battery Installed/AC Power connected
2) Max brightness/Battery Installed/AC Power NOT connected

Those are where we're seeing the problems at. AC power connected = improper throttling. Disconnecting AC Power IMPROVES performance as it runs off the battery, and we're assuming the battery is providing a better power source then the AC adapter is.

Heres the template we're kinda using.
http://forum.notebookreview.com/showpost.php?p=5558878&postcount=3

 

lol, I understand exactly what you mean.

These are my results to date just using 3DMark.

AC adapter - No battery in laptop - Max brightness - Wifi off
3D mark – 6970
SM 2.0 – 2465
SM 3.0 – 3050
CPU score – 2975

AC adapter - Battery in laptop 100% charged - Max brightness - Wifi off
3D mark – 6931
SM 2.0 – 2462
SM 3.0 – 3023
CPU score – 2955

AC adapter - Battery in laptop 100% charged - Max brightness - Wifi on
3D mark – 6947
SM 2.0 – 2446
SM 3.0 – 3066
CPU score – 2918

No AC adapter - Using battery in laptop 96% start, 82% end - Max brightness - Wifi on - Ati Max performance
3D mark – 7032
SM 2.0 – 2519
SM 3.0 – 3090
CPU score – 2863

EDITED TO ADD

No AC adapter - Using battery in laptop 77% start, 64% end - Max brightness - Wifi on - Ati Max performance
3D mark – 7030
SM 2.0 – 2521
SM 3.0 – 3087
CPU score – 2862

Basically it is slightly better on the battery but not by much. I'm just trying another run with the ac adapter again so it is charging it too. If this shows the same I'll have to try and load up the power requirements using things like using an external monitor as well as the laptop screen and some usb devices.

I don't have anything like Crysis so if someone can recommend a way to max the cpu and gpu at the same time I'll gladly give it a go.

 

@khaled

Well, 3DMark '06 *has* to be powerful enough or else we would get the same numbers from battery/AC adapter.

@Ollie

Oh, so no battery and you got that score? Let me try.

Yeah, run FurMark and Prime95 at the same time: should hit some nice temps.

~Ibrahim~

EDIT: About the Turbo Boost and EIST. AFAIK, these are unrelated. Turning one off/on doesn't affect the other. So, yes, you can turn off EIST and still have Turbo Boost.

 

If you want to stress both CPU and GPU at 100% I suggest you to use OCCT, the last version, and run the "power supply test". I never seen more stressful software.

 

Do keep us updated on that. I'm in the same situation as you (and same region) and my 1645 is still in the 'work in progress' stage.

 

So what it boils down to is you don't honestly have the issue that we're having.

Dammit. I'm lost then. The only difference should be that you're AC Adapter is getting a different input power then north american adapters would, right?

So it doesn't entirely rule out the adapter theory, but it still makes my brain explode

Can you install Crysis or Far Cry 2 and run the benchmark loop on and off of AC power and look for stuttering? Man im confused

[EDIT] Never mind i forgot that you said you didn't have Crysis... does the demo have a benchmark loop?

 

Yes, my point is if we are testing the power consumption shortage claim, we need to suck the maximum power that is 100% GPU and 100% CPU under the maximum performance power plan. 3DMark tests each of them alone.
I found that Crysis benchmark alone stress the GPU into full 100% but the not the CPU, that is why i run both of them.
I personnaly got close numbers for 3D mark 06 ~6900 on charger vs ~7100 on battery (as i reported on the old thread), which triggered my doubt on the power consumption. So i decided to consume the maximum power by stressing both CPU and GPU under full stress, CPU stayed at 7x on charger while it worked at 12x under battery which can be interpreted as a lack of power provided to the processor.

 

I have got similar results before, the reason is that 3Dmark is not sucking up the maximum power.
I initially concluded that the result is insignificant, but after running 100%GPU and 100% CPU, i found that the processor is working at 7x under adaptor while it is working on 10-12x under battery. Temperature stayed in a reasonable one 75C.

 

So Khaledseif when you ran this set of tests:
http://forum.notebookreview.com/showpost.php?p=5558925&postcount=6

was it just Crysis looping the bench, or did you have other things running to max everything out? Which programs?