S-XPS 1645 AC Power Throttle Issue Investigation

I also have used Hulu at full screen and have no problem, it works just fine, i don't think your issue relates to the throttling issue when it comes to using Hulu in full screen.

 

Okay I wrote a program that tries to override the clock modulation by continually resetting the IA32_CLOCK_MODULATION MSR anytime the BIOS changes it. This prevents some of the more serious throttling, but multipler is still 7x. I'm not sure how to override the multipler setting, still investigating.

 

Guys,
Ive been tweeting the official blogger at Direct2Dell even sent him the link and asked him forward it to Tech Supoort.
Heres his reply

@symbinizam Yes... and BTW: We'll have new information on @Direct2Dell re: the Studio XPS 16 by Tuesday next week.

So hopefully there is a postive update by tuesday

Ive been constant contact with him..Hope Dell is actually taking a look at the probllems

 

Well my 1645 has been shipped:

Estimated Delivery Date: 12/17/2009 (updated)

So we'll see what happens When i get mine. hopefully something has been done.....highly doubt it

 

eblock12: MSR 0x199 contains the desired multiplier. In the desktop chips, the turbo multiplier is usually added on after this. If the Dell bios is locking that to 7 then your multiplier won't be able to go higher than 7. Throttling the multiplier is usually step 1 of turning your Dell laptop into a slug and then step 2 is when it applies Clock Modulation to really bring it to its knees. Intel lists a Clock Modulation setting of 12 as giving you about 12.5% of full performance. In my testing, that usually translates into a CPU that is working at about 30% of its potential. Combine that kick in the groin with a 7 multi instead of at least a 13 and you are down to less than 15% of the performance that Intel designed into these chips. Not good.

So much for Intel turbo boost on these speedy CPUs. When throttled back like this, my old P3 laptop would be very competitive with it. The RealTemp XS Bench benchmark does a calculation that scales 100% linearly with CPU MHz. Run that benchmark when your CPU isn't throttled and then run it again when your CPU is fully throttled and you'll get a good idea of its actual performance.

I don't think Dell is motivated to correct this. The same thing has been going on in the Latitude E6x00 series for over a year now with no bios fix. They use this method to control heat which is fine but the bios guys have done a horrible job of implementing this "feature".

Edit: By the way, the MSR Tool I posted only accesses the first 4 threads. If you have hyper threading enabled then you won't be able to use that tool to turn off Clock Modulation on all 4 cores / 8 threads. As you know, it's easy to write a quick program to monitor this MSR and to disable Clock Modulation immediately after the bios enables it. The same thing can be done with the multiplier.

 

I've been in the market for a relatively thin-and-light, but powerful, 15" replacement for my xps m1330. The Envy 15 and Studio xps 16 have really caught my attention...but so has this thread. And it seems that the Dell is out of the running. If it's any consolation to you guys, at least you're keeping those of us who research potential purchases ahead of time from throwing our money at a company, and a product, that just aren't doing what they should be. I'm in no hurry, so I'll be keeping an eye on this thread, but for now, it seems Dell will not get my business.

Now to research the Envy....

 

Hey uncleweb, just curious was the e6x00 throttlling issues a thermal one? or mostly anyway. as what i read it seemed like the cooling system was just poorly done etc and is throttling was over protective? It doesnt seem so much of a thermal issue for us as temps are fine on battery when running good. The laptop is effectively throttling to maintain a 90w draw from the adapter. Where the gpu seems to have priority over the cpu. They seem to have made a poor choice of an adapter, and the machine is crippled to match it.

I dunno just throwing that out there, as there probably power management schemes around as well. i wonder if theres other registers that are in the i7 that can point to power issues? i no nothing beyond 8 bit pic microprocessors lol.

 

I think we have our answer coming and people should stop speculating at this point. Stay tuned to Dell's upcoming blog update.

 

The throttling issue on the Latitude E6000 series is an insanely over protective one that doesn't return the CPU to full power until well after it should. That's not the way Intel designs their CPUs. You can run an Intel CPU reliably for hours at full load and at some very high temperatures without it ever throttling. It's the Dell bios guys that think operating your computer at a fraction of its potential to keep it cool is a good idea.

gaah: There's not a lot of speculating going on. tinkerdude showed the world exactly what the bios does on the E6000 Latitude series and it's obvious that these new Core i7 laptops have a very similar over protective scheme managing the thermals of these laptops. It might keep the RMA claims down to a minimum but it's not letting users use the full processing power that their CPU is capable of.

Can someone run my MSR Tool that I posted on the previous page and enter 0x199 in the MSR box and then click on the Read MSR button at idle and while running a single threaded app like a Super PI benchmark or similar. Post a couple of screen shots of my tool and tell me what CPU you're using. Edit: Include MSR 0x198 in your screen shot too. You can run two instances of MSR Tool.

That information will help me come up with a quick fix until Dell decides, if ever, to come up with a proper bios fix and an app that at least shows the throttling that's going on rather than leaving users guessing why their laptop is running so slow.

atlstang: First things first. We need to disable some of this over protective throttling before we can get back to the under powered adapter issues.

 

Okay i just wanted to make sure the difference between the two. As this doesnt look like a thermal issue at all. for its limited to 90w even with a 130w adapte. thats its throttling to keep power from the wall lower then 90w, as we can run external monitor, no battery etc and get a lot better performance on adapter.

heres the link to your msr request, on the 720qm i7. link to note that the 7 there switched back and forth from D to 7 when updating really fast.

 

D in hexadecimal is equivalent to 13 in decimal so that makes sense.

There are three main issues going on here. The Clock Modulation issue that eblock12 was able to detect, the multiplier being throttled back to 7.0 and the underpowered power supply these came with.

I can create a tool to disable some of the throttling that is going on but if the meager 90 watt power supply can't handle this then it might over heat and fail. If you have a 130 watt power supply then this tool might let you use some more of your CPU's capabilities. The tool for Core i7 isn't quite done yet. Probably in a handful of hours. Send me a PM if you're interested in testing it.

 

I would be a little worried about de-throttling the laptop myself without some sort of knowledge that the power distribution hardware on the laptop can handle it...it sounds like a great fix, but I would be really afraid of bricking my laptop.

 

I tried playing some games and overriding the clock modulation helps performance significantly and the system is stable (kept display brightness on maximum). Nothing exploded although given how hot the power supply already gets, (I think someone measured the case at 80 C) I'm not super comfortable raising the multipliers.

 

The first report is in for the new tool.

"IT WORKED!

my Kill-A-Watt shows 146W draw from the outlet, it would never pull over 90 before this. but did the computer get HOT FAST, got to 80c in about 1 min of both the i7 and video card maxed."

Looks like this new tool could cause some serious damage to a 90 watt power supply. Send me a PM and I'll send it your way for testing purposes.

 

I just ran this on my 130w watching it, this fixed the issue. My Kill-A-Watt shows 146W draw from the wall. Display on max brightness and i7burntest running 8 cores max as well as Furmark running. Going to do a benchmark and see how it works... will post with results shortly.


FYI - After 1 min of running this my internal temps hit 79-80c and then the computer throttled the video card to help cool it off. So it’s working correctly.

 

Laptops were never designed to run Furmark and i7 burntest at the same time.

For gaming and normal usage, ThrotlleStop should work wonderful.

 

Great now we can overheat this laptop just like it's designed to.

 

I couldn't figure out how laptop builders were able to use hot Core i7 CPUs. I guess the trick is to throttle them down to 10% of their potential. I don't remember reading that on Dell's website.

 

Does the tool work for the M15X? I have a 150W and 210/240W switchable supply available so I can test with both if the same principal applies for the M15X. The M15X has some pretty good cooling built in so it should be able to handle the CPU at max load for long periods.

 

A little off topic but I once did a test to see how much abuse a Core 2 CPU could take. I started Prime95 Small FFTs to create some heat and then disconnected the CPU fan to see what would happen.

How long do you think it took before it exploded?

Actually, it continued to run just fine for 3 hours without ever skipping a beat. Prime95 was still running without any errors. I finally got bored and stopped this insanity.

Did it get hot? It sure did. For three hours it bounced off the Intel thermal throttle and both cores ran mostly at 98C. Intel CPUs do a fantastic job of managing themselves. In a laptop you'll likely burn your lap long before you will ever burn a CPU.

I almost forgot. It was overclocked 20% and I jacked the core voltage up to create plenty of heat.

 

temp while playing games were fine for me. proc rarely ever hit 70C, and vid card barely hit 75C. a constant 100-110 watt draw from the wall. I dont think the furmark and intel burn test is a fair measure either, although it also did a 130w+ for me too. Everything runs amazingly, seems to load faster and be so much smoother just like on battery. Now we just need to get dell to fix this, now we know what we are missing.

 

Yeah I'm holding off blowing up the 90watt adapter with this until Dell's announcement on Tuesday. Seems like right fix is a new adapter and a BIOS update.

 

ok. problem - it works in windows when stress testing, but after playing a game for 1-2 min for whatever reason the system severily throttles everything, kill-a-wat shows 60-70w draw where it would normaly show 90w without this tool.

games run at 1-10fps and sound is really screwed up.

it looks like something is overwriting it even though i have it running in the background

 

Thanks for the feedback so far.

Siphen: I'll add some monitoring features to this tool tomorrow to see if we can figure out why your computer slowed down so much. The Dell bios might be fighting back.

Update: Part of the problem that Siphen is seeing is likely caused by the way these CPUs are designed to work.

The maximum multipliers for a Core-i7 820 mobile is 15, 15, 21 and 23 when 4, 3, 2 or 1 core is in the active state. Siphen was averaging a multiplier of about 21 while gaming which means two cores are typically in the active state. When 3 cores become active, the multiplier instantly drops to 15 which is a significant change while in the middle of something important. Any background activity on these CPUs can wake up additional cores and kill overall performance in the blink of an eye. Not really the ideal gaming CPU. At least Siphen wasn't down at 7X. I think we've made some progress today but extra heat might also lead to more GPU throttling.

The Core i7-720 has a maximum multiplier of 13, 13, 18 and 21 when 4, 3, 2 or 1 core is in the active state.

 

I will let everyone know how my testing goes once I receive my system on monday. Already have a 130 watt brick waiting for it

 

My Multliplier is 18 and can go up to 21 sometimes, is that normal ?

 

Answered

 

Interesting...by overwriting the max multi, could we possibly turn the 720 into an 820?

 

No, the multiplier is likely locked in the CPU itself so you won't be able to change a 720 into an 820.

 

Is every owner facing these problems? I am planning to order it on wednesday, now I am not sure of what to do? Can anyone help?

 

I wasnt earlier but now am seeing this problem.. I swear its related to key presses. as i can stand still and everything works amazingly.. but when i press a key to move etc, the frame rate drops to like 10fps. I plugged in an external key board and didnt have a problem. ideas?!?!

to vedaantees its a design flaw in the laptop, so if you plan on doing gaming etc and using the gpu/cpu to its abillities, i might wait for a fix as its a lot of money. Supposedly tuesdays someone mentioned dell was going to address the problem, but if its anything short of heres your adapter and bios update i might start researching other laptops honestly.

 

Also 720 has 6 MB L3 and 820/920 have 8 MB.

 

I think they're actually all the same CPU. The 920 is the highest binned in terms of clock frequency. The 820 is the "normal" model, and the 720 has less than or equal to 2MB of cache defective, but has a full 2MB of cache disabled. Anything less than that is discarded. It'd be great for yields, which explains the very cheap pricing.

 

Anybody tried artic silver 5 on GPU and CPU ? Does it make a big difference?

 

Yes, ~5° in GPU & CPU.

 

So, there is no more a 90w cap even with a 130w adapter ?
Could this be fixed with a BIOS update or is it still a misdesign ?

 

En tout cas c'est aussi ce que j'ai compris... Comment ils y sont parvenu ça je n'ai pas tout pigé mais... Donc màj bios + alim 130/150 résoudrait le problème à 100%...

 

Huhhhhhhhh??

 

If enough 90 watt adapters are out in the wild then I'm not sure if Dell is going to be able to release an updated bios. I think it would be too risky if the average user that doesn't read directions could easily install a bios designed to be used with a 130 watt power supply on to their laptop that only has a 90 watt power supply.

If you're lucky, maybe you'll be able to send your laptop back to Dell so they can install a special 130 watt bios and upgrade your power supply at the same time.

The Dell Latitude E6x00 series has had throttling problems for over a year now. I can't see Dell solving this new problem on Tuesday with a magic bios. Running an ATI 4000 series GPU and a Core i7 mobile chip at their full capabilities for an extended period of time is beyond what a typical laptop can deal with in terms of thermal management. Once they move to the much more energy efficient ATI 5000 series and once Intel shrinks the Core i7 mobile CPUs down to 32nm then you might get back to a more reasonable power consumption and heat level. Then they won't have to depend on severe performance throttling at full load.

 

I agree and I think that it would be strange because Dell would have to maintain 2 BIOS branch for the same laptop and as you said I don't think that the "130w" branch would be public. Pretty strange case.

 

-[Deleted]-

 

I think that you have here one of the best solutions! Dell should launch a new bios capable to detect which power supply you use (90W or 130W) and adjust the power automaticaly...

 

I've been following this thread since day one (I've had all the time in the world thanks to a 2 month build time on my XPS 1645). It should be arriving in the next couple days and I'm 99.9% sure I am returning it.

My question is, however, what implications are there on general performance when the cpu is throttled? I can appreciate I will see framerate issues when gaming but can I expect pretty comparable performance to a Macbook Pro (C2D) when doing general computing tasks?

I'm just not clear what the actual processing power limitations are due to this issue.

Thanks.

 

I haven't received mine yet as I'm still waiting for it to be built and delivered, but from what I understand here, these issues only arise under extreme load.

Normal to heavy daily use doesn't - as far as I understand - usually provide a high enough load to throttle the system to under the rated 1.6Ghz (for the 720).

Some games may result in extreme load, and that might result into issues.

I have ordered mine being fully aware of this thread. I have decided I like the potential of this machine enough to take the risk. For my daily usage it will be perfect. The occasional game might be a gamble...

For general computing tasks it will probably be a lot better than the C2D machine.

 

Much faster than C2D laptops in general use

 

FYI, for general computing tasks there has been no issues at all for me it works just fine, and I am not able to compare to the CD2 machine as I have never used it.

 

general its fine, but when using both the gpu and cpu at the same time it pushes the 90w power envelope. people with rgbled screens etc will see it quicker. playing games on it for a while so far and it needs 100-110 on my system alone.

 

Ok... I think we have a temp work around until dell fixes the bios to stop throttling the CPU to keep the system at 90W or lower when a 130W adaptor is connected.

Make sure ThrottleStop is running and Disabling Clock Modulation and Maximize Multiplier Make sure you have a 130W or higher Adaptor, make sure you have an external keyboard hooked into the laptop

With this configuration we need to run some tests, games specifically with ThrottleStop enabled and disabled to see the performance difference.

 

http://www.fileden.com/files/2008/3/3/1794507/RealTempBeta.zip



I added a feature to RealTemp so you can monitor for Clock Modulation. This is a "feature" in Intel CPUs that quite a few Dell laptops seem to be using. When the bios starts using Clock Modulation to control the thermal performance of your laptop, CPU performance can drop like a rock without your knowledge.

RealTemp will help you keep an eye on this now. As soon as Clock Modulation is detected, the Thermal Status area will change and report how severe it is. 87.5% is mild modulation and 12.5% is severe modulation and a severe reduction in performance.

If you see signs of Clock Modulation then you can do a quick (or not so quick) XS Bench benchmark and you'll be able to see how much performance you have lost. This benchmark scales very linearly with CPU performance so when your benchmark score drops in half, that's a sign that your overall CPU performance has dropped in half.

At least now when users feel that their laptop is running like a slug, they'll be able to document it and will be able to see if Clock Modulation is part of the problem or not.

If any of these tools are useful to you then don't be afraid to open up the About... box in RealTemp and click on the Donate button. That feature is badly under used.

 

I'd definitely give Throttlestop a go. Where do I download it from? (The XPS core i7 version - the link to the Maxcore2 posted earlier isn't compatible with this CPU)