M17x-R2 4870/5870 Throttling?

I'm not sure I follow, what would cooling off the PSU achieve? I was told by an AW engineer that they have a bios trip point in place, it's why the system throttled in the first place. They set this limit within a certain margin and its entirely artificial but I don't think it's above 240W. The bios probably allows for temporary spikes above it but not sustained output. The R2 at stock doesn't leave behind much power, I know this because I've measured it with Kill-A-Watt. So IMO, a 300W supply + bios that allowed for it would certainly do the trick. I agree that the 240W is probably sufficient for the OC'ing that we do (as evidenced by the throttling workaround) but for Dell's own safety purposes it might not be.

This is all based on a lot of assumptions though. What I do know is that I recorded throttling on both systems in Vantage and by using the throttling workaround, it stopped on the 4870s. BUT I haven't observed any throttling in Left 4 Dead 2 yet so it may just be Vantage that causes it, all of this is still wide open and left for testing. So if anything, I'd encourage people to start testing games and see if they notice any throttling. At stock clocks or even mild OC's, there should be none at all and the throttling fix for either system (4870 or 5870) should not be necessary. Remember guys, I just put it out there for extreme OC's and just to play around with, there's nothing fundamentally wrong with the M17x-R2 at all. This thread was created by a moderator as a courtesy but I really didn't think the throttling that happens with the M17x-R2 in Vantage was really anything to get too concerned over, that's why I originally posted it in the benchmarking thread as a simple reply. Now that it has everyone's attention, a little more testing with games couldn't hurt.

 

Hello guys!! I just orderd My replacement system. And i went for the 720 CPU, dual 5870, 6 gb memory,500 single hdd and so on.... But i read some pages back that with 540 and 620 CPU you dont get the throttling?? Is it true? I thought the 720 was ok. Should i change my order and get any of the others? I have had 2 xps 1730 and 1 m17x R1 returned. So i dont want to do this one more time if i can do something about it before. Thanks

 

I previously done some tests prime 95 running 4 threads with Furmark running benchmarking in extreme burning mode & logging on GPU-Z.I did these tests with stock gpu clocks & some oc clocks.The oc clocks were 650/1000 & 650/1050 there was no throttling seen.Max value on cpu was 25w,Max temp 58C.The max temp on 4870's was 72c.So with the i7 620M i have seen zero throttling.

 

Can you turn down the resolution and see if that helps the fps? Thats usually how I test for limitations

 

This is my thinking as well, the CPU and GPUs will be fine...it is to protect the cheap power circuitry on the laptop motherboard.

Also, some games may not throttle because they essentially leave the cpu at idle. You will only reach that artificial threshold that joker mentioned if using both gpus and the cpu all at the same time

 

relax - there should not be an issue here with stock clocks. the throttling joker observed was at high overclock settings.

your system should be just fine

 

Indeed. I on the other hand want to push the machine to high limits, so waiting on a workaround soon enough. That or betterndamn Dell drivers.

 

Follow up:

I ran Crysis, Left 4 Dead 2 and Dirt 2 at max settings at 1200P with an insane overclock of 900/1200 and there was no throttling with the dual 5870m setup! This was with stock Dell drivers. I think the problem so far is exclusively limited to 3DMark Vantage. I'm going to let Brian know about this because I'm sure Alienware/Dell wouldn't want potential reviewers marking the machine down because it didn't get some uber Vantage score. I'd encourage the rest of you to do some game tests with GPu-Z logging turned on to see if you observe any throttling at stock clocks and while overclocked. So far I'm very relieved that it didn't throttle in an intense game like Crysis Warhead. The temps got scorchingly hot so I'd caution those that are feeling adventurous to use proper cooling.

 

What were the temps like?

 

It hit as high as 105C with 900/1200 clocks but I didn't have the notebook sitting on a cooling pad at the time and the voltage was set at 1.15v. It makes perfect sense why Dell set the voltage at 1.05v default, the parts simply run far too hot otherwise.

 

Well this is indeed good news!

 

Has to do with energy loss that turns into to heat and efficiency. This is in reference to high OC'ing. A physicist could give a better technical answer than I.

 

Man that is really hot. Isn't it somewhere in the ball park of 115-120 that the hardware starts to fail?

 

do i have throttle i checked hwinfo and my turbo is always 3.3 ghz. getting quiet fan revs up and down just when moving from explorere to desktop etc.
ere picture
http://i47.tinypic.com/281fmut.jpg

 

well im now starting to get MAJOR problems with my R2 its all running at stock speeds on a i7 620m and when running wow as soon as the fans kick in the system slows down and its really choppy got the latest bios and also 10.4 installed

 

is 10.4 better than the dell 5870 driver? im a bit confused with ATI ive donwloaded the only and latest 5870 driver, but does that include CCC or ca i download a newer one?

 

Currently we are using Dell drivers the ATI ones are broken for the 5870x2 at the moment.

 

I have an electrical engineering degree, so maybe I am qualified?

As the temperature increases, the resistance also increases in the components resulting in greater voltage drops across them and more energy is converted into heat. Heat=wasted energy

So as the ac adapter heats up, it becomes less efficient and supplies less wattage/power to the laptop. When overclocking, the power is precious and always in short supply - you need all you can get.

 

What scook wrote is correct but let me also present how dramatic the effect of temperature can be.

On a PSU of 240 watts such as the on in the M17X we have the following happening.

Remember heat increases resistance. Resistance is the reason the adapter is losing power.

Example: first let's find the resistance of the PSU:

M17X PSU:
V= 12 volts
I= 19.5 amps

We know that V=IxR therefore R=V/I

so R=12/19.5 = 0.615ohms

Now let's imagine the PSU goes to 200 degrees celcius. Remember resistance is usually rated at 20 degrees celcius.

R=Ri[1+α(T-Ti)]

R=0.615[1+(3.9e10-3)(200-20)]= 1.047ohms

I=V/R

I=12/1.047

I=11.46 amps

so the PSU at 200 degrees celcius would generate 138 watts instead of 240 watts.

Now we all know the adapter would melt if it got to 200 degrees celcius, I measured the temperature of mine at full load and with some help from me to heat up the PSU, it reached 87 degrees celcius.

Therefore the resistance now becomes .7757ohms

I=12/.7757ohms

I=15.47

Therefore at full load the PSU is generating 186 watts of power instead of the rated 240 watts.

Yes temperature dramatically decreases the power output generated. Now all my calculations are theoretical. It would be good to test the power output at different temperature level which I can do over the weekend...

 

This is an excellent analysis

 

so what your saying is if they improve the cooling on the psu more power will be supplied to laptop

 

Well that is the theoretical explanation but I will confirm with a voltmeter and ampmeter this weekend at different temperatures.

 

I would also like to add that an increase in the temperature will increase the resistance at the coils and the transformers. This leads to drop of the voltage that charges the capacitors and this will cause current rippling because of the rapid discharge of the the capacitors. Consequently, rippling will result to instability and crashes. Also the life and the capacity of the capacitors will reduce by time and temperature.
Note from wiki

about rippling
Ripple (electrical) - Wikipedia, the free encyclopedia

Another thing I would like to mention is that the only accurate way to measure the exact amount of voltage and the rippling (and indirectly the amperes and watts the system draws) is by using an oscilloscope that I currently don't have where I am, otherwise I would have tested it already.

One thought I have but I cannot prove since I don't have the necessary equipment is that the battery helps the current situation by reducing the ripple since it's connected in parallel and improves the stability of the system (the system will draw current from the battery when the voltage of the PSU drops lower from the voltage level of the battery).

Keep in mind that by increasing the available current to the system, the voltage drop on circuits located on the motherboard will increase as well and if it goes beyond a certain level it will lead to crashes. So we will need bigger capacitors and better quality of circuits. All these will increase the cost, so...

 

While on paper this make perfect sense, the power drop on that scale doesn't happen. From direct measurements, I've found the adapter scaled to 240W just fine. If you assume a +20W leeway over the rated 240 yielding 260W maximum and then take 240-186=54W theoretical loss, then 54W-20W=34W deficit that's unaccounted for that was from actual measurements.

 

Joker the other accurate way to measure the current and the voltage (and therefore the Watts) is to use a multimeter at the output of the PSU. How did you measure it?

 

I am going to do the measurements with a voltmeter and ampmeter but my ampmeter has a maximum measurement of 10 Amps which won't work since the M17X PSU goes to 19.5 A.

I need to go buy a new one.

 

How do you plan on doing it? Are you going to do some wiring?

 

Yep, I have no choice but to play with some wiring while the PC is plugged. It should not be hard at all.

 

I used a Kill-A-Watt meter, not the best way to do it but still fairly reliable.

 

I expect the output to be significantly lower than that measured by the Kill-A-Watt at the input of the PSU, we will find out when Joebarchuck will do his measurements.

 

If we knew the efficiency rating of the PSU, it would be a simple task of taking the measured draw at the wall x PSU efficiency to get the actual wattage.

 

Yep, it would be easy but it's going to be as easy to just measure the output. I just need some time and a new ampmeter that goes up to at least 20 amps.

 

Still it would be a theoretical number and it would vary from system to system, but yes it would be nice if we knew it.

 

The efficiency of the PSU has been speculated to be 80% by UncleWebb over in the stutter thread for the M17x

That is probably about right. Any value measured with a Kill-a-watt is before taking into account the drop due to the efficiency of the adapter.

 

Hi guys, the ampmeter that would go to 20 amps was over $100.00. I really did not feel like buying it for this test therefore I will run the test but on a different system.

I am sure the results will be less dramatic but it will give us a clear indication on what the power loss due to heat is.

Sorry for the delay...

 

What 's wrong with 10 amps? I don't see how you would reach that with a single laptop.

 

The PSU of the M17X is 19.5V at 12.3Amps. If I want to truly test the drop in power due to heat, I need my ampmeter to at least be able to check 12.5Amps. Mine stops at 10.

PS: I made a mistake in my earlier post. The PSU is 19.5V @ 12.3 Amps not 12V @ 19.5 Amps.

The reason we want to test this is because Joker believes the throttling is due to lack of power at high OC and as the PSU heats there is loss of power which even more affects the throttling issue.

 

At some places I worked at, we had a high-current probe. You simply used the DVM in DC volt setting, but attached the probe the same way you would measure current. It gave a reading in DC Volts, but it was actually the amperage. One probe could handle 100A. And yes, it was big.

 

One notice, if you use a cheap digital multimeter to measure more than 10 Amperes, you shouldn't leave it on for more than 10 seconds and then you will have to wait for at least 30-45 minutes to cool off for the next measurement otherwise you will fry it.
If you don't have a multimeter with 20A max, you can use two identical 10A multimeters in parallel and add the two amper values
If you want to have a continuous measurement of the current, you need to use an analog amperometer.

 

Thank you for the advice, I will try some measurements over the weekend.

 

It's possible, but chancey. I've seen them fry in less than 2 seconds. You're better off either finding a 20A DVM or a high-current probe.

 

The best way is to use an analog amperometer, like those used on industrial installations that are designed to monitor currents 24/7. (I have a couple of those but not here unfortunately)

 

Looks like I'll be helping you guys on this thread now, too haha. Dell expedited my R2 replacement because I had a conference call yesterday with the lead Dell engineer for the M17x and he wants my specific R1 system sent directly to him for analysis. Already have replacement specs and it's a 920XM and 4870CF. Not at all complaining about the 4870CF vs 5870CF with all the issues surrounding the new cards still. Maybe I'll replace them down the road if things smooth over.

Looking forward to testing with you, Joker.

 

Any news on the throttling issue with the 5870XF?

I was kind of hoping that AW would give us info by now. What is taking so long... Do they just work 20 minutes or every hour also?

 

"I will discuss it with AMD and thanks for bringing it to our attention" are a polite way of saying the check is in the mail.

I don't think the worlds leading graphic card company was simply too busy and forgot about the worlds fastest gaming laptop.

Was it also an oversite yesterday that the level 1 tech I spoke with at AMD had never heard of Catalyst Mobility and a level 2 tech would flat out not discuss it? That sounds a lot like M17x 5870 CF vapourware to me.

I think the best chance everyone has of getting this thing fixed is do what Joker did and call AMD and DELL daily until the issue gets resolved.

As someone has said how hard would it be to include the 5870 CF in the Catalyst Mobility 10.6. Are we now expected to wait a whole month for this fix? Does it really take that long to rectify the situtation?

I posted a question for Jay Marsden yesterday on the AMD forum about why the 5870 was not included in the 10.5 & 10.6 Catalyst release and my comment has not yet been posted. I am still waiting for a response.

I don't think AMD and Dell simply forgot about the issue as this is a common question people ask daily. My understanding is that AMD was pushing hard back in 2007 for all tier 1 manufacturers to be included. I don't think this is an AMD oversight.

Isn't dell the only company offering 5870 CF in a laptop? Perhaps they don't want a generic catalyst moibility driver made public yet that works with 5870 CF in a laptop in order have exclusive rights to this driver for the time being. I think this issue has more to do with marketing than engineering.

 

Sounds like how my replacement went down. My unit went straight to engineering. They did all the setup on my unit and still had issues...so my replacement was authorized on the spot and I got an RMA that went to their department.

I haven't seen a lot of throttling with the 4870's, just some on a few games. Doesn't really matter to me anymore though, I just got a new job that is Apple exclusive, so probably this weekend you will see my unit in the marketplace. I cannot even have a Windows based unit on their network, soon as it is recognized as a Windows machine, the MAC address gets banned...sigh...

 

Originally Posted by hitokiri1
QUOTE: "Well I was talking to tech support finalizing the details of my exchange (warranty transfer) and I mentioned to the tech that I couldn't install ATI drivers on my new system.

He let me know that this is something they know about and they are trying to figure it out. They find that the new drivers 10.5/6 on their test systems were unstable with the 5870's and (like previously mentioned by JOKER) they are working with ATI to figure out the issue. Obviously no ETA and no real resolution but I found it helpful to know that they do in fact know about the issue and are at least trying to resolve it."

 

Guys, I've found a solution to the 5870 throttling in vantage. Problem is that it's a bit of a hassle to implement every time so when I get home I'll try to streamline it a bit. Refernce ati drivers will eventually work on our cards. Dell is very well aware of the problem. I'll ask Brian about it again after I get home tomorrow.

 

Thanks Joker much appreciated.

 

One question that I never saw answered or ever even asked is: does the throttling issues affect only the crossfire configuration of the 5870's or would it affect the single 5870 configuration as well? I realize after reading that it is only happening on ridiculous oc's, but is it possible to have the issue come up in a single setup?