S-XPS 1645 AC Power Throttle Issue Investigation

[unclewebb]

http://www.fileden.com/files/2008/3/...rottleStop.zip

Here's the first official release so people can stop e-mailing and overflowing my PM box with requests now.

Time for more users to give it a try and see what they think.

If this simple program dramatically improves the performance of your laptop then consider donating $5 bucks to encourage future development. This version will expire in a couple of weeks so I hope I don't have to add some nagging messages to the next release of ThrottleStop. I hate them.

I've added Core 2 support including adjustable half multipliers, Super Low Frequency Mode (SLFM) support as well as adjustable voltage ID. Be careful when reducing the VID of your CPU. Lowering the core voltage can dramatically reduce the core temperature at full load but if you go too far, a BSOD might be in your future. I'd suggest lowering your core voltage only one notch at a time while running Prime95 Small FFTs. When you lose stability, you've gone too far.

ThrottleStop includes the ability to lower your maximum multiplier on your Core i7 as well now. This might allow you to disable Clock Modulation and still stay within the 90 watt power envelope. It's up to you to monitor your power consumption to see what is possible.

Happy testing.

[Chris123NT]

Just tried the new version and it works awesome. Made a donation too, your hard work definitely deserves it

[jakertberry]

On battery power these laptops don't throttle as much which means the CPU and motherboard are likely getting more than the 90 watts that the power adapter is allowed to deliver.

For a test why doesn't someone run a consistent load like Prime95 Small FFTs on all 8 threads without using ThrottleStop. Let this test run for something consistent like 5 or 10 minutes and do one test when plugged in and one test while on battery power.

RealTemp 3.50 can show how much Clock Modulation is being used. If on battery power the bios lets the CPU run at full speed without the multiplier sagging down or without any Clock Modulation then that would lead me to believe that this computer is quite capable of handling the extra wattage that a 130 watt adapter can supply.

It wouldn't absolutely prove anything but it might be interesting.

Prime95 running Small FFTs on battery power for 10 minutes (using "High Performance" power settings) and the display backlight maxed:


Prime95 running Small FFTs on 150w AC adapter for 10 minutes (using "High Performance" power settings) and the display backlight maxed:


Prime95 running Small FFTs on 150w AC adapter for 10 minutes (using "High Performance" power settings) and the display backlight maxed and ThrottleStop running:



(Interesting note: when I plugged in the 150w adapter while Prime95 was running, I got a warning that the adapter could not be identified and that the laptop would run in reduced performance mode. The multiplier never went higher than 7. This scenario is not reflected in the results above, as I unplugged, stopped Prime95, plugged back in, and everything was normal)


On a side note, I love how much faster the battery charges with the 150w

[unclewebb]

Your screen shot shows ThrottleStop Enabled but you haven't selected any options so it's not actually doing anything. You need to put a check mark in the Disable Clock Modulation box and one in the Set Multiplier box.

Chris123NT: Thanks for your support.

[jakertberry]

Your screen shot shows ThrottleStop Enabled but you haven't selected any options so it's not actually doing anything.

my bad.. I had unchecked it when I took the screenshot. I only had "Disable Clock Modulation" checked during the test.

[jakertberry]

Just tried the new version and it works awesome. Made a donation too, your hard work definitely deserves it

Seconded. For every word.

[Asherek]

"Fluff" is putting it mildly. Sadly, this thread is full of useless speculation and pointless Dell bashing. That helps no one.

It will be challenging for them to find the useful information, admidst all the crap. I agree that there are some really helpful contributors here, but they often get drowned out by the juvenile rantings of know-nothing folks who think that they are experts.

There's no need to be quite so snarky. This thread has fluff, but it's a forum. There's not much else to expect when we shouldn't have to have been dealing with this issue in the first place. People complain because there's no other venue to rant on. Calling up a customer rep and yelling at them isn't exactly very productive either, but unfortunately that's all we're pretty much left with here in terms of options.

I expect, as a consumer, when I've paid a sizeable portion of money, that I will not be a beta-tester for a company. I expect the product to be working as it was promised, which this laptop definitely does not.

Call it pointless Dell bashing if you want, but unless voices are raised, it seems that nothing tends to get done. The fact that this thread is so large and so well read and posted on hopefully means something to Dell more than the information on it, because it indicates that customer are unhappy (and a sizeable portion of them). Quite honestly Dell should be coming to the same conclusions this thread did without our help. That's what their engineers are paid to do and that's why we shell out big bucks for their products. I feel absolutely zero sympathy for them whatsoever.

[DeathWalking]

Prime95 running Small FFTs on battery power for 10 minutes (using "High Performance" power settings) and the display backlight maxed:
[snip]
Prime95 running Small FFTs on 150w AC adapter for 10 minutes (using "High Performance" power settings) and the display backlight maxed:
[snip]

Prime95 running Small FFTs on 150w AC adapter for 10 minutes (using "High Performance" power settings) and the display backlight maxed and ThrottleStop running:
[snip]

(Interesting note: when I plugged in the 150w adapter while Prime95 was running, I got a warning that the adapter could not be identified and that the laptop would run in reduced performance mode. The multiplier never went higher than 7. This scenario is not reflected in the results above, as I unplugged, stopped Prime95, plugged back in, and everything was normal)


On a side note, I love how much faster the battery charges with the 150w

Can someone explain why battery has a max multi of 12.277 (aka 1 core at 13, 3 cores at 12) versus 150w ThrottleStop's max of 12.007 (all cores running at 12)?

[unclewebb]

As mentioned before, Intel limits the amount of current and power consumption of these processors. Turbo boost is only available when you are able to stay under these preset limits. There is no software available that lets you go beyond the limits that Intel has set. On the Desktop Core i7, many performance motherboard manufacturers have added a feature to their boards that lets the CPU operate beyond the Intel set limits but it's very unlikely that any notebook manufacturer will ever include that feature in a bios, especially Dell.

In his first screen shot, when all 8 threads are loaded, he must be at the Intel limit because it shows that no turbo boost is available and being used. The default multiplier for a Core i7-720 is 12 and his screen shot shows that all 8 threads are running at 12.000. The maximum of 12.277 likely occurred just as he applied the initial full load. During the one second sample period just as the load was coming up, there was some turbo boost available but it didn't last very long. With 8 threads loaded, it's possible to have a maximum multiplier of 13 but only if you are within Intel's power and temperature limits. That number says for that 1 second interval, he was able to get a 13 multiplier for 27.7% of the time and the other 72.3% of the time it was down to 12. That gives you an overall average of 12.277. That's what that number means. Intel provided high performance timers within these CPUs so you can very accurately measure the average multiplier. This method gives a little bit of random variation at idle but at full load it tends to be rock steady, unless your CPU is throttling back and forth between two values like 12 and 13.

The i7-720 will not get any turbo boost when it is fully loaded with 8 threads of Prime so it will be limited to the 12 multiplier. You might be able to use my load tester program included in the RealTemp download to do some testing with. It creates a much more gentle load than Prime so with that you might see the full 13X multiplier when you run 8 instances of Load Tester, one for each thread.

With 8 threads of Prime running and without ThrottleStop, you should be able to go into the RealTemp Settings window and enable some Clock Modulation. Instead of the bios slowing you down, you can slow yourself down using this method. What you should see is that if you apply enough Clock Modulation while testing, sooner or later your CPU will get under the Intel power limitation and your full turbo boost should return so your multiplier will jump up to 13. This will give you a better idea of just what the true limit is before your processor gets throttled down to 12. You might find a certain percentage of Clock Modulation where your multiplier starts reporting some random looking oddball numbers somewhere between 12 and 13. That's a sign that the multiplier is cycling hundreds of times a second based on load. When it is overloaded, it drops to 12. When it thinks it might be able to go higher, it jumps up to 13.

[razrkillr]

This is so confusing... or rather will be confusing. Intel Based i7s with the arrandale microarcutecture will have 4 threads and 2 cores, support for turboboost and smart cache. 4MB of L3 cache. so yes this does seem to be a fair stand up because of the 3.33 you can acheive with turbo boost. Q3 will release a new i7 that has 4 cores and 8 threads. Arrandale has intergraded GPU allong with the 32nm, this will save power + even more power by having the ability to switch from intergraded to discrete GPUs. The low watt draw will help this laptop for sure but even with that trottling will still occur as the lower 1640s with duel cores trottle a little bit compared to the 820QM.

Looking to to future