M17x Stutter

[Elkay]

Excellent overview and diagnosis Elkay thanks,

Ive been too busy with my Job to be able to take many readings with my kill-a-watt, But the highest i've ever seen is 256W running Mass Effect 2 with CPU@ 3ghz and brightness max, Fx on, GPUs stock.

If throttlestop had of been disabled this would of been a slideshow every few seconds!



How did you lock clocks like this?

You can edit the registry manually like I used to do, or simply use:

Powermizer Switch

[Mickbt26]

Thanks but this will only work on 1 card. The other card still downclocks.

[Elkay]

Thanks but this will only work on 1 card. The other card still downclocks.

The first clock starts out downclocked on boot, once you've launched an application that uses both cards and you let the second card get to full speed it won't throttle back down. You may also need to set the Power Management Mode to Prefer Maximum Performance in the NVidia control panel. Try it without that first, though.

[adrian890]

do what i do keep moaning about the r1 and eventually i was given the option (today) to be called back by the replacement team on the 19th to talk about my r2=) i had alot of latency even after a new gpu and motherboard and processor! i cant wait! hoping to ask for CF 5870

[Elkay]

I'm hoping the same happens with me. I'm just being more nice than I probably should be and trying to help them realize what the actual problem is. I don't see any way of them providing a fix in any decent amount of time considering it's both a BIOS and hardware change (300W or more PSU, which Dell currently does not even manufacture) to fix the problem. At the same time, this is giving a little extra time for the 5870s to be officially released which, in my opinion, may be the only complete fix due to their far lower TDP than the 280M (15W less per card).

[Elkay]

webb -



There's the proof.



Interesting that the modulation survived this (also unexplained) core drop.

[unclewebb]

That's a good example of clock modulation stuck on a single core for an extended period of time. The idle CPU is not too hot or overloaded in any way so there is no excuse for this.

Hold Dell's butt to the fire until they come up with a solution to these problems for you. It's unacceptable at any price point.

Some users with defective XPS 1640 laptops have been offered newer models like the XPS 1645 or XPS 1647. Be prepared to fight for your right to a laptop that works at its advertised speed. Don't let them string you along with promises of an updated bios sometime in the future.

[unclewebb]

Elkay: You have all 4 cores locked at the highest multiplier and are not using RM Clock to control the FID / VID so this isn't a problem for you but I found an interesting bug in RM Clock today.

Here's the problem on a QX9650 desktop CPU.



When all 4 cores are fully loaded, two cores are running at the highest 11.0 multiplier like they should be while the other two cores are locked at the LFM 6.0 multiplier which kills performance.

wPrime and HyperPI are a couple of good benchmarks for testing purposes.

Downloads | wPrime Multithreaded Benchmark
Hyper Pi - The Super Pi Front-End

The QX9300 is a very similar CPU so I just wanted to warn users to watch out for this bug if they are using RMC on their laptop with this CPU.

[Elkay]

Oof.. nice find. Yeah, when I saw that RMC wasn't reporting the correct multiplier for me it was enough for me to not use it to do anything more than monitoring.

I can't speak for it's accuracy, but very interesting information this guy's rep supposedly told him. This would make me laugh further at Dell seeing as the 5870 XFire uses 120W versus the 280M SLI using 150W.

http://forum.notebookreview.com/alie...ml#post6254805

[unclewebb]

I found a good article about power consumption of modern GPUs but unfortunately they only tested desktop GPUs. Some interesting numbers.

Power Consumption of Contemporary Graphics Accelerators: Spring 2010 (page 3) - X-bit labs

A mobile 5870 is very similar to a 5770 desktop GPU but at a slightly lower clock speed for GPU and memory.

http://www.amd.com/uk/products/noteb...870-specs.aspx

http://www.amd.com/us/products/deskt...fications.aspx

The TDP numbers that you are using may not be actual power consumption. I don't know if ATI and Nvidia are in agreement on what constitutes full load for a GPU. If one company uses one program to test full load and the other company uses something else then trying to compare TDP numbers might be meaningless. Something to keep in mind. AMD and Intel do the same sort of thing with CPU TDP ratings. Design power and actual power consumption are not always the same.