M17x Stutter

[unclewebb]

You can do the recap Elkay.

I just tested RM Clock. When clock modulation is different from one core to the next, the RM Clock graphs will confirm this problem. It will show zero throttling on a core at 100% clock modulation and lots of throttling on the core that is set lower. It's always nice to be able to confirm data with multiple apps all showing the same thing.

BatBoy: The more data the better. Dell needs to know that this is not just a problem on one person's computer. Clock modulation can be a problem for anyone trying to draw too many watts out of their power adapter. An SLI system will slow down to a crawl and stutter if this happens.

[Elkay]

So I've been away for a couple of days and havent had a chance to catch up on this thread. Looking over the last few posts it appears some progress has been made. Elkay or webb - can you recap?

What it boils down to is that clock modulation is occurring at around 225W according to a Kill A Watt monitoring the M17x power supply. It also appears that just about any game that utilizes more than 2 cores of the QX9300 and also puts a load on both GPUs causes the wattage usage of the laptop to go over 225W fairly easily and cause all of the cores on the QX9300 to drop to 12.5% (316Mhz each at stock clocks) using ODCM (On Demand Clock Modulation), thoroughly killing performance. The clocks will stay at 12.5-25% until wattage stays below 225W for a period of time, *slowly* staircasing back up to no modulation. This occurs in Crysis, Dragon Age: Origins, and Just Cause 2 for me. Another user is reporting Mass Effect 2 is doing the same thing. Using ThrottleStop to "mostly" prevent the modulation, I've seen wattage go up to 257W while doing normal gaming. Using FurMark Multi-GPU you can trigger ODCM to 12.5% instantly at will with the QX9300 and multi-GPU. Basically, in order to fix this problem it looks like there would need to be a drastic reduction in wattage usage (maybe 5870s but Chris already confirmed the R1 would not be getting them), or at the very least a 300W power adapter and a BIOS update to support it. Dell doesn't even have a laptop power adapter >240W, so that solution would take months at the least to come out, and I honestly don't ever see them even doing that since the R2 with the 5870s will stay under the wattage mentioned due to the fact that the 5870 XFire runs 30W lower than the 280M SLI (120W cumulative versus 150W cumulative).

That's the cliffs notes. There are *many* screen shots I've taken documenting all of this, as well as extensive wattage results in various games and with varying system usage/peripherals connected. I've sent my Dell contact half a dozen PMs since last night with all of this information and screenshots but I have not heard back yet.

[Elkay]

You can do the recap Elkay.

I just tested RM Clock. When clock modulation is different from one core to the next, the RM Clock graphs will confirm this problem. It will show zero throttling on a core at 100% clock modulation and lots of throttling on the core that is set lower. It's always nice to be able to confirm data with multiple apps all showing the same thing.

Awesome. I'll get screenshots of this happening tonight, as it happens frequently enough that I'm not worried about being able to reproduce it.

[BatBoy]

Thanks for recapping Elkay and thanks to Webb for continuing his support of this project. I will take a look at your previous posts Elkay for screenshots.

[Elkay]

For simplicity's sake, here's the screenshots I fired off to Dell all in one place with a super-brief explanation:

FurMark Multi-GPU showing clock modulation at will when launching test:
http://img338.imageshack.us/img338/2...throttling.jpg

Recovering after stopping FurMark Multi-GPU:
http://img219.imageshack.us/img219/9...hrottling2.jpg

FurMark single GPU showing no throttling:
http://img225.imageshack.us/img225/2...arkthrottl.jpg

Prime95 running with all cores at 100% showing no throttling:
http://img100.imageshack.us/img100/3...throttling.png

Crysis throttling at stock (also notice one clock stuck at 75%):
http://img689.imageshack.us/img689/1...tlingstock.jpg

More Crysis throttling at stock:
http://img21.imageshack.us/img21/638...lingstock2.jpg

More Crysis throttling showing frequecy of occurance:
http://img571.imageshack.us/img571/2...isthrottle.jpg

More Crysis after exiting game and with Chrome open during game (more frequency):
http://img208.imageshack.us/img208/2...isthrottle.jpg

Dragon Age: Origins throttling:
http://img291.imageshack.us/img291/9...kthrottlin.jpg

Dragon Age: Origins throttling (notice CPU utilization, this is a CPU-hungry game but the game runs flawlessly other than modulation):
http://img156.imageshack.us/img156/3...onagethrot.jpg

Just Cause 2 throttling:
http://img692.imageshack.us/img692/6...ckthrottli.jpg

Windows yelling about insufficient system power for Aero, Alt-Tab out of Just Cause 2:
http://img443.imageshack.us/img443/3...2aerobasic.jpg

My DPC at idle, showing it's just fine:
http://img37.imageshack.us/img37/7169/dpclat.gif

How my GPUs are correctly throttling without a game set to Prefer Maximum Performance in NVidia control panel (NVidia bug that's been around for ages):
http://img571.imageshack.us/img571/5...puclocking.png

How one GPU doesn't throttle back down until rebooting once a game/app set to Prefer Maximum Performance runs and exits:
http://img20.imageshack.us/img20/194...ermizerbug.png

Here's the recap on wattage usage:

Idle (stock with 197.31 NVidia drivers allowing both GPUs to throttle): 63-65W
Idle, 1 GPU locked full (NVidia bug still present when selecting Prefer Maximum Performance in the driver then exiting the program): 83-85W
Idle, both GPUs locked full, stock clocks 1066/1333 FSB/memory: 97-101W
Idle, both GPUs locked full, overclocked 1280/1280: 98-101W

Go Dark mode (FX lighting off): 1-2W difference
Backlighting lowest/highest: 9-10W difference
Full fan speed: 5-6W difference

Prime95 stock (throttled GPU clocks) (highest brightness): 123-125W
Prime95 stock (max GPU clocks) 1.1375V (highest brightness): 135-138W
Prime95 stock 1.1375V (max GPU clocks, lowest brightness): 125-130W
Prime95 undervolt 1.0500V (max GPU clocks, highest brightness): immediate drop to 123-125W

FurMark Single-GPU: 160-178W
FurMark Multi-GPU: 190-233W (immediate and after 233W hit clock modulation starts)

Dragon Age (stock): 180W-220W+ (one scene hovers right at 218-221W stable, then see 224-225W spike from another application doing something and clock modulation occurs)
Dragon Age (stock clocks, undervolted CPU to 1.0500V, same scene as above): 206-208W
Dragon Age + Chrome on Facebook (stock): 215-220W and starts clock modulating to 12.5% 2-3 times a minute
Dragon Age (1280/1280 overclock): 214-215W before fans kicked on, 218-221W after fans kick in (small 223W spike didn't cause clock modulation, after 5 minutes a 225W spike caused clock modulation)

Crysis (stock): 210-220W (first scene, after crashing into the water, swimming around and walking on the beach, 225W spike from outside game caused clock modulation after 3-4 minutes)
Crysis (1280/1280 overclock): stuttered immediately on level launch with seeing no higher than 179W (*very* quick spike perhaps?), then 213-215W before fans kicked on with no clock modulation, 2 minutes fine then as soon as fans kicked high clock modulation occurred

Just Cause 2 (stock): 205-220 hover some places, up to 225W+ and throttling starts in many parts of the game

[BatBoy]

Excellent! Thanks. Unrelated, what display driver are you running - 197.31 chokes on Just Cause 2.

[Elkay]

Excellent! Thanks. Unrelated, what display driver are you running - 197.31 chokes on Just Cause 2.

I was running 197.25 until last night pretty successfully. I was running TS all the time, though. Updated to Dell's 197.31 since I figured they were going to ask me to do so anyway. I also had my 3d clocks locked at full with 197.25, though, to prevent the stuttering which I'm pretty sure was still occurring with those drivers. I'd go back and test them but since I reached out to Dell I'm leaving things as is until I hear back. Have you tried using TS with JC2 to make sure it's not just the modulation causing a problem?

[BatBoy]

The Just Cause 2 performance is driver related I believe. When running 186.82 performance is fine.

[Elkay]

The Just Cause 2 performance is driver related I believe. When running 186.82 performance is fine.

Hmm. I didn't actually play the game other than running around at a save point last night. I'll test more tonight.

My Dell contact also just sent me a message back saying he forwarded all of my information off to engineering but it will take some time to investigate.

[Mickbt26]

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!

I also had my 3d clocks locked at full with 197.25, though, to prevent the stuttering which I'm pretty sure was still occurring with those drivers.

How did you lock clocks like this?