M17x-R2 4870/5870 Throttling?

Well I got mine, and I've been able to make my 4870s throttle at will. Positive note: processor no longer throttle or modulates.

Using a Kill A Watt, I was able to measure the wattage usage going all the way up into the 280s before the ATI cards were throttled by the system down to 300/400, and being locked there. They would stay locked there until the game was closed and stayed closed for a long period of time (haven't been able to measure the time yet, just got my unit Friday and finally fully set up last night). Once this throttling occurs, even RivaTuner has no control over the clocks. I'm wondering if it's the BIOS forcing these clock? Joker.. it even supersedes your clock lock fix on the 4870s, which did actually work up until this throttling occurred. It's obviously another inadequate power problem, just being controlled via the GPUs instead of the CPU this time. I'm at work so I'll post some screenshots tonight when I get home.

More investigation tonight.

EDIT:

I should also note that I can make it do this with Turbo *off* and all overclocking disabled. That means it throttles even at 2.0Ghz.

 

Ok, so I did some testing here, and I see that the R2 flickering thread has been closed. My new R2 has the flickering problem, and I'm already on BIOS A05. I have flickering with FurMark (mostly only if I run Prime95 with it, but happens every so often without it, too), and happens pretty badly with the test scene I was using in Dragon Age to show the R1 clock modulation problem. Can we re-open that thread, or is there a new one that I don't see?

It's definitely a power consumption issue, as temperatures of the CPU cores and the GPUs are well within specs. This is also definitely not a texture issue.. it's definitely the old "flickering" problem that was before A05 for most of you.

EDIT:

Using GPU-Z, I can see what the system is doing. It's periodically shutting down the second ATI card, or at least putting it to 0 load. I'm seeing periodic blips to 2D clocks only on the second card, too, but I think that's from when the load is being put to 0 and not the actual problem.

 

Yeah the 240W supply was never enough for furmark + prime 95 with dual 4870s + 920xm + all the other hardware. I wouldn't worry about that, in real gaming you will never see it.

 

Wrong. See it in Dragon Age. Haven't tested others, yet, but will tonight.

 

If you are seeing it in games you could just have defective cards or are OC'ing. Did you measure the power draw when playing Dragon Age? Last I played that game with my 920 + 4870 xfire setup, there was no flickering at all.

 

Just grabbed it from another thread:

Just a little detail for those that might want to test their R2 system themselves.

Here is a test I did with Dragon Age: Awakening. You won't notice this problem normally.. I'm talking end game characters with uber buffs on all 4 toons (they're all translucent and have weapons with particle effects and different glowing clouds around them). I can post a screenshot of the scene I use to test this with.

Side view of the scene I can make it throttle/crash at will with:



The view of this scene that I can consistently reproduce the throttle or crash (depending on whether throttling is disabled in the VBIOS):



Having all of those characters partially translucent with all of those translucent particle effects really, really taxes the system. The game engine is so optimized though that I get consistent 60FPS even with all of this going on. Not many games will use every bit of system resources available like Dragon Age does. My system will hover in the 240-250W+ range until the throttling or BSOD occurs if I leave this camera angle.

As a side note: I've seen it up into the 270Ws+. Done alone, neither crash it out. Prime95 runs just fine at even at 27x multipliers so the CPU is solid, and Furmark Multi-GPU alone runs alone fine so the 4870s are fine. Furmark only puts the CPU at about 11-13% usage, so it's a good test of just GPUs. Dragon Age I've seen CPU usage over 70% at times, so you're looking at closer to a Prime95+Furmark example.

Another odd finding - I just put the stock throttle clocks back in and wasn't able to make it happen here - though I was able to make the flickering happen about half a dozen times in the past 30 minutes. It's ice cold here compare to my house, though. I'm wondering if there really is some sort of heat problem.. seems weird because my CPU cores only reach 76-80 degrees and the GPUs both stay at about 68-69 degees. Didn't notice any difference in the temps here to home, but will test home in a few.

 

Hi Joe, +1 rep for theory, thanks.

This is very interesting and something I have wondered whether a hot charger would cause power-supply instabilities causing voltages to fluctuate, and also where the higher load causes the supplies to sag. When the voltage goes high, it puts a lot of stress on capacitors. When the voltage goes low, it can put a great deal of stress on semiconductors, because they need to draw more current to deliver the same output. In either case, the heat output goes up and the life [of the semiconductors] will be shortened. I think the SAFEZONE for a PSU to deviate its volts is plus or minus around 5%~6%. Any more then that will cause potential problems. I think a well-regulated PSU can reduce stress on the motherboard's VRM. The VRM than wouldn't have to work as hard filtering and stepping down the voltage. In theory a badly stressed PSU may give too high/low voltage, which would cause the voltage regulation circuits on the motherboard and graphics card to have to work harder which in turn would generate more heat.


Joe have you tested your theory by placing the charger in the fridge and benching? Anyone trying this should allow the charger to warm up to room temperature slowly by wrapping it in something to insulate it afterwords to avoid condensation inside the charger. Anyone into photography knows what I am talking about with cameras on a cold day brought inside quickly.

 

I could test by putting the charge in the freezer but let's be honest it would easily heat up the freezer and probably melt most of the food I have in there and I would have to throw everything away.

I think the best is to just measure it with an ampmeter and voltmeter but I could not find one with the required specs. At least not within a price I was willing to pay for the benchmark.

 

I was just over on the silverstone website and read something interesting. Some of their standard power supplies for desktops are rated at continuous power output up to 40℃ (104F). That would imply your theory is correct and above a certain temperature threshold power output will fall. Too funny about the freezer I never thought about that. Perhaps a cheap styrofoam cooler with ice inside would do the trick as well or place ice ontop of charger.

 

Ok, weirdness. Now I can't get it to throttle with about 30 minutes of testing at home. I did see a few periodic flickers still, maybe 3-4 times. Wattage hovered 244W-254W.

Then I remembered that when I put the throttle-stopping BIOS in place, I unchecked O/C in RivaTuner (I only had it locked at stock clocks, though). Tabbed out of Dragon Age and opened RivaTuner and checked it. Tabbed back into Dragon Age and let it run for about 5 minutes. No changes at all in power output and no stuttering. Tabbed back over to RivaTuner and the exact moment I *unchecked* O/C I got the BSOD with atikmdag.sys error that I was getting when I was blocking the throttling via BIOS.

Oddness. Wonder if something is flaky with RivaTuner and this particular setup for me. Unchecking that shouldn't have BSOD me. I'm running 10.5 drivers currently. I guess I can just leave it as is and hope I don't see any flickering during regular play of the game. Going to check out Crysis.

 

The BSOD you saw from unchecking the clocks is because of RivaTuner. I haven't been able to find a way around it yet. Sometimes it BSOD's other times it doesn't.

 

Alright so I messed around with some clocking. Any kind of overclocking causes the flickering to be pretty obvious in Dragon Age with that scene I showed you. However, if I undervolt the cards .050V, the ever so occasional flicker on that scene is completely gone. Honestly, everything I play runs rock solid on stock clocks, so if that ends up being the solution for me I can deal. I might try messing with some OC with the UV and see if I can get any extra clocking out of it on the stock clocks and still not have any flicker in DAO:A.

On a side note: when I run Vantage even on stock clocks, the first scene has a couple times (pretty consistently at which points each run) where a couple small square or rectangle regions flash black for a frame (not the flickering problem), and the overhead terrain scene has what appears to be forward-backward framerate stuttering (4-5 frames back-forward at times). Is that something with the 10.5 drivers or is there still the possibility something is flaky with my cards? I don't notice any video corruption anywhere else in any games or Furmark.

 

+ REP!! Thanks a lot for your efforts Joker!

On a side note, I have seen some slow down in some games, not sure if it is throttling or not, but everything, including the current OS is completely stock, with BIOS OC of the processor. So, this fix may not help me right now, due to the stock clocks, but I plan on getting everything reloaded manually, and OC'ing so I will definitely use this information. Thanks again!

 

I've run vantage a few times on my system both on the 10.5 and 10.6 drivers and i have not seen what you are describing at all. I also played Crysis last night and had no slow downs. All my clocks are currently at stock.

 

Just an FYI that apparently I'm not the only one with that Vantage problem.

http://forum.notebookreview.com/alienware-m17x/496360-3dmark-vantage-issue.html

I'm really not worried about the cards anymore, at least with regard to that.