[SOLVED] Asus G51J BSoD thread - part 2

Playing devil's advocate here. There's no doubt what you've discovered is happening, not a chance in hell that it isn't. BUT how many of these games even come close to using both the gpu and cpu to 100% at the same time? Because to me it sounds like the difference of just a few watts of power consumption is what we're talking about between a functioning laptop and a BSOD.

Let's say a game does get the gpu to 95%+....is it that likely that any of these games can get the cpu to that high as well...resulting in a BSOD? My point is that if either the cpu or the gpu isn't getting that high as the other one is....well wouldn't that consume less power (at least 3-4 watts less)? Dunno...just figured I'd ask. If I'm way off on my thinking here then please let me know....I'll be the first to admit that I don't know nearly as much about computers as most of you do.

 

i'll try and explain as best i can (i'm no computer engineer, so if someone sees something wrong, let me know).

because the throttling patterns differ in terms of system usage, this would indicate that the change in state is forced and not called. there is most likely no interrupt before it happens as there is no way to predict logically when the power spike would occur.

now, at the instant when the system pulls more power than the mainboard distribution can handle, it will force a drop. the CPU is not expecting this drop and either has a temporarily unstable voltage, unclean current, or both. whichever the case, whatever instructions that are in the pipeline will most likely fail.

direct3d can use the CPU/system memory for some transforms (not prerenders). if even one of those are a load/store to the transforms' memory, the value will be nonsense. direct3d (and WHEA) then report that the expected value does not equal at all what the output value is and trigger a BSoD.

 

I also don't think this is the case. No games I know of push GPU and CPU to 100% each. This throttled issue only happen when you ran furmark and prime95 together.

 

they don't have to. even if only 10% CPU is being used, as long as a d3d function fails on the CPU (unknowingly by the GPU) and the GPU error reporting catches it, it is still flagged as a PCI-E based on where the BSoD call was made.

 

I dont get it. Isn't a bsod issue something that normal warranty covers??

 

I think thalanix has come the closest to understanding the problem. I really don't think it can be controlled with software/BIOS updates.

 

Hello,

I have been following this since the beginning lurking in the shadows of this forum and others. So, I have confirmed this problem on my setup, and this throttling of the CPU seems to be caused by enhanced speedstep. Is there any way to turn this off that you guys have found? I believe XP has shown little to no problems simply because 7 can not seem to leave the CPU alone regardless of manual power options. XP seems to disable this when power options are set to "Always On" negating the crappy throttling (this has been alluded to). So, anyone with XP on this system should chime in and let us know. BTW Thalanix I believe you have hit the nail on the head, however the error may hit the GPU simply because its the next biggest thing pillaging power. In my experience even with heavy overclocking of the CPU the GPU will tend to go regardless of it being the problem due to the power instability, and its sensitivity to it. I do not want to rule out the possibility of there not being a fix we could exact.

 

'Always On' used to lock it to highest multi, but it still throttled for me. now it doesn't even do that, no difference between 'Minimal Power Management'.

my heavy desktop oc'ing only consists of an e8400 @ 4.48GHz/1.41v. the screen would get random blue/grey patterns because the memory couldn't hold, but i guess it could fit the GPU giving as well.

 

Check post 23 out and give this realtemp program a shot as I was able to lock it in 1.6ghz mode, but this still may further us in our pursuit. It isn't ideal, but maybe we can reprogram this , so we have a workaround of some kind:

http://www.sevenforums.com/hardware-devices/8871-w7-cant-stop-cpu-throttling-3.html

 

alright, so using realtemp 3.4 i was _not_ able to lock it in xp or win7. using 3.3 from above it works in win7, about to try xp + testing.

locking works, but throttling overrides it. i get conflicting readings between TMonitor and RealTemp, and based on what i see in temperatures TMonitor has it right. even though RT shows it as 1600, firing up furmark = drop in CPU temps.

 

Good deal Thalanix. Keep me posted as I am anxious to hear the results. With each restart of the computer you will have to reactivate RealTemp 3.30 Beta to lock it back. There must be a way for us to jimmy rig a problem like this to allow us to OC to what we paid for. However, I am a touch relieved to finally see stable FPS numbers in all benchmarking and general gaming albeit a bit slower.

 

same story in win7. RealTemp shows 1600, TMonitor shows 933 (even when locked) but furmark over top of prime95 drops CPU temps instantly.

 

What does CPU-z say for you?

 

Couldn't we disable stepping if we didn't have such a crippled BIOS. I mean the BIOS only gives us like a half of a dozen variables to change.

 

CPU-Z confirms with RealTemp. unfortunately, even with only RT open, as soon as furmark starts rendering the CPU temps drop from 75 to 65.

wouldn't help, the throttling is caused by something else.

it's possible this is only in synthetics, though something tells me if i do a regular-clocked 3dmark run it will stutter at the same intervals it always does.

SM2 test2 still slows down at the usual places, throttling is still there.

 

A couple of ideas to help confirm this theory:

- Underclock 10% or more, test 3dmarks with LCD brightness min and max. LCD brightness should have no effect on 3dmark scores at those speeds as GPU draw will be reduced

- No BSODs should ever have happened on lowest detail settings with *vsync turned on* - this caps fps @ 60 and will mean if the GPU is capable of drawing hundreds of fps (like the GTX260M should at low detail on just about any game) it will be idle a lot of the time

My currently totally hypothetical theory is that as the 150W adapter didn't stop BSODs, it's something to do with power circuitry in the mobo.

Unless, it actually is the 120W adapter not being enough as maybe one of the other influences on BSODs (Intel chipset & openal drivers) was causing BSODs on the laptop tested with the 150W adapter. In any case I'll try and test with something higher than 120W (Xotic was oos on the spare adapter I ordered so I gotta get another one anyway) but don't expect it anytime soon.

Only Asus will know the true state of affairs since I cannot believe they would not do power draw tests before releasing a laptop, at the very least they would have discovered this since. May explain why they've gone very very quiet [/conspiracytheory]

I will test this stuff on mine when I get some time, I still got a hundred things to install. If I have to run a permanent 10% underclock I don't really care that much, so long as it's stable, it's still 3 to 5 times the power of my old laptop.

 

That's the thing I really hate. The quietness of the manufacturer. I'm thinking right about now I want to bring this little problem to the attention of The Consumerist...

 

i sent out a request to anandtech and tomshardware, never got a response. +1 if you do. if it makes it out to the major reviewers, then it could put a fairly big dent in g51jx sales.

 

Wattage fluctuation is still being recorded in HWMonitor. I have noticed the stuttering in synthetics for both 3dmark06 and Vantage. I also notice this in general gaming right before the BSOD occurs. I am unsatisfied with under-clocking solution, also under-clocking does not solve the problem. I find it interesting however that synthetics refuse to BSOD, but playing MW2, L4D series etc. will BSOD and at random times (basically when the fluctuations get out of control and break a certain point, too dynamic to narrow down specific points of occurrence).

 

There is always an alternative, it just may not be the one we were hoping to be able to correct as individuals. I think we will have to come together for those serious enough to exact change as I am extremely displeased with the outcome. Here is to hoping for a fix from ASUS, but I sure as hell will not be holding my breath. It is time to act and make this problem public.

 

Just want to confirm that I just played through COD:WAW with the same settings as last night, and I BSODed within 15minutes. Physx was not the answer, and the power problem is the culprit for sure. My question is, How are we going to demonstrate this issue to ASUS to get them to fix their faulty product? Are we going to point them to this thread or seek out big blogs like the Consumerist? Or should we RMA all of our computers and detail it to customer service?

 

even if we get enough screenshots and data as proof in the wiki, that combined with bigger blogs might not be enough. it's possible they might say "it was designed this way" like the m15x, however bs of an answer it would be.

try playing CoD with 50% brightness (7 or 8).

 

I actually played at 4, and had the keyboard light turned off, which was different from last night which had the lights on and screen at full brightness.

 

Here is another forum post about the W7 throttling. Check out the utility mentioned in post 2:

http://social.technet.microsoft.com...f/thread/b4447bbb-fec4-457f-9ed2-d87965fd6592

 

what you could try, is to find the best combination of underclock/screen brightness settings that will let you run furmark + constant 2.4GHz CPU speed on a core in TMonitor.

call me lazy for not going back a few pages, are those maxed or limited settings?

RMclock is flaky on montevina, doesn't recognize calpella at all.

 

Alright i registered. Now i noticed this with the prime95. I tested 8 threads and i get a solid 1.6ghz in extreme turbo. Note i havent updated everything due to the fact i just found this page and my windows install is unstable(different issue...it was corrupted by me on accident). In my testing they all were 1.6ghz and then i did single core test and got 2.5-3.0ghz and it was jumping around in freg and the multi for the 8 thread was 7 for 1.6 and the single thread was 15(i think)-21. I dont know if this helps but thats what i got. thought it was strange that it was 3 ghz steady the whole way for both. Also i was only running prime95 nothing else and screen was max.


Now secondly, when we do find the cause and if it was asus's fault and they fail to do whats right can't we file a class action lawsuit? I mentioned this to another guy in an email already. I know we cant individually do it because this isn't an expensive enough issue per person but as a class action i think this is doable. If we must take this action i can get a lawyer for free because i am a marine at Camp Lejeune and i can get a free lawyer from base legal.

Anyways i rather not think about having to go this route but i thought i would mention both of these things.

 

running prime95 with one thread and setting affinity to one core = jumping between 20 and 21 multi.

try running furmark alongside prime95, it should flatline all cores.
http://g51jbsod.wikia.com/wiki/ASUS_G51J-A1_0x124_BSOD
if you have time, run the other tests as well and take some screenshots. the more we get on there the more evidence we have.

 

Remember that guy put together the hundred page PDF on throttlegate, lo and behold Dell release a fixed BIOS. Though I reckon they sold a few more Latitude 6400s and 6500s to businesses, that may have had something to do with it.

Unlike throttlegate however this BSOD issue is sporadic and not reproducible. The closest to a common experience across units is stuttering on the 3dm06 test (SM2 iirc)

Is the stuttering in 3dm06 linked with BSODding in games? ie.
* those who BSOD regularly, do you get the stuttering?
* those without BSODding at all, do you get no stuttering?
* if settings are changed that introduce/reduce the stuttering, does this correspond with a change in presence of BSOD or not. (thalanix did putting the LCD brightness way down directly correlate with increased scores for that test?)

Theories are useless without supporting evidence, and because of the sporadic nature of these BSODs it will take time to generate enough evidence to show clear patterns.

 

the stuttering is due to the throttling issue. i listed 5 of the most common times that each 3dmark06 test stutters on under test1 on the wiki, as well as a dozen runs and their scores+brightness settings. in short, lower brightness has a direct correlation with less to no stuttering and higher score.

 

About emailing places about this laptop and asking for help. I will email laptopmag.com. They have awsome reviews on places and i think they have high traffic too. Has anyone tried PCMAG?

 

it was suggested before but i don't think anyone sent anything. only 2 places i tried were anand and toms.

 

I think it was Joker who sent the M15x info to Anandtech too, and they're not interested...I'm REALLY disappointed.

 

i'll post link to this on newegg and zipzoomfly

 

I've put a link here and the wiki link on the boards at arstechnica.com....no responses yet, but we'll see what happens...

 

Might want to change the permissions on the wiki so only certain people can edit it. You don't know who is going to drop by there and decide that they don't like what's being said, and either delete it all or change it. Just a thought.

 

I'm the admin and thalanix has the highest priv level, we can undo any change, it can't be destroyed.

Edit: I will make a mirror too, once I have some time. Having a backup of it would be good, in case wikia.com has some kind of server failure etc.

Edit 2: I decided to do this now, so I have a local backup. I will keep it updated and if there is any reason for it I will make a public mirror.

 

Ok cool. That makes me feel a lot better. Lot's of good info on there.

 

Oh and +rep to thalanix for all his research and troubleshooting on this problem.

 

doh, slipped my mind that wikis get updated. my bad.

I second that +rep.

Has anyone looked at a GPU vdroop issue - does anyone know of a util that logs that?

I am not surprised the mags/sites aren't interested. The journos that aren't lazy are stressed off their nut. They'd only be interested unless you present a cut'n'dry case with clear theory supported with reproducible testing. Aside from that, they want to be nice to the companies that send them new stuff to test as that's what beings the majority of users to their sites.

Maybe we should put G73JH-A1 all though our posts so that search engines will direct people interested in the G73JH to here and they will wonder about Asus' support should they buy a G73JH and perhaps they will consider buying a different model of 17 inch monster gaming laptop with Mobility Radeon 5870 other than the G73JH

 

lol at bennyg's idea....Yeah, when I was deciding between buying this thing and a gateway fx series, I was told that gateway was the one with the crappier support.......so much for that being true.

 

and don't forget the g51jx with the gts360m. we wouldn't want this thread popping up when people google the G51JX-A1 reviews with the new rebadged nvidia 300M series

looks like the asus rep is taking a nice long holiday break: Last Activity: 01-04-2010 05:42 PM. maybe if they're serious about fixing it (gotta hope for something), my hourly updates on there will pay off.

thanks for the reps.

without digital VRM's, i don't think it's possible. none of the usual utils work for that, not even nvidia's system tools.

 

And thalanix, thanks so much for your research and reporting back...+rep for sure!

I've been playing Crysis for the past 2 hrs at high settings 2x AA and at 1280x1024. I've been using the high performance setting in Windows, the keyboard backlight is off, and physx disabled. I also downclocked to 500/900/1200 from 550/950/1375, and while it stutters a bit while loading the map in-game, I've yet to see anything indicative of the Windows message I was getting or the pause before the BSOD yet. I'm sure it'll show up eventually, but not bad so far....

 

Sorry this is probably a stupid post but...nm i figured out what it was. when all four cores are used it cna over cloak by one step to 1.7ghz and then if you use only one core you cna do more sets. But you guys said you say 2.6ghz with one core. I see it at 2.6-3.0ghz. Is anyone else see this? Is cpu-z miss reading?

 

alright i was doing a single thread max heat on prime95 and it jumps all over the place from 12 mutli-20multi??? anyone understand this? I will take videos of this when i go on my 96 Friday and post them. Also when i do the furmak and single core prime95 it goes to half voltage and then i do it without the fur donate it fluctuates from 7 multi-13 multi. Like i said i'll record this stuff for you guys later. I really dont get why the single thread jumps all over the place with nothing else besides firefox running.


*Edit Does anyone want me to run some tests tomorrow when i am at work if i got time? Also learned you cant do turbo on battery mode ^^

 

I was messing around with the computer trying to nail this down, and I got rid of P4G, disabled power service and then turned everything back to default. Noticed HWMonitor processor power went from a max of 45w to 55w max. Weird, thinking it probably just isn't reading correctly.

Update:

Nothing changed, however when I had the power service disabled I got a nice jump in 3dmark06 of nearly 600 points. Audio device becomes disabled after doing this, however the brightness is also low which may have to do with the jump in performance.

 

you mean P4G?

run the test again at different brightnesses to see if the 600 points is mostly due to brightness or P4G

 

FWIW I tested the difference between the lid lights on and disconnected and it makes a difference of all of 1W. I sat in the BIOS setup screen (after waiting, with the fan constant) and it went from 85-86W on my Kill-A-Watt to 84W with them disconnected. I had PM'ed thalanix about this earlier so I wanted to follow up and post the results on the thread.

 

It is due to brightness in this case, and the fact that the audio device was completely disabled alluding to a lack of power when all devices are working as they should (as that is when it freaks out the most).

 

I have finished testing thalanix' theory, and I got these results in W7.

Turbo Extreme, nothing open except RealTemp.

I rebooted after each run.

Top result: Backlight off: 10168
Second result: Brightness at 50%: 10159
Bottom result: Max brightness: 10146



I ran the prime95+furmark test and it flatlines the cores as expected.

Edit: My youtube vid was still 'processing' for over 3 hours, so there must be a problem with it.

Anyways, thalanix I was able to reproduce your results with furmark, it just required waiting 2 or 3 mins till the temps reached a certain level, and then turning the brightness up makes the CPU throttle like crazy, the power consumption drops to around 100W, and then turning the brightness all the way down has the opposite effect. It peaks around 116W at lowest brightness and still throttles a little but is relatively consistent. So it's not a PSU issue, but it's either a power management issue or a power distribution issue...

 

ummmm guys check this out
asus just added a new BIOS update! BIOS 207??? (what happend to 5 & 6?)

http://support.asus.com/download/download.aspx?SLanguage=en-us

though it says

"Fixed Splendid button can’t work"