[Fixed/Workaround] Asus G51J(x) CPU throttling investigation

any time it hits 933 when it shouldn't, it's throttling. BC2 doesn't feel it too much because of the multithreaded optimization. temps do get particularly high if you oc, i've been able to hit 80 on the CPUs and 101 on the GPU running on FHD.medium (585- clocks).

 

OK - but have You experienced throttling in BC2 when running in FHD/high settings without throttling fixed and on stock clocks? I'm planning to check today if the gameperformance is the same after runnig throttling fix, actually I'm waiting for my second backpanel to make some mods, also a copper mod too, but I'd really like to know if it's possible to avoid throttling with 150W PSU. I hope Ryzeki will make some tests

 

the 120w psu is more than capable for anything besides mass oc and furmark+prime95.

the way bc2 is designed (i see it using 4 threads), it will throttle on any clocks. idk how you can play it on high, it's choppy enough on medium as it is.

 

does the fix reset when the system is restarted, or is it permanent?

 

In my experience it resets when the system is restarted, when it comes out of hibernation/standby and it may also be reset if you change from settings. Simply checking and rechecking the box reactivates it.

Peter

 

thanks alot man, I was hoping for a method to make it permanent.

 

No problem - out of curiosity are you planning to get a 150W PSU? I would recommend looking over the reports of PSU shutdowns, info on the temp limit being disabled, and other G51J user experiences though before trying to have this as a permanent fix. It is highly recommended you only set this before launching an app/game that gets throttled and watching your temps/power readings as you use it. Unless we have some kind of limit on this (which I do feel is a good idea and argued for a few pages back) I don't think making this a perm solution is a good idea imho - especially on the G51J...

For reference, here is the documentation that came with the original app used to test the fix for this issue:

"3. What is Prochot.exe

Prochot.exe attempts to use the undocumented MSR 0x1FC bit 0 found on some Intel i7 series CPUs, to disable Bi-Directional PROCHOT for a period of 5 seconds as a means to check if indeed it is the cause of throttling. It should be used sparingly to try and prevent any possible hardware failure from overloading. There is no guarantee that hardware will not fail by Prochot.exe's use and therefore Prochot.exe is to be used at your own risk or not used at all.
"

Just FYI - I too would like there to be permanent - but still safe, but I don't think we're there yet and I don't know if we will ever be. (Permanent meaning you can have this on when you run any combination of apps and games at any time without always being there to watch things and still having the reasonable expectation that your system will still be on and fully functional when you get back).

Peter

 

The CPU is still fully protected when using any app that changes this bit in MSR 0x1FC. The Intel designed throttling and shut down temperature are not changed regardless of how you set this bit. Clearing this bit only eliminates outside sources from initiating throttling of your CPU.

In the testing I've seen so far, I think the Nvidia GPU temperature gets too high before the CPU temperature. I might be able to throw together a special version of ThrottleStop that monitors Nvidia GPU temps so with the ability for a user to choose a safe maximum GPU temp like 90C. If the temp went over that number then the magic bit would be set and once the temp went under that number then the magic bit would be cleared again for maximum performance.

This would allow you to use this new feature with a little less worry.

If the latest version of RealTemp is capable of reading the GPU temperature of your Nvidia GPU correctly then combining some code to create this new feature is a relatively trivial task. Copy / Paste programming is my favorite kind.

Give RealTemp 3.58 a try and let me know what it shows for GPU temps.

http://www.fileden.com/files/2008/3/3/1794507/RealTempBeta.zip

And let me know if you think the above would be a good idea. If so then I'll put it on the things to do list and hopefully early next week I will have something ready to be tested.

 

That's a fantastic idea! I hope it works!

 

well did my first test!!! It didn't crash! I had extreme turbo usb mouse keyboard and cooling pad and full brightness on screen and keys. It didn't crash like i thought it would. This thing definitely is being modded with the copper and back plate.

I hit 104 for gpu and liek 87 for cpu lol
http://img517.imageshack.us/gal.php?g=test1onfirelol.png

 

Another throttle without benchmarks of any kind: I was running a overclock of 560/1400/900 with Extreme Turbo on, and I could see about 20 throttling events in 15 minutes of Borderlands (1000ms readings showing CPU @ 933MHz in rivatuner log, this of course means many many more could have occured).

FYI GPU temp reached 96C max, most of the time 92-95 cycling due to the damn fan not staying on highest!!!

I am currently trying to wrestle with Nvidia System Monitor to set up profiles where it reduces clocks with increasing temperature to keep it at 93C. This is until somehow the fan is able to be turned on highest at a lower temp threshhold...

I've set one up where it overclocks GPU core & shaders up to 15% as utilisation approaches 100% and it works OK.

I was doing it this morning and smelt burning, was crapping myself... then I realised my brother had turned on the central heating after 3 months of summer The temps of this thing are playing on my mind too much.

 

at least yours didn't hit 104 in furmarc and prime95 lol. When i go home on leave i am going to do copper mod and everything. I hope it'll be at low 90s with a prime95 furmac test. I am also running asus clocks lol. I couldn't run standard nvidia clocks and run furmarc. I really want to over clock this but i can't until i do copper mod.

EDIT: I honestly recommend doing that copper mod. It'll save the life span of this laptop and allow you to push it better.

 

it looks like your GPU hit 105 and then started underclocking, or just cutting performance. prime95 + furmark is only a dream for this laptop.

 

not sure what you mean but doesn't BC2 get to these temps. It uses all four cores and heavy GPU so it should produce similar temps right? I am on asus clocks too. If i was at standard clocks i would easily push the 108 nvidia limit. Granted my room is hot as hell because the marine corps has the heat on....its 77 degrees in my room.

 

105 is nvidia's limit. look at your second (left) screenshot, the GPU usage drops even though furmark should have it above 90%.

on 585- clocks, bc2 gets me up to 100 on the GPU and 78-81 on CPU according to hwmonitor, usually staying around 95 and 80.

 

whats your ambient temp? I am assume mine is 80 degrees Fahrenheit since its enclosed area.

BTW i idle at 63 cpu and 67 gpu

 

atm, around 66-70F (18-20C). only hit 100 on the GPU once or twice.

 

It could be the operating environment, or the new thermal paste, or both, but I see idling temps more around 50-60 on the cpu and 60-65 on the gpu.

 

I assume whenever they turn the AC on and its not 77-80 degrees my temps will be better. 10 degree difference is a lot ^^ When i go home on leave i'll be doing the mod and i'll be in a similar temp range as you and i'll record results and post. I am so excited ^^

BTW i havent done the new paste yet. Its at home so i'll be doing new paste and copper mod and drilling the back panel when i am on leave. I'll take new results prior to the mods. Also its only a 24-48 hour tops time period for the paste to break in. I got that info from AS5 site.

EDIT: I was taking a second look at those screenies and i think the thermal limit might be like 101C. I go from 100%ish to 67% or whatever and it stays there as it inches to 104 degrees

 

Hey Unclewebb, that would be great, thanks. It's not so much the temp of the GPU or CPU for me though but the PSU - thalanix was reporting that his system would shut down after doing the prime95+furmark tests for a few minutes and he'd have to unplug and re-plug the PSU to get it going again. I know that is a border case now, but it can be easy to overstress these underpowered 120W PSUs if we leave the system running new and multiple intensive applications at the same time - and if there is a continuous too high power draw over time it can wear out the PSU even if not causing an actual system shut down. With this bit fix combined with normal, though intense, usage I can push the PSU up past 137W and the prime+furmark test had me spiking up to 165W - and this is with a G51Jx that already has a lower heat (and thus power) limit than the G51J.

Thalanix and I had a discussion about this a few pages back now, about my suggestion of using a combination of the CPU and GPU temps to reset the magic bit and then re-engage it when the temps were low enough - and using that as an approximate measure of wattage that we can customize depending on your load (which we could calibrate based on power meters and estimating/observing what other power we draw and how much of a ceiling above that we want to leave without excess PSU stress - and raise if we got a 150W PSU for example). I know it's not perfect, but I think if it's not too much trouble for you to put together such a system then that's a great idea and could help put my mind at ease - so if you can do it then that would be greatly appreciated . There's also the fact that the temp monitor/cut off would act as just that and help prolong the service life of the components while still giving us the option to push things if we really want to - and that would also be very nice .

Thanks,
Peter

P.S. It's late and I'm not doing anything particularly stressful for the system at the moment - but RealTemps looks accurate - it matches up with HWMonitor. I'm currently idling at 56*C with my stock thermal paste and an ambient temp of 19*C though it shoots up quickly with load as expected. I reached the maximum I've let the GPU go after 14 minutes at 92*C when running the stress test with FurMark+Prime95 with a continuous 144W AC power draw and spikes up to 165W - Probably someone with a G51J would have been well over 100*C and probably. My CPU never went over 73*C. I was running my GTS360M at my comfortable and stable 615/2000/1600 overclock.

 

http://forum.notebookreview.com/showthread.php?p=5988399#post5988399

In this discussion I have talked to Soviet Sunrise and he was saying that the psu can put out maybe 140+watts if you keep it cool. He said on his psu he put a heatsink on it to cool it off so it'll last longer and work better. If you put a heatsink on the psu you should beable to amke it more stable and squeeze a good amount more juice out of it. So in thalanix cause he should be able to overclock the 260m and do prim+furmark and not crash as long as the psu is cool. So thalanix might want to get a heatsink and mount it. I'll do the same whenever I want to overclock this thing. It'll be cheaper to get a heatsink and mount it then to buy a new psu. Also Thalanix and Soviet have there own opinions on temps and how long the system will last at 90C for gpu and 80s for cpu. If you looked at my temps they were horrible. I am curious of what uncle webb thinks about the life span of the i7 with these high temps. I am for sure doing the copper mod and back panel to the extreme. I also will do more of what he did over the next year....it'll take awhile to due this lol

http://forum.notebookreview.com/showthread.php?t=416414

 

Should I get a 150W adapter to get the most juice out of the G51Jx? If so, does any random 150W adapter from ebay work? Or should I get an adapter with a even higher wattage?

 

DCMAKER: My standard response about CPU temperatures is that if Intel thought their CPUs were going to self destruct at 90C, why would they set the thermal throttling point to 100C? They build their CPUs for long term reliability. The thermal shut down temperature that they have set is in the 125C to 130C range. Intel obviously has plenty of confidence in their CPUs and that confidence has to be based on warranty claims. If either the thermal throttling temperature or the thermal shut down temperature were causing their CPUs to die then they would lower those limits in a heart beat.

80C and 90C peak core temperatures scare the hell out of us users but those are our own personal fears that are not based on reality. Intel builds some great CPUs that are very capable of thermal managing themselves so that they run at a safe temperature.

CCz_Cataphract: Hopefully within a day or two I will have a new version of ThrottleStop available with this new feature ready to test to better manage the magic bit in MSR 0x1FC.

 

A PSU is one of the most durable components on a machine but it is not untouchable. It will output near it's true maximum output if kept cool but the primary reason why it should be kept cool is for preventing damage. Heatsinking the PSU will not make it squeeze more power out of it.

Frequently taking your GPU up to those temps will slowly degrade the soldering on the daughterboard. The effects of high temperatures and cycling is what will bring it's demise. It's that simple. In your case with your G51, you need to watch those temps more closely as both the CPU and GPU share the same heatsink.

I dare anyone here to ask anyone on [H]ard, OCN, or other dedicated enthusiast sites to run their desktop or notebook CPU's into the 90's, and that includes the i7's. Perhaps someone will be willing to sacrifice a 920 and run an infinite loop on IBT until it breaks providing that it stays within the 80's and 90's. The lifespan of any silicon component is shortened at those temps. Just like PSU's, CPU's are very durable as well and can last well over a decade of service, but if those temps aren't tamed, predicting when the CPU will collapse will be the same as asking when you need to shop for a new CPU.

 

The reason Soviet recommends lower temps is not because of the CPU but rather the GPU and its memory. On most systems it is the same heat pipe cooling both units and its often the hotter cpu at the begining of the pipe. The issue then becomes that the higher temp of the heat pipe from the cpu has the ability to heat up the GPU hotter than it normally would operate. what suffers the most is the gpu memory from the heat. However on the JX its not an issue as there is a seperate heat pipe for the GPU vs the CPU.


Edit.. I C he posted as I was typing..lol

 

Hi there,
I am such a newbie and I have a lot of questions that I want you guys to discuss with me:
First of all, I gonna a new laptop such as G51- X3 with core i5 , GTS 360 but I don't know where I can buy this , I'm living in New York , so does any one can help me?
Then I heard about some problem issues that appear with this laptop , is it harmful to the laptop?
Is it too hot when I play game?
There is a lot of different prices of the laptop , in amazon is 1149 for X3 , I don't know where that I can believe to buy the new laptop
Thanks so much to answer my question.

 

So with this fix Furmark+Prime95 takes the 260M up to 105 degrees? Has anyone tried this with the 360M?

 

it hasn't gotten that high for anyone else so far. It has only been me. My ambient temps are like 80 degrees Fahrenheit, which is offer then most peoples.

 

Tried furmark extreme burning test 1920x1080 4xMSAA+prime95, GPU OCd to 530/800/1325, mouse, cooling pad, brightness (40%) for screen and keyboard on.

Max temp for GPU: 102°C
Max temp CPU: 86/85/86/83 for the cores.
Time- about 10 minutes

Temperature for GPU was 102°C max but then went down to 101 and hovering near 100 every now and then.

I have been using the small OC without much problems, and during games my highest temp so far has been 93°C so far.

Due to the temps, I think I will not push my system further regarding clocks. Maybe I will try modding, but I am not sure... I am more inclined to sell my machine than bothering to mod just for a small increase in performance.

 

That's pretty decent. 102 is a little hot, though. My 8800GT desktop gets to 95 before the fan really starts to kick in. I'm sure a good cooling pad would help.

An un-OC'd 360M will probably run a bit cooler. The reason I was asking is because I think in the future we'll be using our graphics cards for more than just gaming. Project like Luxrender (a 3d rendering engine) and Adobe Mercury are going to make it important to be able to use CPU+GPU without generating too much heat.

 

out of curiosity, can anyone running prime95+furmark and not hitting the 105c threshold on the GPU look at GPU-z to make sure the GPU is still going full speed ahead (after 20 or so minutes)? the card slows itself down at 105c, but that 105c includes a bias (between 5 and 10c usually, the g51vx gtx260m has a bias of 8).

if it still shows it as going 90+% usage, run furmark without AA and at a smaller resolution and look for a drop in frames.

my point of view with the heat wasn't that it's ok, i only mentioned that the g92b hasn't been out long enough to make a good estimate of how long it will last/impact of heat.

 

So basically you want me to run the test for 20 mins? I don't think I can reach 105°C since I reach 102°C fairly quick.. withing 2 minutes and it just stays ther the remainder of the time I leave it on. It actually hovers and sometimes drops to 101°C or 100°C.

I will try later today.

 

run it in extreme turbo too. thats helps raise the heat up too.

 

Yeah I already did on my first run.

 

i'm only interested in the GPU usage or drop in fps. if it only takes 2 minutes, then that's all you need to run it for.

 

Do you guys have manual fan control? If it goes from 102 to 101 it sounds like there's a fan kicking in (or being turned up). I'm wondering if that could be forced higher, or at least forced higher earlier.

 

or thermal throttling on the GPU

 

Ooohh... good point. Shouldn't GPU-Z tell you exactly what's happening, though?

The fan kicking in is what happens on my desktop. It gets to 95ish and then the fan starts making more noise than any computer has a right to make. That's what made me think fan.

I'm planning to get a G51JX soon (in the next week or so). I'll be helping out where I can.

 

on desktop cards the fan can go at any rpm (on my evga 8800 and now xfx 4890 it did), on a laptop it's set states. GPU-z will either show a drop in usage or the application (furmark) will drop in performance.

 

unless you get a really cheap version of a card, the Galaxy 8800GT I put in my bro's desky ~2.5 yrs back which only has the 2-pin fan connector which means its off or noisy-whiny.

btw it basically has the same chip as what I have in this now

 

Hi, I few a few questions about the G51j and throttling... i've used the MSR fix, and it worked... but it sends my GPU to temperatures around 101-102 C. I bought a Zalman NC2000B cooler and it has helped decrease temps by 5-7 degrees (under load via gaming *BC2/Crysis/SCII*) to around 95-96C range for GPU....is it that bad for the GPU? From what i've read its not good but I still have about 1.5 years warranty left so im not too too worried about that. The big question... would disabling throttling void my warranty?


Running: Bad Company 2, Crysis, Starcraft II Beta

Playing these games at 101-102C GPU (no cooling), i've noticed no artifacting or chop for GPU stress as of yet.. but after reading posts, I became worried about the Temps... the thing is that even with throttling enabled I still hit 95-97C for my GPU anyways so with the Zalman cooling im basically back at pre-throttle temps under heavy load. Other than active cooling *Zalman* Is there anything I can do (not including modding since i still have my warranty) to lower the temps? Or should I just game the crap out of it till it breaks and use my warranty up?

Any response would be greatly appreiciated, and btw thanks for the great work and discussion over the past few months, reading these posts has helped me alot.

 

I would also really like to know if temperatures in the high 90s for the GTS 360M is OK since the GTX 260M goes at 100 something for some people?

 

you can do this mod. I am going to do this on leave or right after depending on how much time i got.

http://forum.notebookreview.com/showthread.php?t=414147

http://forum.notebookreview.com/showthread.php?t=416414

EDIT: Can someone add this to the top of the thread as a recommendation

 

Well, my last G51vx always played at 102C and never had problems with it. My current lap goes usually up to 92 and 98 depending if I have an overclock and I have used those temps for several hours, sometimes on a daily basis, without a single stuter, crashes etc. I think you will be fine and as you mentioned, you still have a long warranty for any issues that come.

Hmmm what is your ambient temperature? Some users have low temps but sometimes they also live in places with 20ish Celsius ambient temp. I say that high 90s are not a good indication because the card itself is supposed to run cooler. 90 itself would not be a problem but higher might mean something, you can try calling asus or your reseller for advice.

You are better off making a new topic or bumping those alone, as this thread is exclusive to the CPU issue and not exactly about temps or mods for that. A new "Mod" G51 thread would be better.

And are you STILL an angry unfortunate blah blah despite all issues being "fixed"? haha Come on, even with the BSOD and throttle, I have been playing quite a lot!

 

the reason why i say adding that onto the top thread is because if you play hl2 or BC2 it uses multi-threads which will cause this puppy to run hotter then hell. My temps were high before but after undoing the throttling it has raised temps alot more because the cpu can push itself higher....it would be a good idea just to have the link so people know this possibility exists.

 

Personally I haven't noticed any performance difference with/without throttling workaround enabled in BC2. Would probably be a good idea to run some benchmarks and see if this game actually is affected by throttling at all.

 

I finally added the ability to toggle the magic bidirectional PROCHOT bit to ThrottleStop 2.00 but I haven't done any Core i7 testing yet. Any volunteers?

http://forum.notebookreview.com/showpost.php?p=6077775&postcount=533

I also added customizable profiles so you should be able to set up two different profiles with and without the PROCHOT bit enabled and then switch profiles back and forth based on CPU temperature. After I get some feedback I plan to add the ability to control this bit based on GPU temperature for my Asus G51J friends.

 

Im looking to purchase a G51jx-x1 from newegg, but i see there is some throttling issues. Is this severe? And is there a fix from it directly from Asus? Or are they working on it? And is it confirmed that they are working on it? Or do i have to do the fix in the thread?

 

The throttling is fixed pretty well now (see one of unclewebb's posts in this thread), but the heat is still crazy. I saw the G51J I use hit 102 Celsius on the gpu last time I played with it.

 

Hey UncleWebb,

Was originally going to reply to your PM directly, but though it might be better to post the results publicly and send the PM with a link here as well - maybe someone else will give some of their own feedback as well.

I tested the app with both Extreme Turbo enabled (small CPU overclock) and not (via the ASUS Power4Gear Utility) and it definitely did do the Prochot bit change, I could see FurMark going from no Intel Turbo Boost to full Boost as soon as I unchecked the option under Advanced.

I can confirm if I set the multiplier lower on one of the other profiles it does lower it, and going back to the Performance Profile it raises it again (if the multiplier is set in it).

The system appeared to respond the same if ThrottleStop was enabled or not in terms of CPU usage and Turbo Boost while idle. If I set Battery Mode to 7x mutiplier then it works much like the ASUS Battery Mode in the Power4Gear Utility.

I noticed my CPU temp was at around 58*C-60*C as it idled, so I set a DTS Alarm for 41 and set it to go to profile 4 on hit. I then saw the program go back and forth between them.

Then I set the profile to 3, which only differed by having the Proc Bit checked rather than unchecked. Worked great with FurMark running and a DTS Alarm of 35. Then unset the bit on profile 3 and it stuck in that lower power mode as expected. Reset the bit and it went back up to full performance with Turbo Boost enabled within 2 seconds (and flipped back down right after that).

So it seems to work quite well. Because of the GPU stress (91*C after 3 minutes, no overclocking) I hit 134W during the test (Note: No Extreme Turbo, Wifi or USB devices connected). The CPU Core 0 never went over 70*C and 1 not over 66*C, 2 and 3 peaked at 69*C and 70*C respectively.

Was it tracking the max temp on all the CPU cores by the way, or just 0?

It certainly introduces a throttle like ASUS had, but at an adjustable temperature for the CPU. There's only a tiny cooldown period from what I observed - it did keep my PSU at 134W and under, which is safe, but if I left FurMark going for long enough (or if I had a G51J) then I think I would eventually have a system shut down as the GPU heats up.

Definitely looking forward to the version that includes GPU monitoring as well

Thanks,
Peter