The ThrottleStop Guide

UncleWebb or others, PLEASE HELP!!

I have been having issues with my integrated GPU HD 4000 throttling down very excessively when playing a game for any longer than 30 seconds or so. FPS would be at a comfortable 30-40 fps then drop considerably to 14 or so and hover there, spiking up once in a while. A common sign of throttling. However, I cannot figure out for the life of me how to stop it, or what is exactly causing it, and I am failing at getting throttle stop to show me, however I have noticed some very definite patterns and things, and taken logs, so hopefully you guys can help me decipher the information.

First, my system:

Fujitsu T902 tabletPC
13.3" 1600x900 wacom touch/pen screen
16 gigs of ram ( I do video editing, and effects programs)
512 gig SSD
integrated HD4000 GPU ONLY! no dedicated GPU
and the best ivy bridge dual core I could customize in it: i7 3520M 2.9-3.6ghz

using throttle stop 5.00 for ivy bridge. (at least trying to figure out HOW to use it)
Now, I may have an inadequate cooling system for gaming on here, but I am beginning to wonder if that is in fact the cause of the throttling or not...

There are some very basic things I am very confused about, despite reading as much as I can find on the issue. Here they are:

1.) How is it that my integrated GPU clock cycle is throttling down from 1250 mHz to 650 mHz, (monitored by GPU-z) but the CPU clock cycle remains unaffected? (CPU shows it remains at max or near max speeds during the throttling according to both throttlestop and RealTemp running at the same time...) I thought integrated meant that the CPU clock cycle = GPU clock cycle, and if one throttled the other would follow. But it is clear I was wrong, as the GPU clock cycles drop significantly while the CPU stays around 2.9-3.6 !!?!

2.) When I unplug my computer it tends to run MUCH BETTER. ?? When plugged in and not running throttle stop it stays at a LOW 650 much more consistently and spikes only once and a while up to where it should be, but if I unplug it, the GPU jumps back up to around 14 watts power and clock cycle increases staying more consistently in the high end (1250) and only dropping to 650 once in a while. That is with the even with power saving methods identical between plugged in and unplugged in windows options... Usually, I read about the opposite! Many report dropped FPS when unplugged due to a power limit. WHY ON EARTH would I get more throttling when plugged in????

3.) My CPU is supposed to be getting 35 watts, this is it's TDP, right? So is that shared between the on board GPU as well, which means 14 watts or so is normal for the GPU to be recieving? Thats what GPU-z reads it gets when its performing at its best and not throttling... when it throttles, it drops to 5 watts or so.

4.) I noticed when I opened up the TPL tab in throttle stop, my system displays Limit #1 and Limit #2: I haven't found anything about this chart on the documents... Mine says under limit #1 that turbo time limit is 28 seconds. Is this normal?? What does it all meaaaaan??

5.) MY TEMPS: GPU floats around 84 degrees once it's been under load for a while. I can't tell if the system is throttling when it starts to go over this or not. Also, I am unsure of how to use throttlestop to show me how I would know if my system is throttling due to HEAT, POWER, or CHIPSET(or other manufacturing BS that is way over my head)


THis testing has been done while running skyrim (staring into the distance in one spot), windowed with fraps showing FPS, and throttlestop, GPU-Z and realtemp reporting off to the side.



ANY help and advice would be so awesome. I am able to take it in to get thermal paste changed and heatsink checked, cause it is still under warranty, but I need to know if you guys would even reccomend it, if it seems my system is not actually getting too hot.

thanks for reading my long and troubled noobish rant, let me know of specific test I can run to give more information if you need it... Also, if it helps I can supply logs. Just let me know which ones would be useful while running what settings, etc.

 

I am guessing that the manufacturer knows the thermal limitations of that setup and is throttling it in BIOS.

Hope I'm wrong.

Good luck.

 

What exactly would this mean if that is the case?

Isn't Throttlestop designed to work around that very issue? I'll be perfectly happy to run throttlestop 100% of the time if it could somehow counter that.
How would you go about confirming such a thing?

 

Software cannot counteract anything that is hardware dependant - what ThrottleStop does is allow you to run at the maximum possible (for your system: not MAXIMUM) without throttling.

 

ok, well, I really appreciate your help!

Is there any tests I can run to test your theory? What causes you to suspect this?


edit* ALSO, can anyone confirm that idle temps of 50 degrees celsisus and gaming temps of 86 degrees + celsius is normal for ivy bridge mobile processors??

If not, I will try first having them bring those temps down using hardware fixes: thermal paste, check heatsink and fan, etc

 

When you click on the ThrottleStop TPL button, are the items in that window locked? The bios decides if these are locked or not. If they are not locked, you might be able to increase the turbo power limits and get a little more out of your GPU. I haven't done any hands on testing but I think the power register controls both the Intel CPU and Intel GPU.

 

thanks for the reply UncleWebb,

sadly TPL is locked. Anything else I can try??

 

What values are reported in the TPL window? Post a screen shot if you can.

Manufacturers are allowed to use values that are lower than the default Intel limits and they can choose to lock the power limit register if they want to. That might be what is holding your GPU back.

The only way to unlock this feature is with a modified bios which probably isn't available.

The 28 second value is how long the CPU can operate at the short term power limit before it is throttled back. Long term it operates at the lower power limit. 28 seconds is the Intel recommended default value. This could be adjusted higher if the power register was not locked by the bios.

 

here it is... do those numbers suggest that Fujitsu is keeping things capped down? I am not sure what the numbers really mean, and I got nothing to compare it to.

 

@stoneseeker : I think you should check "Set multiplier", otherwise TS is doing pratcically nothing. Someone correct me if I'm wrong.

GL

 

Your absolutely right, I just don't have it turned on there, and I only opened it to take a screen capture of the TPL.

When I tried using it, I clicked 'multiplier', and then also clicked 'disable turbo'.. this helped keep heat down while gaming and kept the CPU from needlessly going over 2.9 ghz. My problem is not CPU throttling, but onboard GPU throttling. I was hoping TS would have an option to prevent GPU throttling too...

All I can think to do at this point is take it in to get thermal paste re-applied, buy a cooler master ergo stand, and have the heatsink checked. Seems Fujitsu may have built this computer with an inadequate cooling system for the i7. Seems shoddy to me for them to do that, especially when I spent $3200 on this thing. (The Wacom pressure sensitive screen on this big of a tablet, this powerful comes at a premium!)

 

ARK | Intel® Core

The Core i7-3520M is a 35 watt Dual Core processor. Intel recommends that the long term Package Power Limit (Limit #1) be set to 35 and the short term limit (Limit #2) should be set to 25% higher than Limit #1. Your short term limit is 20% higher than your long term limit so that may not be perfect, but it is not completely unreasonable. At 28 seconds, the time limit is set just as Intel intended

You could try using ThrottleStop and drop your CPU speed down even further. At 2000 MHz, it will consume less power which in theory, might leave more power available for the Intel on die GPU so it can run at full speed. Most games are more GPU dependent than CPU dependent.

If Fujitsu had left the power register unlocked in the bios, you would be able to use ThrottleStop to increase the power limits and possibly run both your CPU and GPU at their rated speeds. They probably locked this to avoid any overheating issues.

The Intel datasheet shows that the 35 watt Core i7 Dual Core CPUs can have the long term Turbo power limit increased to a maximum of 48 watts and the short term limit can go as high as 56 watts. I think that would make a big difference to performance when both the CPU and GPU are active.

Unfortunately, with the turbo power register locked, there is nothing ThrottleStop can do.

 

OK, so please correct me if I'm not following you right, but in my case, I have to lower the power used by the CPU because the system will only sustain giving enough power to have the CPU and GPU full juice for a set period of time??
Or is it simply that, the system will reduce power when it senses it is getting too hot? It seems like the latter, because the CPU remains relatively constant, but the GPU clock cycles drop and respond so dramatically, but don't always drop at the same point in time.

If I boot up a game and run GPU-Z it will always show it drops to 650mhz when the temperature gets up to around 84, 86... which happens very quickly. At times I have seen it get up to 94, although thats rare it lets it get that high.

Another reason I suspect that temperature is the culprit of the GPU throttling, is because even when I am just surfing the web and playing some tunes, the laptop has some random spikes in temperature at times of up to the high 70's, and sits at idle in a cold room of around 50-54. It just seems to me that my computer is running around 10 degrees hotter than it should, but it is really hard to find Intel's optimal running temps. 105 that they list is their ABSOLUTE MAX. That is, if it hits 105 or past, the computer will shut down. So I don't think that it is supposed to get anywhere near that, but I can't get any confirmation anywhere on the internet about what temp the processor should be idle at, and what temp is normal under load for a sustained time.

do you guys know where to get those numbers? Maybe I should try calling Intel?

Oh! and why would the laptop run the GPU better when unplugged?! Especially if we are suspecting inadequate power??

Thanks for all your help so far! I'm learning a lot.

 

Calling Intel for some temperature information is the last thing a person wants to do. They usually transfer you over to the "double speak" department and will endlessly talk in circles about what is an acceptable temperature. Core temperatures vary depending on the type of CPU load and the environment you operate them in. The sensors Intel uses are not 100% accurate temperature monitoring devices so Intel doesn't like to talk about specific numbers when it comes to temperatures. All they say is this:



The thermal throttling temperature for most 3rd Generation Core i mobile CPUs is 105C. Intel says that is the "maximum safe operating temperature". Intel CPUs are not supposed to shut down at this temperature. Intel sets the shut down temperature on the majority of their CPUs at 25C above the maximum safe operating temperature. 105C + 25C means Intel would set the shut down temperature to a peak core temperature of approximately 130C.

This scares some laptop manufacturers so they decided to go into the bios and create a shut down temperature lower than the official Intel shut down temperature. If your CPU shuts down and your laptop turns off at 105C, that has nothing to do with Intel.

Intel CPU Turbo Boost is mostly controlled or limited by the values that ThrottleStop shows you in the Power Limits window. For your Core i7-3520M, your CPU is allowed to consume 42 watts for up to 28 seconds. After the 28 second timer runs out, it will be limited to 35 watts. The way a CPU gets itself back under this power limit is it reduces the amount of Turbo Boost which slows down the CPU so it consumes less power.

Intel calls this limit the Package Power Limit so I have always assumed that the package includes both the Intel CPU and Intel GPU. I have not done any hands on testing of this though. Edit: See bottom of this page for my testing.

Turbo Boost can be limited by power consumption or core temperature. In your case, I am not sure what is limiting your GPU. That's why I suggested for you to do some testing. If you lower your CPU speed, it will reduce its power consumption which in theory would leave more power available for your GPU. This might let your GPU continue to run at full speed for a longer period of time.

When running on battery power, your GPU probably gets locked to the lower speed to reduce power consumption so there won't be the spikes of the GPU speeding up and slowing down.

Edit: I did some testing on my desktop Core i5-3570K. It uses an Intel HD4000 GPU and the power limits that ThrottleStop shows you definitely control the CPU and GPU. Using just Furmark in a small 640 x 400 window, power consumption was less than the 60 watt long term limit that I set so the GPU ran continuously at 1150 MHz. When I started 4 threads of Prime95 while Furmark was still running, power consumption was over the limit so the GPU was throttled back to 650 MHz and Intel Turbo Boost was disabled leaving the CPU running at the default 34 multiplier. Time to reboot and go into the bios and jack up the power limits to see if I can make the CPU and GPU run at full speed, even when fully loaded.

Edit #2: In the name of science, I decided to push my Core i5-3570K to the limit.

I raised the turbo power limits up to 120 watts so the CPU would be able to operate at its full potential. Here are the results.

http://img546.imageshack.us/img546/2973/furmarkstresstest.png

Even with the Intel CPU and Intel GPU at 90C, and package power consumption at 102.8 Watts, the CPU continued to use full turbo boost and was running with a steady 40 X multiplier and the GPU was also allowed to run at its full rated speed of 1150 MHz without any dips in speed. In my previous test with the package power limit set to 60 Watts, the CPU was only able to use the default 34 multiplier and the GPU was throttled back to 650 MHz.

If you ever find someone to modify the bios for you and unlock the power register, you can use ThrottleStop to get a lot more performance out of the on chip Intel GPU.

 

WOW. This is by far the most helpful and informative post I have read in any forum in a long time! Thank you so much for your testing and time!!

OK, So I think I have a far better grasp of the situation now. I am still not sure if I am hitting thermal or power constraints everytime, but I'll do some testing with a lower muliplier to see how it performs... I think I may have been witnessing the time limit on occasion, but I also think the GPU is throttling down when it gets over heated, which is hard to say when, but I think its when it gets over 85 degrees. I may be wrong, but it would be at least worth it to have the repair guy do a thermal paste and heatsink job just in case, and see if it makes a difference.

I know the GPU can run at 1250 or 1200 even when the CPU is not in turbo mode, so one thing I could do is stress my system in the same manner you stressed your own the first time: do a GPU test that keeps it running at well under the power limit, (like at 20-25 watt) and see if the system still throttles after a while. If it does, then we know its thermal, does that sound right? I need to know how you knew that when you ran your furmark in a 640x400 window that you were under your power limit. Does TS tell you somewhere exactly how much power your cpu/gpu is consuming at any given time? Did the furmark test have that information somehow built in? If I can nail down how much power I am consuming during a test, I can see how it performs when I know I am eleminating one cause (in this case lack of power) does this make sense? Or is it more complicated than that?

If you could let me know of a VERY specific test I should run to help narrow it down I would appreciate it.

Also, you mentioned being able to modify the BIOS, or find someone who could. Who would be typically approached with this? Rogue Programmers? lol Do any computer repair places offer such a service? I am sort of assuming I would be voiding the warranty if I did do it though... perhaps I could approach Fujitsu and request an unlocked version of my BIOS. haha I doubt I would get anywhere with that though.


thanks again, I'm gonna run some tests!

 

I have been meaning to learn some more about this stuff for a while now so my testing today has been beneficial for both of us.

The power consumption that ThrottleStop reports is the data that the CPU uses to determine if it should be using Turbo Boost or not as well as how much Turbo Boost. I will try to post a few more pictures of some of the tests I did later tonight so you can see what throttling looks like when the CPU or GPU gets throttled back.

Many laptops with a separate Nvidia or ATI / AMD GPU can throttle things back based on the temperature of the GPU and 85C is a common temperature to do this at. Anything is possible. It might be GPU temperature or power consumption or maybe both that can trigger throttling. Some more testing should make things clear.

 

unclewebb, shouldn't your i5-3570K have Lock Power Limit Register unlocked?

Also, are all current ThrottleStop 5.00 installations having the TPL window display Limit #1 and Limit #2 rather than Long and Short, respectively? I ask because I have Limit #1 Package Power Limit set to 45W while Limit #2 has 56W, although Turbo Time Limit is set to 28s on Limit #1 while on Limit #2 is set to 0s.


BTW, congrats on getting a 3rd gen Core i CPU. Given that you mentioned how 2nd and 3rd gen CPUs shared the same features, other than 3rd gen's more granular multiplier control, will any future features be able to trickle down from 3rd gen CPUs to 2nd gen and lower?

 

I wish. This Asus desktop board locks the power limit register in the bios so ThrottleStop can not be used to change any of the power limit values when in Windows. Lucky the bios has plenty of options to control this but I need to reboot anytime I want to make a change. Locking the power limit register is kind of a lame thing to do on an "enthusiast" board. I renamed the headings to Long and Short which is a little more meaningful than Limit #1 and Limit #2. That picture in the screen shot is actually of the next version of RealTemp. I plan to add adjustable time limits to RealTemp for the CPUs that have an unlocked power register but I got kind of side tracked today with my testing program.

For all of these tests, the long term turbo power limit was set in the bios to 60 Watts.

http://img29.imageshack.us/img29/5728/furmarktest1.png

The first test shows Furmark running and the CPU has 1 core fully loaded to help stabilize the multiplier. Furmark barely puts any load on the CPU by itself so I had to run a separate program to load the CPU. Power consumption is at 51.1 Watts which is under the 60 Watt long term power limit. The CPU is being allowed to use full turbo boost (43 multiplier) and the GPU is also running at full speed which is 1150 MHz.

http://img545.imageshack.us/img545/2806/furmarktest2.png

For test #2 I increased the load on the CPU. The average C0% is now 53.4% which shows that 2 of the 4 cores are being utilized. The CPU spreads the load around so all 4 cores are doing some of the work. The GPU continues to run at full speed, 1150 MHz, but the CPU has begun to throttle to keep the power consumption under the 60 Watt limit (58.7 Watts). The CPU is now running at 4000 MHz instead of the previous 4300 MHz.

http://img547.imageshack.us/img547/6294/furmarktest3.png

Picture 3 shows that the C0% is at 78.9% so 3 of the 4 cores are being fully utilized. The GPU continues to run at 1150 MHz but the CPU has been forced to throttle further and is running at only 3600 MHz. Package power consumption is at 59.8 Watts.

http://img255.imageshack.us/img255/6981/furmarktest4.png

The final test shows a similar C0% to the previous test, 77.0%, but for this test, I changed the type of load. For the previous tests I was using my Load Tester program which should be available in the RealTemp download at TechPowerUp.

Download Real Temp 3.70 | techPowerUp

For the final test, I exited 1 instance of Load Tester and I ran 1 thread of Prime95 instead. Prime95 works the CPU much harder which forces it to try and consume more power but the CPU still must try and maintain itself at 60 Watts or less. In this picture, the CPU has completely disabled the CPU Turbo Boost feature so the CPU is running at its default 34 multiplier. The CPU will not be throttled below its default speed so it is time for the CPU Package to start throttling the GPU. The GPU-Z graph is starting to show some choppiness in the GPU Core Clock graph. Instead of a steady 1150 MHz, the GPU is starting to bounce around between 1050 MHz and 1150 MHz. If the CPU load was increased further, there would be more and more GPU throttling as the CPU tried to stay under the 60 Watt Turbo Power Limit.

If power consumption is greater than 60 Watts, the CPU Package will disable CPU and GPU turbo boost but it does not throttle either the CPU or GPU below their default speeds.

It's still possible that the CPU or GPU could be throttled on top of this based on some temperature value set within the bios. At least the above pics show how an Intel 3rd Generation Core i CPU is supposed to work.

 

View attachment GPU-Z Sensor Log - multiplier x21 on battery.txt

Hey UncleWebb,

Here is a GPU-z log file of me playing Skyrim again while running throttlestop set to a multiplier of x21 (roughly 2.0ghz) like you suggested.

It certainly ran a lot better! It maintains the full GPU clock speed of around 1250mhz for roughly a minute or so and then starts to periodically drop to 650 mhz but only for 3 seconds or so, and then soars back up to 1250, and after 5-10 seconds drops to 650 for 3 seconds... repeat over and over. Certainly an improvement over what it was doing before, where it was staying for the most part at 650 mhz, and only getting up to 1250 mhz for 3 seconds once in a while before dropping again.

So any thoughts what this could mean? The log file shows the GPU heat gets up to about 87 degrees or so right around the minute mark, before starting its throttling every 10 seconds loop.

 

The power consumption limit is shared by the CPU and GPU. Slowing down your CPU reduced its power consumption. That extra power was then available for the GPU so it was able to stay at full speed longer. You can try slowing your CPU down further. Maybe 1600 MHz on the CPU will allow the GPU to run at full speed, all the time, and give you a better overall gaming experience.

On my computer, a GPU core temperature of 91C does not trigger any GPU throttling. If you want to figure out exactly what the trigger point is on your tablet, you would need to do some controlled Furmark + Prime95 Small FFT testing like I did above. Those 2 programs put a very consistent load on the CPU and GPU so you can slowly increase heat and power consumption and monitor with ThrottleStop and GPU-Z to see when throttling begins. You could also just run a ThrottleStop log file while gaming. This will give you a record of power consumption. Long term, if your tablet is near or over the 35W TDP limit, it will throttle by reducing the amount of CPU and GPU turbo boost. That's how Intel designed these chips and those limitations have nothing to do with the core temperatures.

If the power limit register was unlocked in the bios and you increased those values, you would probably have a much better gaming experience. Fujitsu must have decided that pushing 90C or 100C in a tablet wouldn't be a great idea for the majority of their customers so they left this Intel Core i CPU feature locked.

Have you installed the latest Intel GPU drivers? They might help but they probably won't make much of a difference because the limitations you are having were built in at the hardware level.

Support for the 3rd Generation Intel® Core™ Processors with Intel® HD Graphics 4000/2500

 

I definitely have all the new drivers... did you get a chance to look over my GPU-Z logs? I'll do a throttle stop log next.

There is something still not sitting right about all this though... My computer shouldn't have to limit the CPU so much in order to not throttle down the GPU so excessively, especially considering this is not a ULV 17 watt processor.
Currently, the weaker ULV ultrabooks are easily outperforming my HD 4000 in GPU clock cycles, and running more stable.
If I compare my unit to an i7 or i5 ULV like the i7-3517UE rated to 17 watt, my computer should be able to produce a much more steady performance, but it seems it's even worse than a ULV in a ultrabook would. Instead of gradually stepping down, or hitting various frequencies in the high range, my computer is either at 1250 or 650 or rarely 350. It doesn't behave like the others when they are exceeding their TDP... I get the impression my GPU in this system is broken.

Check out this really informative article for instance, it shows in depth the problem with ULV processors with HD 4000, and how they throttle and have an inconsistent GPU clock speed due to such a low overhead thermal constraints. AnandTech - Mobile Ivy Bridge HD 4000 Investigation: Real-Time iGPU Clocks on ULV vs. Quad-Core

If you look at the clock cycle graphs at the bottom and how they are throttling, it is more as I would expect it to throttle: dipping constantly under 1200, frequently in the 800-950 mhz range, but never often dipping below that, it's a constant variation very unlike my experience.

"in most games the ULV IVB laptop averages clock speeds that are 50-150MHz lower than its maximum 1150MHz turbo clock"

My situation is far more severe than what he is quoted as a problem. I've never seen 800 or 900 mhz on the GPU clock. It runs either full bore or GPU powered all the way down. It's like the GPU only has on or off, no fluctuations or consistencies. If you look at GPU clock cycles of nearly every other 35-45 watt IVB processor running games they should be consistently at their max or floating somewhere in the 800-900 range like yours did when it began to run out of power. Not just turn off to 650 like mine is.

Something is vastly different with mine, and it is what is bugging me. It could be Fujitsu using a unique power liming in the BIOS, but I have my doubts. In all my research into Fujitsu tabletPC's and laptops from the past 10 years, I have yet to hear of any of them having severely aggressive power saving features, I have heard that of Dell, but not Fujitsu.

My symptoms are beginning to make me think that I have a bum GPU, but I don't know how to test that theory. I would love to find another fujitsu i7 T902 to test with a game and see how it compares, that would really tell me whether or not mine was a special case.

Just to eliminate the possibility, I thought I would upload a picture of the specs on my Fujitsu ac adapter for you guys to look at. I know that throttling has occurde in other systems due to insufficient PSU before, so I thought for the sake of being thorough, I would upload the specs here to see if you guys think they may be too low.



Also, got a message back from another T902 user (with i7 configuration as well) who tried a 640 x 480 furmark test and ran GPU-Z at the same time, and he also has the exact same symptoms as me: GPU clock drops immediately from 1250 to 650 without ever hitting any other speeds. His also reached 91 degrees before doing this. I would love to see evidence of this kind of throttling anywhere else in reviews or users, but I have scoured forums, you tube, and review sites, and all I ever find is every Ivy Bridge processor, ULV or top of the line, ALL outperforming our GPU and not displaying these symptoms. It makes me mad at Fujitsu, and unsure of how to proceed.

 

Looks like I guessed right about this being a Fujitsu specific problem then?

If this is really affecting your use of the notebook: time to sell/trade up.

What other option can there be? Open up the system just to ensure adequate airflow (if that is even enough...)? Hack the BIOS to make the system work how Intel intended? Buy a bigger badder adaptor in the hope of making the system behave?

Nah: get a new system!!! (We'll help you 'build' it. ).

 

haha well, its not that easy to replace... and the problem is more a nuisance and a gripe, as video games are not a priority for me in the larger scheme of things, I just think that for the $3200 I spent in the end, I should have a stable GPU capable of doing some light gaming without throttling .

Honestly, This tabletPC is the nicest money can currently buy, (I've done a LOT of research on tabletPC's) and really there is nothing available on the market with these specs, this size, and a wacom penabled screen to replace it with. Not to mention it is brand new, and I don't think Fujitsu takes returns on custom builds, so I would lose a lot of money trying to get rid of it. Not worth even considering, especially as I am using it daily for work. It really is money well spent for being able to draw/animate on the screen while on the go, I get around 8 hours at least on the two batteries, and that is doing some pretty heavy photoshop work, after effects, and animating in Flash. For $4000 I could buy a Wacom Cintiq 24HD touch, but it's only a monitor, not portable, I would then have to buy a desktop on top of that. So I save money AND get a computer and gain portability this way. The computer is otherwise blazing fast, and I am very impressed with it... I just wish this GPU throttle wasn't there, it would then feel perfect.

Otherwise, for editing and cpu/memory intensive work this thing is amazing. I just wish that Fujitsu left more options in BIOS. It is so packed with security stuff I don't need, ( I have to remind myself it's not aimed at artists) yet it doesn't let me do the most basic tasks in BIOS, and locks me out of any kind of control over the hard ware... typical of a machine aimed at the enterprise/corporate market. It is not a mass consumer device at all that's for sure.

I may try those three options you list if possible. I am considering talking to the others over at tabletPCreview about finding someone to modify the BIOS to remove the limit on the power register. If we succeed in finding someone to do it, we may have to buy a bigger PSU. I'd be fine with that!

 

stoneseeker, the easiest way for you to reduce or avoid IGP throttling is to lower CPU clocks to something around the 2GHz mark, which will then make thermals comfortable enough for the IGP to work with full clocks.

You can configure two ThrottleStop profiles, one for gaming (lower CPU clocks to accommodate full IGP clocks and prevent throttling) and one for working/general usage (stock CPU clocks).

 

Thanks tribaljet. I will do that from now on. It will work at least as a temporary workaround, although it still really limits games like Skyrim where a powerful CPU actually gets utilized by the game, but it is still an improvement. Long term fix would be to unlock the power register on the BIOS it seems, but I might be dreaming there.

 

Hi K' =)

When TS is set to monitor the dedicated GPU, it turns it on every second or so, thus preventing it from idling..
That makes the power draw higher, is there any way to avoid this, but still monitoring gpu temp (for alarm triggering) ?
That's the only thing that holds me from launching TS at startup and keep it running all the time..

Thanks for your insights !

 

lastnikita: I am aware of the ThrottleStop - Nvidia Optimus issue but I haven't had the time to look into it. Trying to find a work around for that is on my things to do list but realistically, it's probably going to be a few months before I get around to it.

stoneseeker: I think I remember reading a long time ago that there is a register in the CPU that determines the Turbo Boost balance between the CPU and Intel GPU. I will do some searching for that article. That might explain the difference in throttling between how your CPU works and how my desktop CPU works when it approaches full load.

If you are bored and want to create some heat, run Furmark in a small 640 x 400 window and run a log file using ThrottleStop. Furmark barely puts any load on the CPU so I would like to see what sort of power draw ThrottleStop is showing, what temperatures and when throttling begins.

If your bios is triggering throttling based on a temperature of about 87C, there's probably not much one can do about that. If you get real adventurous, do the same test with your tablet in the fridge or a freezer. You might have to move way up north to fully use your tablet.

The market is full of half baked mobile products with specs that promise more than they can fully deliver. Manufacturers need to put a little more thought into their cooling solutions.

 

Ok, I'll patiently wait for that to happen, hope it can be solved =)
I could get a setup to automatically switch to performance power mode + lock turbo when gpu is loaded, (temp > 60°), that's why I need to monitor it =)

Wish you the best

 

OK, so I ran throttle stop to log the throttling while running a quick furmark custom bench at 640X480 windowed as you suggested, and watched the GPU throttle after a few seconds into it. You can clearly see the power drop from a consistent 33 down to jumping erratically in the teens. What interests me about these logs, is it never shows it get hotter than 76 degrees! So maybe it isn't hitting a thermal limit after all, but a power limit of some type, which is weird.
Now I'm all lost again.

Unclewebb, did you find that article? Do you think my system is power starved?? View attachment fujitsu 640x480 TSlog.txt

 

I am following the conversation between unclewebb and stoneseeker with great interest.

stoneseeker, I still think this is a Fujitsu BIOS limitation - the less than 76 degrees seems to be a giveaway (the chip can obviously take more than that).

unclewebb's reply should be interesting...

 

is TDP affected by bdprochot for the gpu? say for example if i turn it off, will my gpu usage be more stable?

the reason i ask is bc Mobius suggests disabling bdprochot in his invidainspector OC guide for the gt650ms

source: http://forum.notebookreview.com/gam...guide-workaround-max-clocks-oc-stability.html

 

BD PROCHOT - bi-directional processor hot

This "feature" allows the GPU or motherboard to send a signal to your CPU to tell your CPU that it is too hot even when it is not. This forces your CPU to throttle to the lowest speed possible which kills gaming performance. Not all laptops use this feature. If they do and it was my laptop, I would disable BD PROCHOT.

stoneseeker: I will have a better look at your log file later today but I don't think looking at the numbers is going to solve your problem. Your tablet can not run the app you want it to run at full speed. Without a modified bios and better cooling, this is not going to change.

 

yeah, I guess I am just wanting to understand "why" the GPU clock behaves so differently from every other system with the same or lesser specs and cooling systems. Instead of throttling like every other machine, it seems to have only 2 clock speed modes it can access. FULL POWER or LOWEST POWER. (1250 or 650) No in-between like it is supposed to have.
I suppose could be a strange Fujitsu specific BIOS setting, but I guess I just wanted to eliminate the chance that it was an error in the way that my system was handling load or distributing power to the GPU. I am not knowledgeable enough on my own to figure this out though.

Since I am a visual communicator, here is my GPU clock over time in a chart while playing a game, and another of the same chart with a much weaker 17 watt ultra low voltage i5 in a laptop (ultrabook) smaller than mine using the same integrated HD 4000 playing the same game.



Looking at these charts and all the others I have seen charting many systems and many ivy bridge processors while under load, it really bugs me to see the way my GPU clock behaves in contrast. I've yet to see ANY other system, no matter how crappy and no matter how small behave like my system does when I use the GPU. There has got to be a mistake made in the way they wrote the BIOS or something...

Based on my research of the HD4000 in mobile devices, I know I am not asking too much from my system. But these charts illustrate exactly what I am experiencing that I would consider "abnormal" behavior.

If this issue is beyond the scope of the thread, my apologies. I may have to look into a BIOS mod, but I was hoping there might be someone here who may know what would cause the GPU to act so strangely first.

 

When Intel throttles a CPU due to it reaching the thermal throttling temperature (TJ Max), it usually throttles for periods of time that are measured in milliseconds or more likely, microseconds. It can rapidly adjust the CPU multiplier and voltage so the core temperature is maintained just a hair under the throttling temperature. When throttling happens for blocks of time that are measured in seconds, that's typically a temperature limit that has been set within the bios that is causing the throttling. I don't have any experience with the Intel 4000 GPU throttling. All I know is that my desktop 3570K does not throttle when the GPU is over 90C. Your testing has me curious. Perhaps I will push further until it hits 95C or beyond just to see if there is any GPU throttling. The CPU is supposed to run at full speed up to 105C so I suspect that a properly operating Intel GPU will continue running at full speed too.

 

I've got an Asus Z8PE-D18 server mobo (two Xeon 5650, Win7 Pro) that doesn't allow bios adjustments. Will TS do something/anything for me? If not TS, are there any alternatives at all? I'd hate to waste this mobo by upgrading to a Evga SR-2 just to OC these Xeons.

I use TS 4.1 on my Dell 6400 (X9100, 3458MHz, Win7) and it works great, so that's why I'm asking. Thanks!

 

ThrottleStop does not support 6 core CPUs or dual CPU motherboards.

When you run ThrottleStop and click on the TRL button or the TPL button, does it do anything? Does ThrottleStop even start up? Probably not.

 

Thanks for your informative replies UncleWebb. What program can I use to see what parameters were set in the BIOS? Is there a way I can look at the temp limits they may have set so I can have a realistic idea of when it should throttle like that?? Also, since you've made it obvious its a BIOS setting, how can we know it is a temp limit vs. a power limit?

 

Hello there..
I'm having throttle issue as well with my hp envy 4-1013tx laptop..
It has intel core i5-3317U with base speed of 1.7 Ghz and turbo 2.6 ghz 1 core and 2.4 ghz dual core..

After running prime 95 for about 10 minutes, cpu start to throttle to 798 mhz which is very annoying..
However, this also happen with set multiplier check to turbo..
Here is my screenshot when the throttle happen..


Edit : Now turbo boost not working even i set the multiplier to turbo

 

fatalis26: In your first screen shot, did you try disabling BD PROCHOT. It looks like your CPU is being forced to thermal throttle. This usually happens when a signal is sent to your CPU using the bi-directional processor hot (BD PROCHOT) signal path. This allows other parts of your laptop to send a signal to your CPU which tells it that it is too hot even when it is not too hot. The CPU is then forced to throttle by dropping down to the lowest multiplier. Disabling BD PROCHOT usually fixes this problem by blocking the signal path but it is at the risk of damaging something on your under designed laptop. It's possible that HP is trying to protect something by doing this or maybe they just don't want users complaining that their laptop is too damn hot. I did not engineer your Envy so I have no idea what the reason for this is.

The U in 3317U means your CPU is a low power CPU. Can you post a screen shot of what is in the TPL window. If a manufacturer decides to set the Turbo Power Limits low, this will result in zero turbo boost when the CPU is fully loaded. ThrottleStop is only showing 11.6W so the turbo power limit might be set to 10W or less which would prevent full load turbo boost. On some laptops, this value is adjustable but some manufacturers decide to lock this feature to prevent a user from fully utilizing their Intel CPU. It is up to a user to bring a product like that back to the store and tell them that they can keep it since it is not able to run the CPU at the advertised Intel spec.

stoneseeker: I finally got around to doing the testing that I promised.

I had to really put the spurs to my 3570K. I had to keep bumping up the core voltage while Prime95 and Furmark were fully loading both the CPU and GPU.

http://img27.imageshack.us/img27/6216/torturetest.png

I took it one hair over 100C and it was running as happy as can be. Not a spec of CPU or GPU throttling. The OK displayed in the Thermal Status area of RealTemp confirms that there has not been any thermal throttling of the CPU. It's still 4 degrees away from that since this CPU is rated to run reliably up to 105C. Too bad manufacturers didn't have a little more courage to run the CPU as Intel intended. Intel makes some incredible hardware.

Bios limitations are typically temperature related and not power related. I am just the ThrottleStop guy. You will have to find an expert in bios mods to fix your problem.

 

Hello,i have downloaded your program (TS v5) and use it in my asus A43sv VX 072D

My Spec:
Processor Type = Intel Core i3 Processor
Processor Onboard = Intel® Core™ i3 2310M-2.1GHz
Chipset = Intel® HM65
Memory = 4GB (2x2GB) DDR3 1333MHz SDRAM
Video Type = nVidia GeForce GT540M-1GB DDR3
Win7 x64

i am experiencing throttling as well,when i download your program
i can throttle my cpu to power saver or lock the multiplier until (around) 15x

but when i lock to high value(19-21)it will like nothing happen (and i still experiencing throttling)
i already read the tutorial (also your ThrottleStopDocs.html) to use maximum performance with 100% state
i attached the image with a log file that you can view here: 2012-11-28 - Pastebin.com

 

Thanks, unclewebb. I didn't even try because I expected it wouldn't work on such a system. Thanks for the reply and great software!

 

hey i keep getting WinRing0.dll cannot be started errors and throttlestop wont start. Ive been using it for a while and this only recently happened. Im on version 5 on win7 64bit

 

Thanks for all your testing and information UncleWebb! I'll have to look into getting the BIOS modified, but first I may have to look at asking Fujitsu to unlock the BIOS or make a revision to address the issue (which I doubt they'll do, but its worth a shot) , or getting their thoughts on why it gets so hot so quick and what they can do about it.

thanks again.

 

Hello, i will describe my problem:

I have an Acer 6935g laptop with these specs:
CPU: P8600 2,4 ghz
RAM: 4GB DDR3
GPU: NVIDIA 9600M GT 512MB GDDR3

And when i play Diablo 3 or even League of legends, i have good FPS for the first 30min only (on Diablo 3 like 20min, on LoL like 45min).

And i started my search's to solve the problem, because it was not supposed to happen, and i found this program "Throttle Stop".

Yesterday when i was playing League of legends, when i had FPS drop from 80-100 FPS to 10-25 FPS, i did alt+tab, turned On Throttle Stop in mode "2-Game" and FPS instantly increased back to 80-100 FPS, so i supose this is my problem, called cpu throttling.

Everything is ok, but after sometime , like 1 hour, the PC instantly shutdown, for what i read searching in this thread,its because CPU reachs big Temps. The problem is that my Cpu is never hot, i didnt controlled temps yet, but i am pretty sure that CPU is always arround 30ºC - 35ºC (i will confirm this tonight). Even if they increase abit, i think my BIOS is programmed to interpret "55ºC" or something like that as a "big temp number" , we all know that 55ºC is not high. If when i confirm this, i realise that BIOS indeed doest that, can i change BIOS settings about this? increase "warning number" to something like 80ºC? (All this info i cannot confirm yet, but i feel that)

Anyway, the main problem is the shut down's,anyone with same problem found solutions?

How i prevent PC from shut down instantly?

Thanks for help,
Apokaliptor

 

I think you're confusing your HDD's temps with the CPU temps. No notebook CPU operates at 35C.

I also think it's a combination of your CPU/GPU and HDD temps rising quickly and conspiring to invoke the failsafe so that your system won't melt itself.

Have you recently cleaned out the vents and the fans on your system? Have you repasted your cpu/gpu (ever?)...

Do you have an SSD or a HDD installed? (An SSD can handle higher temps than a HDD).

Do you use a notebook cooler while gaming? What is the ambient temp of the room you're playing in?

Acer's are notorious for running warm/hot. Make sure that the vents/fans are clean and free of dust and you have a good TIM application (and/or the appropriate thermal pads). Again; a notebook cooler is highly recommended.

With all the above taken care of, you may still need to run TS to see consistently high FPS rates - but at least the system won't be shutting down on you.

Good luck.

 

Hello! Thank you for fast reply.

My laptop have a HDD inside. Well yeah, you are right, i think i never found the real problem, but once Throttle Stop solves it, doesn't it mean that the problem comes from CPU throttling?

Answering your questions, my laptop model was never a "hot model", the fan cooler system is pretty ok, and i never had big problems with temps.

Btw, yes i cleaned my fans alot times, i also changed my thermal pad.

I have a alluminium nox boreas cooler, like this one http://www.kuantokusta.pt/img_uploa...6492_3_cooler-para-portatil-nox-boreas-17.jpg

But in the last times i was not using it, it never solved my problem. But, never tried using with at same time with TS, if you think that notebook cooler will prevent pc from shutdown , i hope so

 

Hi unclewebb,

I'm using ThrottleStop v5 & I noticed the reported multiplier has decimal point. So, it safe to say ThrottleStop is not reading MSR 0x198 (Current performance State Value). May I know what MSR register ThrottleStop read to report CPU frequency?

 

kizwan: The method that ThrottleStop uses to determine the CPU multiplier is not a simple process. Being extremely accurate is never easy.

ThrottleStop uses a method that was outlined by Intel in their November 2008 Turbo White Paper. Thanks to rge at XtremeSystems for first introducing me to this paper.

http://download.intel.com/design/processor/applnots/320354.pdf

By continuously comparing 2 timers within the CPU for each thread or core, you can accurately determine what multiplier the CPU is using during that monitoring interval. Any changes in the multiplier due to various C Sleep states or Turbo Boost or whatever are all accounted for when using this method. With these timers running at billions of cycles per second, the results are very precise.

The Intel paper only outlines this monitoring method for the first generation Core i processors. I did some hands on testing and discovered that these same timers were available in the majority of Core 2 processors so I was able to modify this method slightly and get accurate multiplier data from most Core 2 processors as well. This also showed Core 2 - Intel Dynamic Acceleration (IDA) mode being reported accurately for the first time.

One problem I ran into is that some monitoring software was not using these timers in a friendly manner. They are not a protected resource within the processor so any software can read, write or clear these registers whenever they want to. Running a program that does this along side ThrottleStop will totally screw up the results that ThrottleStop shows. That's why I recommend running ThrottleStop by itself. I know it is accurate when it is the only program accessing these timers.

To work around this issue, I switched to some general performance counters within the CPU that were not being used by any other monitoring software that I tested. These performance counters are contained in MSR 0xC1 and MSR 0xC2. The advantage to using these counters is that I was able to get this method working correctly on many of the Atom processors and early Core processors that were not supported by the original method.

Everything was wonderful for a while but then with some of the beta versions of CPU-Z like 1.61.x, Franck started using the 0xC1 and 0xC2 timers for his program. There is nothing wrong with doing that but that meant that ThrottleStop was no longer compatible with these versions of CPU-Z. I was about ready to try and find some new timers again but I am pretty sure that CPU-Z 1.62 has stopped using the 0xC1 and 0xC2 timers. ThrottleStop 5.00 seems to be compatible with CPU-Z 1.62 on the Core 2 and Core i processors that I tested.

 

Thank you very much, unclewebb. The reason I'm asking because I'm trying to figure out how to read/monitor CPU frequency in Mac OS X. As far as I know, there is no tool like CPU-Z for OS X yet (I might be wrong but I look everywhere....sort of). I have DSDT/SSDT dump from a macbook pro 13 early 2011. The SSDT only contain P-States for non-turbo multiplier & there is a claim that due to this, Intel Core processor will not be able to use Turbo Boost in OS X. There is one old app for Mac OS X which is called MSR Tools which can report CPU frequency including Turbo frequency but some people do not trust this app though. It's not open source (as far as I know), so hard to check how it worked.

Quoting from the white paper:-

This explained why SSDT from a pc (laptop) & a macbook has P0 state that only 1MHz higher than P1 state.

 

Hello unclewebb,

I'm having a throttle issue with my ACER W700 (i5-3371U).
After 10 sec of gaming, both CPU and GPU will throttle down to their respective frequencies (1.7Ghz, 350Mhz).

Almost all options are available on ThrottleStop. But I can't figure how to stop that throttling.
I would say it appears when CPU/GPU reach 60°C.


I tried:
- Set Multiplier: 25
- Uncheck BD PROCHOT
- Uncheck C1E
- Set Limit #1: 25
- Set Limit #2: 25
- Check Enable Limit