The ThrottleStop Guide

It is Xtu version 5.1.1.25. giving this message. As you say, I do not think I saw this in early versions of xtu. It does not help to close Trottlestop or close automatically startup.

 

I just re-checked and confirmed I have no problem with XTU 5.1.1.25 and ThrottleStop 8.00 at all. Everything works fine.

I see XTU is telling me there is an update available. Let me download and install it and see if everything still works fine.

 

Yup... still fine with XTU 5.1.2.2 also. No trouble at all here.

 

LoL. Has set Xtu on disabled instead of manually. This is not my day.

 

Quick bug report. When I first reported TS Bench not reaching my 4.5GHz OC and only topping at stock 4.4GHz Turbo Boost, I thought it was something related to the new test load being lighter but I've found out that after each restart, the CPU tops at 4.4GHz rather than 4.5GHz, requiring me to disable and re-enable overclocking in order to unlock 4.5GHz. After doing this, the current TS Bench does reach 4.5GHz.

Also, after setting ThrottleStop 8.00b1 to run at logon, for some reason the Limit Reasons option stops working, saying it can't find LR0101.exe, yet the file is on the ThrottleStop folder and does work. Additionally, what info is supposed to be displayed on the Limit Reasons software?

Could there perhaps be two options on TS Bench, like a light and heavy test? That could cater to everyone and be integrated in the same window.

 

I am glad you got this figured out Papusan. No worries. I have learned not to investigate your bug reports for at least 12 hours.

That bug makes sense. When using the Task Scheduler, there is a bug where Windows tells my program where ThrottleStop was started from which is different than where it is physically located. I need to find a work around for this issue. This is the reason why log files go missing when using the Task Scheduler. The log files are properly created but they end up buried in some sub directory of C:\Windows. I added a feature today so you can redirect the log files to wherever you want them to go. I will try to do something similar so LR0101.exe can be found when using the Task Scheduler.



Kind of nice to be able to redirect log files to a shared memory stick plugged into my router.

LimitReasons tells you the many, many reasons why an Intel CPU with the Integrated Voltage Regulator is throttling. The little buttons at the top of each column clear the results. Intel's 47 Watt mobile CPUs have that tool lighting up like a Christmas tree when stress testing with Prime 95 v.28.5. A properly setup desktop CPU on a decent board should rarely if ever show any throttling. I will post a picture shortly. Dufus did a great job writing that simple little program so I thought it would be nice if more users found out about this very useful tool. The people with 4th and 5th Gen U CPUs should be in for a colorful light show too.

As for your other bug with the 45 multiplier, make sure in the FIVR window that Overclock is checked. If this box was not checked when you first opened this window and then you pushed OK, ThrottleStop assumes that you do not want to overclock anymore so your 45 multiplier will not be available. TS8 beta 2 will check this box for you when you first run ThrottleStop to make it less likely that an overclock accidentally gets turned off. Thanks for the feedback.

 

Oh, wasn't aware that Task Scheduler had that issue. Do you believe it would be of interest to have an autorun feature that would bypass Task Scheduler? And it's quite good that it can read network paths for sure.

Limit Reasons shows nothing on my desktop, laptop still needs to be tested but I imagine the 45W CPU will likely light up as well

Regarding the 45 multiplier, the thing is that I have had Overclock checked on ThrottleStop 8.00b1 from the start, just that it seems not to recognize it and requires me to uncheck and re-check (after accepting each action), with this only happening the first time the software opens.

Out of curiosity but do you find that as ThrottleStop keeps evolving, that there will be specific versions recommended for certain CPU architectures (like 5/6 for Core 2 Duo/quad and older and 7/8 for Core i series)?

 

I prefer using the Task Scheduler when adding start up programs to Windows. I assumed that a Windows function that ThrottleStop uses would tell me where a user had ThrottleStop.exe located. Unfortunately, this function fails when using the Task Scheduler so I have to come up with a plan B. My programming tools are from the Windows 98 era so there are always going to be some minor issues that I need to work around.

I would recommend TS8 for all ThrottleStop users. If you have a 1st Gen Core i or Sandy Bridge or Ivy Bridge CPU, the new FIVR features are of no use but with each version, I always try to make improvements no matter what CPU model you have.

Did you unzip TS8 to one spot on your hard drive and then move the ThrottleStop folder somewhere else? It sounds like your ThrottleStop settings are not being saved. This might be another issue that is related to the Task Scheduler. Windows 8 has a feature that changes the access rights to some files within a folder when you decide to move the folder. This especially seems to happen if you move a folder from one hard drive C to another hard drive D. Wherever you have the ThrottleStop folder located, go in there and delete the ThrottleStop.INI configuration file so ThrottleStop can recreate it next time it starts. You can also check the ThrottleStop.INI properties and make sure that you still have read / write access to the configuration file.

I will look into the multi issue some more but I think it is being caused by the above problem.

When testing, the LimitReasons should look like this.



If any yellow boxes are illuminated, you can click on the heading in that column to clear that information. A yellow box is a log of what has happened. When that box turns red, that means that throttling condition is currently in progress.

Here is an example of my 4700MQ while running Prime 95.

Beginning of Prime95 test shows the CPU is being throttled because it is trying to exceed the 56 Watt Short term power limit (PL2). Later this will change when the PL1 limit kicks in and throttles the CPU down to its long term limit of 47 Watts.





The reason the CPU is throttling is because the CPU Core has reached the TDP power limit (PL1). ThrottleStop reports that the CPU is at 47 Watts. The reduced multipliers across the board show the throttling that is going on. EDP is an indication that there was also some throttling due to the CPU hitting the current limit. When this test first started, it is throttling to 56 Watts initially so there was more current flowing then. The LimitReasons tool lets you watch as throttling progresses. You can also play around with the voltages to help reduce this throttling. Great little testing tool once you understand what it is telling you.

 

I see, and I particularly think of how startup configuration works differently in Windows 8.x and newer so that does seem to be a more logical option.

I only start using any ThrottleStop version when it's on its final folder, which is a version-specific subfolder contained within a main ThrottleStop folder. After unzipping 8.00b1, I changed the current ThrottleStop scheduled task action to launch 8.00b1 rather than 6.00, ran 8.00b1 once to configure it the way I did on previous versions (starting with a clean .ini), ran the notification cleaner, restarted and went from there, with 8.00b1 starting automatically and everything working flawlessly, barring the multiplier issue.
Regarding read/write permissions, no issues there. Also, if it was read-only, it wouldn't allow me to toggle Overclocking on and off.

Limit Reasons showed something interesting. It's always like the following when the system's idling or light loaded:


When I stress test it, with Package Power Short set to 110 (Package Power Long always set to 88), I get somewhat mostly 4.5GHz both both Core and Ring PL2 turn red, with Ring PL2 sometimes going back to yellow while Core PL2 remains red, but both intermitently go back to yellow. When Package Power Short is set to 120 (on ThrottleStop 7 it was never needed), Core and Ring PL2 still turn red but much less frequently.
So, my question is if ThrottleStop 7 handled Package Power differently and if I should keep using 120 for Package Power Short or even increase it. I've tested 125 and never got Core or Ring PL2 to turn red.
Also, should I increase Turbo Time Limit? Is there any harm in using (much) higher than default values?

TPL window:


EDIT: I should note that I'm merely setting clocks 100MHz higher than stock Turbo Boost while using stock voltages as I do want hardware lifespan to be as long as it can be, within reason. I am not particularly looking forward to increasing voltages unless I absolutely have to. I've seen that the system gets BSOD unstable if using stock voltages with all cores set to 4.6GHz.

 

If LimitReasons is flashing red, that means your CPU is throttling intermittently. Even if your Turbo Power Limit values are identical, the amount of throttling can vary slightly from day to day based on core temperature or which way the wind is blowing. I do not know the exact formula that Intel uses but I do know that power consumption is an approximation that is calculated by the CPU. It is not measured and therefore, it is not an exact science. The power limit part of TS8 should be very similar to TS7 so any differences you are seeing are probably just some random variation.

In the long run, it is far more likely that an Intel CPU will become worthless long before it dies. I overclock every CPU that comes my way, God willing, sometimes excessively, with more voltage than any CPU deserves. Intel makes good stuff. My closet is full of CPUs that have seen the apocalypse but surprisingly, they still work perfectly fine. A 100 MHz overclock is not going to hurt a 4790K. Set your power limits to whatever you are comfortable with. Intel Core i CPUs are built like tanks so no worries.

I did lots of testing in Windows 8.1 and the function that used to send ThrottleStop log files into a different dimension when using the Task Scheduler seems to have been fixed. I didn't encounter any problems but I reprogrammed ThrottleStop anyhow. If some version of Windows chokes and can't find the Limit Reasons program, TS8 b2 will open up a file requester so you can point it in the right direction and help it out. If you ever get bored with watching your CPU throttle in Limit Reasons, you can delete LR0101.exe and then reprogram the Limit Reasons button to open up some other .exe file. That's like an Easter egg feature!

 

Reason I asked about power limits is because I recall reading Sandy Bridge CPUs that were overclocked and had their base 95W power limit upped to 150, so I figured that an 88W CPU would have a lower power limit threshold. In that case I'll go with 125 for now, as it seems not to throttle at all but might raise it to 130.

Is there any reason why Limit Reasons starts with EDP and PL1/PL2 on yellow for both Core and Ring, should I be worried about it and what should I change in order to have a blank Limit Reasons window?

So, in theory, 8.00b2 should fix the Overclock bypass (despite checked) at logon?

I hadn't thought about that Easter egg and I love the idea

 

Looking forward to testing TS 8.00 beta 2. Should be even more awesome.

 

tribaljet - I am not sure why EDP and PL1/PL2 are initially checked when you boot up but the same thing happens to my 4700MQ. I think this also happens when I resume from stand by mode. When testing, I just click on the headings and make sure that they are all clear before I turn up the load.

In theory, beta 2 should work even better than beta 1. I will double check to make sure that ones overclock works fine when using the Task Scheduler.

My friend has been testing TS8 b1 on some of his servers.

That's definitely new and improved compared to TS7 or any previous versions.

Edit - One user contacted me and is still having problems with the Notification Area - System Tray icons. I told him that as long as you run the Notifcation Area cleaner program after using TS6 that everything should work OK. Here is how I setup the Task Scheduler and I have not had any recent problems.

http://www.mediafire.com/download/6tnz3ccauuq3ekk/TS6_Win8_Manual.zip

Can someone let me know if this still works? I hate wasting time tracking down issues that I believe were already fixed.

 

Hmm, I see. Well, no sleep or hibernation here, the system's either running or shut down.

Re-checked and TS8.00b1 does indeed show Overclocking checked after starting, unchecking and checking does fix it. It's a minor thing that needs to be done, but it's still there. For reference, here's how the FIVR window is set (both after logon and after re-checking):


It's awesome to see larger systems to work well with ThrottleStop

About the Task Scheduler instructions, they do match my configuration, with the only difference being the "Configured for" being set to Windows 8.1 rather than Windows 8.

 

UncleWebb, just wanted to say I'm using 8.00b1 on a 5820k and it works like a charm. Absolutely brilliant stuff! Thanks for your hard work champ!

 

Hi Unclewebb, great work and good to see some users giving constructive feedback. I was told you were trying to contact me, just popped to let you know you can PM on the other forum, replies might be a bit slow though as I'm not 24/7.

EDP's can happen also due to PL3 limiting, transition in power limit changes and on some systems when using Icc max of less than 256A. Limit reasons was a rough port from MM to MSR to take advantage of a signed driver, don't be surprised if there is a bug or two.

 

You are welcome TBoneSan. Owners of 6 core CPUs have been ignored for too long. I had to rewrite a lot of code to support more cores but it was worth it. Thank your friend Mr. Fox too. He was the guy that lit the fire under me to get me going.

tribaljet - I have been doing lots of testing and lots of rebooting tonight, trying to find any remaining bugs. Not sure about Beta 1 but Beta 2 seems to be working correctly when using the Task Scheduler. The overclock is initialized or maintained based on how ThrottleStop was setup without having to manually click on anything. I also found a work around so the Limit Reasons program is properly located whether you start ThrottleStop from the desktop or use the Task Scheduler. Microsoft can take some of the blame for that odd bug.

Hello Dufus. I have been so busy trying to finalize a version of ThrottleStop that I have avoided looking for you. I finally got caught up on the information that you gave me early last year.

Your Limit Reasons program has been a very useful tool for me. It might not be perfect but it has given me a lot of insight into how power limiting works to control and limit Intel's Core i CPUs. People complain on a regular basis about their throttling U CPUs. Seeing PL1 and PL2 light up in bright red might help explain the mystery.

I read a little bit about the magic 256A number but I need to get caught up on my reading. If you have come across any new information lately, send it my way. I don't have much time at the moment for anything major but I always find a way to make time for the good stuff like the new ThrottleStop FIVR control.

 

I do look forward to test Beta 2

Would it be possible to have detailed description on the Limit Reasons status LEDs? It would be useful overall for diagnostics, but I particularly ask because I have been tweaking a friend's laptop that has an i7-4710MQ with an unlocked BIOS and I've noticed how the Uncore Power LED already starts red and never changes. Core and Ring EDP start yellow as well but increasing power limits seems to have helped, but the odd part is that the values I've entered don't seem to match end power usage (unlike on the 4790K).
Also, his ThrottleStop TDP Throttle is permanently checked, to which I've thought that it might've been the stock power brick (120W) that just can't cope with both the 4710MQ and a Kepler GTX 860M power envelopes. And pushing 3.5GHz on all 4 cores is a bit too much for the laptop's cooling, so I had to tone it down to 3.2 on 4 cores, sadly.

I do have to say, Dufus, that Limit Reasons has been quite useful when adjusting settings and I'm increasingly realizing it pairs exceptionally well with ThrottleStop. Out of curiosity but what is the minimum CPU architecture requirement for Limit Reasons to work? I'm asking because I've seen that Core 2 Duo systems aren't supported.

 

@unclewebb Would it be possible to add voltage control for the VCCIN (ie. the FIVR input voltage)?

The reason why I'm asking is the CPU will actually throttle under high load if vDroop on that rail goes under around 1.62V (default for 4th Gen desktop SKU's is 1.8V). No amount of extra core voltage will help with that throttle reason.

It's not a big deal if you can't becasue anyone who needs a bump in that will most likely be running desktop board and can adjust LLC or offset the voltage by 100mV or so in the UEFI.

 

http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/page-286#post-9983981

 

Well **** I feel stupid now as I actually read that post a few days ago and completely forgot. Looks like its time to attach a debugger to a overclock tool and see what changing VCCIN does.

 

enable support of z3740!!!

 

You know, once there was a time... when people asked for something they said "please". And they would have never had used a "!", let alone three of them...

 

I thought I already did.



Edit - TS8 mostly works OK on my Z3745. This Z3745 is a low power slug. It does not have any throttling problems, it is just slow. Using ThrottleStop results in a performance increase of zero. You cannot overclock these Atoms so supporting these CPUs was mostly a waste of time but I am glad that I did this so you can do your own testing. Hopefully someone can prove me wrong.

You will notice some items that are disabled. Without access to Intel's full documentation, I could not find out if BD PROCHOT or the ability to toggle C1E exists in these CPUs. Same with Chipset Clock Modulation. The power consumption is off by a factor of probably about 512. I will look into this someday but it is not that important.

The multiplier and VID work. The VID numbers that ThrottleStop reports are a little different than what CPU-Z shows. ThrottleStop's numbers are based on the Intel's publicly available documentation and seem to be correct but I don't have anyway to know for sure. Just set the VID as low as you can get away with without crashing. The exact number is not important as long as your device is stable.

 

Okay Unclewebb, I might be the most blind user ever, but I can't seem to get Throttlestop 8 to minimize to the system tray. I've checked and un-checked "task bar", and checked a lot of the options menus. I am just confused about this one O_O.

 

^^Lol yeah same here. Also in Options->Notification Area I'm unable to "check" the Show CPU/GPU/CPU Mhz options. I can change those by modifying the ini and later on uncheck them in the application but even so I'm unable to select them again.
There is some info in the readme file regarding systray but right now I can't test it.

 

OH. Yes I got it working. Once I used that it worked perfectly.

 

Guys,

I'm trying to stop the throttling of Asus Zenbook UX305. ThrottleStop doesn't work on it. I'm not sure but maybe because it has an Intel Core M processor?!

Any comment? Have you guys tried any software to stop throttling of a core M laptop?

Thanks

 

Can you post a screenshot of CPU-Z? Did you try ThrottleStop 8.00? The download link is in my signature.

I was reading in the Lenovo forum about a user that got a boost in performance using ThrottleStop on his M CPU so I think the latest version supports your CPU. As far as I know, the M series are extremely low power CPUs so you are probably not going to see sustained full performance out of these. Intel designs these to run fast for brief periods of time, very brief, and then they throttle. Depending on the bios, you might get a slight performance boost using ThrottleStop but no software is going to turn one of these into a high end notebook CPU.

Edit - Here's the post I found on the Lenovo forum.

https://forums.lenovo.com/t5/S-Seri...ore-thermal-designs-comparison-vs/m-p/2074256

Too bad people don't understand the concept of beta software. If you download ThrottleStop from NBR, how about post some feedback here?

 

Okay Unclewebb my first instance of something possibly off with throttlestop.

I booted my machine to see my wonderful BIOS had reset my current limit to 72A again. TS8 showed me this in its FIVR settings, which I checked and confirmed that it was so in XTU.
I closed TS8, set it to 100A in XTU, closed XTU and restarted TS8, and it still showed as 72A in TS8.

I re-opened XTU and it was showing as 100A in XTU, so I simply set it to 100A in TS8 and saved, but I was wondering if XTU was bugged out, or if TS8 simply didn't update itself between restarts, but if you could clear it up (or check it out) that'd be nice.

 

@unclewebb,

There is any way to stop i5-3230M from throttling at 90*C? I have it set like this:



When it reaches 90*C it downclocks from 3 GHz to 2.6 GHz, and stays like that, to the moment when I stop benchmark or game, then clocks are back to the 3 GHz. I dont know what to do to stop it from throttling.

 

Check Set Multiplier and adjust it as high as it goes. Beware this will cause your laptop to overheat even more, perhaps even thermal shut down at 105C. So I'd fix your cooling first.

 

After setting multipler to max(32 with Turbo) clocks are also downclocking to 2.6 GHz. So I dont know what is the reason of that. I replaced my thermal compounds recently, but it didnt solved my problem. My laptop is Acer v3-571G, and it is known of overheating. Throttlestop is the only solution for it I think.

 

Maybe try Liquid Ultra? worth a try...

 

You have to be really careful when using Intel XTU and ThrottleStop at the same time. These two programs and your bios are all writing to the exact same registers in the CPU.

The first time you run ThrottleStop, it will try to read whatever values your CPU is currently set to. The Current Limit should be in the Power Limits window. If you open up that window in ThrottleStop and it says 72A and you click on OK to exit that window, you have just told ThrottleStop to save 72A as your default value. OK means Save. If you want to look in a window and not save something, do not click on OK. Click on Cancel or the X at the top right corner. The X also means ignore any changes.

If the bios you are using has left the Current Limit register unlocked, you can set ThrottleStop to 100A, click on OK to save this value and then every time you run ThrottleStop, if that register is not locked, it should set your Current Limit to 100A. As long as that register is unlocked, it doesn't matter what setting the bios sets the Current Limit to. As soon as ThrottleStop starts, it should write 100A to the Current Limit register. I know XTU likes to reset ones overclock if it doesn't shut down properly. ThrottleStop always tries to use the settings that you last saved. The ThrottleStop way can be dangerous if you are the type of person that randomly tries different values but it is a lot less frustrating because it should not constantly reset your overclock settings.

To test if ThrottleStop is working correctly, try running it with Intel XTU disabled. Try setting 72A in the bios and see if ThrottleStop is able to raise that to 100A or whatever you have saved. In theory it should work. Let me know what you find out. I really do appreciate your testing. My time for this project is going to be limited for the next while so your testing will help me find and fix bugs quickly.

Are you gaming or stress testing when this happens? Can you turn on the ThrottleStop - Log File option so you have a record of exactly what is happening. You can upload your data to www.pastebin.com and then paste a link here so I can have a look. Your screenshot does not show me anything obvious, except for the Set Multiplier not being checked, so perhaps a log file will provide me with some more clues. Make sure that GPU monitoring is turned on before you start logging data and that you are using the Windows High Performance profile.

 

The problem with this is that it's difficult for me to really notice off the bat if 72A or 100A is actually applied. I would need something that I KNOW throttles me at current limit X but not at current limit Y, and see.

I might test this with XTU benchmark; see if I can force current limit throttling and see how things work.

What I have noticed is that if I save a TS8 register then boot XTU (I don't keep XTU "running" exactly; I always right click and "exit" it when I'm done from the system tray... though I believe the service runs), it shows my previous values saved in XTU. I.E. I set my power limit to 80W and in XTU it said 100W still.

I'll do some testing and see how the cross reference works.

 

Setting defaults in BIOS (I have modified version with advanced menu enabled) and the High Performance Profile solved my problem. Thanks for pointing me at the right direction.

 

What thermal compound are you using? Only Gelid GC Extreme or IC Diamond are good enough nowadays..

 

As I have mentioned, if you are using both Intel XTU and ThrottleStop at the same time, it can get confusing. Don't get me wrong. I like XTU. Great program but.......

I have noticed that if ThrottleStop comes along and changes one of the power limit values, XTU will keep reporting the previous value. XTU does not update itself often enough or at all. It thinks it is the one and only program on the planet that should be changing these registers so it completely ignores changes made by ThrottleStop.

I wrote another testing program called

MSR Finder
https://www.sendspace.com/file/luzete

This program lets you take a snapshot of all of the important registers in your CPU. With this info, I can tell what the power and current limits are set to and whether these registers are locked or not. With my 4700MQ, XTU allows me to change the power limits, supposedly, but when I look at the values in these registers, nothing changes because the power limits register is locked by the bios. The average person playing with XTU does not know about bugs like this so they have no idea what values their CPU is actually set to. No wonder you are sometimes confused. I have tried to make sure that you can trust what ThrottleStop is telling you.

When testing with ThrottleStop or XTU, if you are not 100% certain what your CPU is set to, run MSR Finder, upload the Log file it creates to www.pastebin.com and I will have a look at the numbers and tell you what I see.

 

@unclewebb; This is a benchmark result from Your Trottlestop 8.0, but not tested with a Bga processor Your software works very well... Thanks.

http://hwbot.org/submission/2849827...e_i7_4930mx_2min_47sec_657ms?recalculate=true

 

You are welcome Papusan. Thanks for providing me with the motivation to carry on with this project. Does your 4930MX have any throttling issues when you are not using ThrottleStop? Many of the voltage and power settings are saved into the CPU until you reboot so you might not need to run ThrottleStop when benching. If you do not need to run TS when benching, don't run it. Also turn off some of that stuff in the System Tray. You are only a second behind Mr. Fox. I think you can do it!

D2 Ultima - When testing, run the Limit Reasons program and keep an eye on EDP. Here is the Intel definition.

Run Prime95 v28.5, adjust your current limit up and down and watch for changes in EDP. Changes in ThrottleStop should instantly show up in the Limit Reasons program.

 

This seems like a really good piece of software for my y50-70. Right now it is quite new so don't want start fiddling with the throttling just yet (wanna get comfortable with temperatures etc first).
Right now I have it undervolted (i7-4720hq) -0.70mV using XTU, will this be possible to do with Throttlestop aswell if I wanna start fiddle around with the multipliers?

 

Yup.


Sent from my iPhone using Tapatalk

 

So I'd like to say thank you again for this program, I am using now with Broadwell processor for undervolting purposes. I could manage down -85 mV Offset Voltage and it have little effect on 3Dmark11 physics score. However the TDP level control does not seems to working correctly, I can select only 0 or 1, the important 2 is not select-able;



Actually it might not so important anymore, because Broadwell U seems to be able hold constantly the max turbo speed even without that setting, unlike a Haswell ULV.

 

How's the power usage when holding the turbo for so long? Is it exceeding 15W? Can you do a screenshot of CPU power usage from XTU? if this is holding up, this will be good to hear

 

This is a GREAT piece of software for many Lenovo laptops, especially the Y50 and Y510P. I own the Y510P so I had plenty of hands on time and was able to get to the bottom of its throttling issues. Without informing consumers, Lenovo thought it would be a good idea to disable Intel Turbo Boost soon after you start using the Intel CPU and Nvidia GPU at the same time. That will hurt gaming performance for any games that are CPU dependent. Intel XTU is a fine program but it does not have any option available to fix this problem. ThrottleStop lets you adjust how much Turbo Boost your CPU uses. I think the cooling system in the Y50 is pretty good so depending on your room temperature, you should be able to run full Turbo Boost just like Intel intended. If the cooling system on a laptop is weak or dirty, you can use ThrottleStop to use some Turbo Boost but not all. Very flexible. You can even setup some keyboard shortcuts so you can switch between different CPU profiles on the fly depending on what game you are playing. On my Y510P, ThrottleStop gives me access to 6 different voltages so I have total control of the CPU which is more control than XTU gives me.

The ThrottleStop.INI configuration file holds all of the current settings. If you are already using XTU, you can delete the ThrottleStop configuration file and then when ThrottleStop first starts up, it will read what voltages your CPU is already using and you can go from there. Almost forgot. ThrottleStop also lets me lock my cache ratio to 36 when the CPU is fully loaded. I cannot do that with XTU. The real world value of this tweak is very little but for the guys that love to benchmark, this might be the difference between a gold medal or no medal. Are you sold yet?

Thank you Atom Ant for posting that picture. I don't have access to anything Broadwell related so a simple screenshot answers a lot of questions for me.

You can blame Intel that TDP Level 2 does not exist in your processor. Have a look at the ThrottleStop - Turbo Power Limits window. When the TDP Level 2 line has a 0 in the TDP column, that means that Intel disabled this Level for your CPU. It is good to know that your new CPU is so efficient that it runs fine without needing this level. In this situation, the default TDP Level 0 at 15 watts would be preferred to the TDP Level 1 at 7 watts. Have a look in your Windows Control Panel - Power Options. You might find a similar adjustment hiding in there for Power Limit control.

 

I would not like to install XTU near Throttlestop, but I guess at the front page of Throttlestop "Package Power" showing the same value. I am using LinX 0.65 for stability test, I found it more intensive than Prime 95 and I can check the GFlops numbers to track down if any performance loss occurred. During this test without undervolting Package Power climbing up maximum to 16.8W, with undervolting max to 14.8W;



Unclewebb;
Damn, than it would have been big problem if Broadwell throttling like Haswell, we could not fix the problem. Fortunately transition to 14nm had good effect for power efficiency. I wonder if faster versions of ULV processors can still hold their turbo frequency? Maybe not and waste of money to buy i7 ULV.
Are the other undervolting options (CPU Cache, I/O, IGPU) worthwhile checking for further power efficiency increase or they very marginal?

 

On some CPUs, lowering the CPU Cache voltage is almost as important as lowering the CPU core voltage. The low power U CPUs are going to appreciate this the most. You can use CPUID HWMonitor and probably HWiNFO to view the change in approximate power consumption for each individual CPU component.

For comparison, here is how a 3rd Gen U performs.



Each TDP Level has a default multiplier associated with it. When an early U CPU entered TDP Level 1, the multiplier would get locked to 8 so the CPU would be struggling along at only 798 MHz and usually, it would be running at a power level even lower than the already reduced TDP rating. That's what really killed performance. Hopefully your 5200U will not be so eager to get into TDP Level 1. I have added this multiplier information to TS 8b2.

When only stressing the CPU, both the 3rd Gen and the 5th Gen U CPUs are able to maintain full Turbo Boost. The problem I had was when trying to use the Intel CPU and GPU at the same time. At that point there were not enough watts to go around so throttling would begin. For testing I liked to run something simple like the Heaven benchmark which could be put into a small window so you could keep an eye on how the CPU was handling things. To simulate a more demanding game, add on a single thread of wPrime or Prime95.

The significant increase in GFlops compared to my 3rd Gen U is probably because your 5200U supports the AVX2 instructions. If you have the time, for a fair comparison, could you try running wPrime 32M version 2.10 a couple of times. I am pretty sure this benchmark does not take advantage of the AVX2 instructions so the bench times might be a better measure of improvements in new Broadwell CPU technology.



http://www.wprime.net/Download/

To get wPrime to run in Windows 8.1, I moved it to the main C directory and then right clicked on the icon and I used the "Run as Administrator" option.

 

Than without extra instructions, according to Wprime a 2.5GHz Broadwell just as fast as the 3GHz (or 2.8GHz?) good old Ivy Bridge;




I do not think near working IGP Broadwell U will able to hold turbo frequency, but it might not so important if I have a dedicated 950M Nvidia near. Great to hear lowering Cache voltage will further increase efficiency, looking forward to go ahead and report back after tests. Have many more things to do with this new laptop, now I trying to run it with 2133MHz modules, because it is booting up at that speed, unfortunately crashing loading Windows. Need to go buy more modules...

 

Here's a result from last year for the i7-5500U. IIRC -50mV and 26x multi to keep at the 15W limit.



And about as much as can be done with i7-4700MQ without hitting thermal limits on MSI GE60


Would be interesting to see if the i7-4500U can be OC'd, maybe one day. Pretty sure the power limiting of the i7-4500U could be removed with BIOS modification as long as the manufacturer has a strong VRM.