The ThrottleStop Guide

I’ m trying again....Have you seen the mail I sent to you?

 

Here is where that magic number comes from.

1/64 second = 0.015625 seconds = 15.625 ms

Not sure why things get rounded to 1.001 ms but the above number should be exact.

 

Thanks for the info. I always like it when interesting things show up in my mailbox.

After you reboot, before you run ThrottleStop, can you delete the ThrottleStop.INI configuration file. It looks like you might have accidentally saved a large time value. If you delete the INI file before running ThrottleStop, it should be able to read the default values from your CPU.

For Bay Trail, Intel made some changes to the power limits register. I am not sure if the public documentation is 100% accurate. My Asus VivoTab 8 shows a time limit of 2 seconds as the default value. I will have to double check but I think your tablet shows the same thing.

 

run about 30 minutes, it seems not to get down....temp is 94℃ max.
thx a lot

 

That looks a lot better. The first successful use of ThrottleStop to block throttling on a Bay Trail CPU. Congratulations!

19 is the maximum multiplier when 3 or 4 threads are active so that looks right.

To adjust the core voltage, I like to run Prime95. Fully load the CPU and drop the voltage one notch at a time until your computer freezes or Prime95 shows an error. Increase the voltage a notch or two from there and it should be good. That might help reduce your maximum core temperature.

Edit - 1583 MHz is quite a bit better than 64 MHz. Percentage wise, that might be the biggest increase in performance ever for ThrottleStop.

 

Hi folks,

I do apologize for my ignorance, but I do have a couple of questions. Been playing with TSb2 on my yoga3 14 - i7-5500U; Of course I'm limited by the 15W TDP, so I've been playing in TPL, and adjusting the power package short, I can stabilize the CPU ratio to about 24x a 15watts, albeit the uncore also drops to 24x as well, where as by default, the uncore will remain at 27x while the CPU's drop to 12-15x and bounce around as wattage drops...

One thing I did notice is that if I cycle the pps from a low value, to a high one, I can get turbo, and allowances to just south of 20watts to re-engage for short periods, before dropping back down, so my question would be is there any way to get TS to automatically go back and forth between two pps values, or is there a more efficient way of doing this that I'm overlooking?

cheers!
Steph S

 

yes, it is better than before. the voltage can reduce about 0.1v. If the max ratio gets to 19x, the voltage can set 0.5v(0.63v before). However the power reduces only 0.3w, and in fact the IGP is the most of cpu power(3.5W). I can't adjust zhe IGP voltage...

 

Good job turkey fighter

 

玉坂。。。。
don't call me turkey.... My name is jeff....

 

What is the timing thing and how does it affect battery life lol?

 

Sorry, I only know how to lower the IGP voltage for the 4th and 5th Gen Core i CPUs that use the FIVR.

What is a "short period" of time? I tried to trick a CPU using this method but for me, it didn't really work. Could you turn on the Log File option in ThrottleStop. I would like to see what power you can get your U CPU to run at and how long you can keep it above the TDP limit.

The timer resolution determines how long a task will get to use a core before it gets kicked off that core and put back in the line with all of the other tasks waiting to be processed. The theory is that if you reduce this timing interval, the CPU will constantly be busy scheduling these tasks and it won't get an opportunity to use the low power C States. This will increase power consumption. My testing has shown that is not true at all. You can drop this interval down to 1 ms and it doesn't make a huge change to the C State residency time. An idle CPU can still spend the vast majority of its time in a low power C State like C6 or C7 whether this is set to 15.6 ms or 1.0 ms.

The advantage of a low timer resolution value is less stuttering and perhaps smoother game play. That is why some multimedia apps will drop this down to 1.0 ms. Compared to when Windows was first introduced, a modern CPU can get WAY more done during its time slice compared to way back then. Perhaps the default of 15.6 ms is excessive these days. Perhaps this can be lower or much lower without any negative consequences. ThrottleStop lets you adjust this now so you can do your own testing. D2 Ultima has seen some significant improvements by dropping this down to 1.0 ms. That's why I built this into ThrottleStop 8.

 

Fixed* =D

See, I already tried this. When I closed EVERYTHING on my PC, I tried doing this, and the resulting factor is "none". No programs are listed as adjusting the timer resolution down. When I had chrome and my mouse software and such running, it listed those. So I closed *EVERYTHING*. Mouse software, headset software, steam, throttlestop, etc. I ran CMD to use clockres to test and when I saw 1ms I ran the energy duration limit again and there was nothing. It simply said my PC was set to 1ms (aka 100ns) and that was simply that. No programs demanding it. None requesting it.

 

Hi unclewebb i used this method to find out the maximum stable undervolt offset of the core voltage for my i7-4710mq. Windows freezes at -146,5mV. I rebootet and set -140,5mV. Prime95 runs stable with that. But as soon as i stop Prime95 workers - it freezes immediately with this setting.

I think i found the maximum undervolt that works under full load for me.
But how is the easiest way to find out the maxim stable overall undervolt for all steps between load and idle.

Or is there an other method you suggest to use (static instead of adaptive?) when the laptop runs in Performance profile with PSU connected...

 

unclewebb I presume you still own Y510P
Could you share your profile or undervoltage values for i7-4700MQ(y510p)?

 

short period varies from a second to 5-6 seconds.. I have two examples in the log. on initial startup of an app/benchmark, it can stay above the tdp of 15w for about 30-45 seconds, before dropping back down again, and I've seen it go up as high as 20w.

 

The problem with trying to adjust your voltage this way on a Core i7-4710MQ is that when running Prime95, the CPU will be throttling quite a bit and your multiplier is nowhere near the maximum. I suggested this for the user with the Bay Trail CPU. He was able to run his CPU at the maximum multiplier without overheating or without any throttling.

The big problem I have found with my 4700MQ is that an offset voltage value that is stable at full load will quickly blue screen when lightly loaded because the C States are already lowering the voltage to the bare minimum. This limits me to a reliable offset of only about -40 mV.

What I am thinking about testing is an offset voltage when loaded but default voltage or close to default voltage when idle. This might eliminate the light load BSOD issues while allowing a person to reliably run a larger offset. I am not sure if this is going to be possible. Without some sort of trick like this, the amount of offset you can get away with is very limited.

This trick is not accomplishing anything. You lower the power consumption below the TDP for a handful of seconds and then that allows the CPU to operate above the TDP level for a handful of seconds. The net result seems to be about the same. Intel recommends that the short term power limit be set approximately 25% higher than the TDP limit. On a 15 Watt CPU, your short term limit should be close to 19 Watts. If this is unlocked or if an individual manufacturer set this a little higher, 20 Watts would be possible.

The turbo time limit is somewhat variable. The amount of time you get to operate above the TDP depends on how significant the load is as well as a few other factors. When fully loaded with Prime95, the CPU is trying to operate way over the TDP limit so you will not get the full rated amount of time. When lightly loaded and running just over the TDP limit, you should get more than the rated amount of turbo time. I have tried to trick these CPUs with similar methods to what you are doing but I did not find any real gain doing this.

 

with Prime95, it usually hits its thermal limit way before TDP; it's only when the Intel graphics kick in that it can't maintain the turbo wattage. In any case, disappointing, but just something I'll have to pay attention to/research more when I'm in the market for my next laptop. While I didn't get the yoga 3-14 as a gaming machine, it would have been nice if it could at least maintain at even it's advertised speeds, sans turbo, while under full cpu/gpu load *shrug*

thanks for the help and input unclewebb; the work you're doing is appreciated!

cheers!
Steph S

 

ok i did it another way now, i set my multi to 35 - removed the checkbox in "power saver" and raised the undervolt to -125mV (from -75mV). This works fine for me as a "performance" profile. In settings i put this profile as "use with AC connected". For me its perfekt this way and the massive undervolt results in a -10°C temperature drop in Games (tested with Titanfall). During a 3 hours gaming session my Temps were CPU 69° max and GPU 60° (using Coollaboratory Liquid Ultra and Fans on MAX in my little 13,3" W230SS beast) ;-D - GREAT Tool btw.

 

Hi unclewebb, first of all i want to thank you for this superb program. I've just bought a Toshiba satellite L50-A-1CU wich has a 4700mq processor. After several tests i find out that is EDP limited because the current limit is set to just 39A. I've downvolted it with throttlestop 8 beta2 to -100mV for the CPU, -60mV for the cache and -30mV for the Intel GPU rock solid, maybe it can go lower on cache and gpu voltage but i need time for several tests. The strange thing i've noticed is that with Win 8.1 it arrives at 2.9Ghz on TSBench and when on Win 10 it arrives at 3Ghz with the same settings, also Cinebench gives higher scores. Is this a known thing? Also is there a way to raise the current limit if is greyed out?
Thank you
Alex

 

I have checked out this wPrime + Heaven benchmark together and the CPU part does not seems to throttling. The Nvidia GPU was disabled, so it is sure the Intel IGP were stressed. Probably the IGP part was throttling, since the end of the benchmark I got about 8FPS lower score as without stressing the CPU;

 

Download link for ver 8.00?

 

Unclewebb's Signature xD

 

See:
http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/page-301#post-10004545

 

You're welcome. Good to hear that you are able to run a significant under volt without any issues.

If the current limit is grayed out, that means the bios has locked this register. The only way around this is with a modified bios which can be dangerous or not possible if a mod bios is not available. Some of the beta versions of Windows 10 are not fully functional so it wouldn't surprise me if they were leaner and ran a little faster. I don't think there will be a huge difference in performance between Win 8.1 retail vs Win 10 retail final. After a year or two of patches, Win 10 performance will probably be very similar to Win 8.1. The only advantage I can see is the newer OS might take advantage of the newer CPU instructions like AVX and AVX 2.

I think the 5th Gen U CPUs are an improvement. There seems to be less of the severe throttling issues that the earlier U CPUs used to do. Having one of these paired with an Nvidia GPU is a good idea. Even if the Nvidia GPU is not very powerful, it frees up a few precious watts for the Intel CPU which can significantly improve performance.

 

Yes, pairing with an efficient maxwell GPU is really good combination, i play GTA 5 1080p, most of settings at high, some very high and I get solid 30+ FPS. The best i can enjoy this with no higher than 40db system noise, thanks to low heat generation.
Many thanks to you as well, the program is great, you provide excellent support. With the program I lowered voltage for CPU, Cache, also for IGP, seems I save additional 2-3w power consumption. I will be happy to give you a little donation to support your good work, just private me the way you would like it . Thanks!

 

I have been using 8.0b1 with my Core M 5Y10c (Broadwell-Y) Asus T300 Chi for a few weeks.

The T300 Chi has very interesting behavior. On battery the long-term limit seems to be 4.5W and I definitely notice TDP throttling. Nothing in ThrottleStop can change this, not even the TDP level control (which is unlocked but seems to be ignored). Loading all four threads with TS bench I sit at around 1.9GHz.

With the power adapter plugged in, the long-term limit is 9.5W so long as the Intel DPTF drivers are not installed. Without DPTF, the 5Y10c will sit on 9.5W all day long, and it never hits its thermal limit. There is no CPU only load (not even Prime95 on all threads) which can make the CPU TDP throttle. The case heats up to around 48C in this case, which is painful if you hold the device for a long period of time.

With the 9.5W limit, you can make the 5Y10c in the T300 Chi throttle with CPU + GPU load reasonably easily.

With the DPTF driver installed, there appears to be a temperature-controlled governor which lowers the TDP. My guess is that DPTF is measuring the surface temperature of the device and decreasing the TDP limit to keep the surface temperature low. Disabling BD PROCHOT has no effect on this behavior, but it's easy to work around by just uninstalling DPTF.

With DPTF the T300 chi usually settles on 5.5-6.5W after 5 minutes or so of high load. At that level CPU throttling is still pretty minor, but CPU + GPU loads throttle a lot.

The good news is that my 5Y10c undervolts very nicely. I was able to get -100mV on the CPU and -85mV on the GPU, which buys a couple hundred extra MHz under CPU + GPU loading.

 

Can you post a pic of your settings?
I'm trying to see if I can help with a Dell venue 11 pro.
Would be great to see your undervolt settings.
Thanks

 

Dear Unclewebb, Intel XTU changes multiplier between Base and Turbo clocks of my CPU and it works. CPU is rare ES. Throttlestop only changes multi if I set it between 8x-25x range.

Is it possible so I would check what registries XTU changes so the Throttlestop would add this ability to this ES CPU? Again, I understand that my CPU is very rare scenario so you may not want to add its support but if it is easy for you and I can get all info you need...

 

What CPU model do you have? PM me a CPU-Z screenshot if it is a secret. Intel XTU has access to a register in the chipset that ThrottleStop does not have access to so I might not be able to fix this for you. Send me some screenshots of ThrottleStop and XTU so I can have a look.

I was just reading about that so I am pretty sure your guess is right. Device surface temperature seems to be the latest way to trigger throttling. Thanks for sharing your results.

 

Thought I'd drop in to say thanks again - I ran ThrottleStop to get 2.4 GHz on my T7500 back in the late-2.x, early-3.x days, but it was only intermittent since at that time I had to turn off EIST to enable Dual IDA mode, and it usually wasn't worth it to lose the battery life for an extra 200 MHz. Still, it was cool to get 2.4 GHz occasionally on a nominally-2.2 GHz, multiplier-locked CPU.

Tonight I finally upgraded my 2007 Inspiron to a Core 2 Extreme X7900. I say finally because I tried to in 2013, but got some of the cheaply-packaged Newegg thermal paste on the mobo, and the mobo quit working. Now that I have some properly-packaged Arctic Silver from Micro Center, I did the upgrade. And both the CPU and the newer ThrottleStop are nice improvements. The ability to have dynamic frequency while still having the OC is very nice indeed, as is that it remembers the clock on reboot. Though the last one is a bit of a double-edged sword when it remembers a clock that's too high for the current voltage. But all in all, very nice improvements over the years.

And I'm pretty happy that I'm able to get 3 GHz on air on a laptop that shipped with 2.2 GHz as the max option. Based on wPrime 2.10 testing on the X7900, that gives me a significant 36.5% CPU boost over the 2.2 GHz that was my usual max on the T7500, which is quite nice (and in line with the mathematically expected). Probably should've benched it on the T7500 before the swap so I had 2015 benchmarks for it rather than 2010, but oh well. 3.2 GHz works at idle, but not load, and 3.4 GHz is too much even for idle. Upping the voltage might get me to 3.2 at load, but I probably won't do it. Temps are high enough at 3.0 GHz, though no worse than I expected. And I've already got a newer laptop, so this was mostly for the sake of the upgrade and tinkering.

Great work, and I'm glad to see it still gets updates!

 

That's OK, I'm already happy that you decided to look at it.
It's QBC1 Genuine Intel(R) CPU @ 2.60GHz (ES) 55W 8MB Cache so I safely call it 3920XM. CPU-z report is attached. I set 37x in XTU to show that it works there.

 

There is a register in the chipset that ThrottleStop does not have access to that can be used to limit the maximum multiplier. Dufus wrote a small program that checks this register and sets it to the maximum value to make sure that it does not get in the way of overclocking. Try running MaxMulti first and then run ThrottleStop.

MaxMulti
https://www.sendspace.com/file/comwv8

In your ThrottleStop - Turbo Ratio Limits window it shows, Turbo Overclocking - Locked and then it says Unlimited. What this means is that yes, you do have a CPU that is similar to a XM CPU which supports Unlimited overclocking but the bios has set a lock bit so after you boot up, the maximum multiplier cannot be increased. I think that even when using Intel XTU, this will prevent you from increasing the multiplier higher than 38.

Hopefully MaxMulti + ThrottleStop will let you get up to 38 without needing XTU.

Do not start XTU and then run ThrottleStop. Can you show me a screenshot while running the TS Bench test so there is a load on the CPU. Adjust the Set Multiplier value to 32T and take a screenshot so I can see this problem.

 

1. Sure I can but I can tell you what gonna be. Multiplier will stay at 38x if I adjust value to 32x, 33x or whatever between 26-37.
2. Also disable Turbo checkbox works fine so I can set 26x using it. but I am not sure if some other functions work fine like TDP or else. Don't recall for sure though.
1. 2.

Yep, you are correct. Without unlocked BIOS I would get 30x or 32x max (don't recall now). But if I set Max Turbo multiplier in BIOS it works.

It tells "Allowable multi already set to Max".
My IMO is that this CPU expects some different commands or most likely expects them in different registries. If I need to scan or analyze something you just tell me. I thought that I should compare registries at stock versus registries after I set some multi via XTU.

 

When using ThrottleStop, make sure that you are using the Windows High Performance profile in the Control Panel - Power Options and that the Minimum processor state is set to 100%.

The CPU-Z Report you posted seems to be when your CPU is idle. Use XTU and set your multiplier to a value less than the maximum and then start CPU-Z. Load your CPU with something simple like a couple of threads of Prime95 and then click on the Save Report (.TXT) button with the CPU loaded and post that.

I am pretty sure that my 3rd Gen Desktop CPU can be controlled with the ThrottleStop - Set Multiplier option so I am not sure what is happening with your CPU. Your next CPU-Z report might show some details.

 

Did all of that: power plan, 35x, 2 threads of Prime95, report.txt

P.S. unclewebb, is it possible to add additional Turbo voltage? CPU-z says VID supported. That would be really cool.

 

The ThrottleStop screenshot you first posted shows that Extra Turbo Voltage is available and that you have that set to zero. Mr. Fox uses this on his 3920XM or 3940XM and it seems to work. Does that ThrottleStop feature not work on your CPU? I will go check out your CPU-Z data.

Code:

    MSR 0x000001AD        0x00000000    0x26262626
    MSR 0x00000198        0x000022E8    0x00002300
    MSR 0x00000199        0x00000000    0x00002600

The 2 important MSR registers that ThrottleStop uses are 0x1AD and 0x199.

MSR 0x1AD shows your turbo multipliers. 0x26 hex = 38 decimal so your four turbo multipliers are all set to 38.

MSR 0x199 is the multiplier request register. This also shows 0x26 or the 38 multiplier.

MSR 0x198 shows the current multiplier and it shows 0x23 = 35 so those other 2 registers that usually control your multiplier are not in control of your CPU.

The only other register that I know about is the one in the chipset that I don't have access to. When using XTU to set your CPU to a reduced multiplier, can you try running the MaxMulti by Dufus program again. Just to see if this makes any difference. Just curious. I gave it a good try but I don't think I can solve this problem for you. I guess you will have to continue using XTU.

 

Yep, when I do this Dufus' program asks if I want to set it to max. In this case it becomes 38x until I set something in XTU again. This doesn't affect TS.

I must have told, it uses another microcode if that changes. And yeah, Voltage doesn't work too (this one really sad), as TDP and multi.
Thank you for trying anyways. I appreciate very much.

 

I wonder if the fact that it's an engineering sample is the reason for this... @Mr. Fox are you able to get full control over your CPU with TS8 on M18x R2? I believe you don't have an ES chip. If you have contorl and @James D does not, then it might be the CPU itself that's the issue. Process of elimination is sometimes the best method of determining the cause of a problem! =D.

 

The register in the chipset that XTU has access to can be used to limit the maximum multiplier. Your testing confirms that is what is going on.

The Extra Turbo Voltage setting is in the same register as the one that contains how much Turbo Overclocking is available. Turbo Overclocking higher than the maximum set in the bios is locked so it looks like your Extra Turbo Voltage is locked too. I am not sure if it is your bios that is responsible for locking this or perhaps it is just because your ES CPU is a bit of a hybrid. CPU World calls the QBC1 a 3720QM probably because the base frequency is the same.

http://www.cpu-world.com/cgi-bin/CPUID.pl?CPUID=24762

With its 55 Watt TDP and default 38 max multiplier, I agree with you that it has more in common with the 3920XM.

 

Yes, I have an OEM chip and I have full control of my 3920XM. I'm more inclined to think the issues @James D is having could be system BIOS related rather than CPU. I have a 2920XM QS CPU in my M18xR1 and it also works as expected with ThrottleStop. My 4930K in my Clevo has limited control. I have to set my max values in the BIOS and adjust down from that in ThrottleStop. I think that might be due to limitations in my Clevo BIOS, not the CPU or chipset, or ThrottleStop.

OK, we posted replies at the same time. If @James D is running that "3720QM" that is supposed to be a 3920XM ES chip, I have seen several examples in this forum of people that bought that CPU and never got 3920XM level performance from it. It must be an early ES or something, because it has limitations that most XM CPUs do not have. Not saying the chip is no good, just different and with some limitations that 3920XM users do not experience.

I still think the behavior he is seeing could be some crippling in his system BIOS. If the OEM never intended for an Extreme CPU to be installed, the BIOS might not provide full support for it and could introduce limitations.

 

Possible, but I'm still not sure.

Mainly because your 2920XM is a QS, or Quality Sample, which is basically the end-user product. Engineering samples are far older and may not even be feature complete, and he said he had an ES CPU. I was looking up what QS and ES meant since I know some users on this forum have bought them cheaper, and those were the differences I found out.

That is most likely indeed due to your BIOS. I think Clevo has been steadily improving in all areas over the years. Killing Optimus for its high end machines, giving more and more controls to the BIOS and better upgrade compatibility via GPUs etc... all of those are steps in a very good direction. I simply hope they sell enough that they decide to keep doing this. Too many people are voting with their wallets that thin, throttling machines are the popularity.

 

My 4930MX is an ES chip and I have full control of it with ThrottleStop. I think it's that early 3920XM CPU that is misidentified as a 3720XM based on the weirdness I have seen from others using the same CPU, but definitely could be missing or locked registers in the BIOS, etc. No way to know for certain unless he tries a real XM CPU in the motherboard to see what happens.

 

True. All ES chips aren't the same either, so no way to really tell. But if you definitely have control with a 3920XM and I have full control with a 4800MQ, I think it stands to reason that him not having full control lies somewhere other than throttlestop.

 

Yes, I don't expect it to be as good as OEM per Hz or per max clocks achieved and I can live with some additional limitations. I just wanted to know if some of those limitations can be bypassed by ThrottlStop because it bypasses a lot + wanted to achieve at least 4GHz because others managed way higher on the same CPU (but using Alienware though). Latest is achieved by additional voltage and this one is locked by BIOS if you ask me.

P.S. I certainly would use more calculations power but it's not worth for me considering how much it would cost. Heck, I would sell whole package right now if I had slower but working laptop

 

The leatest ThrottleStop does not support ATOM Z3736F……

 

Can you post a screenshot of CPU-Z?

Does ThrottleStop 8.00 start up? Is there an error message or does it just not work correctly? Post some more details so I can have a look.

 

When open the ThrottleStop800, there is a message that your CPU does not support.
But I don't have this tablet. It's others.

 

My intel i7 4510u is tdp limited to 15W. IT runs for about 45 seconds before tdp throttling takes place. Some fiddling around and testing showed i can stay at 2.8ghz without tdp throttling if i can increase the tdp to 17W. How can I do it in throttle stop ?
This question might sound stupid to most of you, the reason im asking this question is the pic called supercharged.jpg in throttle stop folde. I cant for the life of me find that screen when I can adjust the td. Someone please help ?

 

Tell your friend that ThrottleStop 8.00 supports his CPU. Older versions of ThrottleStop do not so he needs to update ThrottleStop.

ThrottleStop has many different options available depending on the CPU model you have. The screenshot above will only show up if you have a 1st Generation Intel Core i CPU. You have a 4th Generation CPU so you should never see that screen. That's why you cannot find it.

For your CPU there should be a button called FIVR and another button called TPL. That stands for Turbo Power Limits. Have a look in there and post a screenshot of that.

 

Acer VN7-571G with unlocked bios. Changed short term and long term power and set time to 128s (bios limit) these settings along with undervolting helped me get up to 55 seconds at 2.8ghz (was around 25 sec before).
So in short i want to move the tdp limit to 18W to get full 2.8ghz out of this proccy. I tried to find the msr correspoding to tdp but, u know..ima noob .
TDP throttling is seriously bottle-necking my gtx850m in games.

Google " Dell venue 11 Pro i5 overclockED !" Thats exactly what i want to do.