The ThrottleStop Guide

Hello good can overclock e8335 2,93mhz with throttleStop?


pd:My configuration acer aspire 6930g nvidia 9600 gt (625mhz 1550mhz 900mhz) cpu core 2 duo E8335 2,93mhz 4gb ram

Thanks

 

I don't believe the E8335 has an unlocked multiplier. It's not an Extreme CPU. The only way to overclock it is to use a program like SetFSB.

 

You're saying ThrottleStop monitor window doesn't show actual CPU state which is wrong.

I beleived all of those screenshots were more than enough to demonstrate the issue but apparently they were not for you. Please check the last screenshot again. That's an instantaneous view of the multiplier at 7 while the encode process took place. So why wasn't it 7 at the time of the fifth screenshot?

Would you please let me know how I can verify my CPU idling at multiplier 7?

Oh! By the way I was changing power profile in order to better demonstrate the issue for you and particularly by your request - minimum of 100 remember?
I understand how Minimum and Maximum Processor State work and how they suit me.

 

For your screen shots 1, 2, 3 and 4; it doesn't matter what Set Multiplier value in ThrottleStop you are using because it will be ignored at idle. What controls the multiplier at idle on your CPU are sleep states like C1E, C3, C6. You don't have C1E enabled so that means your CPU has cores and threads entering the C3 and C6 sleep state. If you have an option in the bios to disable C3/C6; your CPU average multiplier would no longer be bouncing around at an average of approximately 11. With the Minimum and Maximum processor state set to 100%, and with C1E/C3/C6 disabled, the average multiplier should be at the maximum which is 16.

Picture 6 has the Maximum processor state set to 5%. At full load that is limiting your multiplier to 7.00. The C3/C6 sleep states are not used since the CPU is spending most of its time in C0.

I agree with you that picture 5 looks odd. The Minimum and Maximum processor state values you are using might be confusing Windows.

To try to understand what's going on in picture 5, could you try using the High Performance profile and set the Minimum to 5% and the Maximum to 100%. Run ThrottleStop in monitoring mode and post a picture of what that shows at idle. When set like this, if your CPU is fairly idle, the minimum multiplier should be close to 7.0. If it's not, that to me is a sign that there is a problem with Windows.

After that test, try setting the Minimum and Maximum to 100%. ThrottleStop needs to be set like that. Turn ThrottleStop on, check the Set Multiplier box and set that to 16. After that, check the Power Saver feature. In theory, that should give you an average multiplier of 7.00 when your CPU is idle.

Edit: Here are a couple of examples from a Core i5-650.

By setting the Minimum and Maximum processor state to 100%, ThrottleStop can be in control of your multiplier. With the Power Saver option checked, the CPU is using its minimum multiplier which on this CPU is 9.00. It's very steady at idle and not wandering around. C3/C6 are enabled.



The other option is to not use ThrottleStop and let Windows 7 manage the minimum multiplier. By setting the Minimum processor state to 5% and the Maximum to 100%, the CPU will also be able to use its minimum multiplier which is 9.0. In this second example, ThrottleStop is in monitoring mode only and C3/C6 are enabled. Try those two tests and see what your CPU shows.



I know Windows 7 is more aggressive than previous operating systems and tries to transition quicker to a higher multiplier when lightly loaded but using either of the above two methods should let your CPU use the minimum multiplier. I usually use the High Performance profile but for my example I decided to try the Balanced profile and it seems to work correctly too.

Edit #2: This picture shows what happens when you disable C3/C6 in the bios. No more wandering multiplier at idle and if you set the Minimum to 100%, you actually get 100% + some turbo boost.



The default multiplier for this CPU is 24. When you disable C3/C6, you only get a maximum +1 turbo boost up to 25. You need to enable C3/C6 to get access to more turbo boost and the 26 multiplier when a single core is in the active state.

 

This is why I worry about the issue. It's my Latitude D500 and all I have to compare with the Inspiron. It runs Windows XP SP3 and idles the CPU at lowest multiplier very rapidly and steadily.
Then I tried the Linux distros and my Inspiron achieved the very same - but your comments on this have been understood already, thanks. I still beleive Penguin is right though.

Thanks a lot for taking the time for this! It's much appreciated.
I'll be back soon with results of you recommended tests.

 

As I mentioned, Windows 7 has changed. The multiplier is much more dynamic when lightly loaded compared to Windows XP. A small load on XP and the multiplier will stay locked at the minimum. That same load on Windows 7 will cause the average multiplier to start to increase.

I believe the reason for this change is that modern processors don't need the multiplier to be held to a low value to save power. Here's a test at idle that shows virtually no difference in power consumption whether the CPU multiplier or VID is high or low.

http://forum.notebookreview.com/6369641-post5060.html

The 45nm Intel CPUs drop the VID to a lower value internally at idle so the typical settings that ThrottleStop or RM Clock let you control have zero impact on power consumption. I think Microsoft realized this with Windows 7.

As for CPU-Z, here's what the author has to say.

What he doesn't mention is that CPU-Z wasn't 100% accurate for previous Core 2 generations at idle either. It doesn't use the internal timer method that ThrottleStop uses so it does not always report the multiplier accurately at idle.

Some of what happens at idle internally is ignored for consistent CPU-Z validations. Keep that in mind when trying to get to the bottom of this. It drove me crazy for a long time too since there are so many tools out there that are not in agreement.

My methods are based on the November 2008 Intel Turbo White Paper. I figured I might as well go to the manufacturer to see what the proper method is.

edit: ThrottleStop does not support your Pentium M 745 but with a few changes I might be able to get it to work on that CPU. If you would like me to try that then let me know and I'll send you a version that you can do some testing with. I'd be interested to see it working on the older stuff.

I went back and forth through all these cross linked threads trying to find out who Penguin is and what he had to say.
I can't disagree with the Penguin if I don't know what he said.

 

In fact I have noticed CPU-Z has issues of its own but I could verify equal behaviour also with I8kfanGUI. That's well reputable monitor.

On my Inspiron with ThrottleStop and CPU-Z running, changing Minimum Processor State in the power profile also changes the clock reported by CPU-Z at idle, mismatching ThrottleStop report.

I see all or most of your screenshots have the C0% at numbers much below 2 when idle. I've spent much time looking at ThrottleStop on my Inspiron while supposedly idle and C0% oscilates around 4, never actually decreasing below 3. Is that an issue?
I have Core Parking disabled. Do you recommend that?

Have you noticed Picture 5 of my earlier post was when idle? If you did then, doesn't that contradict that " what controls the multiplier at idle on your CPU are sleep states like C1E, C3, C6."

Of course I will more than gladly test ThrottleStop with the Pentium M if you like.

 

My i5-650 has a fresh install of W7-SP1. It doesn't have antivirus software or much of anything else running on it so that's why the C0% numbers are very low at idle. You can see in my last screen shot that when the CPU is running with a 25 multiplier at idle, the C0% numbers are extremely low. This is a sign that the CPU is able to quickly get the background tasks done and then go back into the C1 state so it ends up spending very little time in C0 at idle.

When I manually lower the multiplier at idle, the CPU is slower so it has to spend more time in the C0 state to get its work done processing background tasks. It's an interesting and meaningful number but is often times ignored because it is not the same as what the Windows Task Manager shows. What these two load meters are measuring are not the same, especially at idle.

I noticed that your picture 5 was at idle and I couldn't come up with a reasonable explanation for that either. My guess was that Windows was getting confused with your 1% and 5% power number requests. I'm looking forward to seeing your results when you do the two tests I mentioned with more normal Minimum and Maximum processor state values.

I haven't done any testing of Core Parking. Some users are convinced that it should be disabled. With a Core i3 that doesn't use turbo boost, it's probably best to leave it disabled but don't quote me on that.

I'll PM you a link shortly to a version of ThrottleStop that tries to support your Pentium M. I have no idea what might work. It might report nonsense so post a screen shot so I can have a look. Some of the registers in these early CPUs are in the same location as the later Core 2 CPUs so it's possible that it might work.

Pretty much all early software read the multiplier from a single register. This method is not accurate at idle and can result in a number that is either too low or too high depending on whether software does a Sleep(0) just before reading the multiplier register. If software doesn't do this, the multiplier might be reported too high because the CPU is actively working just before that register is queried. If software does a brief sleep just before reading this register, the CPU might not have time to get back up to its normal speed so it ends up reporting a multiplier that is too low. It's an impossible situation. The only accurate way is to use the high performance timers and to calculate the average multiplier during a monitoring interval. Intel recommends 1 second intervals but early software didn't use this method on the Core 2. CPU-Z has started using this method on the newer Core i CPUs but I don't believe it follows this method exactly so it ends up reporting a multiplier that is consistent but may not be 100% accurate when the CPU is lightly loaded.

 

​

​

​

 

Thanks o770 for showing me all those tests.

I have absolutely no idea why your i3-330M behaves completely differently than the i5-650 I have. Did you do a fresh install of Windows on this computer?

What I posted has always worked for me and for other users I've helped. If you ever swap some parts or find a solution, let me know.

 

Just want to confirm what o770 is reporting (on my i3 350M based Asus U30Jc).

This is also a fresh install Windows 7 x64 Ultimate - installed on January 21, 2011.

 

Thanks for confirming that tilleroftheearth.

I've tested Core 2 mobile and desktop and that Core i5-650 desktop I posted and I've got some testing below of a new Sandy Bridge Core i5-2500K. They all worked the same. I just assumed that they all worked like that but I guess I'm wrong.

Here is a new Core i5-2500K that has a minimum multiplier of 16.

http://img440.imageshack.us/img440/6760/tsidle100.png

http://img338.imageshack.us/img338/996/tsidle5.png

In terms of power consumption, it probably doesn't make any significant difference at idle.

There is a way to measure the percent of time in C3 and C6 which I've been thinking about adding to the next version of ThrottleStop. It's very easy to accurately measure this in any Intel Core i based CPU. This might help explain what you guys are seeing that I'm not.

 

Hello.
Sorry for interrupting your conversation, but would testing my computer with SetFSB to see if the TME Unlock pin mod would be sufficient for overclocking be a safe procedure? I ask because it says the computer will freeze up if that method would not work for overclocking, and I want to make sure that I can just do a hard shutdown, reboot, and be fine again.

 

XSpiritusX, hello. Welcome to the forums!

Dear unclewebb and tilleroftheearth
I've just reinstalled Windows - fresh from Microsoft disc like previous installation, started ThrottleStop before anything on first logon and it's just the same.

C0 still never below 3 and reaching 5 percent.
It's a very light session without Search and Indexer, Defragmenter and such things, now even without anti-virus, amounts to 23 processes and less than 700MB of working memory at idle - makes me proud , please check it out:

​

 

Here's about as low as my i5-650 goes with C3/C6 disabled so it is using the highest multi at idle. The faster the CPU, the less time it needs to spend in C0 to get the background tasks done.



Edit: Here's the same CPU in the same closed case after enabling C3/C6 in the bios. Cores and threads are now spending time completely asleep. In C6 the core voltage drops down to zero so there shouldn't be too much core heat then.

You can tell something is going on because look at the idle core temperatures compared to the first picture. They've dropped significantly. Idle C0% is up which in my opinion shows that when a core is awake, it's having to work harder now because the CPU has been slowed down with lower multipliers and sleep states. Try turning off C3/C6 in the bios if you have that option. There is also a C7 sleep state but this CPU doesn't seem to use that one. Maybe C7 is only for the mobile CPUs.



Hopefully I'll have some time tomorrow to add C3/C6 percent monitoring for Core i processors to ThrottleStop.

 

Thank you very much, o770!

 

ThrottleStop 299.2

I think I finally figured out what's going on o770.

The AUTO setting on the Asus board I'm using only enables the C1 and C3 sleep state. I was assuming that the AUTO bios setting also enabled C6 at the same time but it wasn't doing that. When I enabled C6 manually, the i5-650 I'm using had the wandering average multiplier at idle just like your CPU does.

The wandering multiplier at idle may not look good but CPU power consumption is minimized when the CPU is using the C6 sleep state so this is a good sign that C6 is being used.

The easy way to check is by running the new version of ThrottleStop that includes C0/C3/C6 monitoring. That helps explain a few things. Time for some more testing tomorrow.

 

I've disabled C6 because of it's impact on SSD (Vertex2) 4k writes/reads and access times. Now, if I keep this on the background will this show C3 and C6 at one specific time and/or how much it's actually spent there? ie: I'd like to confirm it doesn't use C6 but does use all other states up to C7 (being a mobile platform).

 

There are two timers that automatically increment whenever a Core i CPU is in either the C3 or is in the C6 state. Software can sample these timers once per second and determine what percentage of time the CPU was in the C3 or the C6 state during the last second. That's what ThrottleStop reports now.

These timers run within the CPU regardless if ThrottleStop is running or not so sampling them once per second does not put any load on the CPU. If ThrottleStop shows C6% at 0.0 when idle then your CPU is not going into the C6 state.

Here's an example of a 2500K.



The interesting thing is that this new Sandy Bridge CPU has no trouble using the idle multiplier when it is idle and using the C6 state but the older Core i CPUs might have a bug that o770 has discovered. Time for some more testing.

If a CPU is not using C6 then it should show that it is spending most of its idle time in C3. I'm going to play around with a register I found last night so maybe ThrottleStop can change this on the fly without having to reboot.

 

unclewebb,

just wanted to report that with TS2.99.2 my i3 350M is showing the C6 state at 80% (as I knew it was being used (er... not used! )... from the battery life I get out of this little U30Jc!).

Also, if that is your SB machine you do know about the defective H67/P67 chipset issue with Intel, right?

See:
Intel Discovers Bug in 6-Series Chipset: Our Analysis - AnandTech :: Your Source for Hardware Analysis and News


Although I don't really have a need for TS, I just want to say thanks for all your work on it.

Cheers!

 

I just heard about that Intel bug today. Luckily, I'm not an early adopter so that screen shot is from my friend's machine. He wasn't too happy.

Now that I know some more about C6, I'm going to try and come up with a ThrottleStop option so C6 can be toggled on and off. That might be useful when gaming or using a laptop with a limited number of bios options.

 

I'm sure a few people who run SSD's would love that option!

(Disabling C6 state).

 

Do you know why disabling C6 appears to disable C3 also (I can disable/enable C1E, C3, C6, C7 in BIOS but everything else is enabled except C6)?

I'd been running fairly aggresive power settings as it's always using the adapter. Using those settings, when I re-enabled C6 in BIOS and checked the C3 and C6 % in TS, it was showing 0 for C6 and few % for C3. Multi was going up and down so I know C6 was being used/enabled.

But if I then reset the power settings back to default (C6 still enabled), I see C3 being very high and C6 being also used. All this I think is normal.

But if I then retain those default power settings, disable C6 in BIOS and check what's going on, TS is showing 0 for both C3 and C6 all the time. I can't see anything saying C6 disables all other power saving functions..?

Also, what is odd is that even though C6 was not being used (as per TS), though it was enabled in BIOS, the SSD was 'slow'. When I disable it, SSD speeds go up. So I guess C6 is used by something else besides the CPU also which isn't showed in TS.

 

I think there are some C-State bios options that don't work correctly or are misleading. I was surprised when I discovered that the AUTO setting on the Asus desktop board disabled C6. It's definitely disabled. TS shows 0.0%, the register that keeps track of this isn't changing and the Kill-a-Watt meter I'm plugged into also confirms that.

The multi going up and down doesn't guarantee that C6 is being used. C1E or C3 can cause a wandering average idle multiplier too. I don't own a mobile Core i CPU so perhaps it was C7 that was being used or something else was going on.

If you want to see if your CPU is using C7, you can try downloading my MSR Tool.

MSR.zip

The register you need to check is 0x3FA
Enter that into the MSR Number box at the bottom and then click on the Read MSR button about every second and see if that register is increasing or not. On my desktop CPU, that register reports 0 all the time. 0x3F8 is the register that counts the time in C3 and register 0x3F9 is the register that counts how many CPU cycles are spent in C6.

This info comes directly from Volume 3B of the Intel public docs. The MSR numbers are listed in Appendix B.4 for Nehalem CPUs and newer.

It's possible that the C6 option in your bios also disables C3 at the same time. This might be a bug or by design. Is there a separate C3 option in your bios?

After you disable C6 in the bios, try using the MSR Tool and show me what MSR 0xE2 shows you. That register might be blocking your CPU from going into C3 when C6 is disabled in the bios.

I did some testing and having C6 enabled or disabled didn't make any significant difference to my SSD hard disk performance while testing with CrystalDiskMark. Disabling C3 and C6 and setting the Minimum processor state to 100% brought the idle multiplier up to 25.0 on the Core i5-650 I have. The 4K read speed went up by about 40% and the 4K write was up by about 20% but the other tests I did seemed to only show normal random variation from one run to the next.

Maybe post some pics of your tests. I'm still learning about this stuff myself but I'm pretty sure that what TS reports for C0%, C3% and C6% on the Core i CPUs is accurate.

 

Any way I can increase performance on my core 2 quad QX9100? When on 4 cores, i was able to hit 3.0 ghz stable. I tried disabling 2 cores, and more overclocking, but it didn't even get pas 3.0 ghz. how can I increase my performance? I unlocked multiplier/vid

PS: it doesn't have temp sensors...........

 

Currently, I have these options in BIOS available:

- CX states (covers C0, C1, C2, C3, C6, C7)
- Enhanced C-States (C1E)
- C6
- C7

Using the MSR tool, I see C7 is never used as far as I can see. When I re-enable C6 I can see both C3 and C6 values moving both in TS and via MSR tool. As there is no C3 by itself, it may explain why disabling C6 also disabled C3 (MSR confirmed it's value is not changing from zero).

You may not find any performance differences as you need a mobile platform running PM55 (or the newer 6 series going by some posts). Using CDM and looking at the 4k writes/reads, this is what I get before disabling C6:
Random Read 4KB (QD=1) : 13.381 MB/s [ 3266.9 IOPS]
Random Write 4KB (QD=1) : 26.934 MB/s [ 6575.7 IOPS]

When I disable C6, it jumps up to:
Random Read 4KB (QD=1) : 29.045 MB/s [ 7091.0 IOPS]
Random Write 4KB (QD=1) : 68.164 MB/s [ 16641.7 IOPS]

Also, again it may be because there aren't enough C states to choose from, but disabling C6 is the only state that removes 'wondering' Multi and 'boosts' the SSD speeds where they should be (previously I had to disable CPU idle which gave similar boost but raised the idle temps too high while this method doesn't raise the).

If you do find a way to incorporate more C states into TS, leaving wondering Multi with the 4k boost.. that would be the best .

 

ThrottleStop_299.3

I added a C6 Off option to this version. On my desktop board, it doesn't seem to be any advantage to hard drive or benchmark performance. It does let the multiplier settle down at idle if you disable C6 but actual power consumption goes up a watt since the CPU is not in the low power C6 state anymore.

Something new to play with. Any mobile feedback would be great.

error-id10t: You might have to enable C6 in the bios and then try this new C6 Off option in ThrottleStop. I believe that if you disable C6 in the bios then this new option will have zero effect.

technos: How high have you raised the VID? When you don't have any temp sensors, it's hard to judge if you are already pushing it too hard. It's usually heat that limits stability when overclocking a Quad in a laptop. If you can find a way to improve cooling, you might be able to overclock it some more. Post a screen shot so I can have a look. Nothing wrong with 3.0 GHz out of an ES processor.

 

Hey unclewebb,

I'm wondering if there's any way to stop my i7-2720QM from clocking down to 800 MHz? In some games (especially League of Legends) it results in small lags when the CPU feels unchallenged and thinks it's not needed at full power, but the other second goes in turbo mode again, why can't it just stay at least in the standard clock of 2200 MHz while playing a game?

Isn't it strange anyways that I only see the CPU in the energy saving clock of 800 MHz or in turbo mode and never in the standard clock of 2200 MHz?

Looking forward to your solution.

 

I was going to try this for you unclewebb, but I can't download from the link provided.

Has it been uploaded yet?

 

Yeah, I wasn't too happy but luckily I have 4 SATA3 ports so my SSD and 2 HDD's are on them already.

I'm a desktop guy, what's the benefit of disabling C6 on my SB build? I didn't see any problems with C6 enabled and my Corsair F80.

 

If you can download 299.3 (or you have the option in your BIOS), disable C6 and see if it makes a difference on your system.

On the HM55 chipset (mobile), effectively disabling it made a MSE 'Full' scan go from 4.5 Hrs (idle) to 1.5 Hrs on something like 4.5 million files (U30Jc - i3-350M + 8GB RAM) while surfing the net and running Outlook with an 8GB+ file at the same time on an Inferno 100GB (SandForce) SSD.


Asix,

Look up a couple of posts to find the link for 299.3 (hope it works for you - it isn't for me, still) and see if that fixes your lags!

 

I've uploaded TS2.99.3 here. Maybe it works now.
I'll try tonight to disable C6 and see what's good or bad.

 

Thank you burebista!

(I hope unclewebb doesn't mind...)

I have enabled TS with C6 off and will report back with any interesting findings.

 

I don't know what exactly disabling this C6 thing does, but it doesn't solve the lags and doesn't stop the CPU clocking down some of the cores to 800 MHz while playing a game. :/

 

I think you are expecting or hoping that your CPU will do something that Intel did not design it to do. With something like a 920XM Extreme CPU you can lock the multiplier but you can't do the same on Intel's non-Extreme CPUs.

Have you set the Control Panel -> Power Options Minimum processor state to 100% yet?

Try running a ThrottleStop log file while you are gaming. Keep track of what time you start and what time you finish at so the log file will make more sense to me. Also enable the More Data option. After you are done gaming, exit ThrottleStop and look in the ThrottleStop folder. Upload ThrottleStopLog.txt to Free File Hosting Made Simple - MediaFire or anywhere convenient and post a link so I can have a look.

A Core i7-2720QM has a default speed of 2200 MHz. When you are playing a game, there should be enough load so that it can maintain that speed. If you ALT+TAB out to the desktop while gaming, these CPUs can slow down so fast that you will never see the true speed your CPU is running at. The log file should show what's going on.

Disabling C6 and maybe C3 is designed to help with the SSD read/write speed problem that the original Core i7 mobile chipset has. It wasn't designed to fix the problem you are having so that's why it doesn't do anything for you.

asix: The lags you have when gaming are likely caused by your GPU going into a power saving state. Intel CPUs are a lot smarter and faster at getting up to full speed than GPUs are when coming out of a sleep state. Try monitoring your computer with GPU-Z so you can see if your GPU MHz are sagging down while gaming.

 

ThrottleStop 2.99.5

Just playing around with blocking the CPU from entering the deeper sleep states. During testing, one user with an SSD reported zero benefit from doing this. Another user with a mechanical hard drive said this little trick made his laptop feel faster during normal use.

This new feature is only for Core i CPUs. The new button can be adjusted to limit what C sleep states your CPU will be allowed to access. This feature might end up only being useful to a very limited number of users so if it doesn't increase your benchmark scores by a 1000 points, no need to complain. If it provides any benefit to you then let me know.

 

unclewebb would it be possible to unlock the full VID range so that we can undervolt as well?

 

Unfortunately the Quads don't support Super Low Frequency Mode or the lower voltages that go with SLFM so their minimum voltage is higher than a similar Core 2 Duo. The minimum in the Quads is usually about 1.05 V. ThrottleStop lets you access that but it's not possible to go any lower than that.

 

I've been lower before, but I forgot what program I used. I was running in the .8 range on my Q9100

 

unclewebb,

Just a quick update:

With 2.99.5 and Max enabled for the C-states, I only got about 3.5-4hrs of 'work' out of my U30Jc - but it was noteworthy how responsive and cool running the system was (~46 degrees for cpu cores).

With C3/C6 disabled, my temps hit 71 degrees (only on 2nd core) and the system was no snappier than the MAX setting (this was with 299.4, of course).

With 299.5 and C1 enabled is that effectively the same as C3/C6 disabled?

I noticed that the temps were getting up there with that setting...

With Max, am I right to assume that you're enabling C6, but not C3?

 

Great program Uncle, thank you! I'm trying to optimize an Alienware M17xR2 (Core i7-720QM & Windows 7) for audio recording and It may be what I was looking for.

I have a couple of questions:
- when you start the program you said somewhere that it automatically turns off clock modulation and chipset clock modulation & sets the multiplier to its maximum. Does it make those changes permanently? I ask because before installing your app I had lots of cracks during audio recording and now it runs almost perfectly (and I only started the app in monitoring mode). I have rebooted the computer a couple of times and still no cracks and pops during audio recording so I guess the clock modulation and chipset clock modulation are still turned off...

- Could you have a look at this thread C1E, C3, C6, EIST, Speedstep, Turbo Boost, Core parking - Gearz.com and tell me what you think about it? I would be happy to have the possibility of turning off C1E, C3, C6, EIST (Speedstep) Turbo Boost & Core parking from your app because my BIOS doesn't allow me to do it (hmm seems that ThrottleStop 2.99.5 is going in this direction, cool). If it works it will make a lot of people in the audio world happy ;-)

Thanks in advance

 

The minimum allowed VID for a CPU is locked into the CPU by Intel and that number varies from one CPU to the next, even with the same model number. I just checked the Intel Spec Update sheet and for a Q9100, it shows:

VID = 0.850-1.100 V

It is certainly possible that your Q9100 had a very low VID in the 0.8 range but most Core 2 mobile Quads don't. The typical LFM VID voltage is 1.00V for many of the 45nm mobile Core 2 Duos and 1.05V for many of the mobile Core 2 Quads.

What is the lowest value that TS lets you select? The VID adjuster in ThrottleStop is limited by what your CPU supports for minimum VID. It can't go any lower than that.

tilleroftheearth: That new button in ThrottleStop tries to limit what C States your CPU uses. If you set this button to C1, it tries to redirect C3 and C6 requests to C1 so your CPU will be in a more ready state and might feel more responsive. By blocking requests to C3/C6, this prevents your CPU from idling down and conserving power. That setting is not recommended if you are trying to maximize battery run time.

The C3 setting redirects any C6 requests to the C3 state. The C6 or C7 setting is how most CPUs are usually set so these deeper sleep states can be used to conserve power. The MAX setting is one I dreamed up. I found a way to redirect C3 requests to C6 so you will probably see little to no activity in the C3% monitoring column when using this option. In theory, a CPU consumes less power in C6 than in C3 so the MAX setting should minimize battery power consumption or I guess maximize your battery run time.

The Kill-a-Watt meter seemed to show this but I'm also well aware that a cheap KAW meter is not accurate enough to reliably show a difference between C6/C7 or MAX. They are all very close. On my Core i5-650 desktop CPU, the C1 setting consumed about 10% to 12% more power than the C6 setting so in theory, C1 should result in a more responsive processor. I'd only use C1 when plugged in and only if I wasn't too concerned about global warming.

I'm just happy that one user finds this new option interesting. I don't think any benchmarks are going to show any repeatable difference so the only advantage is if it makes your laptop feel more responsive, then it probably is.

 

After you reboot, any changes that ThrottleStop has made will be gone until the next time you run ThrottleStop. I'm not sure why it is working so well for you.

Setting that new button in ThrottleStop to C1 might help with audio recording. Let me know.

The M17x-R2 doesn't have any throttling problems that I'm aware of. I guess it can happen during some extreme situations but in normal use and even when pushed to the limit, it's usually OK.

EIST (SpeedStep) and Turbo Boost are OK so I'd leave those on. Try turning off C1E and setting the new button to C1 and see how your laptop runs.

Also run a log file while in monitoring mode so that you will have a record of your CPU's performance in case there are any problems. There's no need to check either of the two Clock Modulation boxes unless your laptop actually has a problem. That's why it's good to monitor for this problem first.

I've heard about Core Parking but I haven't done any tests to show if it is good or bad. I'm not totally convinced that it is a bad thing because the fewer cores in the active state on a Core i CPU, the higher the multiplier can go and the faster the CPU will run. I'm not sure if disabling Core Parking was promoted as a solution when maybe that was not the real problem or maybe Core Parking in reality doesn't live up to Core Parking in theory. I don't have a modern laptop to test on so you will have to try running your laptop with Core Parking enabled and disabled and see if it makes any difference.

 

Most of people doing audio recording just look for stability, responsiveness and low latency so keep up the good work buddy, we need you ;-) I'm gonna test this weekend and post the results. Thanks

 

Let's say I want to test EIST & Turbo & C States OFF, I just have to uncheck those settings in your app right?
And what about the "Multiplier" setting, do I have to force it (12 for my Core i7-720QM CPU if I don't want Turbo)?
Thanks again, I'll run a log and see what happens

 

That statement is not true and that's not how Turbo Boost works. The first post you pointed me to has a few misconceptions like that and makes C State comparisons between the early Pentium 4 CPUs, Pentium III mobile CPUs and the Core i7. They are not the same. Hyper threading is also much improved in the Core i7 architecture.

There has been a lot of misinformation spread about how Core i7 works because there are a lot of monitoring tools that are not 100% accurate. Based on that, people start believing stuff that isn't even happening in their CPU. ThrottleStop follows the Intel recommend method to accurately monitor what these CPUs are doing internally. Most other software does not. Here's my favorite quote:

I look forward to the results of your testing this weekend. It would be great to find out that the new C1 button is just what you were looking for.

Edit: You can use the Disable Turbo option if you don't want to use any turbo boost or you can use a Set Multiplier setting of 12 and that will do the same thing. I totally disagree with turning Turbo off for most applications. You can click away and turn on and turn off whatever you like. When you turn off EIST, ThrottleStop won't work 100% correctly. EIST is not the evil that it has been made out to be in the forums. Most users don't even understand what its purpose is before deciding to turn it off.

Post a screen shot of how you have TS set up when you post a log file or let me know what you were clicking on.

Also go into the Control Panel -> Power Options and set the Minimum processor state to 100%.

Djangology: I put a post in that thread you mentioned and I also posted a link to TS 2.99.5 so the music guys can check out the new C State limiting feature and see if it is useful to them. I'm stirring up the old hornets nest over there.

 

Throttlestop lets me go 1.05 at the lowest. CrystalCPUID and CPUgenie let me go much lower. Again, isn't there a way to toggle the controls off so I can manually set the VID?

 

I'd have to say that Crystal CPUID and CPUGenie are either not accurate or they know something that I don't know. Try using CPU-Z to report VID to see if these ultra low values are actually being used. Just because a program lets you select low values doesn't mean that the CPU is able to run at those values.

The Quads don't support SLFM so most of them have a minimum VID of 1.05. That number is set by Intel at the factory and it is impossible for any software to go below that Minimum. Can you post a CPU-Z screen shot showing a VID number less than 1.05? The only way to go below that value is when your CPU enters the deeper sleep states like C3/C6 but you can't run those voltages when a CPU is in LFM or HFM mode. The CPU ignores requests like that.

Edit: I just tried Crystal CPUID. It uses the desktop VID formula for my T8100 mobile CPU so the numbers it is showing are meaningless. It has a great big list but the really high and really low numbers all have an (x) beside them which means the CPU does not support those values. You can't depend on that information because it is not accurate.

 

I have a i7 q820 and been using throttle stop.
i was wondering what the effect will be if i assign certain cores i.e 0,2 to a game and enable throttle stop to turbo what will happen?? thanks

 

Zedman 5526: That's a good question. I always thought that if you had a game or app that wasn't overly multi threaded and if you used the Task Manager Set Affinity... function that you might be able to get your game to run a little faster.

If you locked a game to the first 4 threads using Set Affinity..., those usually correspond to the first 2 cores so that would allow the other two cores to spend more time asleep which should give your CPU more turbo boost.

That's the theory. In practice, I don't think there will be a significant difference. Tasks get shuffled around rapidly that the CPU does a good job as is at maintaining a high turbo boost level. I haven't seen any real world testing of this so maybe you can try to find some gaming benchmarks to prove whether there is any advantage to doing this. You might want to try running a ThrottleStop log file while testing to see if a higher multiplier is being maintained or not.