The ThrottleStop Guide

cool thanks for the quick reply, I will try. could you give me a link to set affinity and also if I enable log where is it stored?? also does the fact that I ve stoped core parking matter?? sorry for the 20 questions but what does ts bench do??

 

When Task Manager is open, click on the Processes tab and then right mouse click on the .exe file that you are interested in. Choose the Set Affinity... option and that will open up a small window where you can decide what threads in your CPU that task is allowed to run on.

If you enable Log File in ThrottleStop, this will create a file called ThrottleStopLog.txt in whatever directory or folder you have the ThrottleStop program located. It's best to exit ThrottleStop when you are finished testing so all data gets written out to that file. During normal use, this file is only updated once per minute because I'm not a big fan of apps writing tiny amounts of data to my hard drive continuously.

I'm not sure how Core Parking will affect your results. You could try testing with Core Parking enabled and disabled if you want to compare.

The TS Bench is just something I quickly put together that crunches some numbers and will give you a general idea about how fast your computer is running. It's not the end all and be all of benchmark programs by any stretch. I just wanted to include something that a user could test with while adjusting ThrottleStop. wPrime is a better multi threaded benchmark to use when testing but some users had trouble getting that to start correctly so TS Bench was designed to be an option if you had that problem.

The TS Bench is very dependent on how much background stuff you have running on your laptop. wPrime should give more consistent results from one run to the next. For the kind of testing you're doing, you might want to try a 3DMark sort of benchmark or maybe an actual game that has a built in benchmark option.

 

There is a little bit of disagreement in the M15x forum about how turbo boost works in Core i CPUs so I asked my friend ahmadmud to do some testing for me to try to prove this.

He has a Core i7-920XM but it works the same internally as any other Core i CPU. The only difference is the XM CPUs have an unlocked turbo multiplier as well as unlocked turbo TDP/TDC turbo power settings.

The point I was trying to make is that all 8 threads use the same multiplier when threads are in the active state. You can have a different average multiplier in these CPUs when individual threads and cores are going to sleep but when cores and threads are in the active state working on a problem, they all utilize the exact same multiplier.

I thought a good way to try to show this is with the wPrime benchmark. It takes a calculation and divides it up into 8 parts so each thread can work on a different part of the problem at the same time. The default multiplier is 15 in these CPUs. Any multiplier higher than 15 can only be obtained when using the turbo boost feature of this CPU. ahmadmud has set his maximize 4 core turbo boost multiplier to 25. That means if the CPU is operating within the temperature and TDP/TDC turbo power limits, all 8 threads will be able to use the full 25 multiplier.

Here are some screen shots:

TDP@90W TDC@80A BCLK=135.0 MHz
http://img830.imageshack.us/img830/7458/captu11re.png

The first screen shot includes a slight overclock so the base clock is running at 135.0 MHz instead of 133.0 MHz. The CPU appears to be trying to exceed the turbo TDP or TDC limit. When this happens the CPU will start to rapidly turbo throttle. The CPU can cycle turbo boost on and off hundreds of times a second when you reach this power limit. Turning turbo boost off, lowers the power consumption limit below the set values and when the CPU sees that it is back under the limit, it turns turbo boost back on. Intel includes high performance timers within their CPU so software like ThrottleStop can read these timers and determine the average speed each individual thread is running at. You can see in the picture that it is reporting the same average multiplier for each thread.

If only 4 threads were being allowed to use turbo boost, this would show up in the amount of time each wPrime thread is taking. That's not the case. Each thread is accomplishing the same amount of work. That shows that all 8 threads really are running at the same speed.

TDP@99W TDC@90A BCLK=135.0 MHz
http://img703.imageshack.us/img703/2127/captu22re.png

The second screen shot shows the same computer running at a base clock speed of 135.0 MHz. The difference now is that turbo TDP has been increased to 99W and turbo TDC has been increased to 90A. This allows the CPU to use turbo boost more often. ThrottleStop is now reporting an average multiplier of 24.95 which is very close to full turbo boost. It's hard to say if the TDP limit or the TDC limit or maybe even the core temperatures of 95C and 96C might be just starting to limit the amount of turbo boost. It can be any of these factors.

The wPrime times confirm that all 8 threads are working at a similar speed. If 4 threads were only running at the default 15 multiplier while the other 4 threads were mostly running at the 25 multiplier, wPrime would show a huge difference from one thread to the next but that's not the case. The only difference is normal random variation from one thread to the next.

The next two screen shots are very similar to the first two. The only difference is that the base clock has now been raised from 135.0 MHz to 136.3 MHz. As you overclock these CPUs, power consumption goes up and at the same turbo TDP/TDC settings as before, the average multiplier will decrease slightly.

TDP@90W TDC@80A BCLK=136.3 MHz
http://img89.imageshack.us/img89/1894/90742363.png

TDP@99W TDC@90 BCLK=136.3 MHz
http://img41.imageshack.us/img41/4946/captu33re.png


wPrime 32M Final Score
http://img28.imageshack.us/img28/4509/wp32new.png

There's no way in the world you could run a score like that if only 4 threads were being turbo boosted. You need all 8 threads running at full turbo speed to run numbers like that.

Thank you very much ahmadmud for supplying me with this data. It helps show that ThrottleStop is accurate and is correctly reporting that all 8 threads are turbo boosting. It also helps users understand why ThrottleStop is such a powerful tool when used on one of these Intel Extreme CPUs that are unlocked. By adjusting the turbo multipliers and the turbo TDP TDC limits; you can end up with a CPU that can run, fully loaded, more than 100% faster than the speed a Core i7-720QM can run at. The 920XM and 940XM truly are extreme.

 

Yeah the 920XM partnered with Throttlestop is a powerful beast of a cpu just wish we could cool it more .One of the Alienware guys "DR650SE" done WPrime 32M test in 6.599 with his i7 920XM @ 3787.7.9Mhz.

 

is there a link for v3.0

 

Basic, question. The program seems to be locked for my M4500. I know others have used it. How do I unlock?

 

Unclewebb please incorporate different TRL for different profiles. Would be so useful and a donation will surely be on its way!

 

shakennstirred: I haven't released 3.00 quite yet. More projects than time at the moment.

jerryyyyyy: What CPU model do you have? You need an Extreme Core 2 or Core i before you can unlock the multiplier or turbo boost feature. The 720QM is a locked processor and besides monitoring it, there isn't that much ThrottleStop can do with one of them.

widezu69: That's a good idea but it will probably be awhile before I add that feature. Donations are a four letter word so I've been told not to discuss that on NBR.

 

ok no probs, thought it my be out as i seen the version number on a pic lol

 

ThrottleStop 2.99.6

-fixed regular Core 2 CPUs being recognized as Extreme.

 

Thanks for the response. Looking forward to it If donations don't mean much then obviously rep will have to do

 

no problem man, and in fact, i should be the one thanking you for creating such wonderful program...

it's really a must have for all extreme cpu owners

 

2/10/11

Hi, I have questions about this great program called ThrottleStop.

I have a Gateway notebook that's running a Core 2 Duo T7600 @ 2.33 Ghz. It used to have a Core 2 Duo @ 1.73 Ghz and I replaced it. The BIOS in this computer blows so I'm not able to do much with all the cool settings in ThrottleStop (TS).

With TS OFF, CPU-ID tells me the factory settings are:

Core Speed - 997.9 MHz (sometimes it will change to 2300 MHz or so)
Multiplier- x 6.0
Bus Speed- 166.3 MHz
Rated FSB - 665.2 MHz

When I turn TS ON and bump up the multiplier to 10, CPU-ID tells me:

Core Speed - 1663.2 MHz (does not change)
Multiplier - x 10
Bus Speed - 166.3 Mhz (same)
Rated FSB - 665.2 Mhz (same)


What's going on here? Is the resulting 1663.2 MHz the speed of each core? Or do I have a 2.33 Ghz CPU that is somehow only running at 1.66 Ghz when clocked?

I know the performance on the CPU has increased because I do pro-audio, and I am now able to record to record at lower latencies without dropouts. I can see that the CPU meters that come with my audio software (Cubase) aren't hitting in the red anymore.

Thank you very much!
Derek
[email protected]

 

Intel® Core 2 Duo T7600

A T7600 is designed to run at a 14 multiplier and a 166 MHz bus speed. When set like that, you are looking at approximately 2.33 GHz. If you use ThrottleStop and only set the multiplier to 10 then you are limiting the maximum speed of your CPU to 10 x 166.3 MHz = 1663 MHz. Why are you doing that?

Why did you stop increasing the multiplier when you got up to 10? Keep going higher so your CPU can run at its Intel designed speed. Some of the T7600 CPUs are actually T7600G CPUs. These have an unlocked multiplier so you can go much higher than 14. If the latest version of ThrottleStop in my signature shows the UNLK button (unlock), then you likely have the more desirable T7600G.

Your bios will rarely limit what ThrottleStop can do. TS is designed to control your CPU whether the bios is willing to help out or not.

When asking for help it is always a good idea to post a screen shot of ThrottleStop or any other software you are curious about. You can upload images to ImageShack® and then copy and paste the link to the Thumbnail here. Everyone has a Paint program on their computer. All you have to do is push the PRINT SCREEN button on your keyboard (Top Right), open up Paint and select Paste to paste an image of your desktop into Paint so you can save it as a jpeg file and then upload it.

Edit: Here's an example of a T7600G running with a 20 multiplier.
The T7600G is a beast.

http://img291.imageshack.us/img291/6256/throttle.png

 

Well first off, thanks for responding so quickly! I really appreciate that.

The reason I stopped the multiplier at 10 was because the when I got to 12, the computer blue-screened. I tried again with the same result... once I hit 12, blue screen. So I backed off. The picture I sent shows it set at 11.

ImageShack® - Online Photo and Video Hosting




Uploaded with ImageShack.us

 

Here is another link:

ImageShack® - Online Photo and Video Hosting

Thank you very much,

Derek

 

DeeRock:

The reason your cpu is blue screening at 2.33ghz is because you have the voltage set to 0.95. You should be able to run it at about 1.1250 volts without a problem.

I have a t7600G, and 2.33 is perfect at 1.1125 volts, but use 1.1250 just to be safe.

When you have throttlestop off, the cpu is using its own voltage management, which is between 1.0375V-1.300V, depending on the multiplier used at that time. When you force the multiplier using throttlestop, you need to change the voltage yourself to ensure that it will cope at the desired level. Therefore just change the voltage to 1.250 and it will no longer blue screen you.

Hope this helps

Matt

 

unclewebb, first, many thanks for great program! Two questions.

1. My BIOS (P67 + 2500K) have 3 turbo boost options: "long duration maintained" = 1..56 seconds", "short duration power limit" and "long duration power limit". I haven't found the first option in the StopThtottle. Why?

2. From what MSR you reading CPU's power consumption? How about accuracy? This number equals only core power consumption or also includes "uncore"? Datasheets told nothing about it.

 

DeeRock: Take aliveanddead86's advice and give your CPU some more voltage. That will fix the BSOD and then you will be able to run at the maximum multiplier. It's always best when playing around to maximize the VID voltage first and then increase the multiplier. After that you can do some stress tests like Prime95 and then you can slowly try lowering your VID voltage one notch at a time. Do some more stress testing at each level and then try to lower it another notch. Try downloading the latest ThrottleStop version which should properly detect the T7600G CPUs just in case you have one of those. There was a bug in some versions of TS that was recently corrected.

Probably because I don't have any hardware and I've received close to zero feedback about any of the other features for Sandy Bridge that I added. ThrottleStop is free but new hardware is not so I've decided to end further development. I'm not begging for donations which isn't allowed on NBR anyhow but here's the math that in the long run kills many freeware utilities.

Free software = Can't afford new hardware = End of development

There are a few lock bits in the Sandy Bridge CPUs that the bios can set. If the power limits and time limits are locked by the bios then there isn't any Windows software that can do anything about that. If this is the case, you would need to rewrite the bios for ThrottleStop or any program to get access to these power and time limits.

When you click on the TPL button and open up the Power Limits window, are these values locked? If they are locked them the time values are also locked in the CPU and can only be changed when you are in the bios.

I've been waiting for feedback before adding adjustable power time limits. My thought was that adjusting the time limits is not important. The important thing to do would be to adjust the power limits. If you could adjust those high enough then the time factor should not be an issue at all. I was just trying to keep things simple for end users.

Here is the power consumption register that ThrottleStop reads.
It is fully documented in Volume 3A and 3B of the latest Intel docs.

http://img228.imageshack.us/img228/213/energystatus.jpg

It's very easy for software to read this register, say once every second, and then determine power consumption in watts.

burebista did some testing for me and we discovered that register is not always accurate. To determine power consumption, you need to measure actual voltage. This power register is a nice addition by Intel but it seems to use the VID voltage as an approximation for actual voltage. That might be a fairly close approximation on a laptop but on desktop motherboards where you can go into the bios and crank up the core voltage, that approximation becomes less than perfect. I asked for more info on the overclockers.com forum and on xtremesystems but got zero feedback.

Intel knows the accuracy of this new register but I won't hold my breath waiting for them to tell the whole story.

 

Unclewebb... I know NBR doesn't allow donations n such... But you should setup a little website with a little download page and the option to donate... Man I'd be first in line to give you some dough... Your awesome... And id love for you to have the top/newest tech ...

One quick question though... So because these values are read-only if you will... and intel has given even the etreme SB mobile CPU a tdp cap (I read jokers post on his site) this means that the first gen i7's (920/940xm) will be much faster when OC'd right? As the tdp settings aren't locked on those...

Do you think we will ever be able (or you I should say lol) to unlock the tdp ceiling put in place by intel on the sb CPUs? Or is it clear-cut downright impossible? Thanks bro

 

Ok kool that worked. The CPU is now able to hold a multiplier setting of 14, which brings the speed to 2.3 Ghz. The multiplier won't go any higher. Now as you know, the T7600 (no G) is sold as a 2.3 Ghz CPU. Did I gain anything by using TS?

Thank you!

 

The only thing you really gain is the ability to undervolt your cpu, thus reducing the heat created when at full load. Unfortunatly as it is not the unlocked version, you do not have the ablility to overclock.

 

Ok, so if I replace my current CPU with a T7600 G, I should be able to run it up around 3 Ghz, right?

And if I get a T7600G, what should the voltage be? 1.1250 as you indicated above?

 

I know the 2920XM Sandy is unlocked by Intel but what you can do with one of them depends on whether the bios writers lock them or not. Dell could choose to give you a limited amount of adjustment in the bios and then set the lock bits so software like ThrottleStop won't be able to do anything with them after that. It's up to Dell to decide.

I think Intel learned a lot by what TS was doing to some of their CPUs so decided it was time to close some of the loop holes I found.

I'll have to wait and see but the days of unlimited fun like you can have with a 920XM or 940XM might be over for Sandy Bridge. When a user with a 2920XM contacts me, I'll get him to read some registers so I can see if anything can be done. The current version of TS should handle them OK but that's only in theory. I need to see some 2920XM hands on testing.

DeeRock: Use as much voltage as you need to so that your CPU is stable. The VID voltage is also fully adjustable up to 1.50 V on the T7600G. Lots of room to play.

Here is what dreams are made of. The T7600G with sSpec number SL9U5

 

I personally can run my t7600G (I got it for £100) at 3.0ghz, with 1.3875 volts. As I have a XPS M1710 I can BIOS overclock up to 3.16ghz, however, this is just not stable for me, even using 1.5 volts it just freezes.

If you can get a T7600G, then go for it, the performance increase is amazing, but the temperatures can get pretty high. My XPS throttles if the temp is high 70's for more than 5 minutes, therefore I use 2.66Ghz, at about 70 degrees when encoding, and 2.83 for gaming. I hardly ever use 3.0 GHz as it just throttles after a few minutes due to the BIOS. If I set the rating to 3.0GHz is the BIOS, then as soon as it hits 85 degrees it throttles back, so for me its a choice of 3.0ghz then throttle, or change the multi. myself but not to exceed 75 degrees for too long.

As unclewebb states, the sSpec is SL9U5, just have a look through eBay, and maybe sell off your old cpu? I got £80 for mine, which means all this extra power cost me £20, which is not too bad.

Cheers

Matt

 

huge thumbs up for that program

I used RMClock for long time until today. What made me switch - ThrottleStop uses much less CPU to function (i.e. max of 1%), compared to RMClock that I've seen hit 5% often (in the Task Manager). As I'm doing performance improvement these days I happen to notice that last night .. so it seems like I gained 4% CPU power by switching to ThrottleStop (kind of sounds like auto insurance add, lol)

... using v.299 of yours, no problems of setting it up. Just had to dig up what some of the options mean, and you Unclewebb have done wonderful job explaining

small question though
- at x13 Normal my CPU works at 1.025V fine (gave errors at 1.000V)
- at x14 (or IDA) it wants much more for some reason - currently at 1.075 to be stable. (at 1.050 it crushed while running Super_PI).

can it be done so that the max normal multiplier gets the first voltage, while the IDA multiplier gets the second voltage ? Like separate VID setting for the IDA ? ... because for now it seems that if I want to run IDA I have to set the higher voltage, and considering the fact that it's not in use all the time I end up running the higher voltage for my x13 Normal multiplier ... i.e. not that efficient anymore.

big thanks

P.S. My 4 profiles are:
- Performance
- Performance + IDA
- Power Saver (x6 max)
- Test Profile

 

I think everyone worries too much about maximizing efficiency when they really don't have to. Check out this comparison.

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

Mobile 45nm Core 2 CPUs do a great job of reducing power consumption internally at idle regardless of your FID/VID settings. Running them fast is a good thing because the faster they get their work done, the more time they can spend in the power saving C3 and C6 sleep states.

Single IDA is good but Dual IDA is even better.

http://forum.notebookreview.com/win...c-acceleration-ida-both-cores-core-2-duo.html

RM Clock can't do that.

If you're plugged in, live a little and use the IDA multiplier. If you're on battery power and want to improve efficiency, set up a battery only profile with the lower core voltage and with IDA / turbo boost turned off but you will probably find that the actual savings are minimal. I added battery power consumption monitoring to ThrottleStop to try and show this.

It's common for any CPU to need more voltage when you reach the last IDA multiplier. The only option at the moment is to go back to RM Clock. It lets you use different voltages for IDA and the HFM multiplier but it burns more CPU cycles so you probably won't be any further ahead.

 

What feedback you are waiting for if time limits not present in program? You don't know msr number? Can I help you? Even if register locked by BIOS, you can show current settings, for example if time limit=auto. Its interesting info.

You are right. I change Vcore from auto (1.27) to 1.35, but "energy status" msr still show same numbers.

 

I didn't see the point of displaying the turbo time limits if these values are locked by the bios and can't be adjusted but I see your point that it could be interesting information. I'll consider adding that feature.

The Power Limit Control MSR is well documented by Intel so that's not a problem.

Thanks for confirming my suspicions about the new Energy Status register. It was a good idea by Intel but isn't accurate if you adjust the vcore in the bios.

 

My BIOS changes only Time Window for Power Limit #1, #2 always zero.
But I dont know how to convert msr to seconds.
For example, values for 28 seconds:
MSR_PKG_POWER_LIMIT=800083B0 00DC82F8
MSR_RAPL_POWER_UNIT=00000000 000A1003
According to datasheet the result is 20.8 seconds. Maybe BIOS selects number not greater then specified, next 32.

 

I'll decode your MSR registers so maybe other users can follow along.

MSR_RAPL_POWER_UNIT=00000000 000 A1003

0xA in bit[19..16] is 10 so your power limit time units are in 1/(2^10) seconds. That's 1/1024 which is ~976 µs (micro-seconds) which is ~1 ms (millisecond).

The 3 at the end of the POWER_UNIT number means that power is in units of 1/(2^3) watts or in increments of 1/8 watts.

MSR_PKG_POWER_LIMIT=800083B0 00DC 82F8

The first 8 in the last 4 digits shows that Power Limit #1 is enabled.
0x2F8 = 760 / 8 = 95 watts

Limit #2 is usually set to 25% higher than Limit #1.

MSR_PKG_POWER_LIMIT=8000 83B0 00DC82F8

The 8 in 83B0 means Power Limit #2 is enabled.

0x3B0 = 944 / 8 = 118 watts

The Intel formula for time limits is rather confusing to say the least.

MSR_PKG_POWER_LIMIT=800083B0 00 DC82F8

0xDC is in bit[23:16]. I'll convert that to binary to make it a little easier to follow:

Code:

 1   1   0   1   1   1   0   0
23..........................16

bit[23:22] are both set so that is the value 3 so,
F=1.3

bit[21:17] = 0 1 1 1 0 b = 14 decimal

(2 ^ Y) * F = ( 2 ^ 14 ) * 1.3 = 16384 * 1.3 = 21299.2

To convert that to seconds you need to divide by 1024

21299.2 / 1024 = 20.8 seconds

We both got the same answer so we must be doing something right.

If your bios shows 28 seconds then based on Intel's formula, I'd have to say that your bios is wrong or is setting your CPU wrong.

The Time Limit #2 is all zeros so,
F=1.0
Y=0

(2 ^ Y) * F = ( 2 ^ 0 ) * 1.0 = 1 * 1.0 = 1

That tells me you can only exceed your upper 118 watt power limit for one time unit which is only 976 µs or about 1 ms. For 20.8 seconds you can run between 95W and 118W but after that, turbo boost will decrease if it has to so long term you will be forced to run under the 95W TDP limit.

That's how I understand this. Let me know if you disagree or if I screwed a calculation up.

MSR_PKG_POWER_LIMIT= 800083B0 00DC82F8

The very first 8 in this value shows that bit[63], the lock bit, is set so software like ThrottleStop can't change either the power limits or the time limits. For the mobile CPUs, I think only the 2920XM will have these values unlocked. Some Gigabyte desktop boards have this MSR unlocked while Asus and Asrock seem to lock it by setting bit[63].

 

Hmmm. 324644.pdf dont have POWER_LIMIT_2_TIME and also says:

That you think about that?

Turbo algoritm works not exactly as you described, at least for Sandy.
Example:
BIOS long power limit = 50W, short power limit = 150W, time = 28 seconds (actually 21?). DTS was -50 during the tests.

winrar benchmark:
MSR_PKG_ENERGY_STATUS=35 (far below 50)
FID drop from 34 to 33 (turbo turned off) after 60..65 seconds.

TS build-in benchmark:
MSR_PKG_ENERGY_STATUS=42
FID drop from 34 to 33 after 30..35 seconds.

As you see, turbo algorithm dont use MSR_PKG_ENERGY_STATUS (or use it very specific) and time window dont exactly equal time limit value.

 

TS BUG
http://savepic.net/522824.png

 

Thanks for the bug report. I've seen a few screen shots like that but I've never been able to duplicate it. I'll have a closer look at my code.



As you know, the MSR_PKG_ENERGY_STATUS MSR is not 100% accurate because it is not based on actual core voltage. Intel might not use this information to determine turbo boost or it might build in a margin of error to be on the safe side. Maybe Intel assumes a 5W margin of error. I've seen one Sandy Bridge mobile CPU that was running at about 40W when long term stress testing even though the Intel TDP limit was 45W. The long limit was 56W but it seemed to be limited to about 50W for a short period of time before turbo throttling to only 40W.

Can you show me where that quote came from? If you can, post some pictures of your tests showing ThrottleStop so I can see what it is showing. Try to include a ThrottleStopLog file too. It will be easier to see when the load goes up and when the throttling begins.

 

Howdy unclewebb, been playing around with ThrottleStop now that i have repasted my CPU.
I can get my WPrime 32m score down to 8.4 with the following settings

Turbo Ratio Limits = 26 for all four cores
TDP/TDC = 90

I can up the TDP etc but it seems to be capped by the ratio limit, is that correct? so basically my 920xm wont go any higher?

 

betc: If you are running wPrime and the average multiplier that ThrottleStop reports is 26.00, then increasing TDP or TDC isn't going to make that go any higher. You are already getting maximum turbo boost. The only way to go higher is to go into the Turbo Ratio Limits window and increase some of those limits to 27 and start testing again by adjusting turbo TDP / TDC and see what average multiplier you get. I can't remember seeing anyone running wPrime with the 27 multiplier on all 4 cores reliably. I know 5150Joker used to run with the 26 multiplier on all 4 cores but his core temperatures were getting up over 90C so at this level, you are getting close to the practical limit.

 

Quick question.

I'm using ThrottleStop v2.95.

I have dual IDA enabled in BIOS and in TS. I noticed that on my T9300, the max VID I can set is 1.2125 in TS. I know the T9300 can do 1.25v. I'd like to know what is the limiting factor. The program or my BIOS?

I ask, because I plan on doing a hardware overclock of the PLL soon. I'd like to be able to take the CPU up to 12.5x 266mhz FSB = 3.312ghz. But I'll likely need 1.225v-1.250v for a stable O/C at this FID.

 

The maximum VID is set by Intel when a CPU is manufactured. I believe ThrottleStop is reading this maximum value correctly.

Intel® Core 2 Duo Processor T9300

VID Voltage Range 1.000V-1.250V

Not all CPUs have the exact same maximum. It varies from one CPU to the next, even with the same model number.

My T8100 also lists this exact same range but the maximum IDA VID is locked by Intel at 1.200.

 

Well if it varies from CPU to CPU why bother even posting such specs up? My T7500 isn't stable anywhere near 1.000V with RMClock trying to undervolt the idle, yet people can get it under 1.000V..

Edit: Would ThrottleStop be better for undervolting?

 

Well, I'll just have to hope it will work @ 1.2125v when @ 3.3ghz. It might, considering I'm running 2.7ghz(13.5 FID) @ 1.0625v.

 

Likely will be OK since you want to run 12.5x266, non dualIDA. Your current dualIDA mode requires a considerable increase in voltage.

 

I posted the specs because I was trying to understand where he was getting his 1.250 number from.

If you have RM Clock set up and it does what you want it to do then there is no need to change to ThrottleStop.

 

Yep, that's where I got the 1.25v figure from. Intel.

Incidentally, my other T9300 has a max of 1.2125 also.

But as Nando says above and I tend to agree with, I should be fine with that specific max for what I intend to accomplish.

 

cheers for that uncle web. im starting to get that the limit is already getting hit. its jDP90 and 26 multipler. well not 90c anyway, around 80c. looks like thats it though

 

​

Unclewebb

Hello. Have you ever checked ThrottleStop against Windows Perfmon?
Are Perfmon counters inaccurate?

Have you found information on the wandering multiplier when the C6 state is operational?

By the way do you remember my tests with Linux keeping the lowest multiplier at idle?
Intel PowerTOP shows C3 is the deepest state CPU reaches - very much like Perfmon on Windows.

​

P.S., if interesting - probably the same you get with your i5-650 system.
This is CPU under full load and (of course) both monitors agreeing:
YouTube - CPU stats
Not as much demand:
YouTube - CPU stats

 

No, sorry I wasn't referring to you, but Intel's ark site.

I wish to undervolt it even further, and I've hit a wall with RMClock.

 

Tsunade_Hime: Both RM Clock and ThrottleStop are ultimately limited by the minimum and maximum VID that Intel sets for the CPU. On some laptops, RM Clock doesn't let you access the SLFM voltages. If that's the problem, you can try ThrottleStop.

o770: I've never compared ThrottleStop to Windows Performance Monitor. It looks like every sleep state C3 and deeper all gets reported as C3%.

This is the information that ThrottleStop reports as C3% and C6%.



Unfortunately, my Core i5-650 test rig was picked up by its new owner so I don't have any hardware to test my theories on. What I seemed to figure out was that individual threads and cores can still go into the C3 or C6 state even when you try to block this by clicking on the C6 button in ThrottleStop and setting that to C1. That button seems to block the entire CPU package from entering the deeper sleep states but it does not stop individual cores or threads. The information reported by these counters is not comparable to the information being reported by Windows Performance Monitor since it is showing individual cores while ThrottleStop is reporting the sleep state of the entire CPU package.

My best guess is that when the CPU comes out of C6, the CPU automatically goes up to the highest multiplier before idling back to the lowest multiplier, going into C6 and then waking up and repeating this cycle. That might give you an average multiplier that wanders endlessly back and forth at idle as it cycles from low to high.

I have a suspicion that this could be a bug in the original Core i design. With Sandy Bridge, you can use C6 and the multiplier is rock steady at idle. The minimum multiplier for a Core i5-2500K desktop CPU is 16.

 

Thanks Matt and UncleWebb! You guys inspired me so much I went and bought a T7600G. Right now there is a guy on Ebay who has a few of them (299.95 USD) I need to have my computer operating at it's highest potential for my audio work and I don't believe in cutting corners when it comes to my tools. I'll be consulting this page for your info about settings after I install it. I'm excited!

Thanks again so much for your innovation, information and kindness!

Derek

 

Hrm okay, I think I will try ThrottleStop for my Latitude 13 as it has the ULV Core 2 processor, and that doesn't work with RMClock.

 

Tsunade_Hime: On the Alienware M11x with an SU7300, Dell turns off the EIST bit in the processor and then locks it in the disabled state whenever you overclock it in the bios. Once EIST is disabled, ThrottleStop can't adjust the FID or VID anymore. As long as your Latitude 13 isn't doing that, ThrottleStop should work OK.

DeeRock: I can't wait to here what speed you can run your new T7600G CPU at. Hopefully it puts some new life in your old computer.