ThrottleStop - Performance Adjustment Tool for Core 2 / Core i CPUs

[Movieman]

I got him..
He had all of 2mins and now he's history.

[unclewebb]

Doesn't anyone have a laptop? Doesn't anyone find ThrottleStop useful? It's now finally possible to overclock Extreme mobile CPUs by 35% and jack up the voltage VID with or without bios support. I find that kind of useful.

Head to the first post in this thread to learn about the new features.

[Hondacity]

longer battery life....!!!!

thanks uncle!!!

will test tomorrow..i have a dell core2duo

[Hockster]

Would this work for desktops with unlocked processors?

[unclewebb]

It works to unlock the desktop Core 2 Extreme CPUs so you can increase the multiplier when in Windows but I haven't added support yet for the new 655K and 875K CPUs. I don't have any new hardware to test on so it might be a while before I add this feature.

Hondacity: Depending on what CPU you have, you might be able to use ThrottleStop to overclock it a little. Check out the links in the first post for more details. You can also reduce the core voltage and temperatures but that's not as much fun as going faster.

[unclewebb]

ThrottleStop 2.55
http://www.mediafire.com/?zzmn0wwzktz

My friend somebody over on TechPowerUp sent me some info about adjusting the multiplier on the new 655K and 875K CPUs. I've tried to add what I learned to this version of ThrottleStop but without any hardware, I have no idea if this will work or not. Just right click on ThrottleStop and select the Max Turbo Limit menu item and it should make sense after that.

Edit: This new feature seems to be working as intended and it might also work on the Core i7-965 and Core i7-975 Extreme CPUs. It should let you increase the maximum multiplier similar to what ELEET lets you do.

[Chrysalis]

interesting, so this tools allows lower voltages at idle whilst not using auto voltage in bios?

[unclewebb]

Core i CPUs do not support adjustable core voltage. Only the Core 2 CPUs support this. If you are using a Core 2 desktop CPU, this will only work if your motherboard supports this feature and it is set up correctly. Usually you would have to put the voltage on the AUTO setting in the bios. A Core 2 mobile CPU should work without needing any adjustment in the bios.

[Chrysalis]

unclewebb they do, my i5 does it.

1.15v on load and 0.85v on idle.

[Brama]

unclewebb they do, my i5 does it.

1.15v on load and 0.85v on idle.

My core i7 too.

[unclewebb]

What I should have said is that I don't know of any Windows software that allows adjustable core voltage on the Core i CPUs. On Core 2 CPUs, there used to be a voltage request register that software could write to which would ask the CPU to change the voltage. This register has been changed so you can't do that anymore. There might be another register somewhere that you can use to control voltage but Intel says that isn't possible anymore on the Core i CPUs.

[PiLsY]

Any chance of adding atom support or is that a totally different kettle of fish? I can oc this 1600 nicely at stock volts but its pointless in a netbook - dropping the vid would be awesome...

[unclewebb]

The Atom CPU doesn't have high performance timers inside the CPU that ThrottleStop depends on. These timers are either disabled or don't exist in the Atom. I might be able to create a hacked version of TS for you to check out but I probably won't do a general release. I'm not sure if it will be useful for you but I guess it won't hurt to try and support your Atom. Maybe tomorrow. I'll send you a PM when I have something for you to try.

[PiLsY]

Awesome, thanks dude . Anything to try and get a bit more battery out of it. I can get about 6 hours browsing at the moment - it never gets used for more than that.Have tried dropping cpu core clock but without dropping volts too it makes little difference. Not sure how much difference the volts would make either tbh, but given how little power it consumes even 5 watts or so dropped would help.

[unclewebb]

If you are using your netbook for surfing the net, I don't think dropping the VID voltage is going to make any difference. Most people don't realize that these CPUs are designed to enter a lower voltage state at idle. Lower than the VID setting in RM Clock or ThrottleStop lets you access. I'm interested to see what difference it makes so I'll try to come up with something you can test but I don't think you're going to see any significant difference in battery life.

[PiLsY]

I was thinking if I can lower the default vid then it will lower the idle voltages by the same amount too? Or do C-States have fixed voltages? I always thought it was set by VID...

[cstkl1]

I've been thinking about combining a few of my tools into a new program to give better thermal control over laptop / mobile CPUs.

I'm not yet sure what's possible but I did some testing tonight on a T7200 which uses a VID of 1.2750 at full load. When running a single instance of Prime95 Small FFTs, the core 0 temperature was up to 86C.



CPU-Z doesn't report the actual core voltage for this laptop, just VID, so actual core voltage was likely slightly less than that due to vdroop. I adjusted VID as low as it could go and I ended up at a VID setting of 1.0625 volts.

One thread of Prime95 was still running as I dropped VID lower and lower and lower and look at the results. In under 4 minutes, Core 0 had dropped 20C from its maximum of 86C down to 66C with this one simple tweak. I think most laptop users could appreciate a simple app that gave them this much control.



I didn't run a gazillion hours of P95 but it was obvious that this CPU is capable of running reliably at full speed without needing anywhere near the default voltage. You would save a lot of power and battery life when watching a movie or whatever without sacrificing any performance.

My plan would be to support Core 2 mobile CPUs including the 45nm series like the P8400. I like the possibilities but I need some positive feedback before deciding how much effort to put into this new project.

What do you think?

Magic word.. P8600.. i am sold..

please oh mighty one..

[unclewebb]

I was thinking if I can lower the default vid then it will lower the idle voltages by the same amount too?

I found that this isn't true. At idle, many of the 45nm CPUs including the Atoms, depending on bios support, go into a deeper sleep state where the traditional VID value is not used at all. The CPU uses a different VID value that may be controlled by some register in the CPU but this magic register, if it exists, is not documented by Intel in the publicly available documentation.

That's why I don't think being able to control VID is going to make any significant difference to your Atom, especially at idle. It might help a little when loaded. I'll try to get around to finishing that Atom version I promised a few days ago so you can test this out.

[PiLsY]

Ah crap . Still happy to test it out for you if needed - doesnt sound like its going to get me the 4.5hours battery life I was promised on the box though. Bloody liars!! Have to run the cpu at 50% max to get over 3 hours.

[stasio]

ThrottleStop 2.77

http://www.fileden.com/files/2008/3/...rottleStop.zip

[GrossmeisterB]

I have a Laptop with T5500 intel-CPU, I want to lower power consumption (especially when gaming that thing gets really hot!!), so what are the settings I should use?

[JohnZS]

I have found an unusual issue with ThrottleStop Kevin.
It might be localised to my system as it uses an ASUS motherboard (which has ACPI issues etc), but it is something to look into.

With a QX9650 it is possible to change the multiplier within windows using your fantastic ThrottleStop 2.77 tool,. HOWEVER performance degrades when Mhz goes upwards.

Initially when I reported this to you the majority of my testing was done using an old command based memory benchmark (membench). We both decided back then it wasn't a good test to use as it was probably not even 7 aware, yet alone 7 x64.

Anyway, I decided to test using a CPU bound game called Grand Theft Auto Episodes from Liberty City.

The test results are rather odd.

Here is the first test with the QX9650 @ Stock

Statistics
Average FPS: 47.56
Duration: 36.50 sec
CPU Usage: 75%
System memory usage: 38%
Video memory usage: 98%

Graphics Settings
Video Mode: 1920 x 1200 (60 Hz)
Texture Quality: High
Shadow Quality: High
Reflection Resolution: High
Water Quality: Very High
Texture Filter Quality: Anisotropic x8
Night Shadows: Off
View Distance: 25
Detail Distance: 40

Hardware
Microsoft Windows 7 Ultimate

Video Adapter: NVIDIA GeForce GTX 295
Video Driver version: 259.31
Audio Adapter: Speakers (Creative SB X-Fi)
Intel(R) Core(TM)2 Extreme CPU X9650 @ 3.00GHz

I now RAISE the multiplier to 9.5 (3172Mhz)

Statistics
Average FPS: 46.23
Duration: 36.54 sec
CPU Usage: 75%
System memory usage: 39%
Video memory usage: 98%

Graphics Settings
Video Mode: 1920 x 1200 (60 Hz)
Texture Quality: High
Shadow Quality: High
Reflection Resolution: High
Water Quality: Very High
Texture Filter Quality: Anisotropic x8
Night Shadows: Off
View Distance: 25
Detail Distance: 40

Hardware
Microsoft Windows 7 Ultimate

Video Adapter: NVIDIA GeForce GTX 295
Video Driver version: 259.31
Audio Adapter: Speakers (Creative SB X-Fi)
Intel(R) Core(TM)2 Extreme CPU X9650 @ 3.00GHz

Notice the score has dropped slightly, I then decided to RAISE the multiplier again to 10.

Statistics
Average FPS: 43.95
Duration: 40.50 sec
CPU Usage: 75%
System memory usage: 39%
Video memory usage: 98%

Graphics Settings
Video Mode: 1920 x 1200 (60 Hz)
Texture Quality: High
Shadow Quality: High
Reflection Resolution: High
Water Quality: Very High
Texture Filter Quality: Anisotropic x8
Night Shadows: Off
View Distance: 25
Detail Distance: 40

Hardware
Microsoft Windows 7 Ultimate

Video Adapter: NVIDIA GeForce GTX 295
Video Driver version: 259.31
Audio Adapter: Speakers (Creative SB X-Fi)
Intel(R) Core(TM)2 Extreme CPU X9650 @ 3.00GHz

File ID: TLAD-Benchmark.cli

Again another dramatic drop in FPS.

I then re-ran @ Stock with NO Throttlestop (by rebooting my PC) and got a score of 47.58 (very similar to the initial run of 47.56).

Do ASUS have something funky going on in their BIOS? Or is this my horrid c0 QX9650 deciding to give us a headache again.
John

[unclewebb]

JohnZS: My friend burebista discovered something interesting when testing with SetFSB. The internal Windows timers that some benchmark programs use can get screwed up when you start overclocking.

http://www.hardwarecanucks.com/forum...tml#post383693

I don't think this problem applies to your QX9650 but after seeing that test, anything is possible.

When you use ThrottleStop to adjust your multiplier higher in Windows, on some motherboards, that might be changing timings within your chipset so memory bandwidth performance decreases. I haven't seen that problem when testing with a QX9650 on the older Intel P965 chipset. If the Windows timers are not being screwed up by adjusting the multiplier then I would have to assume it is a bug or limitation of the chipset you are using.

Can you post or send me a link John of a freely available benchmark program that shows this problem on your computer? I'd like to run that bench on my computer just to see if this problem can be isolated.

I think the reduction in memory performance that you measured earlier is real. Another user with a QX9300 mobile CPU which was on a board with an Nvidia chipset showed the same thing when ThrottleStop was used to increase the multiplier from within Windows. In his tests, the increase in CPU speed more than made up for any decrease in memory performance so overall benchmark performance increased even when memory performance decreased.

Edit: Here's a testing program I wrote to make sure the timers in your computer are running at the same speed when overclocking. If your internal timers get out of whack, some benchmark programs won't be reliable. The long term timer ratio reported by this program should be 1.0000

WinTimerTester
http://www.mediafire.com/?220h640tyxiwwpi




ThrottleStop 2.80
http://www.fileden.com/files/2008/3/...rottleStop.zip
http://www.mediafire.com/?89u0sctu15vn9g2

[JohnZS]

Thanks Unclewebb
I shall have to give that a test later. IF it is a windows timer thing I may have a fix as I managed to fix setFSB back in the days of the 865P maintaining PAT when over clocking days.
Windows has a switch /usepmtimer which you can add to the boot ini to force Windows to use the HAL ACPI timer instead of it's own timer, this enabled me to overclock and keep PAT back in the days of the Pentium4 S478 Northwood processors.
I think with Windows 7 I will need to use bcdedit as I recall Boot.ini editing was laid to rest by Microsoft, however I may need to consult with technet to find that out...
John

[unclewebb]

Thanks for the info John. I read about the /usepmtimer fix for Windows XP but I don't know how to make that happen in Windows Vista or Windows 7. Let me know if you get this figured out. I think there are a lot of programs that can get screwed up by this bug because everyone assumed that Microsoft had fixed it. It's pretty dumb when the most accurate way to measure time in Windows can become completely inaccurate as soon as you make an adjustment with SetFSB.