Awesome! I should really start reading up and installing PhenomMsrTweaker now. Thank you so much! :)Quote:
Originally Posted by kink
Printable View
Awesome! I should really start reading up and installing PhenomMsrTweaker now. Thank you so much! :)Quote:
Originally Posted by kink
You're welcome. Oh and I forgot to mention that custom C&Q uses only P0 and P1, not P0-P3 like standard C&Q on K10.5 CPUs (like yours).
Thanks for looking into it for me, kink. I think it may be time to bust out my #2 and try my luck at shading. :)
any K8 & Turion Ultra version? would be nice :up:
Nope, I've finished working on this project, I don't find it that interesting anymore. K8 support shouldn't be that hard to implement though as technical docs are available too.
well, k8 version would be exellent for Turion laptops
and Athlon 64 desktops :yepp:
Hi Kink, this is really an awesome tool you've created! :up: Exactly what I need. Thank you so much!!
I just have a few questions.
1. Do I need to install the AMD Cool n' Quiet driver for this to work? I don't remember installing this driver when I installed Windows XP on my machine. But I installed the AMD Processor Driver v1.3.2.0053 from AMD's site and PhenomMsrTweaker seems to be working fine. Is AMD Processor Driver the same thing as the AMD Cool n' Quiet driver, just a different name?
2. What happens if you do not enable Custom Cool n' Quiet? Will the Standard Cool n' Quiet take over?
3. How do you define "CPU Load"? Is it the same as the "CPU Usage" percentage that you see in Windows Task manager? So if you set an Overall CPU Load Threshold of 23%, does it mean that the performance state will kick in when the CPU usage showing in Task Manager goes higher than 23%?
4. I'm now looking at the modified 3-performance states version of PhenomMsrTweaker, and it has these three threshold settings that I'm not sure I understand:
Overall CPU Load Threshold (%) High/Low: (the default is 32%)
Overall CPU Load Threshold (%) High/Middle/Low: (the default is 32%)
Overall CPU Load Threshold (%) Middle/Low: (the default is 32%)
Can you please explain how these settings work? Like, if I want the high performance to kick in when the CPU usage is more than 70%, the middle performance to kick in when the CPU usage is 30% to 70%, and the low performance to kick in when the CPU usage is below 30%, what values should I enter in these settings?
Again, thanks very much! Your work and help are greatly appreciated. :bows:
Hi there.
1. Nope, no driver required. Your driver is the C&Q driver.
2. Yep, at least as long as you re-enable C&Q in the BIOS menu and select a powersaving profile in control panel -> energy options (or the like). You need the AMD driver for that to work.
3. Almost - it's the complete CPU usage, including kernel time (time spent in OS kernel mode).
4. I haven't taken a look at the mod yet, and probably never will, therefore I can't help you. The labels seem to be rather crappy though, I wouldn't understand it myself.
Thanks, man! :)Quote:
Originally Posted by kink
So, since the CNQ driver isn't needed, I can even uninstall it and everything will be fine? Is there nothing at all in Windows that needs it if you're not using standard CNQ?
BTW, I'm not exactly sure what the "Make custom P-state settings permanent" option is supposed to do. It seems to set the system to a fixed P-state which never changes even if the load drastically changes, in effect disabling P-state transitions even if custom CNQ is enabled. Is that what it's supposed to do?
Yes I guess the driver is only required for standard C&Q, having read the description on the webpage.
If you choose to make custom P-state settings permanent, the customized settings (multipliers and both voltages per P-state) are applied automatically during boot-up and when resuming the system from stand-by mode. This is only useful if you customized the settings beforehand, obviously. It shouldn't interfere with custom C&Q, but again, I haven't tried the mod.
Thanks! :)Quote:
Originally Posted by kink
Is it normal for my core voltage to fluctuate in CPU-Z? My CPU voltage is set to 1.3V in the BIOS but it varies from 1.304V to 1.336V in CPU-Z. This happens whether or not I have PhenomMsrTweaker installed. With PhenomMsrTweaker installed, this is during P-state 0 in which I have the CPU VID set to 1.3V.
Absolutely normal. Under load, the voltage is usually slightly reduced, the amount depends upon the number and quality of the voltage regulators on the mainboard..
Thanks! :) I was starting to worry that my power supply is bad.Quote:
Originally Posted by kink
quick question: How would one know what an appropriate P1 (power saving) state would be?
I also want to know the answer to this question. So, Kink, do you just set a CPU multi that you would like to be at in the P1 state, and determine by trial and error what lowest CPU voltage you could go before the system crashes, then bump it up a little (kinda like the reverse of overclocking)?Quote:
Originally Posted by Rezag1000
Let's take a specific example. In my case I have a Phenom II X3 720 BE which I overclocked in the BIOS with the following settings:
HTT Reference Frequency = 250 MHz
CPU Speed = 3625 MHz using 14.5 multi
CPU Voltage = 1.4375V
NB Frequency = 2500 MHz
NB Voltage = 1.2625V
HT Frequency = 2000 MHz
My P0 state settings in PhenomMsrTweaker are:
Core Multi's = 14.5
CPU VID = 1.4375
NB VID = 1.2625
If I want my core multi to be 7 (for 250 MHz x 7 = 1750 MHz CPU speed) in the P1 state, what voltages would you suggest for CPU VID and NB VID in this state?
I would really like to get this program to work on my machine. The one thing that is keeping me from overclocking my CPU (again), is that Cool n' Quiet works so great when idle. And on my motherboard I can't change multiple and keep Cool n' Quiet.
Anyhow, I installed the x64 version of the program. And as soon as I execute the exe file, it just says " PhenomMsrTweaker stopped working". I have no idea why, I've looked through this whole thread, installed Visual C++ Redist 2005 and 2008, x86 and x64. I also tried the x86 version of PhenomMsrTweaker, still nothing. I do also have .NET Framework 3.5 SP1.
My computer:
Phenom 2 x4 940
Windows Vista x64 Service Pack 2
The error message:
Quote:
Description:
Stopped working
Problem signature:
Problem Event Name: CLR20r3
Problem Signature 01: phenommsrtweaker.exe
Problem Signature 02: 1.2.2.0
Problem Signature 03: 49c1aac3
Problem Signature 04: PhenomMsrTweaker
Problem Signature 05: 1.2.2.0
Problem Signature 06: 49c1aac3
Problem Signature 07: 127
Problem Signature 08: 16d
Problem Signature 09: System.Exception
OS Version: 6.0.6002.2.2.0.256.1
Locale ID: 1053
anyone find this programe not working as well as the default CnQ?
while I adore the availability of options, it seems to switch between the two states alot more as compared to the default CnQ
anyone got a different setting on the CPU Load Interval and all?
I have my CPU load threshold at 70%. This means that the performance state kicks in only when the load goes higher than 70%. So most of the time I am at power saving state and only shift to performance state when really necessary.Quote:
Originally Posted by wilflare
@Rezag1000 & mtl777: There's no right answer to your question about an "appropriate" powersaving state as it completely depends upon your needs and liking. The Phenom I halves its speed in P1, my X4 9950 defaults to 1.3 GHz @ 1.05V; the Phenom II uses 4 states, the lowest P3 state is set to 800 MHz at some voltage < 1V - correct me if I'm wrong, I think I've read it somewhere. I personally set myself a voltage limit (1V currently) and try to use the highest stable frequency (currently 6x262 MHz = 1.57 GHz). 4x262 MHz = 1.05 GHz seems to be stable at 0.85V, but I prefer wasting some minor Watts and having that extra 50% speed in P1 (=> lower percentual CPU usage, P0 doesn't kick in for very short loads).
Make sure to use tested stable settings for P1 too, don't assume P0 kicks in immediately when there's load. I use PhenomMsrTweaker's custom C&Q, set the power scheme to 'Powersaving' so that only P1 is used and then run a Prime95 session (large FFTs) for at least 2 hours. I suggest to set the NB VID to the same value as for P0 because the NB runs at the same speed all the time (that's mentioned on the webpage).
/edit: Just to demonstrate the usefulness of separate multis in P0: I'm currently running 11.5|12|11|11.5x262 MHz = 3|3.15|2.88|3 GHz @ 1.425V after having figured out that the second core performs best of all and is able to compensate for the "bad" third core. The result of many painful Prime95 sessions and crashes... :D
Hi Kink, good to hear from you again! :) Is there a limit set by AMD as to how low your core voltage can go in P1? Like, can you ever go below 0.7V?Quote:
Originally Posted by kink
In my case I need my NB VID to be 1.2625V at P0 in order to make my 3625 MHz (250 x 14.5) CPU overclock stable. Do I still need 1.2625V NB VID at P1 considering that I'm underclocking my CPU to 1250 MHz (250 x 5) at this state? I thought that raising the NB VID was necessary to support (1) the CPU overclock and (2) the higher NB frequency due to the overclock. But in P1 state, (1) is eliminated (since I'm underclocking), and only (2) remains (since my NB frequency is still the same), so wouldn't that require slightly less NB VID then?Quote:
Originally Posted by kink
Hey,
hmm I don't recall reading something about a minimum Vcore - I think the CPU gets unstable pretty quickly at voltages below 0.8V. Additionally, there's not much point in undervolting below 0.8V as you're only going to save 1-2 Watts, I guess.
I don't think the NB VID needs to be raised to support core overclocking, at least I haven't experienced that myself. But you are obviously free to experiment, just make sure to check for stable operation (use a somewhat memory-intensive test like large FFTs).
I was just wondering if AMD set a lower limit as a safety precaution. I guess it's not possible to damage the CPU by going too low with the voltage but I wanted to be sure.Quote:
Originally Posted by kink
I have tried 1.2V NB VID but no stress tests yet, just normal PC usage. I find that the machine very occasionally freezes, like once a week. I'll try large FFT tests next. Thanks! :up:Quote:
Originally Posted by kink
Thank you kink. Great tool! :)
Use it on Win7 64.
Hello Kink,
I would like to thank you for this great app. It is really a marvel.
I was simply wondering if you could integrate a 3rd state that goes between P0 and P1. The reason I ask is because scrolling some websites is laggy since it sticks to P1, and I have tried many settings tweaks that resulted in continuous switching.
I am running Windows 7 64-bit RTM and AMD Phenom II X3 720 BE.
I would really appreciate if you could let me know if it's possible, and if you have any other tips/recommendations.
Thanks,
Why not adjust your p1 state to accomodate your browsing!
I'm running a 940BE and running it at 1.4ghz @ 1.05v. This runs p95 and OCCT linpack stable and havent' had ANY issues with it.
On a very basic w/c system i get 28C idle and 40C load on my p1 state, compared to 32C idle and 48C load on my p0 state (3.5ghz@1.35v).
This is running my cpu-nb vid @ 1.325v and cpu-nb @ 2520mhz.
Also adjusting the amount load settings might help aswell (or manually setting it to high performance/power saving just make sure you have the phenommsrtweaker app running).
My 2cents anywayz
Hello.
I recently installed Windows 6 64 bit and now the program works!
However, I want to use three states or more, because my overclocked state got some higher vcore than the unclocked one, obviously.
W7 x64
Phenom 2 x4 940
I downloaded this modded version:
http://www.hardwareluxx.de/community...&postcount=171
Seems to work well, however, I don't understand the "Overall CPU load threshold" options, here they arem the different states are called Low, Middle and High:
Overall CPU load threshold (%) high/low: 24
Overall CPU load threshold (%) high/middle/low: 24
Overall CPU load threshold (%) middle/low: 24
Could someone explain the differences? Right now my states are at 800 MHz, 3 GHz and 3.6 GHz, and when idling or just using the webbrowser it constantly jumps between state 1 and 2, which shouldn't be needed.
I used K10Stat before, which I rarely had to jump to state 2. And it doesn't make sense that it should, 800 Mhz is plenty for that.
Anyone who might have an idea?
Hello Glaucous,
I have tried the 3-states modded version and I have the same problem as you except that I have toned it down a little bit.
I did not fully understand the way the 3 CPU load thresholds work since I have tried many settings and I was not able to conclude anything major.
I am running a Phenom II X4 720, and my settings are the following:
High:15x, 1.3250v
Middle: 8x, 1.1000v
Low:4x, 0.8250v
Overall CPU load threshold (%) high/low: 60
Overall CPU load threshold (%) high/middle/low: 55
Overall CPU load threshold (%) middle/low: 45
Let me know what do you think.
Also, what about K10Stat? any good?
I registered pretty much solely to say thank you to Kink for this terrific program.
My gaming framerates in WinXP tripled compared to stock CnQ with a recent Regor 240 system. More details here.
I do have two questions/items for Kink (or others) (and apologies if they were answered already in the six-page thread):
• First off, MsrTweaker seems to work fine whether I disable CnQ in the BIOS or not. I am assuming the only potential problem with NOT disabling it would be that if I accidentally selected minimal power management in display properties, the official CnQ and MsrTweaker could conflict, correct?
• Secondly – MsrTweaker really shows relative values - for example, I wanted a voltage of 1.2 and 1.1 for CPU Vcore. MsrTweaker does fine with my Regor if I don't mess with the BIOS voltage - it drops the stock 1.425 by 0.225 to set 1.2 at the CPU. However, at one point, I had the BIOS voltage lowered by 0.15 or so, and setting MsrTweaker to 1.2, it still dropped that by 0.225, resulting in 1.05 at the CPU - bottom line, any BIOS voltage correction will be added or subtracted from MSRTweaker's selected value.
Finally, I have not tried and have no reason to try the tri-state version that was asked about above and Kink said on the previous page the labels were confusing, but I have a WAG as to what the settings should be – but I could be totally wrong.
For tri-state, you really only need two transition values – i.e. low to medium at 40% CPU load, and medium to High at 75% CPU load (Kink’s file only has one value for low to high).
A previous post said the default values were 32, 32, and 32 – which comes pretty close to adding up to 100% - but not quite.
I am thinking those values would make the processor transition at 32% and 64% usage, and if you wanted my example above of 40, and 75, you would set the values from top to bottom at 25%, 30%, 40%.
I.e. – I suspect the bottom number is processor load before the medium transition, the middle number is additional load until the high transition, and the top number is just 100 minus the other two values.
But I don’t speak or read German, I am not the author, and I haven’t tried the program, so be careful with anything I said.
Again - thanks to Kink for the original version.
I have switched to K10Stat, and it allows for far more customizations, and 4 states. Had MSRTweaker had that, I would have sticked to it.
K10Stat also has the option to choose which load method to consider. Highest core value, average value above all cores, lowest value...
Kink, that is a nice little program you wrote.
It is certainly more user-friendly than the other one I've been trying to use.
However, for some reason it doesn't work flawlessly on my hardware (PII x3 710/Gigabyte ma-790x-ud4/WinXP x86)
What I want is to enjoy benefits of both overclock and CnQ. OC set up in BIOS right now is HT 255 / 13x / Vcore 1.375 - it boots up and works 24/7 in this mode if I want it to. There're a few lower power modes that also work fine, the lowest being 4x / Vcore 1.0
However, PhenomMSRtweaker (and neither the other program, cnqsvcOC) seem to be able to switch between these modes without a hitch.
I've entered the numbers into the program, but when I try to activate either, it only seems to change the voltage and not the multiplier:
http://keep4u.ru/imgs/b/2009/09/22/1...e98c077c1a.jpg
http://keep4u.ru/imgs/b/2009/09/22/d...b394118526.jpg http://keep4u.ru/imgs/b/2009/09/22/4...fddcb5d16b.jpg
cnqsvcOC had the same problem: voltages switched just fine when it switched p-states but the multipliers went haywire and the CPU never ran at 13x.
Ideas / advice will be greatly appreciated :)
I'm an Athlon novice, but a couple of things I noticed:
You have to click "Apply" on the main screen and then "Make Settings Permanent" and "Apply" on the Service screen for it to work properly.
If you don't have a Black Edition CPU and 13x is more than the stock multi, it probably won't work.
13x is stock, I wrote that I have it set up like this in BIOS
It's running at 13x255=3.3GHz right now
Not positive on this, but ...
Make PMT takes over for CnQ, so first make sure you have something besides Laptop on Minimal Power Management in your Windows power settings.
The first screen looks correct, so click Activate on P0 and then Apply, then click Activate on P1 and Apply. Then click Service and click "Make Settings Permanent" and Apply.
Now if you right-click and select High-Performance you should should be at 13x, Power Saving you should be at 4x, and Balanced, it should shift with CPU load.
If that fails, I'm not sure what to tell you to try ...
Do you have CnQ enabled, and does CnQ work normally without PMT loaded?
Both activating customized pstates manually ("Activate" buttons) and allowing the service to do it only switches
Vcore and not the multiplier (once it's down to 4x). The program isn't that hard to master :)
> Do you have CnQ enabled, and does CnQ work normally without PMT loaded?
I keep it disabled, but just checked to refresh my memory and it doesn't work (likely because I increased Vcore in order to overclock the CPU)
Of course it worked with the stock numbers and I believe it worked with an OC which didn't involve increased Vcore, but that wasn't
high enough :)
I don't think I can help you more - I ran stock clock and decreased voltage to keep the PC cool. PMT worked fine for me, but other than dropping back to the stock clock and testing if PMT works with that, I don't know what to suggest.
@bobruto: If I understood correctly, the multi is actually changed - at least once, down to 4x, but then raising it back to 13x fails. Try to use 'Auto' as multi in the BIOS menu and see if that works, if I recall correctly there were some reports about C&Q not working correctly if the multi is set manually. Always depends upon the BIOS implementation...
@Tiger-Heli: Thanks, glad to hear there's another happy user. About your questions:
1) You almost got it right. The problem is that if the active power scheme allows CPU throttling (such as the default 'Balanced' scheme, not just the powersaving one), standard C&Q is active if enabled in the BIOS, conflicting with custom C&Q and resulting probably in a mess (haven't tested it myself).
2) It's not about absolute or relative voltages - you change the voltage IDs (VIDs), not the actually delivered voltages. They may correspond to the actual voltages (more or less), but that depends upon your BIOS and mainboard. In your case, the BIOS adds a fixed offset to the voltage IDs; my BIOS (ASUS M3A78-T) alters the voltage IDs for P0 (but does not allow to customize the other P-states). This topic has already been covered in this thread, I think.
Multiplier - That makes sense as I think mine was set at Auto.
1) We might be saying the same thing here, but from what I could test of it, the Windows (XP) power saving state made a difference. i.e.:
- CnQ enabled in BIOS, non-throttling state (Home Office, Always On, Presentation), PMT worked fine whether in Balanced, Power-Saving, or Perfomance.
- CnQ enabled in BIOS, throttling state selected (Laptop, Minimal power usage), you might have conflicts - IMS, the Windows CnQ would over-ride PMT, so you might not have all cores locked, etc.
Since there is little reason to select a power saving state with PMT running, it didn't make much difference, but better safe than sorry.
2) It probably was covered in the thread and the distinction was lost on me. With BIOS VCORE (VID?) set to normal, CPU-Z values are VERY (0.002) close to PMT requested values.
1) Yep, I forgot that you're using WinXP ;)
2) Your BIOS VCore entry should actually be named 'VID offset' as you actually set an offset applied to every requested VID (difference between the selected VCore and the default one). Mine, on the other hand, should be called 'P0 VID' as that's what it controls. It would be nice if open-source BIOS implementations replaced the two monopolists; the community could deliver excellent stuff (see projects like openwrt for routers). But artificial product differentiation by castrating available BIOS options and the like seems to be one of the most important aspects for mainboard producers, sadly.
Open source BIOS would definitely be interesting - I'm amazed for example, how much better your app works than AMD's CPU throttle handling.
changed it to Auto, same thing :(Quote:
Originally Posted by kink
Once it dropped down to 4x, I made the 2 states in PMTwkr the same, setting them both to 13x/+0.1V but to no avail.
Where do I look now? In Gigabyte's customer feedback section? :)
UPD: went to the gigabyte site, typed in a long post describing the situation, but before submitting it, decided to try
other multipliers for P1.
And what do you know, it works just fine if you use anything above 5x!
So apparently 1.27GHz @ 1.0V is stable, PMT works like a charm and I'm a happy man
Thanks everyone for listening!! ;)
Hey, does anyone have any idea how to get this to work in Windows 7 x64 RTM? The older version without the service works fine, but the newer versions with the installer just instantly close with a "phenommsrtweaker has stopped working". Any ideas? =(
Stabwound, I am also using Win7 x64 RTM, and I have no problem with Tweaker v1.2.2. Maybe something is wrong with your windows installation?
Another happy user. Works great in WinXP 32-bit. Thanks, kink!
550BE @ X4 (B50)
# consec samples for P0: 8
seems to control jumping from P-state to P-state fairly well. Is it harmful to the CPU/mobo to have it fluctuate between the 2 states, say, twice every second? If not, then it's probably better to lower the # of samples for P0 back to default.
3.3Ghz @ X4 1.328V idle
800Mhz @ X4 0.992V idle
Do we have an intel equivalent tool which runs on windows 7/vista 64? I tried RMClock but that doesnt work, perhaps requires me to login as admin. I tried disabling UAC but still didnt work :(. For my AMD system I use k10Stat.
I was wondering ...is it possible to make 2 or 3 icons on desktop for each
p-state so that I could manually select the p-state
What I do in WinXP is put a shortcut in the 'startup' folder and set the program to run minimized. Then you can right-click the icon in the task bar and select the desired p-state.
Thanks ...this seems to work ...at least in theory (I'm using theQuote:
Originally Posted by Bkobe7
3 p-states mod with Vista x64).
When set in this way (BIOS C&Q disabled, custom C&Q disabled), and
set to PowerSaver the system starts in Performance mode:confused:
What I would like is that the system starts in PowerSaver mode and
only when manually set enters Performance mode. Can this be done
...also is it possible that the PowerSaver mode is applied only once
Windows starts (to avoid any possible crashes dorring boot up)?
Damnnnnnnnnnnn lovin' this new 555:
http://richierich1212.smugmug.com/ph...4_ZNYjg-X2.jpg
Voltage sipper.
I can't get it working on PII 955 + GA-790FXTA-UD5 + W7 x64. When i start program windows pops this stop error:
Description:
Stopped working
Problem signature:
Problem Event Name: CLR20r3
Problem Signature 01: phenommsrtweaker.exe
Problem Signature 02: 1.2.2.0
Problem Signature 03: 49c1aac3
Problem Signature 04: PhenomMsrTweaker
Problem Signature 05: 1.2.2.0
Problem Signature 06: 49c1aac3
Problem Signature 07: 127
Problem Signature 08: 16d
Problem Signature 09: System.Exception
OS Version: 6.1.7600.2.0.0.256.1
Locale ID: 9242
Yeah I couldn't get it working on an early Win 7 beta build as well. It may have to do with drivers or something. Many users report the same issue.
This is Pro retail version, everything is up to date. I've booted from another disk with my old system (W7 x64 RC1) and everything is working well. :confused:
I am running just fine on Win 7. I don't use the automatic feature. Instead I just use 2 states. Idle and surfing LP state and then full speed for gaming, etc... I just run it at startup and switch from the tray.
I ran P-MSR-Twkr very smoothly for however long I had Vista 64, and actually was using the modded version with (3 or 4) P-states. I was running it fine after clean-installing Win 7 Pro Retail 64, too, but maybe a month later, it stopped running, giving me just the "... has stopped working" or whatever it is, and I couldn't ever get it going. I switched to the Custom CNQ instead and that seems to run just fine, even as a service, it's just not as slick or feature-rich as PhnmMsrTwkr. :(
Anybody know if support for the X6 (Thuban) will be available?
I think the author doesn't support this program anymore...and someone has used K10 stat instead. You might want to try that.
Hi folks,
I've just released a revamped new beta, supporting all 5 P-states, Thuban, and a highly configurable custom C&Q.
See the sourceforge page.
Please uninstall any previous version before installing the new one, and provide feedback. Have fun :)
thanks for sharing kink! will give this a try :up:
Getting it now, thanks for the update. I was hoping for one.:up:
Windows 7 x64. Haven't had PhenomMSRTweaker installed in a long time. Was using CNQSVC, but stopped the service and set it to manual start, which effectively uninstalls it.Quote:
Originally Posted by kink
Currently in BIOS CNQ is disabled, I believe.
The PhenomMsrTweaker exception is "PowerReadFriendlyName()" and behind it, or rather on top of it, is an unhandled exception in .NET Framework. Here is the stack which the .NET debug message is showing:
Code:************** Exception Text **************
System.AccessViolationException: Attempted to read or write protected memory. This is often an indication that other memory is corrupt.
at OpenLibSys.Ols.ReadMsr(UInt32 index, Int32 core)
at PhenomMsrTweaker.K10Manager.GetCurrentPState(Int32 coreIndex)
at PhenomMsrTweaker.Form1.timer1_Tick(Object sender, EventArgs e)
at System.Windows.Forms.Timer.OnTick(EventArgs e)
at System.Windows.Forms.Timer.TimerNativeWindow.WndProc(Message& m)
at System.Windows.Forms.NativeWindow.Callback(IntPtr hWnd, Int32 msg, IntPtr wparam, IntPtr lparam)
Thank you, works awesome on my system 2!!Quote:
Originally Posted by kink
3.3GHz setup:
http://richierich1212.smugmug.com/ph...7_CCa87-X2.jpg
Thanks for the feedback! @richierich: Does the Turbo still work if you enable custom C&Q? You can easily check that by setting the P-state bounds to 1-1, which simply puts all cores into P1. Now check with the GUI, CPU-Z or Overdrive whether P0 gets activated sometimes during a single-threaded load. /edit: Should be easier to detect if you fix the process causing the load to a single core (the GUI displays the current P-state of core 1, refreshed every 250 ms, so fix it to core 1) - open the task manager, right-click on the process, and set its affinity to CPU 0.
@cyclometric: Thanks for the info. The exception code is now fixed (no more secondary exceptions). I'm surprised the PowerReadFriendlyName() fails, I thought 256 bytes would be enough for every power scheme name - it should work with the hotfix provided below (only contains the GUI exe, simply overwrite it). It also fixes a silly bug which caused the GUI not to start if your max multi is locked (non Black-Edition) and < 13.
Thanks for the updated version.
thx for this great program, i like it!
its tools like these that power conscious overclockers love.
btw i see the theres a dropdown for different windows power profiles. does CnQ act differently between them? like the load point at which the cpu jumps from one pstate to another is different based on different power profiles?
i know that some games which are duel core optimized, dont push 4 cores enough to push to the fastest pstate, i actually get better performance running at max speed, instead of letting CnQ do its thing. but it seems that Balanced, and High Performance, with the same settings (5% min 100% max) acted differently
Well, the default C&Q is a black box - its buried somewhere in the Windows driver and affected by the BIOS initialization. Phenom I is only equipped with 2 default P-states (the K10 architecture provides for 5 P-states), and the cores are individually managed (unganged C&Q). This lead to a performance drop of 25% for single-threaded apps on my machine (compared to no C&Q) because a thread hops very quickly from one core to the other (Windows scheduler). So every time it hops to a new core and the core was previously idle, the core processes the thread for some time at its low speed until C&Q detects the load and switches to high performance. By that time, the thread may already be about to hop to the next core...
Phenom II uses 4 default P-states (5 if it supports Turbo) and all cores are kept in the same P-state at a time (ganged C&Q; Turbo is an exception) - but only if using Vista or newer (as far as I know). By doing so, thread hopping is not a problem anymore (well, it may still not be optimal for caching), but the CPU will probably consume more energy compared to unganged C&Q.
By using custom C&Q, you have full control over what's happening.
And remember that Windows power schemes do not only affect the CPU, but a bunch of other power-related settings too which may influence overall system performance. Also, the percentage range for the CPU is not that finely grained (it's probably only a way to limit the range of allowed P-states for default C&Q).
good insight, btw, what stupid reason did windows have for the thread scheduling jumping bananas thing
can this tool force 16x multiplier on the 1055t? without turbo... so 16x multiplier for all cores
hi,
it seems a nice program. ı want to try it but I think it has some resolution problems because I can't see all options and I can't expand it, either. the desktop is at 1920*1200 but it is the same at 1440*900. ı will attach some screenshot and you will understand what ı mean..
This is due to your DPI settings (aka font scaling). I've tweaked the GUI for normal 100% scaling, didn't think about that. I may fix it in the future. Thanks for telling. :)
@zaraza: I'm working on it, it should be possible.
Bugfixed version 2.01 out. Even DPI aware now :). Supposed to handle hardware Turbo more clearly too (please report - should also allow you tu use higher multis in non-P0).
Download
Me again, notifying about release 2.0.2 :) - now supports setting the max number of cores in the Turbo state when using custom C&Q (also allows you to disable the Turbo). Higher multis should be allowed now, e.g. for 1055T. Could someone owning a Thuban check whether the Turbo works as expected when using custom C&Q? I still have my Phenom I 9950, can't test it myself...
/edit: updated the first post with some undocumented functionality (fan management)
Quote:
Originally Posted by kink
Thanks very much for this fix, kink. I am happy to be back to using PhenomMsrTweaker instead of the CNQOCSVC... your tool works much better, and has enabled my machine to get a low temp of 25 @ idle, and still runs great fully oc'd to 3600 (200x18) on my PhII 940. Hip hip hooray! :up:
So I can keep my processor at 16.5x?? Awesome. I'll have to try it out later tonight.
@richierich:
Yes, that should be possible according to the AMD docs. There's a multi-limit for non-P0 states in case Turbo is enabled. Additionally, there's a hard multi-limit for all states, so you may not be able to use anything higher than 16.5x because 1055T is no Black Edition CPU. But you should always be able to use up to 16.5x if you disable the Turbo in BIOS. In case Turbo is enabled but not locked (according to the docs, that depends upon the CPU), PhenomMsrTweaker should also allow you to use the max multi in non-P0 states by temporarily disabling the Turbo while applying the settings and then re-enabling it. Please report whether it works as expected. I'd also be very interested to know if the Turbo gets activated when using custom C&Q...
Well until Gigabyte fixes Cool 'N Quiet on my board I won't be able to test it with my 1st system.
Why not? I think you can also disable C&Q under Windows by editing the power scheme (set CPU min and max speed to 100%).
Don't have that option in Windows.
They are only available if C&Q is enabled in BIOS. What's the problem then, the same as particle reported about? C&Q still active even if disabling it in BIOS?
Yep Gigabyte hasn't totally disabled Cool 'N? Quiet on most of their mobos for Thubans yet. Hurry up already.
So have you tried to enable it in BIOS and then restrict the speed to 100%? In case that also doesn't work, I could hack a special version of the tweaker which restricts the allowed P-states to P0 only (if at least the Turbo can be properly disabled, that is :)).
yeah CnQ is windows dependent, just modify your power profile to be a minimum of 100%, and your locked in at max speed (not sure if this isnt true for Thuban)
Just and FYI - On my MA790FXT-UD5P, regaurdless of power options (set to high performance in Win 7) it still throttle as if CnQ was on even though it was off in the bios. I will check the 890GX again and see if this was the case on it as well.
Please read more carefully: we already know that C&Q isn't properly disabled if you choose to disable it in BIOS (for these Gigabyte boards, another user on another forum mentions strange behavior on a Asus M3N-HT Deluxe as well). It'd be interesting to know if C&Q gets disabled by ENABLING it in BIOS and then restricting the CPU's speed to 100% in the Windows power options (for your currently active power scheme).
Tried that in high performance, CNQ is still overriding Windows setting. Gigabyte problem.Quote:
Originally Posted by kink
Okay thx for the report. You could still try to customize all P-states to the same settings, so that C&Q still switches between the states, but since they are identical, there'd be no actual change. Custom C&Q cannot be used obviously. But by testing that, you could confirm whether the 16.5 multi works in non-P0 states.
Tried switching all multi to 16.5x in all p-states, hit apply, no go on my 2nd system. They all revert back to stock multis.
Thx again. I assume the Turbo is locked then (there's a CPU-dependent lock regarding the number of boosted states (1) and the number of idle cores for the Turbo to kick in (3 for hexacores)). It's probably unlocked for Black-Edition CPUs. So I cannot temporarily set the number of boosted states to 0 to circumvent the multi-lock for non-P0 states. If the "Turbo cores" setting in the service dialog is grayed out, the Turbo is either not supported or locked.
kink, thanks for updating PhenomMsrTweaker.
so a 1090T would let you set up to 6 boosted cores? if so thats another reason why that chip is so great
Nope, at least one core needs to idle. But of course you can "boost" all 6 cores - by simply disabling the Turbo and using the Turbo settings (or optimized ones) in P0. Simple as pie. :)
/edit: richierich, did you not perform the Turbo-multi-test with the latest version?
New version out - v2.0.3. Some minor changes, e.g., the Turbo cores setting is now independent from custom C&Q (since the Turbo doesn't work with custom C&Q anyway).
kink: Thanks for all your hard work man! Very user friendly (as usual :up:)
It could just be a quirk with CPU-z, or it could be that the new MsrTweaker is tuned for Thuban cores...
The multi's seem to work fine, but according to CPU-z my volatage isn't being adjusted with a 965(C3).
I'll DL the new version and see if it helps (could very well just be a quirk with CPU-z, haven't tested it with a Meter yet)....
I'm going to pick up a 1055 this weekend and I'll be sure to try your software on it!
Reguardless... Your the man! :clap:
I'm sure I'm not the only one who appriciates all your hard work and support!
Hey Dave (I assume ;)),
thx for your feedback. It's words like these that keep PhenomMsrTweaker going. ;)
It's not optimized for Thuban since I have to rely on other users to test it on Thuban, I still have my X4 9950. ;) Everything works as expected on my system, so it may really be a CPU-Z problem. You could also use Overdrive or Core Temp. In any case, when you apply the settings in the main window, the settings in the GUI are automatically refreshed so that you should immediately notice any unexpected behavior.
Hi Kink and thanx for supporting this excellent proggy!
I have a problem tho, i was using phenommsr before, but for some time i was using K10Stat, however it freezes my system from time to time and i havent been able to fix that so back to phenommsr on new version :).
However after installing it (XP x64 SP2) i cant get it to run.
Any ideas anyone ?Quote:
---------------------------
PhenomMsrTweaker exception
---------------------------
Value of '1,55' is not valid for 'Value'. 'Value' should be between 'Minimum' and 'Maximum'.
Parameter name: Value
at System.Windows.Forms.NumericUpDown.set_Value(Decim al value)
at PhenomMsrTweaker.PStateControl.LoadFromHardware()
at PhenomMsrTweaker.Form1..ctor()
at PhenomMsrTweaker.Program.Main()
---------------------------
OK
---------------------------
Oh, and about using turbo multipliers, it SHOULD be possible, one user here with a 1035T and some beta bios was able to set Trubo multis for all cores with turbo disabled.
I think I know the cause of this exception. If a P-state is not initialized by your BIOS (such as probably your P4 state), its corresponding register may be set to 0, leading to 4x multi @ max voltages (1.55V). For safety, I check the voltage limits of your CPU (you cannot select higher or lower voltages in the controls), so it seems like your voltage is restricted to something less than 1.55V. So your P4 state is not valid, hence the exception.
What you can do to circumvent this is using the service to first set P4 to valid settings, then you will be able to start the GUI and configure everything. So navigate to HKLM\Software\PhenomMsrTweaker in the registry and create a string value named 'P4'. The string itself consists in 4 (or your number of cores) 32bit hex values separated by '|', e.g.. '38005840|38005840|38005840|38005840' (these here set the multis to 4x, CPU VID to 1V and NB VID to 1.2V, so they should be safe). Now open the list of services and restart the PhenomMsrTweaker service. This should apply P4's settings, after which you should be able to start the GUI.
Yes, you can set the CPU manually to the max multi that the Turbo can use. So if your chip has a default 14x and turbo is up to 16.5, then you can set ther CPU up to this 16.5.
Turbo itself is very handy also, here is a screen shot to show what I mean.
CPU is running 3.6ghz,
NB 2400
HT 2400
HT Ref 300
RAM 1600 8.8.8 24
Can you guess the turbo speed?
http://3800z24.info/Phenom/CrossHair...turbo-Pi1M.jpg
From what i know Turbo multi isnt settable on anything below 1090T so i guess 5ghz :P .
-
Kink, i fondled a bit phenommsr, coolnquiet and my power profiles.
Firstly i will say that this 910 chip with this foxconn behaves a bit odd im not sure bios understands this cpu correctly(it boots 2ghz Nb and on AM2+ it should boot 1800 as one thing)
Second, when i set home/desk (default) power profile in windows, coolNQuiet doesnt work at all, i have to set laptop.
And i think you are right about something wrong with P4, it doesnt seem to work at all.
I didnt noticed at first but even as phenommsr gives an error, if i dont click "ok" in error window, phenomsr works ,and if i set P4 voltage below P3 it starts to work properly (sort of).It would be nice if there was a command option to disable maybe this P4 checking before program starts ?
I tried the trick you suggested, opened regedit found correct string, added P4 ,modified it with a string you supplied ,rebooted even, but error persists :/.
All in all P4 doesnt work ,it never gets that low even when i manually fix this error problem (have to do it on each reboot).
Im a bit tired today so i will investigate further tomorrow ;-)
Oh well, C&Q is pretty crappy on Windows XP, based on my experiences with Phenom I and other users' reports including the Phenom II. Standard C&Q should clock the cores individually, which leads to a notable performance decrease. If this still holds, you're most likely better of by replacing the standard C&Q with the custom one, which also handles P4 then appropriately if you wish to. Feel free to perform some benches comparing standard C&Q with some custom C&Q settings and let us know. :)
:p: Turbo is 4.35ghz on 3 cores. And yes, you can change turbo on 1055t up to 16.5.Quote:
Originally Posted by RaV[666]
So, if I use Turbo in MSRTweaker, do I need CnQ on in bios? If so, does CnQ need turned off only for the custom CnQ option? Was going to try some testing with turbo and wanted to make sure I have the config right.;)Quote:
Originally Posted by kink