Interesting discovery when undervolting T61/T7500...

Message

Redmumba · #1 Post by **Redmumba** » Wed Jun 27, 2007 12:42 pm

So this is rather odd.

I've been undervolting my new laptop, stress-testing and finding the best voltages, and I've discovered something very interesting. While Intel lists the maximal multiplier to be 11.0 (2.20ghz), I discovered that both CPU-Z and RMClock have an elusive 12.0x multipler. When I force 12.0x, the processor will actually jump to 2.4ghz...

12.0x 1.275v
11.0x 1.250v

Anybody else discover this with their processors as well?

tebore · #2 Post by **tebore** » Wed Jun 27, 2007 1:35 pm

Nice an unlocked Chip. So you got a T7700.

Redmumba · #3 Post by **Redmumba** » Wed Jun 27, 2007 2:05 pm

Oh really? It seems to have "trouble" staying at 2.4ghz, though... as in, it keeps switching back and forth between 2.2 and 2.4. I'm almost certain this is a matter of conflicting power management (I'm still getting the hang of Lenovo's power management utilities), since I had RMClock running as well.

The one thing I can't seem to do (and this seems common), is undervolt the processor better than what Lenovo's software does. The funny part is, Lenovo's own software can undervolt the processor to 0.850V at 8x, but I can't get it below .937 at ANY multiplier in RMClock. Mixed messages I found through my Google searches say that its a software issue, while others say its a hardware issue.

Also, for some odd reason, NHC doesn't like the T7500 (or not mine, at least). It does not display Multiple/Voltage information, or allow their modification. Could this be because the processor is "unlocked?"

Does anybody know of any alternative undervolting utilities?

awolfe63 · #4 Post by **awolfe63** » Wed Jun 27, 2007 3:27 pm

The newest mobile core duo processors have a feature where when only one core is active, it can be overclocked by 1 step. This should be automatically handled by Speedstep drivers.

Pascal_TTH · #5 Post by **Pascal_TTH** » Wed Jun 27, 2007 3:43 pm

awolfe63 wrote:The newest mobile core duo processors have a feature where when only one core is active, it can be overclocked by 1 step. This should be automatically handled by Speedstep drivers.

True, all Santa Rosa CPU can have +1 for the CPU multip when only one core is used. That's a kind of dynamic overclocking. It's really a nice feature.

Redmumba · #6 Post by **Redmumba** » Wed Jun 27, 2007 4:53 pm

awolfe63 wrote:The newest mobile core duo processors have a feature where when only one core is active, it can be overclocked by 1 step. This should be automatically handled by Speedstep drivers.

Ah, good to know!

I don't suppose you can shine any light on the lower-end voltage limitation, eh?

sokos · #7 Post by **sokos** » Wed Jun 27, 2007 6:30 pm

how do u underclock or overclock your TP's ?

#8 Post by **tomh009** » Wed Jun 27, 2007 6:37 pm

Well ... you can't really directly underclock or overunderclock -- but you can underundervolt with either RMClock or NHC.

http://cpu.rightmark.org/products/rmclock.shtml
http://www.pbus-167.com/

Redmumba · #9 Post by **Redmumba** » Wed Jun 27, 2007 9:15 pm

...and I don't recommend that at the moment. I'm not sure if Core 2 Duo just isn't fully supported or what, but neither one of them is doing a "great" job of undervolting. NHC doesn't allow any sort of voltage manipulation whatsoever (at least not on here), and RMClock won't allow you to set a voltage lower than .937--while using the included Lenovo/IBM software will automatically underclock it to 8x at .850V.

That's why I'm wondering if anybody knows any alternative programs.

I'm also not sure why the "lowest" underclock setting is 8x, when its reported to go as low as 6x...

tebore · #10 Post by **tebore** » Thu Jun 28, 2007 7:54 am

Since Yonah Intel has locked out the lowest voltage allowable. The support has to be in the BIOS. RMClock used to just read the registers from the CPU and go from there. Now Intel left it up to the Manufacture to allow lower than 0.937. This is why Lenovo can go lower.

Intel said it has something to do with how the OEM designs their systems. Like using a thicker PCB would allow for a bigger undervolt but if you're cheap like Acer then you won't be able to push that much. They'll only undervolt if complete idle too.

josh999 · #11 Post by **josh999** » Sun Mar 09, 2008 12:48 pm

Redmumba, (and anyone else who cares to help

)

I just got the same configuration as you (T61, with Core 2 Duo T7500 CPU).

But, my numbers are significantly different/higher than yours. CPU-Z reports 1.313V @ x 11 Multiplier. And, the lowest I can get using Thinkpad Power Savings is .963V @ x 8 Multiplier.

Any ideas why my "minimums" using Thinkpad Power Savings are so much higher than you saw?

Also, the last post in this thread is over 6 months ago. Has anyone discovered the best utility for under-volting the T61/T7500 combination? Or, is it best just to leave this processor to the Thinkpad Power Savings?

aaa · #12 Post by **aaa** » Sun Mar 09, 2008 12:57 pm

The 0.85v is called SuperLFM I believe, which is used with FSB reduction (8x teeny fsb instead of 200). I thought RMClock supported it... maybe it was a beta.

josh999 · #13 Post by **josh999** » Sun Mar 09, 2008 1:07 pm

Thanks for the reply!

.... I am using the latest (1.44.1) CPU-Z. Any ideas what I should use for a second opinion?

Redmumba · #14 Post by **Redmumba** » Sun Mar 09, 2008 1:15 pm

Right. The newest RMClock beta supports the Santa Rosa platform and all the joys that it brings. So underclocking the 2.2Ghz chip to 800Mhz is a reality. Plus, it also supports the Ultra-Low Frequency, which lets you drop it down even lower.

josh999 · #15 Post by **josh999** » Sun Mar 09, 2008 2:10 pm

I don't see a beta version of RMCLOCK. I have a version 2.35.0 built on 2-29-2008. I donloaded it, and am running it "side-by-side" with CPU-Z. But, I am confused!

I have set my Thinkpad Power Settings to Lowest Maximum CPU speed. RMCLOCK says my CPU is at 8x, 798Mhz, and .85V.

CPU-Z reports 8x, 1596Mhz, and .963V.

Who is right?

aaa · #16 Post by **aaa** » Sun Mar 09, 2008 2:23 pm

CPU-z is wrong... it doesn't support Santa Rosa apparently.

Enigmatist · #17 Post by **Enigmatist** » Sun Mar 09, 2008 3:16 pm

Redmumba wrote: The one thing I can't seem to do (and this seems common), is undervolt the processor better than what Lenovo's software does. The funny part is, Lenovo's own software can undervolt the processor to 0.850V at 8x, but I can't get it below .937 at ANY multiplier in RMClock.

What Lenovo software are you referring to in this case? I havn't been able to find any undervolting software from lenovo, only the power manager...

josh999 · #18 Post by **josh999** » Sun Mar 09, 2008 3:56 pm

I think that he IS referring to the Power Manager. The voltages will vary with the Maximum CPU setting.

One more question for the "Undervolting" experts out there. When running RMCLOCK, Thinkpad Power Manager is still running and trying to change many of the same settings. So, who's "in charge"? Is there a way to disable Power Manager?

Enigmatist · #19 Post by **Enigmatist** » Sun Mar 09, 2008 11:37 pm

josh999 wrote:I think that he IS referring to the Power Manager. The voltages will vary with the Maximum CPU setting

As far as I know, Lenovo Power Manager doesn't allow you to modify the voltages...or am I missing a whole section of Power Manager's capabilities.

Furthermore, what programs can I use to undervolt the CPU with best results under Vista 64 - BIT. Emphasize on the 64 bit, a lot of the programs suggested dont seem to work with 64 bit systems or are there ways around that problem too?

dozer · #20 Post by **dozer** » Wed Mar 19, 2008 2:03 am

Enigmatist wrote:
josh999 wrote:I think that he IS referring to the Power Manager. The voltages will vary with the Maximum CPU setting
As far as I know, Lenovo Power Manager doesn't allow you to modify the voltages...or am I missing a whole section of Power Manager's capabilities.

Yes, I too would like to know where this magical-menu is. All I've seen in Power Manager is the usual stupid/simplistic stuff that's in regular windoze 'power' control-panel applet.

thanks!

pae77 · #21 Post by **pae77** » Wed Mar 19, 2008 2:05 am

The magical menu is in Vista's power options, although as stated above, it only allows you to set minimum and maximum CPU power levels for a profile, which indirectly affects the voltage the CPU will run at.

dozer · #22 Post by **dozer** » Wed Mar 19, 2008 3:01 am

ahh...thanks.

I thought they were talking about the Thinkpad Power Manager.

I run windows 2000...so I'm not privy to that high-tech vista stuff...

Enigmatist · #23 Post by **Enigmatist** » Wed Mar 19, 2008 10:42 am

I checked out the power options in vista, it doesn't actually do all that much in terms of undervolting.

So essentially there is no stable software to lower voltage in a vista x64 system?

pae77 · #24 Post by **pae77** » Wed Mar 19, 2008 4:42 pm

I keep hoping that Notebook Hardware Control will come out with a version that will run on Vista x64 machines.

NIXin · #25 Post by **NIXin** » Wed Mar 19, 2008 5:51 pm

Mine T7300 is very easily undervolted!
I actually found the minimum stable voltages for all of the multipliers and my temperatures decreased significantly.
Here's the list:

Type - FID - VID
SuperLFM - 6x - 0.85V
Normal - 6x - 0.85V
Normal - 7x - 0.85V
Normal - 8x - 0.85V
Normal - 9x - 0.875V
Normal - 10x - 0.9375V

Whoa! 0.9375V Instead of the standard 1.2V at full speed. That's really nice.

My CPU temperatures were 50*C when idle and when undervolted it's like 40*C. Stress - before undervolting it would go as high as 88*C after running Orthos for a couple of hours and right now only about 64*C. That's a huge difference. I wonder why they set the default voltages at a so high level.

dozer · #26 Post by **dozer** » Wed Mar 19, 2008 11:48 pm

Because any batch of IC's has quite a large spread of parameters when made.

So, a company using those chips can either test every one of them, and set up each machine specifically for that chip only...which costs a lot of time and money...

...or, they can just set ALL machines up for the WORST chip in the batch. That's fast, easy, and cheap for them.

The only downside is that virtually all laptop owners' machines are running at least somewhat warmer than they need to be...and some owners' laptops are running a LOT hotter than they need to be. (outliers on the parameter-distribution curve)

pae77 · #27 Post by **pae77** » Thu Mar 20, 2008 6:41 pm

Redmumba wrote:...and I don't recommend that at the moment. I'm not sure if Core 2 Duo just isn't fully supported or what, but neither one of them is doing a "great" job of undervolting. NHC doesn't allow any sort of voltage manipulation whatsoever (at least not on here), and RMClock won't allow you to set a voltage lower than .937--while using the included Lenovo/IBM software will automatically underclock it to 8x at .850V.

That's why I'm wondering if anybody knows any alternative programs.

I'm also not sure why the "lowest" underclock setting is 8x, when its reported to go as low as 6x...

Well, I've just started playing around with RMClock and I think I've discovered the reason for at least some of the issues you noted in the above quote. In RMClock, for some reason, the default configuration found in the "Advanced CPU Settings" menu is set to "desktop." If you change that to "mobile" by checking the appropriate box, the CPU voltages available suddenly drop to the expected 0.8500 and things seem to work and display more correctly, that is more as one would expect. I hope you will check this out and report back whether this helped resolve anything.

Also, just for anyone that might be interested in running this program on Vista x64, one way to do it is by disabling Vista x64's irritating driver signature enforcement. This can be done at boot by pressing F8 as if you were trying to boot into safe mode. You will be presented with a list of boot options. The last one on the list (iirc) allows you to boot vista with driver signature enforcement disabled. Unfortunately, everytime you boot you would need to go through this F8 procedure, but since I don't reboot all that often, it isn't that big of a deal. Hopefully, eventually the program will get a signed driver or someone will come up with a permanent way of disabling driver signature enforcement which I find to be intrusive and a hassle for no benefit as far as I can see.

dozer · #28 Post by **dozer** » Fri Mar 21, 2008 1:06 am

ahhh....good info on the 'desktop' default in RMclock...thanks.

Interesting discovery when undervolting T61/T7500...

Interesting discovery when undervolting T61/T7500...

Who is online