M11X clocking observations.. Speedstep etc..

I have noticed if you boot the M11X up without the overclock setting in the bios and the speedstep turned off, your multi is set to 6x and clock speed is 1.2ghz @.9V. With speedstep on your multi is set to 6.5x and clock speed is 1.3ghz @1.013v AND you can use crystal CPU to turn it back to 1.2ghz and change the voltage down to .9 while in windows.
Now with the overclock setting on, the speedstep switch seems to be disabled or bugged. So you get to windows with a 6x multi and .9V with speedstep on or off, this would explain why some people may have troubles with their cpu wonking out with the extra 300mhz. This is a serious bug really, to run with the extra 300mhz they should bump the voltage to AT LEAST what the 1.3ghz uses. I mean it is really just a pinch more for stability.

So, here it is in short:

overclock OFF, speedstep OFF = 1.2ghz (6x, .9v)
overclock OFF, speedstep ON = 1.3ghz (6.5x, .1.013v)
overclock ON , speedstep OFF = 1.6ghz (6X, .9v)
overclock ON, speedstep ON = 1.6ghz (6x, .9v) <--- no change (bug?)

 

The reason for that is due to Alienware's inadequate configuration of the overclock option. Speedstep is completely disabled in overclock mode. As we don't have a statement from AW regarding this, we can only speculate, and below is what I think.

It is my belief that they disabled Speedstep in OC mode to enable them to "lock" the 6x CPU multiplier and voltage in place.

• Why would they do that? They didn't want the CPU increasing to it's standard 6.5x multi under load resulting in the higher clock of 1.73GHz.
• Why didn't they want the CPU to go to 1.73GHz? Heat issues possibly, maybe it was reducing battery life too much for their liking.
• But why is the voltage so low then, isn't it supposed to be increased when a CPU is overclocked to attain stability? Yes, it should be. But, undervolting a CPU reduces heat output and increases battery life. In such a small form factor, where battery life and heat dissipation are already concerns, it seems they had to undervolt to come to an acceptable compromise.
• But it runs at normal voltage in standard mode (i.e. OC off, SS on), so why is it decreased for the OC? In addition to the voltage, the extra clock speed of the CPU will increase power draw and heat output too. They may well have decided that running the CPU in OC mode with normal voltage was simply too much.

The reason I say inadequate is because I think it was poorly implemented. If the system was unable to cope with the extra heat output and power draw from an OC'd CPU at normal voltage, then don't offer the option. I don't think it's fair to M11x owners to offer the option of an OC at the expense of system stability. We've already had issues with M11x's in OC mode (albeit only a couple) most likely due to the fact that the CPU is simply not capable of reaching that high of a clock on such a low voltage.

Most owners are not experiencing issues (which is good) but as all chips overclock differently (some better than others), we may see more and more instances of decreased stability while OC'd.

There has been mention of a BIOS fix by members but I've heard no official statement (or any official info at all in fact) in relation to it. Even still, if the reason for the undervolting and lowering of the multi was due to heat and power draw issues, then a BIOS update certainly won't fix that. If system stability does become a real concern, then the only option for AW is to remove the OC option altogether.

The reason why a lot of members are annoyed by the way the OC mode actually functions, is because Asus seem to have absolutely zero issues regarding heat and battery life running a standard voltage overclocked SU7300 at 1.73GHz all the time. The only possible reason I can think of is that the power draw and heat output of the GT 335M in an 11" chassis is so much, that even increasing the CPU's voltage and clock speed slightly in OC mode may have put it over the edge.

 

Well part of the issue is dell probably isn't ever testing the laptop with the OC option enabled during build and QC, so it gets to us without "real world" testing (real world as in, how we will actually be using it).

meh.

 

It could simply be a bug. When it comes to Intel chips, there is always so much headroom for overclocking, especially with these low powered ones. I love to overclock, so much so that I have tore the M11 down to the cpu heatsink to locate the PLL

They really should rename the speedstep option in bios to something else as the way they use it is not anything like speedstep. The regular range of multi's and voltages are not even represented when you use the speedstep utils. Windows cannot even manipulate the chip like it is supposed to via speedstep.
I suspect some engineer back in the core Taiwan, China or Korean company that makes these, figured the thermal profile and cpu performance was fine at 1.6ghz. It does bug me that they figured if the machine runs fine at 1.6ghz @ .9v, why on earth a bump in voltage at 1.3ghz? For battery performance they should have lowered the V to .9 as well. Maybe even lower for 1.3ghz and 1.2ghz. Or it could be a bug in the bios for the first models out. Like we have never seen that before. I have owned nearly 20 laptops now over the last 15yrs, seen that a few times. Not a lot but a few for sure.

 

I know it's a new machine, new BIOS and all that, but I can't really see this as being a BIOS bug. It definitely seems to me that it was done purposely (considering the amount of things that are just plain wrong with the configuration). But, I do hope you're right. The issue definitely needs to be addressed by AW.

I agree about Speedstep, they've chopped and changed it In stock form, the SU7300 CPU has two P-states of 4x and 6.5x. In standard mode, the SU7300 should drop to a multi of 4x (800MHz) when idle, not 6x (1.2GHz). Also, as you've mentioned you're unable to manipulate Speedstep in Windows, which you should be able to do. So they've screwed with Speedstep, the CPU multipliers and the voltage.

If I get a M11x, what I would like to do is not to even go near their disastrous "overclock" option in the BIOS but leave it in standard mode instead (unfortunately with multi's of 6x and 6.5x). Then, hopefully I will be able to use SetFSB to adjust the bus speed (eventually) and RMClock to set the voltage to what it should be. I saw your thread on the M11x PLL which I'll be keeping a very close eye on. Thanks for your efforts so far.

Oh yeah and I meant to say thank you for going to the time of putting up the table in your first post. Changing the BIOS settings and restarting the machine four times in succession just to read CPU-Z is a right pain!

 

Poorly implemented is correct. It was very intentionally included in the bios and then mentioned in the manual. The smart thing to do would have been to make it an option at order entry. They could have charged a few bucks for it, enabled it and fully tested it at the factory, and everyone would be happy. They could have marketed the option, and deflected a bit more criticism of the underpowered cpu.

 

Can someone give CPU True Speed a try?
http://www.geocities.com/izenkov/product-truespeed.htm | Direct Download Link: http://www.geocities.com/izenkov/Zip/CpuTrueSpeed-1.8.exe

 

Unfortunately not all of the people are like rpg-XPS and me and a few others that are a "bit" annoyed about the whole OC equilibrium that AW is serving us, most of the buyers saw the nice shiny videos from CES and thought that this laptop is the best thing after the slice of bread. I am really hoping that in a short future they will eventually find out that there are other people that expect the maximum quality from a brand like Alienware and not some sort of mediocre and suspicious product with questionable performance and reliable quality.

 

So let's recap:

CPU True Speed shows 1.73 (1733.394 MHz)

Windows System Properties also reports 1.73 GHz
(Processor: Genuine Intel(R) CPU [email protected] 1.73GHz)

CPU-Z is reporting 1.6 GHz

Everest: ?

HWiNFO32: 1.599 GHz

 

Well, all in all the laptop is pretty much as advertised. There really isn't any other little netbook that can touch it for the price, I mean I got 8gigs of memory, the higher chip and 320gig drive with mine for a total cost of 1100$. I call that a pretty good deal considering it plays everything really well.

 

There's no denying that. Considering what you get, it's a very well priced laptop.