Is your processor Running slower than you think it should be ? Or at a different multiplier than it should be ?

[Markfw]

There are two things that are common in many Intel processors, they are designed to save heat, and power by lowering the multiplier and the core voltage of the processor at times that the "power" or extra speed is not required. These CAN be disabled in your BIOS, and in some cases it is recommended (overclocking is one example), but if the processor needs the speed, it will "wake up" and utilize all the speed it has, so its not required that you disable it in the BIOS. If you want to know more, read the links below. You can also ask other specific questions here

Here is Intel's C1E and other sleep states documentation

Here is Intels EIST documentation.

This I did not know, but from a thread below, I give credit to Hans:

Originally posted by: hans007

Originally posted by: pRada
Wow... thank you very much guys. Is this C1e and EIST/Intel Speedstep in the bios figuration? How exactly do I disable it?

if you are using windows, just go to the power management thing in control panel and change it to no power management.


c1 halt is only working with "minimum power management" or higher profiles. why would you want to disable it?

Other comments like this are welcome, but please be sure of your facts, and NO negative comments please.

 

[aigomorla]

Well sometimes its not always 100% hardware. Software will sometimes give wrong readings so make sure you double check with different software to see if its indeed hardware and not software.

Coretemp 0.95 (example) had a bug which doesnt display correct overclock if you use a multi lower then stock.

They fixed it in the latest revisions, but incase someone see's an old 0.95 screenie with 9x450, be very skeptical, its more like 8x450.


Programs which display Overclock:

Cpu-Z
Motherboard Utility, like uGuru, AsusProbe, ect...
Coretemp (MAKE SURE ITS THE LASTEST VERSON!!)
Everest

 

[Markfw]

Great point aigomorla !

 

[WT]

FYI, but the newest version of CPU-Z incorrectly displays the multiplier on my MSI P6N Plat board .... I have speed stepping disabled in the BIOS and I noticed my multi was bouncing between 6x and 9x with an e6600. Doing a quick Google search showed that the new version does indeed have an issue, and rolling back to CPU-Z 1.38 or earlier will correctly list the locked multiplier.

 

[aigomorla]

Originally posted by: WT
FYI, but the newest version of CPU-Z incorrectly displays the multiplier on my MSI P6N Plat board .... I have speed stepping disabled in the BIOS and I noticed my multi was bouncing between 6x and 9x with an e6600. Doing a quick Google search showed that the new version does indeed have an issue, and rolling back to CPU-Z 1.38 or earlier will correctly list the locked multiplier.

heh this is very important!

Thanks for the heads up!

So as i said always double check software. Sometimes you'll be suprised how much 15 min of double checking can save you hours of rebuild/reinstalling OS.

 

[Markfw]

I must say, I think this thread is more informative than the last 50 on the same subject....This is exactly what we should do more here...

 

[n7]

Hah, good thread idea.


Now let's see how many newbies completely ignore it & post threads on it anyway :laugh:

At least we can link them to here anyway

 

[blanketyblank]

This thread needs to suggest a way to actually verify multiplier when EIST and C1E are enabled. I'm greedy and would like to have an overclock and power saving, but I just can't tell if it's working for me or if enabling these options is permanently locking me to 6x. I've tried running cpuz while stress testing and it still show 6x.

Are there any more reliable tools for testing? I suspect either a software bug or my motherboard.

 

[Markfw]

Originally posted by: blanketyblank
This thread needs to suggest a way to actually verify multiplier when EIST and C1E are enabled. I'm greedy and would like to have an overclock and power saving, but I just can't tell if it's working for me or if enabling these options is permanently locking me to 6x. I've tried running cpuz while stress testing and it still show 6x.

Are there any more reliable tools for testing? I suspect either a software bug or my motherboard.

OK, Please post your full system specs, and the version of cpuz. Notice above, the post on bugs in versions after 1.38, try that version also. Until we know all the specs, your bios setting, and what its reading, its hard to say. This used to be simple, but now some of these issues are making it more complicated, hence the single thread here to focus on the issues.

 

[blanketyblank]

Yes, just tried cpuz version 1.38 (Did you mean even earlier?) with the same result.
I've added an image as well. I'm running the latest BIOS for my board which was supposed to be released on 2007-12-16 and fixed some c1E funcion (I'm thinking it didn't).

My specs are currently:
Q6600 (should be at 2.7 Ghz 300mhz x 6-9 ideally)
Abit IB9
BFG 8800 512 Mb GTS
2 Gb Geil Value DDR 2 8000 5-5-5-15
X-Fi Xtreme Gamer
Windows Vista 32 bit

Here's the bios updates for my mobo:
http://www.uabit.com/index.php...l=334&product_name=IB9

Here's a pic of CPUZ 1.38 while stressed using prime95. 1.44 looks the same.
pic of CPUZ

 

[Markfw]

The only thing I can think of, is set your bios for disable all( c1e,eist,vanderpool, etc...). Then multi x6 Then run superpi. Write down the 2m time. Then set bios to 9x, and run superpi again (2m) and write down the time. The re-enable c1e and eist, and run superpi again, and see what cpuz says, and what the time is. I would be interested in the result.

It does sound like you may have a bios issue.

 

[blanketyblank]

woohoo I fixed my problem through a software program!!
After posting on abit's forum someone suggested a tool called crystalcpuid which let me set my multiplier to 9x when the system is under load. CPUID gives me the correct multiplier now while I'm running prime and then ramps back to 6 when I shut it down.
Since software can fix the problem this leads me to believe potentially vista is at fault for not controlling the multipliers.

 

[Markfw]

Originally posted by: blanketyblank
woohoo I fixed my problem through a software program!!
After posting on abit's forum someone suggested a tool called crystalcpuid which let me set my multiplier to 9x when the system is under load. CPUID gives me the correct multiplier now while I'm running prime and then ramps back to 6 when I shut it down.
Since software can fix the problem this leads me to believe potentially vista is at fault for not controlling the multipliers.

Well, I think we can agree, its bios or Vista !! Anyway, glad to hear your issue is fixed, and maybe this can help others solve the same or similar problem.

 

[BadOmen]

What if the multiplier reads ok in all tools, but the FSB shown at cpu-z and the bios is different than the one at coretemp and the System section of Windows control panel?

(the control panel one doesn`t show the formula, just the result (3072), which is the same as coretemp)

Here are the numbers:

384 X 8 = 3072 according to coretemp and windows

The BIOS and cpu-z report 340 X 8 = 2720

I posted it in another thread because it's not the same thing discussed here, but here it goes. Trying to avoid cluttering.

 

[aigomorla]

Originally posted by: BadOmen
What if the multiplier reads ok in all tools, but the FSB shown at cpu-z and the bios is different than the one at coretemp and the System section of Windows control panel?

(the control panel one doesn`t show the formula, just the result (3072), which is the same as coretemp)

Here are the numbers:

384 X 8 = 3072 according to coretemp and windows

The BIOS and cpu-z report 340 X 8 = 2720

I posted it in another thread because it's not the same thing discussed here, but here it goes. Trying to avoid cluttering.

??

im very lost on this, all three programs are reporting different things?

Well simple OC laws dictate always follow the higher one. Can you post a screen shot of this with all three programs?

 

[BadOmen]

Somehow I fixed it.

I went to the BIOS to reduce vcore and change my conservative overclock from 8x340 to 9x310 (the thing is still damn hot).

I took the chance to enable C1E. I don't remember, but I think TM2 and EIST were already enabled by then.

After that, Coretemp and cpu-z began showing the same FSB.

Funny. The multipliers were always the same in both software, although eist and c1e were supposed to mess with them depending on the cpu load, right?

 

[nyker96]

excellent idea to sticky this topic, seen too many questions on this already here. I wished the m.b. side can sticky a permanant mb recommendation thread.

 

[flexy]

correct me if i am wrong....your thread-title suggests there are real life siutations where a CPU is running slower than it's suppsoed to be, and you say that in conjunction with the speedstep/EIST etc. settings in BIOS.

The point is that in "real life" this is a non-issue - since throttling would only occur when extreme overheating of a CPU appears, something like a few degrees over max. specs. This would only apply in extremely overclocked, extremely BADLY thermal situations and with LOTS of Vcore.

Intel Speedstep/EIST etc. are the equivalents of old-school AMD "Cool'n'Quiet"...this is, CPU has lower multi and vcore when idle, but *immediately* clocks up to 100% speed and Vcore voltage once the the slightest load is applied on the CPU. This is a good thing and works allright. As you also say, you can very well all those features have activated in bios. On my pretty overclocked quad those bios thermal management settings dont influence the overclock at all.

In Bios, however, do NOT set a fixed Vcore, but set Vcore to AUTO (!) and then set "VCore AUTO Add" and some value like 108%, giving your desired CPU voltage in case you overclock and need more VCore. Do NOT set a fixed voltage.

Reason: If you have Vcore on "AUTO" and set "Vcore Auto ADD" value then Vcore is actually lowered *also* when idle - if your Vcore is set fixed in bios Vcore is NEVER lowered! <---- i also just found that out.

No problems with great overclock...and at the same time max possible power saving when idle.

 

[Markfw]

flexy, the whole purpose of this thread is to eliminate the countless threads that get created when speedstep changes somebodys multiplier to run at 6 instead of 9, or speed at 1800 instead of 2400, and they post "why is my machine running slower than it is supposed to ?

It gets old, and this is to inform the masses which may not be as savvy as you, that this is normal.

 

[flexy]

I see. You're talking about people setting multis and FSB, and then they boot into windows and see the multi is lower (because they're idle).

 

[BadOmen]

And in the end, I still couldn't get one thing: are EIST and C1E generating the same thing (saving energy, reducing vcore, heat, etc) by doing different things? One by flushing the cache when the processor is idle, the other putting the processor to sleep until necessary?

Are they never in conflict?

 

[Markfw]

Well, the way I read the Intel documents, C1E is this:
"C1 - Auto Halt: core clock is off. C1 is a state where the processor is not executing instructions, but can return to an executing state essentially instantaneously. Some processors also support an Enhanced C1 state (C1E) for lower power consumption."

While EIST is enhanced speedstep, changing the multi to 6 instead of 9 (as an example) and lowering voltage. I didn't see anything about cache.

 

[BadOmen]

Ouch, sorry, the thing that flushes the cache into memory is this one:
Enhanced Intel® Deeper Sleep with Dynamic Cache Sizing.

It was in that document too.

So, both of them (EIST and C1E) can work simultaneously with no known conflict, right?

 

[Markfw]

Originally posted by: BadOmen
Ouch, sorry, the thing that flushes the cache into memory is this one:
Enhanced Intel® Deeper Sleep with Dynamic Cache Sizing.

It was in that document too.

So, both of them (EIST and C1E) can work simultaneously with no known conflict, right?

Ask Intel ! I just pass on what they say... And I never use them. So Intel says so, so it must be so, right ?

 

[BadOmen]

Oh, they never mention both names together in one conversation.
Probably some religious tradition.
:laugh: