FSB not downclocking

EDIT: Solved! I used HWiNFO32 as suggested by Dufus and it reports the dynamic FSB is in fact working. CPUz and Everest are a little less accurate for my machine.

I just realized my FSB isn't downclocking properly, ie: it's always at 266MHz (x4=1066MHz). I can manually OC/UC the max FSB by using SetFSB, but on battery at least, shouldn't it be dynamically changing the FSB on its own?

 

its always supposed to be 1066MHz.. your multiplier has to downclock.

 

IIRC both the fsb and cpu multipliers downclocked to give me lower clockspeeds (and therefore less power consumption). CPU-Z used to report slower fsb's if I wasn't under full load.
Although, speaking of multipliers, I can't go below 6x. On my P8400 I could get 3x (LFM/SLFM?) regularly.

I ask because I've lost about an hour of battery life since I upgraded my CPU.

 

That's because the E series has 44W TDP which is higher than 25W TDP... also is an apple processor so downclocking might be different.

 

The processor itself is downclocking its multiplier fine, but I thought all Socket P CPUs have the ability to dynamically change the FSB on demand.
It's a 35W E0 stepping version and is identical in capability to a T9800.

 

i don't know.. as far as i know , FSB nver downclocks.. only the multipliers do.

 

Watch your FSB using CPUz when your computer is idle (and not at 100% cpu performance) and you'll usually see your FSB go nuts, changing frequencies many times every second.

Anyone with an E series socket P CPU who can confirm this problem or otherwise?

 

On my laptop (Dell Inspiron 1545) I have a celeron 900 (R0 stepping. Identical to E0 but with 3MB and not 6MB l2 cache) and it downclocks using the FSB. On my laptop with any processor that has speedstep though, the FSB remains unchanged and it only changes clockspeed by changing the multiplier. I haven't seen both multiplier and FSB throttling happen at the same time on my laptop.

 

ah see ,.. the E series and P series have speedstep so that's why FSB doesn't downclock.

 

My T5550 only lowers the multiplier, not the FSB. AFAIK all the recent and current CPUs only lower the multiplier.

 

The Vostro laptop which I chucked a T9300 in could clock down using both multiplier (EIST aka Speedstep) and FSB (SuperLFM?). The FSB and idle frequency would drop to 100MHzx8 (from 200MHz FSB/800MHz QDR and x6 idle multi). Which confused some programs that didn't dynamically read FSB (would report as "1.6GHz" based off x8 multi and 200MHz default FSB).

However I found it almost never did in real usage because the few background processes of 2-3% CPU utilisation were enough to keep processor use above the load threshhold. That's what I think went on anyway based on spending far too much time fiddling with RMCLock and other monitoring programs.

The i7 quad doesn't alter BCLK, stays at 7x133MHz at idle, however it can shut off "unused cores" but again the 2-3% background (too many Win7 aero + gadgets) means that doesn't happen as often as it could.

 

Wow guys, thanks for all the replies! Much appreciated.

@Trottel: Yeah I've heard Celerons have part of their cache and speedstep disabled. I didn't know they varied the fsb though. Good to know.

@Sean: Speedstep is for the CPU multiplier only, not for the FSB. My SpeedStep is working fine, FSB is another story.

@sgotgeta: I could have sworn they did both, as the FSB is linked to more than just the CPU, so it's a major power saver outside of just processing. Maybe I'm mistaken.

@bennyg: This is exactly what I could swear my P8400 did. I have a little display in my laptop that tells me the current CPU speed and it would sometimes drop to just a few hundred MHz. Everest and CPU-z would show the fsb switching all over the place idling, and locking at full speed when under load.


I would think that even with SpeedStep, a 35W cpu is probably draining my battery faster, too (vs my old p8400's 25W). But almost an hour less of runtime away from the wall? I feel like that's a bit off, and my FSB is always locked at 266MHz.

It could be background processes as you guys mentioned though. Any ideas on how to read the fsb in safe mode? CPUz and Everest won't start correctly.

 

TDP is more marketing and shouldn't be adversely different between the CPUs, especially at idle. Try using RMClock to undervolt your CPU.

 

dtd00d download and run the freeware HwInfo32. It should tell you if system supports dynamic FSB and if it is enabled or disabled.

 

@sgogeta: Yeah true, though his one definitely idles hotter than my P8400. I should look into undervolting at idle though, that's a really good idea. I'll fire up RM clock when I have some time and can risk BSODs. Thanks! +rep

@Dufus: Hey! I think you're onto something! I have three different states, and this makes a lot more sense now:
-SLFM (DynFSB) 133MHz * 6 = 800MHz
-LFM (Min) 266MHz * 6 = 1.6GHz
-HFM (Max) 266MHz * 11 = 2.93GHz

CPU-z reports the same frequencies but different multipliers (for example, instead of 133MHz*6 it says 266MHz*3). I guess it was a matter of program accuracy. And I also guess this means you guys found my problem and I can rest easy now. As I'm typing this I 'm watching the FSB fluctuate between 133 and 266, and every so often I catch it in a state inbetween the two. I'm going to attribute the rest of the battery life loss to a hotter CPU at stock voltages with double the cache to power.

Thanks again guys for your respective contributions! Have some rep!

 

Sorry to revive this thread but I figure I'd post my findings:

Intel Core 2 CPUs do in fact have Dynamic FSB frequency switching (Called DFFS), but mine does not, nor does it have IDA. However, I was able to undervolt my CPU down to 1.1V, which helps immensely on battery. While it'd be great to have the best of both worlds (DFFS and UVing), I'll settle for a win and a loss. I'm getting another 15-20 minutes of battery, and it no longer consumes more than about 30W idling. So while it may switch between one or two different FSB levels, it does not support as many switching levels as a typical Core 2 CPU.

I suppose the lack of IDA is another blessing in disguise, since I can run a sustained overclock a little further than with it. Ah well.

Thanks again guys for earlier. Figured I'd post this for other people's reference.