OC with speedstep - possible?

[neiroatopelcc]

Hello. I know sufficient about c2d overclocking to make stuff stable and working. But I've always turned off c1e and speedstep in order to get there.

But what if I want speedstep enabled?

I read somewhere that the ordinary haltstate only cuts the fsb from the circuit and thus only marginally lowers power consumption. So I was hoping on speedstep to do the trick. But I've asked on a local board, and they tell me that if I change the cpu voltage from auto speedstep'll not touch voltage, and if I leave it on auto and oc the bios will raise the voltage and again turn off speedstep.

So. Can I do it somehow?

My plan is to oc an e6600 to 333x9 on stock cooling and pray that it won't crash when speedstep kicks in and lowers the cpu to 2ghz - but if it only lowers the multiplier it won't really affect power consumption in any notable way...

Edit: specs:
e6600 with stock cooling
antec sonata 3 with 500w atx 2.2 psu
ga-p35c motherboard (the one with ddr2 and ddr3 slots)
8600gt
2gb corsair pc2-6400cl5
2x250gb sata in raid 0

 

[Dahak]

Quite often with speedstep,as with AMDs coolnquiet,leaving it enabled when overclocking causes system instability.I don't remember exactly why it does this,but it has to do with original clock speeds of the identified processor.At any rate,it's been my experience that with overclocking,bothe need to be disabled .Otherwise the system when it goes to clock down,becomes unstable and hence your computer crashes or freezes.Goodluck.P.S.It does effect power consumption by lowering the voltage to the cpu,thereby reducing power needs,thereby reducing power consumption.

Dahak

AMD X2 5600+ @ 2.8ghz(stock)
M2N32-SLI DELUXE MB
2 GIGS DDR2 800 RAM
THERMALTAKE 850WATT PSU
7950GT KO(WAITING FOR MY OTHER TO COME BACK FROM RMA)
ACER 22IN. LCD
SMILIDON RAIDMAX GAMING CASE
80GIG/250gig SATA2 HD's
XP MCE

 

[jackcrackerman]

Speedstep *IS* Achievable while overclocking depending on how far your pushing it. Speedstep works by dropping the core multiplier and the voltage... so if you have your FSB too high, when that core voltage drops due to speedstep it will cause a system crash. So really, the only way to find out with your current settings if your system will run stable with speedstep is to overclock to a stable level, turn speedstep on and see if it can stay stable in its energy conserving state. There are several C2D overclocking guides that explain this in more detail as I am... only a noobie regurgitating what I've previously read.

 

[neiroatopelcc]

dahak - I know a fair bit about it already. The amd stuff doesn't work, because the windows driver amd made doesn't account for overclocking.
With intel though, you don't install a driver to do it, windows just 'magically' does.

@ both - I know it can run while turned on. What I've heard though, is that it will only lower the multiplier, making the system run slower but not lower the voltage of the cpu - thus consuming almost the full wattage at 2/3 the speed, making it useless.

What I'm actually asking is :
1) Is it true that overclocking the fsb will make speedstep not touch the vcore setting? but only multiplier?
2) And do I enable only speedstep or also c1e? c1e makes the core skip some fsb clocks, thus potentially lowering clockspeed by some amount I don't know. Presumably lowering fsb from 266 to 200 under stock conditions or something along those lines. Will that work with fsb oc'ed? if so, it'll greatly increase chances of the cpu handling the effective low-power state without crashing (since it may reduce the cpu to 266*6 @ 1.1v or so)

Does anyone have a concrete yes or no answer, or a guide which does?

 

[jonny_ftm]

Speedstep or rather EIST as labeled on new MB, will only affect CPU multiplier, not the actual FSB clock, thus lowering the CPU speed. It won't affect the vcore. Just test by yourself to see...

C1E, will lower the vcore on idle state.

Enabling both is needed for optimal power management

On the P5B Deluxe: enabling both at my high vcore: C1E no longer triggers. So, at high vcore, most MB (AMI Bios) will disable C1E. Check for your setup to see if it's the case (don't use CPU-z to monitor vcore)

In my case, enabling both, even that only my CPU frequency is clocked down (multiplier from 9 to 6), my system is unstable on long use (random crashes)

So, as said before: overclock with C1E/EIST disabled. Enable them when you reached your liked stable overclock. On daily long term use, if you have crashes, disable them, if not, you're lucky and should keep them to be "green eco label"

 

[neiroatopelcc]

I can't possibly see why it should become unstable when it clocks down to 6 multiplier, if it doesn't touch the vcore ... it's not like it's stressing the cpu more after all.

Anyhow, if the gigabyte and the asus are anythign alike I suppose it'll also leave the voltage alone on the gigabyte one ... as I feared.


As for trying - now that's the hard part. I've got approx 8 days to build and set up the system - and I haven't even ordered most parts yet.... and after those 8 days I hopefully won't be seeing the system again. That's why I wanted to collect some knowledge beforehand, given the lack of experimenting time.

 

[neiroatopelcc]

What?? cool & quiet bugs in windows ?? On all my athlon64 I never had problems leaving it activated and overclock to the max of the machine.... Don't say it bugs if you didn't tried it.

What I've heard though, is that it will only lower the multiplier, making the system run slower but not lower the voltage of the cpu - thus consuming almost the full wattage at 2/3 the speed, making it useless.

Your right it will only modify the multiplier but it is far from being useless if it lowers your temps when idle. That is what someone leaving it enabled is thinking: saving some heat when idle and lotta power when needed.

My old athlon 3500+ worked with the cool'n'quiet so long as I kept it within 10% of specs, but beyond that it would bluescreen when the amd driver loads - because the driver is not considering the possibility of a vcore change.

With the opteron I had after that one you couldn't even oc 1% before it crashed on cool'n'quiet - and ever since I've made sure to not activate it.
So I HAVE tried it, and I HAVE failed making it work. I even tried 3 versions of the driver - even though the newest was hopelessly old.
I don't hate amd or anything, but cool'n'quiet does not work when overclocking.

Edit: and idle temps shouldn't be reduced by anything more than 1-2% if you're only changing multiplier ....

vcore makes the signal strength more stable - and produces more heat - multiplier clocks the cpu - higher vcore is needed to increase the fading signal strength when oc'ing the cpu. It's the vcore supplied that generates heat, not the actual speed.
Also the northbridge heats equally much if it's running 333 or 266 - as long as you don't change mch voltage.

In english : vcore and cpu state determines heat production, not cpu speed

 

[jonny_ftm]

That's not totally true
vcore is the main responsible for heat production, but at same vcore a 3.2GHz C2D will generate much more heat than a a 1.8Ghz one

You can also try that yourself, lowering the speed, while keeping vcore high

Speed -- > heat. Heat is energy, energy is power. So yes, you will definately consume more power on high cpu clocks for same vcore. Now, to see the real gain, you will need an external wallet mesuring device to check power drawn from your PSU. Tom's guides recently edited a nice article where they measured a 15W decrease at stock speed with the C2D extreme when enabling ESIT/C1E. They sadely didn't test ESIT alone

Someone with the right material, should enlight us, but the gain should be marginal: less than 50% of 15W: 7.5W at the best of the best

 

[jonny_ftm]

Interesting link:
http://www.extreme.outervision.com/psucalculatorlite.jsp

For my setup and overclock (see my sig), my CPU power consumption is estimated at 180W. Only lowering my speed with EIST (6*346=2076Mhz), if their estimation is true, I can reduce power by 60W (120W consumption)

Of course Tom's guide was made at default speed, and we all know overclock needs too much power, especially some high 72% overclocks or above like mine

I also tested the Tom's guide setup: X6800, with ESIT/C1E enabled: the calculator shows a 20W decrease at 1600MHz (6*266)

So, yes, the calculator looks to correspond to real testing, so maybe on high overclocks, the gain can be significant.

I'll try to enable again ESIT alone and check stability, If I can gain 50W, than why not

 

[neiroatopelcc]

but can someone explain how one can save 20W or more by simply clocking down the system at idle?

If the system in general ran 1.8 instead of 3.2 I'd understand why it generates less heat - because the various circuits are more likely to be used when more operations are used in the same timespan (at higher clock) .... but at idle it wouldn't be using any circuits more often, cause it'd just execute the few commands it has to and won't exactly have a huge queue of commands needing to be executed - that could alter the heat output with a higher clock .... that's my understanding of it anyway.

Can someone explain how I'm wrong? and it's entirely possible that I am....

 

[jonny_ftm]

I think you're partially true: the largest gain should be on load of course. But even at idle, it should be some gain: the idle temperature is the reflect of power use: at same vcore, heat increases at idle with speed, so do power use (it is the power used that generates the heat). There is a nice thread here showing the relation between temperature/FSB/vcore

By the way, I tested the Speedstep technology alone without C1E on my current overclock, so far stable, with temperatures at idle lower by about 3-5°C

 

[mrmez]

SS does work with OC... usually

E4300 1.8 oc to 3.1

Normally...
200x9=1.8 or 200x6=1.2 With speed step

Now...
350x9=3.15 or 350x6=2.1 With speed step

It turns down the multiplier, so as long as ur not running the lowest muli... like 450x6... then SS 'SHOULD' work

 

[neiroatopelcc]

okay ... a 3-5C decrease in temp should be an indication to speedstep indeed doing something .... unless the ambient temp was 3C lower than normal ....

 

[apache_lives]

my E6600 @ 8x400 (3200) underclocks to 2400 at idle - stock speed (bar the fsb) most of the time to me is rather cool

 

[jonny_ftm]

Yes, ambiant temperature was identical (digital thermometer)
On my previous settings, I was sometimes unstable with SS. It depends, you always have to try and see. I didn't keep it enough time now, so I will let it on until I have crashes

 

[Zorg]

The P35 can OC to 1333 with no increase in Vcore. That may aid in EIST.

All of the CPUs listed above in our table are 1066 FSB processors, but all ran fine at 1333 FSB at default multiplier and default voltage. Of course this is the FSB frequency Intel will be introducing on their soon-to-be-announced processors. This little side effect will make the P35 with DDR2 a favorite overclockers' board with current Intel Core 2 Duo and Core 2 Quad processors. A $189 E6420 can perform even better than an E6700 just by selecting a 1333 bus on P35 and leaving everything else at default. Likewise, a $500 Q6600 will outperform the ~$1000 QX6700 with just a bus speed change.

This little side effect will certainly be noticed by Intel. We have to wonder how fast the 1066 processors may start disappearing with this kind of free, painless overclocking available with the new P35 boards.

http://www.anandtech.com/memory/showdoc.aspx?i=2989&p=1

Kinda renders all previous chipsets useless aye?

 

[neiroatopelcc]

Now that's just brilliant news, considering I ordered a P35C board for the computer in question (for ddr3 support, futureproofing)

 

[jonny_ftm]

Please consider this carefully.
Read full article also
The overclock limit will be the same for now, since the P965 is not the limit for actual C2D, rather the CPU is a limit: you won't overclock higher
Most current MB, will support the 1333 FSB new CPUs with a bios update

Except DDR3 (which have no real advantage for now over DDR2), no any benefit

 

[Zorg]

Please consider this carefully.
Read full article also
The overclock limit will be the same for now, since the P965 is not the limit for actual C2D, rather the CPU is a limit: you won't overclock higher
Most current MB, will support the 1333 FSB new CPUs with a bios update

Except DDR3 (which have no real advantage for now over DDR2), no any benefit

The limit may be the same, but can you get an OC to 1333 with just a FSB change on a 965? Are you advocating the 965 over the P35? :roll: By the way since you read the article carefully, you know that you can use the P35 with DDR2 and get the same OC as a gift.

 

[Zorg]

Now that's just brilliant news, considering I ordered a P35C board for the computer in question (for ddr3 support, futureproofing)

I was going to ask what board you bought. But I thought I would just post the info. I like that board with the DDR2 and DDR3 slots. Although, you do give up 1394 and some other stuff. I'm still trying to decide.

 

[bigblack]

Ok, my two cents.

I'm a lucky guy that can oc the E6600 @ 2997 333X9 @ 1.22V(24 h Orthos +100% Tat load stable). Using EIST and C1E, only the multiplier goes to 6, resulting a 1998Mhz frequency. Vcore stays at 1.22V, temps remain exactly the same (Tat idle 51-50C, Tat 30 mins 100% load 67-65C, with ambient being 31C). Of course @ load i get the full 2997 speed. So i don't know what happens to power consumption with EIST and C1E enabled but my core temps are exactly the same, so i suppose i don't really conserve any energy. Could be wrong though...

 

[jonny_ftm]

What I mean is FSB 1333 is 333*4
P965 can easily reach 480MHz, and for the good boards get over the 500MHz

I'm not advocating a P965 is better than P35. But, P35 boards when they'll start shipping, like with all previous chips generations (ALL, no exception) including the P965, they will be immature. It took years to the P965 and 975X chips to get the performance they have, and they actually all largely clock above 333MHz.

What I felt important to say for a noob rading your advice:
- Don't switch your P965 board for an expensive P35 now: your gain in overclock will be 0 until intel makes chips clocking 600MHz FSB, and providing the P35 can support it, which I doubt

- DDR3 modules are still immature and don't offer any significant gain over DDR2 modules

Maybe the P35 is the future, but it is definetely not even the near present

 

[Zorg]

What I mean is FSB 1333 is 333*4
P965 can easily reach 480MHz, and for the good boards get over the 500MHz

I'm not advocating a P965 is better than P35. But, P35 boards when they'll start shipping, like with all previous chips generations (ALL, no exception) including the P965, they will be immature. It took years to the P965 and 975X chips to get the performance they have, and they actually all largely clock above 333MHz.

What I felt important to say for a noob rading your advice:
- Don't switch your P965 board for an expensive P35 now: your gain in overclock will be 0 until intel makes chips clocking 600MHz FSB, and providing the P35 can support it, which I doubt

- DDR3 modules are still immature and don't offer any significant gain over DDR2 modules

Maybe the P35 is the future, but it is definitely not even the near present

I was not saying that people should dump their 965/975/680i mobos to get the benefits of the P35. I was just saying that for an OC of 25% you would probably be able to use the EIST, and he stated that 333 was what he wanted. This doesn't even address the painless OC to 333 with no increase in vcore that I posted earlier. I agree, there is always the concern of being an early adopter for anything related to computers, and anything else for that matter. I guess you should read the article again because you missed some important points.

While the purpose of this review was to compare DDR3 and DDR2 performance, something completely different emerged from the memory bandwidth tests. Namely, the memory controller on the P35 is definitely an improvement over the P965 memory controller. This is evident whether the P35 is running DDR2 or DDR3 memory.
Unbuffered results show the same basic pattern as buffered results in this case. Here DDR3 is clearly the best performer at the same slow timings at DDR2-800, with DDR2 on the P35 behind about 3% and DDR2 on P965 about 12% lower. DDR2 is still faster at the better timings available with current DDR2 memory.

Understand I am not advocating DDR3, which should already be apparent from my previous posts. The only reason that DDR3 is in this quote is because DDR2 on the P965 and P35 are benched against it.

In Standard/Buffered memory bandwidth, the P35 (Bearlake) chipset is providing a 16% to 18% improvement in memory bandwidth compared to the P965.

This is a significant increase at any OC.

We were really surprised at the gaming test results. We really did not expect the bandwidth improvement of P35 to have much impact on gaming results, but Far Cry showed a 2% to 5% improvement in performance just comparing P35 to P965 under the same conditions. It really didn't matter whether P35 was running DDR2 or DDR3; the improvement was essentially the same.

What's not to like?

Even considering the early adopter/immature chipset concerns, if you are going to be buying a new mobo there is no good choice other than the P35 for a non integrated graphics solution.

Edit:grammar

 

[neiroatopelcc]

@ jonny_ftm :
1) no 965 board can oc to fsb 333 without changing anything else - at least I haven't seen such board yet.

2) I doubt that p965 boards can EASILY reach 480fsb as you say. My best friend's got a GA-P965-DS4 and even with the mosfet cooling he can't get it stable past 460fsb (rev 1.0) ... so there are exceptions. Not that 460 in any way restricts us from taking his e6400 to the limit, but still. They won't do 480-500 they'll do 450-510
3) ddr3 doesn't give any benefits yet. Correct. But even at horrible timings they're as fast as d9 chips, so they are the future. In two years time they'll be payable, and by then the p35c board I ordered will still be in use, cause the computer I'm upgrading has been running since 2001 - every fu<king day.

@ Zorg :
I'm running a ga-p35-ds4 on my own system - can't get my e6600 past 375x9 stable (with memory at ddr2-1200) due to cpu limitations. However my board 'only' goes to 510fsb stable. I don't know why I can't get 533 as I had hoped, but I really just can't. (running a tr 120 extreme with 2 fans)
If you wish I'll tell you how much more or less I can get out of the p35c board - since it'll also be running with an e6600 - but with stock cooling and some inferior corsair xms pc2-6400 (mine are crucial pc2-8000)

@ all
How much does the g33 graphics suck? would it be able to pull vista and inventor 11 at acceptable levels for a workstation
Edit: By workstation I mean students needs for starting to learn cad stuff.