• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Speedstep/EIST and C1E - Does it impact performance?

D

DarkScythe

Limp Gawd
Joined
Mar 7, 2008
Messages
367
Hello,

My E8400 is slated to arrive today and as excited as I am about that, I still have some research to do lol. I'm planning to do a "minor" OC to 3.6GHz with it (9x400) but I'm concerned about a few things.

I'd like to enable the power saving features for obvious reasons, and I believe C1E and EIST/Speedstep (I assume they're the same thing) are the only 2 power saving features that are usually recommended. However, I also have a Gigabyte GA-EP35-DS3P board which comes with DES/Dynamic Energy Saver which I've read causes a performance hit when active, so I'm thinking of disabling it. I can't find any information on whether or not C1E/EIST does the same thing and impacts performance or not though. Can anyone shed some light on this?

Thanks in advance.
 
It does not.

When the CPU is under low load < 5%, it stays throttled down.. but when it is under a little more (don't know exact %), it instantly throttles up..

If your cpu idles at 2-3% due to a server, it will stay throttled, but if it spikes, during that spike it will throttle up.

I verified this with CPU ID.
 
It impacts performance minimally. The savings in power are definitely worth it and you won't notice it.
 
Okay, if it does not or very minimally affects it I guess it's ok..
The Gigabyte DES was reported to really cut into your FPS or something.
My LCD is quite old, so my maximum gaming resolution is limited to 1280x1024 which I believe is still a CPU-bound resolution.

My CPU just arrived, so I'm quite excited now to be holding it. Q745A849 is printed on it, whatever that means. In any case, I believe EIST/Speedstep works by lowering the multiplier to 6x when idle, correct? And C1E lowers voltage?

Does this mean that in order to use C1E, you must set voltage control to automatic? People say the mobo's are usually too generous with voltage when set to auto, but I'm not sure how it will handle the voltages when I overclock.

Thanks again
 
EIST and C1E can both lower the voltage, I believe; EIST does it when the load is low, whereas C1E is a low power state the CPU can go into when it's idle. To be honest I'm not sure if EIST+C1E is any better than EIST alone though.

You do have to have auto voltage on for the voltage to be lowered, and yes, motherboards tend to overvolt overclocked CPUs (if it's not overclocked, they'll give the chip its VID, although admittedly it might run okay at less than that, but I wouldn't call it overvolted per se). This unfortunately means that you don't get as much power saving as you might have, once you set a fixed voltage. It would be nice if motherboards let you set the voltage it should drop to on EIST/C1E, but I don't think they do.
 
Do both EIST and C1E need to be enabled at the same time or can you pick and choose?

I was under the impression that EIST only lowered multiplier, not voltage. If automatic voltage adjustment needs to be enabled, then that's pretty bad for overclocking since it could potentially give the CPU too much voltage and fry it..
 
You can enable EIST and C1E separately.

EIST will only lower the multiplier if you have set a fixed voltage. If you have the BIOS voltage control set to automatic, it will lower the multiplier and the voltage. It works with or without, but it stops changing the voltage if you specify one.
 
So then that means I can still save some power by leaving EIST on to lower the multiplier after I manually set the voltage to its default, correct? If that's the case, would C1E have any effect, since voltage is not set to automatic control?

Some quick googling also appears to say that I might have lucked out and got a proc from batch Q745A which seems to be one of the "better" ones.

In any case, would this thread be better suited under overclocking? I originally thought it would be under Intel Processors since I'm talking about Intel proc technology, but I think it's the overclockers who really have to deal with it..
 
Just leave them both on. The voltage won't change if you set anything other than automatic, but it doesn't hurt to leave them both on anyway. And the Gigabyte power savings stuff only works at default speeds - if you overclock it is disabled.
 
I just did some quick googling, and yes the voltage won't change if I set it manually, but if EIST/C1E can only change the multiplier, is there still any real power savings? I thought the bulk of the power savings was in the reduction of the voltage.
 
It's both. Power is proportional to clock speed and by the square of voltage, both minus some overhead that stays the same with speedstep like I/O.

If for example you have a 9x multiplier chip that scales from 1.3v down to 1.2v at a 6x multiplier, you would save:

over 33%* of the power at 6x vs 9x multiplier
~15% by the drop in voltage from 1.3v to 1.2v

Both together is of course ideal. I have both EIST and C1E enabled on my overclocked Q6600. At 3GHz, the voltage is ~1.275v and at 2GHz the voltage is 1.200v. I have to slightly bump the voltage at 3.1GHz.

* you actually save more with C1E since the CPU drops to even lower power states at idle and the CPU isn't running at 100% with the 6x multiplier. Newer stepping 65nm dual core chips use only ~8W at idle (Extended Halt/C1E). I can imagine that 45nm can match or beat it with C2E.
 
The lack of voltage change when overclocking seems to be an issue with Gigabyte boards (and possibly some other brands as well, but definitely not all of them). My Abit IP-35E goes from 1.264V to 1.344V when the multi changes from 6x to 10x by running Prime95, and the voltage is set at 1.3425V manually in BIOS.
 
Well, my biggest worry is that the motherboard or C1E/EIST will choose to overvolt the processor.

Most people appear to be able to go to 9x400 on stock volts or a very very minor bump, but say that if left on auto, the mobo will give it way more than needed, which is potentially dangerous IMO. As well, if the FSB remains at 400 and the multi drops to 6x, we're still a little bit above 6x333 stock, and I'm afraid that it will drop the voltage to the 6x333 level making it so that 6x400 isn't receiving enough power.

I would love to have all the power saving features enabled, but I also don't want it killing anything accidentally.
 
If the motherboard has been updated with the latest bios compatible with 45nm chips, then it should take into account the reduced voltage needs of the wolfdales and move the autovolts accordingly. My last chip, an e6400, with a 965 board (DS3) jacked up the volts to the 1.45 region when on auto. That obviously would be excessive for the 45nm (although some OC'ers take it up that high or more).

It may still be the case with some motherboards that are compatible with 1333fsb chips (which include the 45nm), and the wolfdale may boot up and work but its not optimized for it and could give it more volts than necessary when on auto. My DS3L was like that when I first bought it, it worked right away with my e8400, but I promptly DL'd the latest 45nm compatible bios before I even contemplated OC'ing it. Even then I put vcore on manual, if you know its working range, then no need for autov.

Re C1E and EIST, I always had them on for as long as I remember. Do some synth benches (SANDRA and 3DMark CPU test) and you will find virtually no difference in performance.
 
I will check what BIOS version my mobo has, although I picked this board partially because I heard it had Wolfdale support out of the box.

At this point I know I will set a manual vcore while enabling EIST/C1E, but my problem is that if I set the vcore at a manual setting, that effectively reduces half (or more) of the potential power I could be saving only because of fear that it will not assign correct voltages after an OC. Even if it is optimized for 45nm processors, I'm not sure if Gigabyte takes an OC'd Wolfdale into account for their automatic assignments, and even then, people say they give more than necessary.

So say at stock (I don't know what stock is, but let's say it's 1.2v) then maybe it will auto-assign 1.3v to be safe. At idle it'll underclock and give it 0.9v. Assuming it scales with an overclock, let's say we need 1.4v for it. Gigabyte might auto-assign it 1.45 or more which I hear is very bad for the Woldales, and when idle, it might lower it to 1.0v or something which might not be enough juice for the processor to run at 6x(whatever it's OC'd to.)

If I am completely wrong in this assumption, feel free to correct me because this is what I THINK happens, or would happen.
 
There is obviously communication between mobo makers and Intel re compatibility for the 45nm chips. I dont think major brands like Gigabyte, Asus, etc would miscalculate voltage to these chips if their bios updates say its 45nm compatible. And I would have no problem putting it on auto at stock speeds or a mild OC, but where I would draw the line is if I intend to do 4ghz or more.

On the retail box it say 1.225v as default, so that would be a nice figure to put into the bios and should cover a nice OC (depending on the chip) to 3.6ghz or even slightly more. Mine is at 1.275 (1.232 cpuz) doing 3.6, most other chips should do better than that. Unfortunately mine is not one of the better weeks/batches (Q748) but is still nice and cool at that OC.
 
You must log in or register to reply here.
Back
Top