E8400 is amazing!

[Vicotnik]

some one here in norway did 3Ghz at 1.016v, primestable 4 hrs, as the minimum voltage.

Mine doesn't boot into WinXP at 1.05v.
1.10v seems stable though (done some Prime95 and QuickPAR testing).

Can I trust CPU-Z v1.43 btw? Voltage is set at 1.1v in the BIOS, but CPU-Z reports 1.072v.

[djkest]

Okay, I tested some power settings, it doesn't QUITE make sense yet.
CPU load was stressed with Prime 95 25.5 small FFTs.

3.60 GHz, 1.25625V, 75W idle / 121W load
3.0 GHz, 1.2250V, 74W idle / 105W load
3.0GHz, 1.150V, 73W idle / 107W load
3.0GHz, 1.10V, 73W idle / 95W load (-0.1V on RAM)
On that last run the Vdroop was 1.06V.. is the Norway guy measureing bios voltage or Vdroop?

I was having some problems for a while, these 2 microsoft files kept loading that would take 15% cpu load and I had to keep cancelling them to get true idle power consumption numbers. With my old video card, old power supply, I think I was idling about DOUBLE this..

I might try something more drastic like 2.66GHz, 1.05V or something.. who knows. It's hard to be stable at low voltages. It's not about your cooling anymore, I think it's more about how stable the power is coming from the power supply and motherboard. Also varies from chip to chip.

You can easily see that the idle consumption is low, because large changes in the top end do not effect the idle consumption much at all. This is very good I think for a powerful and fast machine.

Any other requests? I aim to serve.

[smilingcrow]

I'd like to see a really low cost quad core. $150 or so. Would be nice.

It’s only likely to happen this year if AMD are forced to lower the prices of Phenom due to them being uncompetitive. Otherwise I don’t see the incentive for Intel to lower prices much on their Quads.

[djkest]

It probably won't happen. They dont' even have an 45nm wolfdale dual cores for $150. Right now, the E8200 is going for about $190. We might see that go down to oh... $160 or $170 by the summer time. The E8500 is supposedly a beast, because it is the creme de la creme of their wolfdale lots. Someone got 4.8GHz on air w/ one, already.

BTW, @ 1.06V Vdroop it was prime stable for an hour... so yeah, that works. How low can you go?

[ryboto]

that's pretty awesome power consumption, wish I could afford it.

[ghettojiggalo]

Okay, I tested some power settings, it doesn't QUITE make sense yet.
CPU load was stressed with Prime 95 25.5 small FFTs.

3.60 GHz, 1.25625V, 75W idle / 121W load
3.0 GHz, 1.2250V, 74W idle / 105W load
3.0GHz, 1.150V, 73W idle / 107W load
3.0GHz, 1.10V, 73W idle / 95W load (-0.1V on RAM)
On that last run the Vdroop was 1.06V.. is the Norway guy measureing bios voltage or Vdroop?

I was having some problems for a while, these 2 microsoft files kept loading that would take 15% cpu load and I had to keep cancelling them to get true idle power consumption numbers. With my old video card, old power supply, I think I was idling about DOUBLE this..

I might try something more drastic like 2.66GHz, 1.05V or something.. who knows. It's hard to be stable at low voltages. It's not about your cooling anymore, I think it's more about how stable the power is coming from the power supply and motherboard. Also varies from chip to chip.

You can easily see that the idle consumption is low, because large changes in the top end do not effect the idle consumption much at all. This is very good I think for a powerful and fast machine.

Any other requests? I aim to serve.

are you running these tests on a passive ninja or no?

[djkest]

Depends on your definition of passive. See pic on page 1. I call that semi-passive, others would call it just PASSIVE. There are fans close by that have a large influence on it, lets just say THAT.

[wsc]

Not to start a battle or anything, but I would completely disagree with the assertion that you are running passive. There is significant airflow across the CPU HS from something other than convection, ie, it is actively cooled. However that's not to say it isn't capable of running "truely" passively cooled.

Just my $0.02, take it with a grain of salt. The 8400 is an awesome CPU and I am planning on replacing my ancient socket A machine with one.

[smilingcrow]

Okay, I tested some power settings, it doesn't QUITE make sense yet.
CPU load was stressed with Prime 95 25.5 small FFTs.

3.60 GHz, 1.25625V, 75W idle / 121W load
3.0 GHz, 1.2250V, 74W idle / 105W load
3.0GHz, 1.150V, 73W idle / 107W load
3.0GHz, 1.10V, 73W idle / 95W load (-0.1V on RAM)

Legionhardware
has a review of the E8400 that looks at power consumption at various speeds from stock to 4.4GHz; the data shows GHz / VCore / Power – Idle / Load / Difference:

4.40 / 1.450 / 191 / 283 / 92
4.20 / 1.450 / 186 / 268 / 82
4.00 / 1.400 / 183 / 242 / 59
3.60 / 1.225 / 179 / 219 / 40
3.00 / 1.225 / 162 / 204 / 42

(Note: The VCore values at 3 & 3.6GHz have been amended as there was a typo in the article).

The relationship between the 3.0 and 3.6 GHz figures seem anomalous compared to the overall trend. I don’t see why the idle power draw should rise by 17W when the voltage remains unchanged!

The temps that the retail chip they tested reached were very low and they managed 3.6GHz with the Intel box cooler and it only hit 55C. 4GHz should be fairly easily achievable in silence I imagine with a decent setup although beyond that I wonder whether the small die size of the chip might make dissipating the heat problematic. At 4.4GHz I estimate that it’s putting out as much heat as a 65nm Q6600 G0 at ~3GHz but with less surface area.

[NeilBlanchard]

Hello,

Well, at 3.6gHz it is running 20% faster than at 3gHz -- so that would be the 17watt difference, right?

[smilingcrow]

Well, at 3.6gHz it is running 20% faster than at 3gHz -- so that would be the 17watt difference, right?

Surely the idle power draw is affected much more by Vcore than clock frequency which is why 17W seems anomalous. The difference between 3.6 & 4.2GHz is also 600MHz and the idle power difference is only 7W even though the VCore is raised by 0.09V in this case when it is supposedly static between 3 and 3.6GHz. It just doesn’t add up to me.

[mcoleg]

smilingcrow, the power draw does rise with frequency, if not as much as with voltage.

on the other hand, the higher the frequency, the bigger the voltage droop so it might more or less equalize the power consumption between, say, 3.0 and 3.6. on some motherboards it's more prominent so the difference will be smaller; other boards buffer vcore so the difference would me more prominent.

[smilingcrow]

If you look at the temperature data above it shows that the idle temps at 3 and 3.6GHz are very similar and then there’s a relatively large jump from 3.6 to 4GHz as the VCore is increased.

I take this as possibly verifying my hypothesis that there is an anomaly in the power data and I think the 3.6GHz power draw at idle should possibly be 169W and not 179W. Otherwise the idle temp at 3.6GHz should surely be closer to that recorded at 4GHz rather than that at 3GHz.
Another explanation may be that in the process of over-clocking the system they reached a point whereby they had to increase various motherboard voltages which resulted in a one off increase in power consumption which distorted the results.

[mcoleg]

hehe, wanna see something fun then?

here's what i recorded with p5e-vm and e6600 so i could have some reference numbers prior to sticking e8400 into it:

orthos small

p5e-vm

e6600

1.3v vcore ? default in bios

__________________________

3.2 - 124w idle; 164w load

idle 45-46c (700rpm)

load 57-59c (2000+rpm)

cores approx. equal.

m/b 43

__________________________

3.0 - 121w idle; 159w load

idle 55-56c (400rpm)

load 64-65c (700-1000rmp)

cores approx. equal.

m/b 45


====================

i needed just the power figures so i didn't bother putting cpu and exhaust fans on the controller so i could have same rpms all over the place.

[continuum]

Wow... awesome!! Now to wait for the Q9450's...

[djkest]

Those power figures from legion are crap, unless they turned EIST off. The difference in idle power consumption between 3.0 GHz and 3.8GHz was only 2W for me. Although, they are also severely overvolting the CPU, even at stock clocks, which makes no sense to me. Stock voltage is 1.225, why do they have it at 1.36V??? They must be high. You should only use as much voltage as you NEED to get it to run. Heck, I can run 1.10V at 3.0GHz, thats the voltage they should use for that speed, IMO.

[smilingcrow]

Those power figures from legion are crap, unless they turned EIST off.

Don’t forget that it is an over-clocking guide and over-clockers often tend to turn off features such as EIST and C1E as a matter of principle.

The difference in idle power consumption between 3.0 GHz and 3.8GHz was only 2W for me. Although, they are also severely overvolting the CPU, even at stock clocks, which makes no sense to me. Stock voltage is 1.225, why do they have it at 1.36V?

Just because your E8400 is rated at 1.225V don’t forget that Intel CPUs have a wide spread of voltages and for the E8400 the official range is 0.85 – 1.3625V; see here.

You should only use as much voltage as you NEED to get it to run. Heck, I can run 1.10V at 3.0GHz, thats the voltage they should use for that speed, IMO.

Suggesting that everyone should under-volt is unrealistic and you may turn out to have a particularly good chip so 1.1V at 3GHz may not be typical.

Okay, I tested some power settings, it doesn't QUITE make sense yet.
CPU load was stressed with Prime 95 25.5 small FFTs.

3.60 GHz, 1.25625V, 75W idle / 121W load
3.0 GHz, 1.2250V, 74W idle / 105W load
3.0GHz, 1.150V, 73W idle / 107W load
3.0GHz, 1.10V, 73W idle / 95W load (-0.1V on RAM)

I find your results very surprising which suggests that the 45nm process is very different from the 65nm one when it comes to under-volting. Only 2W extra at idle between the two extremes is very different from what you’d get with the 65nm process.

Some people suggested that the difference in the wattage figures that Legionhardware recorded at idle was more down to the differing clock speeds. I just tried disabling EIST/C1E on my system and it showed how little wattage is affected by altering clock speed alone and keeping VCore static. For my E4500 M0 with a fixed VCore the difference between 1.5GHz at idle and 2.75GHz is only 4W. The Legionhardware data shows a difference of 17W between 3 and 3.6GHz with the same VCore. When you look at my data and djkest’s data it suggests that Legionhardware’s figures are anomalous assuming that they are just referring to the CPU and that all other parameters are unchanged.

[Schlotkins]

A little off topic, but do we know when the 45 quads are coming out now?

[KenAF]

A little off topic, but do we know when the 45 quads are coming out now?

Yes. The first week of March. I forget the exact date, but it's around the 7th.

[Conroy]

Okay, I tested some power settings, it doesn't QUITE make sense yet.
CPU load was stressed with Prime 95 25.5 small FFTs.

3.60 GHz, 1.25625V, 75W idle / 121W load
3.0 GHz, 1.2250V, 74W idle / 105W load
3.0GHz, 1.150V, 73W idle / 107W load
3.0GHz, 1.10V, 73W idle / 95W load (-0.1V on RAM)
On that last run the Vdroop was 1.06V.. is the Norway guy measureing bios voltage or Vdroop?

I've finally had time to do tests as well, e8400/ultima-90/radeon 3870/antec solo. I have a 7v 1200rpm slipstream exhausting, and a 1200rpm s-flex on the ultima 90. My default voltage is 1.225v.

[email protected] and [email protected] are the mininum voltages my system can handle without crashing or giving prime95 errors.

I used prime95's "in-place large FFTs"

3.6 ghz, 1.2250v, 77w/125w.
3.0 ghz, 1.2250v, 73w/113w. (stock setting)
3.0 ghz, 1.0375v, 70w/98w.

idle core temps @500rpm s-flex
3.6/1.2250 - 28/39
3.0/1.2250 - 24/35
3.0/1.0375 - 22/34

load core temps @500rpm
3.0/1.2250 - 44/49
3.0/1.0375 - 37/44

load core temps @800rpm
3.6/1.2250 - 48/52
3.0/1.2250 - 43/48
3.0/1.0375 - 35/44

load core temps@1200rpm
3.6/1.2250 - 46/51
3.0/1.2250 - 41/47
3.0/1.0375 - 35/44

I don't really trust these temperature readings, especially the low-end numbers, since I think my room temp is around 21, but there you go.

Oh, also, when I say 1.225v, cpuid actually reports 1.216v, and when I say 1.0375v, cpuid actually reports 1.024v.

By the way, Vista 64 is a pain in the butt. neither easytune nor rmclock work properly because they use drivers that are unsigned.

[smilingcrow]

I've finally had time to do tests as well:

3.6 ghz, 1.2250v, 77w/125w.
3.0 ghz, 1.2250v, 73w/113w. (stock setting)
3.0 ghz, 1.0375v, 70w/98w.

That looks impressive especially since you have a VGA card that will consume 15 – 20W at idle. Take that out of the equation and your looking at ~80W at load @3GHz when under-volted.

Were you able to determine the maximum voltage that can be provided by using software utilities? I suppose it might well be the 1.225V default!

I don't really trust these temperature readings, especially the low-end numbers, since I think my room temp is around 21, but there you go.

What were you using to measure temps? I’ve seen a lot of reports suggesting that certain Wolfdale CPUs have problems reporting accurate DTS temps for lower values and sometimes the temperature reported are static at say 43C at idle and only increases when under load and the temp rises above this value. Something to keep in mind if anyone has unusually high idle temps with these chips.

[cmthomson]

What were you using to measure temps? I’ve seen a lot of reports suggesting that certain Wolfdale CPUs have problems reporting accurate DTS temps for lower values and sometimes the temperature reported are static at say 43C at idle and only increases when under load and the temp rises above this value. Something to keep in mind if anyone has unusually high idle temps with these chips.

Remember that DTS supplies a negative number, degrees below the (secret) throttling temperature of the CPU. In the first generation C2D chips, throttling was reported to be 85C (although some programs used values as high as 97C), and the second generation C2Ds were reported to be 100C. Well here we are with another generation. Maybe the throttling temperature is so high, and the idle temperature is so low, that the value reported by DTS maxes out...

[smilingcrow]

Remember that DTS supplies a negative number, degrees below the (secret) throttling temperature of the CPU. In the first generation C2D chips, throttling was reported to be 85C (although some programs used values as high as 97C), and the second generation C2Ds were reported to be 100C. Well here we are with another generation. Maybe the throttling temperature is so high, and the idle temperature is so low, that the value reported by DTS maxes out...

I like your way of thinking but the way the problem is being reported suggests to me that this isn’t the issue.
The DTS values are I believe stored in registers that are directly accessible via Windows so unless they are less than 8 bit (256 steps) they should have plenty of headroom.
CoreTemp reports the Tjunction Max for Wolfdale as 105C which is only 5C greater than for more recent 65nm C2Ds. Whether this is accurate it may be too early to say.

[Conroy]

Were you able to determine the maximum voltage that can be provided by using software utilities? I suppose it might well be the 1.225V default!

I'm pretty sure easytune could set voltages higher than the default, although I didn't do much with higher voltages intentionally. I believe that easytune set voltages in the 1.3x range when I was playing with it and turned on the mysterious "turbo" mode. I know that when the bios set it to a higher voltage, I could use easytune to set it back down. I'll try some more in a few days. RMClock wouldn't load at all in vista 64. Do you know of any other software to try? I may have to just install an xp partition next weekend.

When my accelero gets in this week, I'm also going to try swapping out the s-flex with a 92mm nexus to see what the effects are. I guess that's not of any interest to this thread though.

What were you using to measure temps? I’ve seen a lot of reports suggesting that certain Wolfdale CPUs have problems reporting accurate DTS temps for lower values and sometimes the temperature reported are static at say 43C at idle and only increases when under load and the temp rises above this value. Something to keep in mind if anyone has unusually high idle temps with these chips.

CPUID's hardware monitor was the only vista 64-compatible program I could find that would give me core temperature readings. I saw the problem with the static idle temps on other boards, but I did not experience it on my machine; I just have an abnormally low idle temp on one core.

[smilingcrow]

I'm pretty sure easytune could set voltages higher than the default, although I didn't do much with higher voltages intentionally. I believe that easytune set voltages in the 1.3x range when I was playing with it and turned on the mysterious "turbo" mode. I know that when the bios set it to a higher voltage, I could use easytune to set it back down.

Does Easytune doesn’t allow you to set individual voltages for each multiplier as RMClock does? I’ve assumed it’s designed more for setting a single VCore that doesn’t change.

RMClock wouldn't load at all in vista 64. Do you know of any other software to try? I may have to just install an xp partition next weekend.

There’s a 64 bit version of CrystalCPUID.

[Conroy]

Does Easytune doesn’t allow you to set individual voltages for each multiplier as RMClock does? I’ve assumed it’s designed more for setting a single VCore that doesn’t change.

Yep, single VCore. It's not as useful as RMClock would be.

Thanks for the pointer to CrystalCPUID! I'll try it out.

[smilingcrow]

Those power figures from legion are crap, unless they turned EIST off. Although, they are also severely overvolting the CPU, even at stock clocks, which makes no sense to me. Stock voltage is 1.225, why do they have it at 1.36V?

I checked with the author and it was a mistake, where it says 1.3625V it is actually 1.2250V.

[Beyonder]

Looking at this thread, I can't help but wonder: how feasible would it be to run an E8400 system with a decent video card on a picopsu? I'd love to run a minimalist system (say, 3850?) with an E8400, one of those Western Digital green drives, and a good video card for my primary system...

[smilingcrow]

Looking at this thread, I can't help but wonder: how feasible would it be to run an E8400 system with a decent video card on a picopsu? I'd love to run a minimalist system (say, 3850?) with an E8400, one of those Western Digital green drives, and a good video card for my primary system...

No chance of that working as the PicoPSU 120 is only rated for 84W continuous output on the +12V rail.

[djkest]

You can get a 120W pico PSU. Maybe if you had one of the new 36xx radeon cards coming out soon, this CPU, and a WD green hard drive, you could do it.

Anyway, interesting to see your results so close to mine! I just crashed after 15 min of prime95... testing my 3.8GHz numbers. I may have to mount a fan on the ninja, it's starting to get a little toasty.