920 D0 3850a795, 4.2ghz@0.92v ?

[S&M]

what do you ppl think, should I keep this cpu ?

DFI DK X58
INTEL CORE I7 920 D0 3850A795
OCZ BLADE 2000MHZ 7-8-7 3X2GB

this is just first run , all voltage is set to auto ...
0.92V when Idle, 1.21 when load

memory timings not set, also testing with stock cooler for now ...

http://resources.iopanel.net/guides/SM/920D0.JPG

[cstkl1]

disable c state
run linx
run 8x32m .

rethink ure thread title
then decide.

[ChaosAD]

And you should also upgrade everest to v5 for more accurate readings.

[roller11]

Hard to understand what's going on with the validation, .928 Vcore is inconsistent with 21x. That is, if .928/1.21 volts is idle/load, then at .928 should show 12x multi and EIST is enabled. The screen shot suggests that isn't the case. So was EIST enabled and the voltage was really 1.21 during the superpi run? Or was the run made with vcore = .928 (i.e. loaded), and thread title is wrong? Need clarification, but if you managed a 32M superpi run at 4197 Mhz and .928 vcore, then this is definitely a great D0 and you should keep it.

[cstkl1]

Hard to understand what's going on with the validation, .928 Vcore is inconsistent with 21x. That is, if .928/1.21 volts is idle/load, then at .928 should show 12x multi and EIST is enabled. The screen shot suggests that isn't the case. So was EIST enabled and the voltage was really 1.21 during the superpi run? Or was the run made with vcore = .928 (i.e. loaded), and thread title is wrong? Need clarification, but if you managed a 32M superpi run at 4197 Mhz and .928 vcore, then this is definitely a great D0 and you should keep it.

no he didnt
his vista is running on power saving mode or balance mode
that means min proc is at 5%
and max at 100%
with eist enable.. u get results like above

look at his 16m score
i can do that same result on 32m at 4.0-4.2ghz

[unclewebb]

cstkl1: He is using XP and not Vista so the Minimum processor state doesn't apply.

When using XP, to fully enable Speedstep you might need to set your Desktop processor to Portable/Laptop in the Power Options -> Power Schemes.

http://img528.imageshack.us/img528/4057/xppower.png

When you do this, EIST will be fully enabled and CPU-Z and RealTemp should both report a multiplier of 12.0. Neither program in the screen shot shows that so that means that EIST is not fully enabled.

S&M: Looks like a nice D0. Show us some more results at full load running Prime95 Small FFTs to see what voltage you can run at when loaded.

[S&M]

@cstkl1 - first of all, I was asking if it's possible , that this is the voltage ...
second thing is , this cpu temps are very high , at bios, when everything is set at stock , I see around 48-50 celsius degrees, this is with
the Core i7 stock cooler, is it possible ? bad reading ? heatsink not attached good ?

[unclewebb]

The Core i7 OEM cooler is completely inadequate when doing any serious overclocking but the Core i7 temperature sensors are usually pretty accurate, especially core 0. What's your air temperature and is your case open or closed? Your temps don't seem too wrong.

1.21 volts when loaded sounds about right when set to AUTO. I haven't heard any complaints about that board reporting the wrong core voltage and if it's dropping to 0.92 volts at idle then that part of EIST / SpeedStep seems to be working. The only way to know 100% for sure is to use a volt meter.

[cstkl1]

@cstkl1 - first of all, I was asking if it's possible , that this is the voltage ...
second thing is , this cpu temps are very high , at bios, when everything is set at stock , I see around 48-50 celsius degrees, this is with
the Core i7 stock cooler, is it possible ? bad reading ? heatsink not attached good ?

the voltage is because of ure eist
and its not anywhere near stable

the d0's are not fussy on HT voltages
so no suprise at all.

stock coolers are not good enough for any form of overclocking for d0's
my c0/c1's were good for stock vcore
but d0's arent.. have to undervolt them .

[cstkl1]

The Core i7 OEM cooler is completely inadequate when doing any serious overclocking but the Core i7 temperature sensors are usually pretty accurate, especially core 0. What's your air temperature and is your case open or closed? Your temps don't seem too wrong.

1.21 volts when loaded sounds about right when set to AUTO. I haven't heard any complaints about that board reporting the wrong core voltage and if it's dropping to 0.92 volts at idle then that part of EIST / SpeedStep seems to be working. The only way to know 100% for sure is to use a volt meter.

dude
dont just look at cpuz
look at the other reported multi's and speed

his eist is working just fluctuating because of the superpi or some background proccess.

somemore i suspect he has triggered the tdp/tdc protection also.

so eist, tm , c-state , tdp/tdc .. u get something like the above screenshot.

[Ancient_1]

dude
dont just look at cpuz
look at the other reported multi's and speed

his eist is working just fluctuating because of the superpi or some background proccess.

somemore i suspect he has triggered the tdp/tdc protection also.

so eist, tm , c-state , tdp/tdc .. u get something like the above screenshot.

I have been through this and what unclewebb says is true. once you set it to portable/laptop the readings will be right in those apps and the muti bounce will become more stable.

Here is what my multi's looked like before I changed the setting to Portable/Laptop.

I am doing a few hours of encoding right now or would give a shot of those programs all showing the correct values.

http://webpages.charter.net/chew_toy/Ci7/MultiI.jpg

Here is how they read after changing to Portable/Laptop

http://webpages.charter.net/chew_toy/Ci7/MultiN.png

[DJSUB]

voltage change to very low in idle if you enabled C1 or CX in the bios.
same with the clockspeed. if you check cpuz only few mhz lower than the original you set.
but when you try to look at it in realtemp & coretemp you will see that your clockspeed is getting more lower clock & the multi also changes everytime.
even the tdp will go down in coretemp.

if you want to know the real auto voltage in that clockspeed you should disable C1 or Cx in the bios.

[unclewebb]

dude
dont just look at cpuz
look at the other reported multi's and speed

his eist is working just fluctuating because of the superpi or some background proccess.

somemore i suspect he has triggered the tdp/tdc protection also.


dude, I did look at his screen shot and noticed that 8m54.750 for a 32M SuperPi run is normal for a Core i7 at 4.2 GHz so no it doesn't look like he triggered tdp/tdc protection.

Here's how ReverendMaynard did a while ago at 4.2 GHz (9m03.972):
http://forums.extremeoverclocking.com/attachment.php?attachmentid=126682&d=1228783060

I'm not sure what cooling the Reverend was using in his screen shot but S&M's idle temps are also very similar to ReverendMaynard.

The older version of Everest he is using might not correctly support Core i7.

EIST / Speedstep has two parts. One part lowers the core voltage at idle and the other part lowers the multiplier to 12.0 at idle on a Core i7. With RealTemp reporting 0.0% Load, he is definitely idle and there aren't any significant background processes going on. If he fully enables SpeedStep like I showed how to do in the post then his multiplier will also drop down to 12.0 at idle which CPU-Z and RealTemp will both correctly report.

[sparks.oc]

On 4200MHz and PI I have: 9 minutes and 2.945s (PI mod 1.5XS). So the PI calculation seems ok.

[eva2000]

S&M are you setting vcore via vid + vid special add or just via vid in bios ? try with just vid and no special add vid % added to it.

32M pi does seem a bit off but that can be down to tweaking/tuning

http://fileshosts.com/intel/DFI/DFI_UT_X58_T3EH8/results/eksupremelt/920_3836a756_1/corsair/2000C7GT/231/19x/10xmem/190b_dof_210_1050_78720_rfc60_rw10_wtr5_rrd5_rc31_ faw31_rtpa_1.3v_10023_1.53v_1.8v_129125/spi32m_08m48s610ms.png

[unclewebb]

Nice time eva2000 at that speed.

I just noticed one difference between SuperPI mod 1.5XS and HyperPi. When you run SuperPI 1M, the calculation gets shared between all cores. In HyperPI, the calculation gets locked to core 0 when you are only running 1 thread. It's not much of a difference for 1M but the difference might be more when running 32M.

Given that was S&M's first SuperPI run on his new D0 without any memory tweaking, it looks like it's performing as expected.

[roller11]

dude
dont just look at cpuz
look at the other reported multi's and speed

his eist is working just fluctuating because of the superpi or some background proccess.

somemore i suspect he has triggered the tdp/tdc protection also.

so eist, tm , c-state , tdp/tdc .. u get something like the above screenshot.

Even if EIST is fluctuating, vcore should track multi. So when vcore fluctuates down to .92, multi should track to down to 12x, but he got .92 at
21x multi. At least that's what I've observed on all the 8 different x58 boards I've tested, I always have EIST enabled and C states disabled/enabled, never seen a high multi with a low Vcore (or vice versa).
If power plan is 'high perf', EIST is disabled. If plan is 'power saver' multi and vcore are always low. if 'balanced', EIST works correctly unless you have a gigabyte or ASUS board.

[Zucker2k]

Chips seems impressive; but let's see some shots with all cores at 100% load and cpu-z in the background. Should help answer some doubts.