Sign in with
Sign up | Sign in
Your question

I7-860 OC evaluation

Last response: in Overclocking
Share
Anonymous
a b K Overclocking
January 10, 2010 9:01:27 AM

Background

Have been working on this OC for 2 days now, mostly since i am new to this. I believe i have a stable OC now (had temp issues in the start).

I would be very grateful if someone could look at my OC to see if it is ok, which OC to choose, and whether or not it can be improved further.

My main issue earlier was high temp, but this was solved by turning of hyper-threading.

Specs:

CORSAIR FULL TWR OBSIDIAN SERIES 800D
COOLERMASTER Hyper212 Plus
CORSAIR Power Supply 850W HX
CORSAIR 4GB (KIT) DDR3 1600MHz/ CL7/ DOMINATOR GT
GIGABYTE GA-P55A-UD6, P55, Socket-1156
XFX HD 5970

Results :

Stock cpu without OC (Hyper-threading and turbo on)

Idle:

Load:

OC1:
QPI Clock Ratio : Auto
BCLK Frequency : 160
Turbo : On
Hyper-Threading : On
System memory multiplier : 10x
CPU Vcore : Auto
QPI/Vtt voltage : Auto

CPU Core Speed : 3520
Memory Frequency : 1600




My observations : OK temp, Low core frequency, everything on and stable

OC2:
QPI Clock Ratio : 32
BCLK Frequency : 175
Turbo : Off
Hyper-Threading : On
System memory multiplier : 8x
CPU Vcore : 1.225
QPI/Vtt voltage : Auto

CPU Core Speed : 3674
Memory Frequency : 1400




My observations : High temp, Good core frequency, Hyper-threading still on. Was not able to OC further cause the temp got above 70 o it went unstable with lower vcore.

Decided to turn off hyper threading and experiment with turbo and BCLK.

OC3:
QPI Clock Ratio : 32
BCLK Frequency : 170
Turbo : On
Hyper-Threading : OFF
System memory multiplier : 10x
CPU Vcore : 1.20
QPI/Vtt voltage : Auto

CPU Core Speed : 3740
Memory Frequency : 1700




My observations : OK temp, Good core frequency, high RAM speed, perhaps abit high processor power usage (110W)

OC4:
QPI Clock Ratio : 32
BCLK Frequency : 180
Turbo : Off
Hyper-Threading : OFF
System memory multiplier : 8x
CPU Vcore : 1.20
QPI/Vtt voltage : Auto

CPU Core Speed : 3780
Memory Frequency : 1440




My observations : OK temp, Good core frequency, "Low" RAM speed, ok processor power usage (95W)


FINAL CONCLUSIONS:

So i guess the question is, do i need hyper-threading considering i will use the rig for gaming? Hyper-threading gives high temperatures as seen in OC2. They are below 70 though so might be ok???

If i turn of hyper threading the temp drops significantly as seen in OC3 and OC4 to below 60. I can run the BLCK at 180 without turbo or at 170 with turbo and get the same core frequency. at 170 i use higher RAM multiplier then with 180. Both temps are ok, but 180 has a lower cpu power usage.

So please help me, what should i use, and should i alter anything else?



More about : 860 evaluation

a c 126 K Overclocking
January 10, 2010 9:44:20 AM

well I've done some gaming tests with hyper-threading and without hyper threading and in most cases HT enabled made very little difference, but in Race driver GRID I gained an additional 20FPS AVG and also the Min FPS shot up.

Tests done with a Ci7 920 @ 4ghz + HD5850 Stock. I'll try and find the exact numbers for you, but like i said, in most games there was very little improvement.

Im wondering should I keep HT on at 4ghz or take it off and go for 4.3ghz as like yourself my OC is also limited by temps.
Anonymous
a b K Overclocking
January 10, 2010 11:41:24 AM

Quote:
Turning off ht is kinda like hitting the fuel saver button on your car isn't it?


Tbh i dont think so. Cause no games today can use 4 cores, let alone 8. So in most games it shouldnt make any difference, thats what i have been told.

Most overclock guides also tells you to turn HT off.
Related resources
a b K Overclocking
January 10, 2010 3:15:42 PM

If you did a little research you could have saved $80-90 on your CPU and just bought an i5 750 instead and a better cooler.

I like having all 8 threads running, never get bogged down, I paid for them I'm using them, just IMHO.

Anonymous
a b K Overclocking
January 10, 2010 3:19:54 PM

Here is my latest try :

BCLK frequency: 180
Memory multiplier : 8x
QPI Clock Ratio : 32
Turbo : Off
Hyper-Threading : On
CPU Vcore : 1.25
QPI/Vtt voltage : 1.24
RAM voltage : 1.64

Seems to run stable (1 hour prime blend and 30 mins small FTT tests)

Hyper threading is now on, but get some high temperatures.



Maks temp during prime stress is now 70-72 (depending on if case is open/closed). The problem is i dont know if this is ok or not. Some say to aim for below 60, while a guy on overclocker says this is more then fine considering its prime. Dont know what to belive......

January 10, 2010 5:35:59 PM

70-72c should be fine. You have to consider that this is at full P95 load. Which means that this is the hottest your cores will ever get. (under same conditions and ambient temp) Normal use will be a lot lower. The most important thing is to be stable! That means a small and large FFT's test for at least 14 hours each in prime95. Some even say 24 hours. I say if its going to stay at that setting for 24/7 use, then yes 24 hours. Better to be safe since you are going to leave it there. JMHO.
January 10, 2010 6:22:53 PM

Three suggestions:

1. Don't use "Auto" for voltages. I'd guess theres a good chance by setting vtt at "auto" you are running in excess of intel's voltage max spec (1.21v). Now, some folks are doing this intentionally, others are keeping their overclocks within the intel spec. Its an interesting debate, but don't avoid the decision and let some "auto" function make it for you. Of course we exceed the Intel specs on cpu clock rate as the basic goal but all of us have concluded that the clock max settings are more influenced by matching the available hardware to the markets willingness to buy. The voltages specs, however, appear to some of us to be set by engineering, in which case, consequences to exceeding them may be real.

2. Hyperthreading: It depends on your applications. Evaluate them for impact. I find my graphics modeller makes great use of threading, and therefore I leave it on. You may be happy without.

3. Prime 95 isn't telling you that your oc is stable in only a 3 test period. I don't really know how long you are running it. I was studying my box initially by short periods of prime 95 testing just to get a feel for it. I was pretty sure I was stable at 3.6 with very low vcore, but long term prime 95 testing proved I wasn't. I had my first failure on that build after 2 hours or so. But the instability was real, as I had seen an aborted boot once (only once, which of course I just raised my eyebrows and rebooted). I was able to stabilize my 3.6 but only after raising vcore to the point that I declared my oc was finished at 3.6 until I improved my cooling.

Cheers,

Bob
Anonymous
a b K Overclocking
January 10, 2010 8:11:57 PM

Thx for the reply bob, was helpfull. To follow up i got a couple of questions.

1) On which voltage is the max spec 1.21v? The cpu is rated for 1.45 vcore max i believe.

2) which voltages should i set manually? Now i am running with
vcore at 1.25
ram at 1.65
ptt at 1.2

Should i manually set other voltages? I have not done long term primes yet (max of 1 hour) but i have a feeling i might get unstable. Cant increase vcore any more or temps will be to high, so guess ill try a new burn with same voltages at 175 and hope for the best.
Anonymous
a b K Overclocking
January 10, 2010 8:34:54 PM

one more thing, do you still use speed step or similiar to make cpu ramp down at idle?
January 11, 2010 1:52:01 AM

You have a lot of room between your vcore and max...I think its around 1.45 or so. You mustn't raise your voltage on ram any higher. If ptt meant vtt, then you are over voltage, intel's spec max is 1.21. see http://download.intel.com/design/processor/datashts/322... table 7.4.

I've paid no attention to ramping down at idle, so you can teach me that side of it!

B0b
Anonymous
a b K Overclocking
January 11, 2010 3:49:15 AM

Ramping down at idle will make the multiplier drop when the cpu is able. If i have it on the multiplier will drop from 21 to 9 during idle. Its called differently on different mobos. Some call it speedsted, on my gigabyte, its something to do with reducing Cx states...

didnt mean ptt, mean QPI/vtt voltage for the uncore.
Anonymous
a b K Overclocking
January 11, 2010 3:58:37 AM

however, dont know if that is what is meant by vtt, cause i have read in a few guides that many mean the QPI/vtt voltage should be the same as vcore voltage during over clocking.
a c 126 K Overclocking
January 11, 2010 8:56:50 AM

im 99.9% sure your CPU will not ramp down at idle if you OC, But dnt take my word for it, just test it out yourself to be absolutely sure.

My QPI and Vcore voltage are not the same, and they don't need to be. My QPI is 1.313V and my Vcore is 1.28xxxV
Anonymous
a b K Overclocking
January 11, 2010 2:07:50 PM

Seems like i now have a stable OC 180 (3780Mhz) with hyper threading on. Done 8 hours of small FTT, starting 8 hours of large now.

My pc didnt ramp down on its own, i had to change the setting in bios from auto to enable. But my pc will now ramp down when idle.
January 11, 2010 8:24:53 PM

whats your voltages?
February 9, 2010 12:02:22 PM

Finally an i7 860 thread!

I have mine OC'd to 3.6ghz with 1.21v...probably a bit higher than it needs, but I have to get to testing it. Just got to overclocking it last night. Will do more testing this week.

Suggestions for suitable core voltage?? Temps are 36C idle, will have to run prime 95 after I find my ideal OC profile..
February 9, 2010 4:39:33 PM

I'm continually studying my build, after several months I'm getting a good feel for it. I'm now benching with LinX, which produced many more errors than either OCCT or Prime95. OC that appeared stable before proved not to pass the LinX testing. However, OC that does pass LinX testing has also passed under the other benchmark proggys, so now I use LinX exclusively.
I've recently been working with the 21 multiplier, and exploring bclk in the 170s. I'd previously ignored this range because it produced relatively low memory speeds (1392). As some compensation, I've tighted the memory timings to 7,7,7,24. I leave LLC off. I've learned quite a bit about llc now, and know that if reports that involve vcore don't include this information, it leaves room for substantial misunderstanding. So, with LLC off, my vcore is now at 1.3125 in bios, with a 1.232 remaining as vcore under 100% load. My Vtt is 1.21. My Vdimm is set at 1.6v. Vpll is set at 1.86v. Hyperthreading is Enabled. At 174x21 (3.654ghz) I pass LinX for 50 minutes, and coretemp reports max temps to be 73, 69, 68, 73. Running OCCT for 1 hour produced lower temps at 70, 66, 68, 71. I suspect this may be my 24/7 settings, although I'm always studying the overclock. I don't want the temps to go any higher for a long term build. I watched temps during a 3d rendering last night, in which coretemp registered 100% activity in all 4 cores, with 8 workers showing on the rendering, temps reached the low 60s. I suspect this is the very worst case real life application, so the system should be good for a long life. Time will tell.
February 9, 2010 11:04:24 PM

My 860 runs stable @3.89GHz for more than 4 hours in P95.

Since I don't do heavy encoding or anything that would run it to the max for that long I consider it more than stable for my means. I've got all the energy-saving features on as well so the cores drop their multi's down to 9 when not in use. Been running this setup for a few weeks now and the only errors I've had in daily use have been while OC'ing my GPU.

Note: The four hours mentioned above is simply the longest I've let P95 run. I've got *** to do and haven't had an opportunity to let it run longer; the chip may well go beyond that time frame in P95
February 10, 2010 12:28:01 AM

Deke,

You can find out in 1 hour what LinX reports for your stability.

Can you tell us
1. do you enable LLC?
2. Vtt=?
3. Vcore in bios =?
4. What is your multiplier?
5. What is your temps on your hottest cores during 100% load?
6. Using hyperthreading?

Its fun to discuss how far we can push, but also good to describe in detail so we learn the difference between a really hot, high power dissapating oc, vs a nice 24/7 type oc.
February 10, 2010 5:24:55 PM

I leave LLC on and actually have had great luck leaving my voltages all on auto besides CpuPLL (1.9) and VRAM (1.56). I was manually changing the voltages but actually got better results letting the mobo decide.

I didn't document the max voltages while running the last test but did note that Vcore never reached 1.4

Multi at 21 (which the computer always jumped to 22 since I didn't physically disable turbo) and baseclock at 177.

With HT on and all cores/threads blazing away the max temp any core reaches is 74 in P95
Idle temps drop below 20 most of the time.

If you would like more specifics on what my mobo is running voltage-wise I can run another test later today and try out LinX
February 10, 2010 6:14:51 PM

THanks Deke, you might indeed want to check your voltages. Setting voltages to auto is normally a bad idea, as the mobos notoriously tend to ignore Intel specifications for maximum vtt (1.21). Still, you could be fine...hope so! LinX will be a bit of an eye opener, I suspect. By using LLC, you are defeating an Intel transient voltage protection design. No big deal if your vcore is quite low, but no real value gained either. With LLC enabled, the software forces Vcore at load to come close to vcore in bios. What is not considered here, is the transient spikes that occur when the voltage demand changes. Damage can occur if those spikes exceed safe voltage levels. By DISABLING LLC, you now have a system in which the voltage set in Bios becomes the "not to exceed" voltage for your core. This gives you control, the voltage drop and load voltage droop is designed to provide safety room for transients. What you'll find is that your Vcore as measured at 100% load may be hard to reach with LLC off. Translated, that means you could be experiencing damaging transients with LLC on.
February 10, 2010 6:34:23 PM

Okay. So I'm starting up some new testing and have already done something stupid procedure-wise: I picked up LinX and the new version of RealTemp.

First, I noticed LinX runs hotter than my previous P95 runs so I went to double check and P95 is claiming to be hotter as well under the new RealTemp (LinX about 10C hotter; P95 ~7C)

Now I've got alarms setup through the bios that I know have worked well in the past of alerting me to high CPU temps but they aren't going off when RealTemp shows values higher than what would trigger the alarms.

Is there a setting I have to flick in the new RealTemp to have accurate values? I'm tentative to run anymore grueling tests without knowing I have at least somewhat accurate temp readings.


And I'm aware of the Vtt and LLC issues.
February 10, 2010 8:40:51 PM

I don't know realtemp, I've used Coretemp, and I've just accepted the temps are right. What had you been using before to measure temp?
February 10, 2010 8:44:02 PM

A previous version of RealTemp
February 11, 2010 11:36:53 AM

Bob,

I've never tried running memory timings at anything other than Mfg recommended, even whilst overclocking.

Any advice on this? I am currently running my 860 at 3.6 ghz, I have the CAS 9 version of Gskill's Ripjaws DDR3 2000mhz. Currently I'm just using the XMP profile in BIOS, as I have yet to have set aside time for a serious overclocking session. I'd like to tighten up the timings a bit, cas8 or even 7 if possible--but it will have to be done on a time where I am not competing in a gaming league at night. Likely this Sunday I will give it a shot.

Any help appreciated!

For reference:
Win7 Pro 64
Asus Maximus III Formula
i7 860 (3.6ghz currently) with Xigmatek Dark Knight cooling
4 GB Gskill ddr3 2000mhz CAS9
ATI HD 5870
Corsair 850HX psu
February 12, 2010 4:23:26 AM

What timings does the xmp profile substitute? My timings are basically imitating the XMP profile for my ram, but I'm not actually setting xmp. I'm not a memory expert, but I did figure since I'm forced to underclock my ram, I might as well tighten the timings, and the xmp profile seemed a likely place to start.

Cheers,
Bob
February 15, 2010 12:44:49 PM

My XMP timing (I will have to double check this) are 9-9-9-27 running 2000mhz (per specs).

I will have to double check that to make sure those are all true, but that is the profile. I'd really like to see it at CAS7, but I realize it isn't that likely unless I were to forfeit a lot of speed. Will have to find the "sweet spot"...hopefully you guys can help!
!