Sign in with
Sign up | Sign in
Your question

Core i7 920 D0 OC w ASUS P6T Deluxe V2. BIOS examples - comments?

Last response: in Overclocking
Share
June 21, 2009 11:49:23 PM

All,

I am getting close to being satisfied with my system. I never OC-ed before and this was something I wanted to try, so I made sure I got the D0 to have the newer revision. I've been reading for the past 2 days on what has worked for people and wanted to share my experiences. Please comment on any of my findings if you believe I'm incorrect. I'm a noob at OC-ing. I have to say, I should have started small and worked my way up, but I went in reverse order, just to see if I could get it stable at 4Ghz first.

Before I go into the OC details, I want to add that I have a TRUE 120 Extreme RT1366 with 2 Thermalright 1600 RPM fans. I also have the following memory (6GB G.Skill DDR3-1600 PC3-12800 CL 8-8-8-21)

http://www.newegg.com/Product/Product.aspx?Item=N82E16820231247

When I initially built the system, I had to set the timing of this memory properly.



In the ASUS P6T BIOS, when you set it to DDR3-1600 it is important to know that the BIOS will set the voltage to 1.8V (thanks to TurboV for showing this) if you leave the DRAM BUS Voltage to [Auto]. You have to manually set this to 1.60 to avoid running the memory at 1.8V since Intel warns about going past 1.6V. In the images in my 3 examples below, you will find how I manually forced the memory to 1.6V while retaining the 1600Mhz setting. Memory posts fine at 1600 with the adjusted voltage.


First Attempt:
Ai Overclock Tuner: Manual
CPU Ratio: 20
Intel Speedstep Tech: Disabled
BCLK: 200 (200 x 20 would give me 4 Ghz)
DRAM Frequency: DDR3 - 1603Mhz (Once you change the CPU Ratio and BCLK, you have to manually tone this back to the 1600 range)
CPU Voltage: 1.33125 (When left at stock, system would post but prime would BSOD after 5 minutes. Increased this and just matched the settings for QPI/DRAM Core Voltage)
CPU PPL Voltage: 1.80
QPI/DRAM Core Voltage: 1.33125
DRAM BUS Voltage: 1.60




I ran Prime95's small FFT test to check temps and was not happy with the results and stopped it after an hour. While the CPU idled around 40C to 42, temps climbed into the 80s after about 50 minutes.





Please comment on these temps. I had them in the low 70s for the first 30 minutes. I then put the side panels on and left all case fans at the speeds I have them set at, which is about 50-60%. Cranking all fans to 100% seemed to not make a huge differences besides noise. What about these voltages? Would this reduce the life of my CPU given I usually run my PC 24/7?


Second Attempt:
No screen prints for the second attempt. I just changed
CPU Voltage: 1.27500 (It was not stable at 1.22500, so I picked this as it was somewhat in the middle)
QPI/DRAM Voltage: 1.27500 (Lowered this as well as I read this one can affect temps more compared to CPU Voltage. Correct??)
All other settings same as First Attempt

I ran Prime95 for about the same time with temps maxing out around 76. So overall about a 10 degree difference in a similar timeframe, and with no errors. So from this I can only conclude that the lower voltage helped drop the temps at load. At idle, the temps were 39C to 41C so no significant differences. I realize an hour of Prime95 is not enough to provide details on stability but I'm just comparing temps at different settings. I also want to get some comments here on settings I used as I don't want to harm the CPU long term.


Third Attempt:
For my final attempt, for now, I just wanted to see what I could do at stock voltages. I had seen many comments on this site and others mentioning that just bumping the BCLK to 160 would be all that was needed. I want to add that with QPI/DRAM core Voltage set to Auto in the BIOS, that TurboV was showing this as 1.37500 while I read this should really stay under 1.35. For that reason, I set it manually.
Ai Overclock Tuner: Manual
CPU Ratio Setting: Auto
Intel Speedstep Tech: Enabled
Intel Turbo Mode Tech: Enabled (to get a multiplier of 21 if needed)
BCLK Frequency: 160
DRAM Frequency: DDR3-1604Mhz (Had to manually set this again after setting BCLK)
CPU Voltage: 1.22500 (I set it manually to ensure this value would be used)
CPU PPL Voltage: 1.8 (I set it manually to ensure this value would be used)
QPI/DRAM Core Voltage: 1.22500 (I set it manually to ensure this value would be used)
DRAM Bus Voltage: 1.60




Running Prime95 again for an hour showed me temperatures nearly the same compared to stock settings. Temps never went past 68C. Idle temps are 35-39. I', assuming this is nearly the same compared to stock speeds due to no voltage changes. what's nice though is the increase in speed with virtually no additional changes.





Overall, I'd like to get comments about these temps and similar setups. Are these temps fine? How about the voltages and impact on CPU life? Is the TRUE 120 giving me enough cooling? I believe I installed it as needed with enough AS5 paste. Let me know.

More about : core 920 asus p6t deluxe bios examples comments

June 21, 2009 11:57:02 PM

What ver. of Prime95 are you using? IIRC the newest one supports multi core.
June 21, 2009 11:59:53 PM

AKM880 said:
What ver. of Prime95 are you using? IIRC the newest one supports multi core.


25.9
Related resources
June 22, 2009 12:01:25 AM

Try running one test, I'm interesting to see if it stresses all the cores/threads.
June 22, 2009 12:07:33 AM

AKM880 said:
Try running one test, I'm interesting to see if it stresses all the cores/threads.


Can you be a bit more specific? As I said, newb to this. I just figured that since the temps on all cores went up in testing (see images) that I was testing all cores.
June 22, 2009 9:56:05 PM

So my idles temps of around 40, is that fine at stock voltages running at 3.2 Ghz? Do I need to re-mount the true with more thermal paste? How about the load temps of upper 60s at this voltage/speed?
June 23, 2009 1:34:46 AM

so no concerns either about core 1 being 2 - 3 degrees cooler compared to 0, 2 and 3 at idle? It may just be that the AS5 paste did not spread evenly.
June 23, 2009 4:35:55 AM

As much as it annoys me your individual core temps wont be the same, a few degrees different is expected. 3 of my cores are normally the same and one is always 2-3 degrees warmer. In process of just doing mundane things on your PC the core loads wont be even and thus more active core should be warmer.

This is what I have been told anyway, as I too thought it could be my AS5 application not being even.

AKM880 - Im not sure what you mean either, in the OPs pics you can see the 8 workers running in prime95 so will be loading on all cores. Also if only one was being maxed the temp of 1 of them would be noticeably warmer.
June 23, 2009 6:32:22 AM

D3LTA09 said:
As much as it annoys me your individual core temps wont be the same, a few degrees different is expected. 3 of my cores are normally the same and one is always 2-3 degrees warmer. In process of just doing mundane things on your PC the core loads wont be even and thus more active core should be warmer.

This is what I have been told anyway, as I too thought it could be my AS5 application not being even.

AKM880 - Im not sure what you mean either, in the OPs pics you can see the 8 workers running in prime95 so will be loading on all cores. Also if only one was being maxed the temp of 1 of them would be noticeably warmer.


Thanks, that makes me feel better about my TRUE 120 Installation. Would hate to re-do it but would if I had to. What about the temps and voltages during my 3 tests. Any comments?
June 23, 2009 8:58:42 AM

Your voltages and related temps seem quite normal, I got about a 10c drop in max temp when I went from 1.30 back to stock 1.20vcore at 3.5ghz as at that clock you can run stock voltages as you have already found out. Its good you still kept monitoring your voltages when set to auto as this can be a key problem people have. I set nearly every option manually (most to stock settings) just to avoid suprises like you have encountered, like qpi voltage increases.

In terms of safe temps for 24/7 as already mentioned under 70 is a good target, I have gone from 3.5 to 3.8 to 4ghz but found 4 just too hot and a bit unstable so atm im at 3.8 with vcore at 1.30 and matching qpi, cpu pll at 1.88. I also have dropped my ram multi down (and as its linked uncore multi too) so even though my ram is capable of 1600mhz (actually 1800 if raise my timmings and command rate) Im only at 1200ish atm with lower timmings and command rate.
My temps are are in the low to mid 60's at max load after complete thermal saturation.
But I think I will got back to 3.5 with stock voltages as no need to generate extra heat and stress on my chip for no performance difference.
a b à CPUs
a b K Overclocking
June 23, 2009 3:30:00 PM

Note: For the i7 920 keep temps below 68C. NOT 70C.

See:
CompuTronix said:
The Tjunction Max specification of 100c is for throttle and shutdown overtemp protection ONLY. The limiting thermal specification, which is the value shown in Intel's Processor Spec Finder, is Tcase Max (CPU temperature), NOT Tjunction Max (Core temperature), which is a common misconception among many users.

Core i7 specs:

Vcore Max 1.375
Tcase Max (CPU trmp) 68c
Tjunction (Core temp) 73c

For more details, click on the link in my signature.

Comp :sol: 

http://www.tomshardware.com/forum/252869-29-long-settin...
June 23, 2009 3:42:16 PM

Hello guys,

I do not mean to hijack this thread, but i have one question for you guys.

I have 920 @ 3.8. cpu vcore is 1.33 but qpi is 1.220. it is perfectly stable. but i see people set both vcore and qpi the same. should i keep vcore and qpi the same?

Please advise me.

Thanks.
June 23, 2009 5:33:25 PM

In responce to the earllier question I am running @ all voltages accept memory on auto which runs my vcore@ 1.07v and 37watts on the low end and 1.33v and 147watts @ 100% prime testing. and memory is @1.6v stable even with timings at low 6-6-5-18 and overclocked to 1400Mhz!

June 23, 2009 11:14:24 PM

htoonthura said:
Hello guys,

I do not mean to hijack this thread, but i have one question for you guys.

I have 920 @ 3.8. cpu vcore is 1.33 but qpi is 1.220. it is perfectly stable. but i see people set both vcore and qpi the same. should i keep vcore and qpi the same?

Please advise me.

Thanks.


I have mine the same but it comes down to stability, you want the lowest voltages for the highest frequency so if you are stable at that id be inclined to leave it. I have read threads where people were talking about quite large OC's of i7 who seemed to think that if you kept these 2 voltages within a certain degree of difference between each other it was better for your chip and OC but none of them had any real factual basis to prove these claims. However im definitely not a expert OCer at all so Im sure someone will know more about this?
June 24, 2009 6:03:22 AM

im running core i7 920 at 3.0 ghz and CPU Volts 1.0... DRAM BUS Voltage im running it at 1.64 and QPI/DRAM at 1.35 maybe i need to lower down the QPI/DRAM but i dont know to what lol and Timings i dont know much about that.. they are currently at 9-9-9-24 maybe someone could help me with that.
June 26, 2009 11:22:31 PM

I guess my main concerns are temperatures and voltages and impact on the life of the processor. My 4.00 Ghz at 1.275 seemed stable and went through an hour of Prime with no issues but I am not sure if I liked the temps in the mid 70s, especially after seeing comments here to keep it under 70, or even 68.

A voltage increase to 1.275 seems small to me so I'm just curious as to whether this affects the life of the CPU at all. Or is the temperature going to be the issue more so compared to the voltage?

At the same time, with the TRUE 120, with 2 fans around 1600RPM, my idle temps are the same for stock speed compared to my current 3.2Ghz, which are around 35-39C. That seems normal from what I see on this forum and other forums. However, is that good for the TRUE with the push/pull setup? I have my AC set to 79F, which is 26C.

Should the TRUE do better at load in my 4.0 Ghz test at 1.275 voltages? I installed it as recommended using a line of A5 paste and I watched where the yellow triangle was to get the direction correct. I just see others posting lower temps with higher voltages on several forums so I'm just curious at to if I applied enough AS5.
June 27, 2009 8:17:28 AM

Sounds about right in terms of idle and load temps at those ambients, TRUE's have a notoriously uneven surface from what I have read, alot of people lap them to get the most performance out of them. Might be something to consider, personally its something I wouldnt be bothered to do and the reason I have never bought one, if Im outlaying good coin on a heatsink I expect it to have a nice surface.
June 27, 2009 3:47:05 PM

So is it odd for one core to always be 3-4 degrees cooler? I have been wondering if I should remount the TRUE just to see if that would resolve that. Not sure if it's even an issue.
a b à CPUs
a b K Overclocking
June 28, 2009 2:30:48 AM

cjl said:
That's Tcase though, wheras core temps will be 5C higher in most cases. So, you can have core temps up to 73C by that spec.

Yeah your right. I stand corrected. Should have bolded the Core temp (73C) and should have said to stay below that.
June 29, 2009 1:16:16 AM

I took the TRUE off today and it looks like I did the AS5 just fine. I did not remove it due to one core being different but because I wanted to try the washer mod to get it on more tight. I used a washer about 2 mm thick.
I have it back on now and at idle the difference is maybe 1 degree. At load, I see only a 3-4 degree difference. Is that due to the AS5 still needing to cure?

Anybody see a problem with using a washer and impact on the MB and/or Chip due to additional pressure? My TRUE and CPU are not lapped.
July 1, 2009 9:07:21 PM

Why dont you buy IC Diamond 7 Carat thermal paste it lowered down my temps. and im running at 3.3ghz on load it does not go past 58-60 also i changed my heatsink to True 120 didn't lap it or anything but the weird thing is that with my dark knight running at 3.0 i was at idle temps of 39 and load temps of like 70+ using AS5 and then when i switched to True 120 and IC Diamond and my idle temps went to like 40-45 which is wierd but on load temps thats were i see the difference that it doesnt go past 59-60 running prime.. do idle temps matter??
July 5, 2009 1:47:51 AM

I'm going to stick with the AS5 for now. Just wanted to post some more details on the differences with the washer added.

At 4.00 Ghz (BCLK: 200, CPU Ratio 20), using stock voltages is not possible. Pushing them to 1.27500 for CPU Voltage and QPI/DRAM Voltage got me to about 75 in CPU-ID Hardware monitor, 78 in Real Temp 3.00 after running for about an hour. Seemed stable. Changing both these voltages to 1.25625 provides about 2 degrees difference and it had not crashed in 45 mins. Not sure how CPU-ID works versus Real Temp (not calibrated) but the difference always seems to be about 2 degrees.

I also did some testing with setting BCLK to 190 and CPU Ratio to Auto with Turbo enabled to get 3.8 standard and about 4.0 when turbo kicks in. Obviously, temps are the same as the last test at 4.00.

Setting the BCLK to 180 and CPU Ratio back to 20, gives me max temps of 71 in CPU-ID and 74 in Real Temp.

With my PC being on 24x7, I'm just curious how people feel about these temps. The computer would never be running at 100% for a long time so hitting the mid 70 range would be rare at best. With a BCLK of 190, with speedstep and turbo enabled, the processor would run at about 2.28 when not busy (190 x 12), and would be in the 3.80-3.99 (190 x 20 - 190 x 21) range when busy with temp reaching the mid 70s.
Otherwise, using a BCLK of 180 may be safer to stay closer to the 70 range.

Just looking for some opinions.
July 5, 2009 9:00:03 AM

Here is my latest settings with voltage @ 1.38 Vcore and 1.64v memory everything else on auto stable for 12 hours of benchmarking with temps from 40c-60c all day!

and here is my 3dmark06 score at same settings in windows xp sp3
http://service.futuremark.com/compare?3dm06=11380133
July 6, 2009 2:57:14 AM

Pointdexter5
for the 4ghz overclock at 20X200 or 21X191

increase PLL voltage to 1.88, and you may be able to lower vcore and vtt (QPI/DRAM on Asus) once you do this - but one thing at a time. If this helps stability and you push further then IOH needs to be at 1.2 instead of auto
July 6, 2009 6:51:25 AM

seversphere said:
Pointdexter5
for the 4ghz overclock at 20X200 or 21X191

increase PLL voltage to 1.88, and you may be able to lower vcore and vtt (QPI/DRAM on Asus) once you do this - but one thing at a time. If this helps stability and you push further then IOH needs to be at 1.2 instead of auto


The 20x200 seems to be working well with the current settings, one hour run in p95 with no crashed. The main issue for me is temperature and knowing where to stop. If my 4Ghz reaches 75-76 after an hour, I know it's just a few degrees above desired so just looking for opinions on this. I can stay under 72 at 3.6 - 3.8 so maybe pushing it to 4 may not be necesarry.
November 1, 2009 11:19:38 AM

Just a note on thermal compound. It's really important to not put too much on! In a perfect world, with both the top of the proc and bottom of heat sink milled perfectly, compound would insulate. Metal to metal is the perfect conductor. In the real world, microscopic imperfections prevent perfect contact. A VERY thin layer of compuond (which is essentially metal paste) would fill these gaps. When I put mine on, I use a creditcard to scrape it almost entirely off. Give that a try and let me know if your temps drop. Cheers from Canada.
November 1, 2009 1:51:08 PM

Rotor888 said:
Just a note on thermal compound. It's really important to not put too much on! In a perfect world, with both the top of the proc and bottom of heat sink milled perfectly, compound would insulate. Metal to metal is the perfect conductor. In the real world, microscopic imperfections prevent perfect contact. A VERY thin layer of compuond (which is essentially metal paste) would fill these gaps. When I put mine on, I use a creditcard to scrape it almost entirely off. Give that a try and let me know if your temps drop. Cheers from Canada.


Although you are right about a thin layer of paste and i like the credit card application better than the bag on finger crap i see people use it gives you a nice even thin spread using a card. He does not need to reseat his heatsink he needs to keep testing his cpu and system will eventually reveal to him its comfortable max overclock. A 4.0 is great but its like wearing a diamond grill and walking around smiling at everyone. its just for bragging not really necessary or even helpfull in any real applications other than random crashes when you truly load your system. i would stay away from those 80's man give yourself a 3.5 and get your memory tweeked and stop listening to these guys brag about 4.0 they r sad and lonely and require that you envy them and the unstable 4.0 overclock that they run. :hello:  :hello: 
November 1, 2009 1:53:53 PM

OH ! .... " AND I FARTED" (comp is sad)
November 2, 2009 3:38:54 AM

FREERAVEN said:
Although you are right about a thin layer of paste and i like the credit card application better than the bag on finger crap i see people use it gives you a nice even thin spread using a card. He does not need to reseat his heatsink he needs to keep testing his cpu and system will eventually reveal to him its comfortable max overclock. A 4.0 is great but its like wearing a diamond grill and walking around smiling at everyone. its just for bragging not really necessary or even helpfull in any real applications other than random crashes when you truly load your system. i would stay away from those 80's man give yourself a 3.5 and get your memory tweeked and stop listening to these guys brag about 4.0 they r sad and lonely and require that you envy them and the unstable 4.0 overclock that they run. :hello:  :hello: 


I agree. Sounded like there might have been too much compound. Just an idea. I have a very similar rig with the exception of dual saphire 4890hd's and a corsair liquid cooler on the CPU. I'll be following the posts with interest. I'll be happy at the best settings for 3.2 or 3.5
a b à CPUs
a b K Overclocking
November 4, 2009 7:41:10 AM

Shadow703793 said:
Yeah your right. I stand corrected. Should have bolded the Core temp (73C) and should have said to stay below that.


θja=θjc+θch+θha

θja: junction-to-ambient
θjc: junction-to-case
θch: case-to-heatsink
θha: heatsink-to-ambient

Due to θha, I think you can have a few degrees more.

EDIT: NVM, I am wrong as the specified temp is case temp instead of heatsink temp.
November 6, 2009 4:22:32 AM

Quote:
In the ASUS P6T BIOS, when you set it to DDR3-1600 it is important to know that the BIOS will set the voltage to 1.8V (thanks to TurboV for showing this) if you leave the DRAM BUS Voltage to [Auto]. You have to manually set this to 1.60 to avoid running the memory at 1.8V since Intel warns about going past 1.6V. In the images in my 3 examples below, you will find how I manually forced the memory to 1.6V while retaining the 1600Mhz setting. Memory posts fine at 1600 with the adjusted voltage.


I'm using 6x2gig CORSAIR DOMINATOR ram cards with the proper timing and speed (1600MHz 8-8-8-24) but when I lower the voltage to 1.60-1.66V I cant get windows to boot up properly what could I be doing wrong?

http://www.newegg.com/Product/Product.aspx?Item=N82E168...

Aside from the RAM timing and speed I have set all the other BIOS settings to auto.

!