i7 920 overclocking thread

[{core2duo}werd]

Originally Posted by Patch

I'm hitting a very hard wall.

Base clock of 220 Mhz.

Period.

With tinkering I'm getting various settings more and more stable at 219-220 Mhz bclock, but it won't go past. Even lowering the multi down to 12. Vcore of 1.45 is fine for 220X21 (turbo on) to get some benching at 4.6Ghz, but I can't even get through early post with a higher baseclock - even with 1.55 Vc. Can't even get far enough for an error screen.

I'm on water and have yet to even see my core temps higher than 80C.

Once, I got into Windows for a quick validation at 221 bclock for 4.64 Ghz.

Is ANYBODY getting a base clock higher than 220 on their 920? Even with extreme cooling?

I'm thinking 4.64 ghz will my chip's absolute limit. Though I'm not complaining

i wish my wall was at 220 lol mine is at 180...

[Rosaki]

W0w nice
Can i know how to post picture?Ty

[stargate125645]

Alright, since there is a vast amount of information scattered around everwhere, I will attempt to combine it into one post. Help me fill in the question marks and give me other voltages to add that help with and/or limit overclocking. Any other recommended temperatures are also welcome.

Max. Temperature Estimates:
- Tj_max is 100 degrees C. This temperature should not be exceeded; most try to stay below 80 degrees C for 24/7 operation.

Max. Voltage Estimates:
- CPU voltage (V_CC) should not exceed 1.55V, but most prefer to stay at around 1.3625V or lower to mitigate chip degredation.
- Memory voltage (V_DDQ) should not exceed 1.85V, and it should be within 0.5V of the uncore voltage. The default uncore voltage is 1.15V, from which we can use the 0.5V maximum differential to obtain the oft-referenced 1.65V maximum memory voltage.
- QPI (uncore) voltage (V_TTA and V_TTD) should not exceed 1.35V.
- PLL voltage (V_CCPLL) should not exceed 1.89V. Some have reported keeping it at 1.8V (default) for overclocking, while others have safely increased it to 1.88V. Some review sites have used beyond 1.9V in their overclocking experiments.

Recommended Voltages:
- ICH and IOH voltages may need to be increased to as much as 1.3V.

"Max. Temperature Estimates" should not be exceeded during operation of the chip. "Max. Voltage Estimates" are the voltages most likely to be changed for overclocking, and as a general rule should not exceed the listed estimated maximums, which are supplied by Intel documentation. "Recommended Voltages" may need to be adjusted for overclocking; however, no maximum voltage estimates have been obtained for these variables. As a general rule these voltages should not be adjusted unless necessary.

Liability Note: I take no responsibility for how this information is used. You use the information contained within this post at your own risk!

[Extreme Newbie]

Originally Posted by stargate125645

Alright, since there is a vast amount of information scattered around everwhere, I will attempt to combine it into one post. Help me fill in the question marks and give me other voltages to add that help with and/or limit overclocking. Any other recommended temperatures are also welcome.

Max. Temperature Estimates:
- Tj_max should not exceed 100 degrees C, but most try to stay below 80 degrees C for 24/7 operation.

Max. Voltage Estimates:
- Core voltage should not exceed 1.55V, but most prefer to stay at around 1.3625V or lower to mitigate chip degredation.
- Memory voltage should not exceed 1.65V austensibly, and it should be within 0.5V of the core voltage.
- QPI voltage = ?.

Recommended Voltages:
- PLL voltage can remain at 1.8V (default), but many have safely increased it to 1.88V.
- ICH/IOH voltage may need to be increased to as much as 1.3V.
- VTT = ?.

"Max. Temperature Estimates" should not be exceeded during operation of the chip. "Max. Voltage Estimates" are the voltages most likely to be changed for overclocking, and as a general rule should not exceed the listed estimated maximums. "Recommended Voltages" may need to be adjusted for overclocking; however, no maximum voltage estimates have been obtained for these variables. As a general rule these voltages should not be adjusted unless necessary.

Liability Note: I take no responsibility for how this information is used. You use the ifnormation contained within this post at your own risk!

I am happy to see that someone has decided to try and put all of this information together in one place. Rep+ to you for getting the ball rolling.
What do you think about starting a new thread with this information? I am worried that this will get lost in here.

What I can add is this:
Max QPI voltage is specified at 1.35v
Memory Voltage is 1.65v max but should be within .5v of QPI voltage not core voltage.

[stargate125645]

Originally Posted by Extreme Newbie

I am happy to see that someone has decided to try and put all of this information together in one place. Rep+ to you for getting the ball rolling.
What do you think about starting a new thread with this information? I am worried that this will get lost in here.

What I can add is this:
Max QPI voltage is specified at 1.35v
Memory Voltage is 1.65v max but should be within .5v of QPI voltage not core voltage.

I have no problem starting a new thread.

I have added the QPI information you provided (I presume that is from one of the tables you linked to before)? Any other voltages I should add? Also from that table, is the V_DDQ the memory voltage? I have it listed separately right now in my summary.

Regarding the memory voltage being within 0.5V of the core voltage, I took that from this:

While we were fearful about the memory overclockability of Core i7 CPUs we've found this not to be the case - we've hit 1,600MHz without even breaking a sweat on pre-production BIOSes and companies are already launching triple channel memory kits that are hitting 2,000MHz at a record low 1.65V. You can push the memory voltage, but Intel performance guru François Piednoel was keen to stress that the CPU voltage must be kept within a 0.5V potential difference to memory.

With CPU stock voltages at sub 1.2V this only allows an upper safe maximum of ~1.65-1.7V, however increase the CPU voltage to 1.4 or 1.5 (or more) with some extreme cooling (remember these CPUs are rated at a 130W TDP at ~1.2V) and you give yourself more breathing room to 1.9-2.0V on the memory.

Source: http://www.bit-tech.net/hardware/cpu...tecture-dive/3
Did I read it wrong? That second paragraph is what makes me believe they mean the core voltage.

[Patch]

Originally Posted by laxrunner

Anything above about 217 takes work. The highest I've ever gotten mine is right at 222, but it was in no way stable and gave BSOD instantly if I tried going to even 222.5. I've seen higher a couple times, but if I recall correctly those were at ~1.9V to the ram and 1.6V or so on the QPI and NB. Not sure what the Vcore was but those voltages are a lot more than you want to be pushing on a regular basis unless you have cash to low on new parts all the time.

Hmm...gives me food for thought.

At Bclock 220 and turbo on/HT off, I can get into Windows and surf Firefox (but nothing CPU intensive). That's using Vc 1.45, VTT +220, PLL 1.87, 1.65 to ram, IOH 1.3. Since my 1600 mushkins tend to give me constant C1 beeps (they ring in my head now) when clocked over 1730 despite lots of power and loose timings, for higher CPU clocks I run the RAM at a lower multi (end up in low 1300's mhz) with tighter timings of 6-6-6-16.

Sooo....when trying to get bclock to 222 I've stepwise adjusted parameters individually, in different combinations, and all of them together up to the following levels:
Vc 1.55
VTT +320
PLL 1.95
IOH 1.34
RAM 1.66

I've upped the SB volts some and upped the IOH subcategories
I've adjusted CPU and IOH skew in different combinations up to 200ps.

Taken Turbo on and off
Tried different combinations of PPM, EIST, CXE on and off (keeping CXE off helps all my higher clock combos)
Tried disabling QPI fast link

With 222 bclock I still can't even get to a beep or a blue screen to direct me to a problem to address. I just get a brief power on/shutdown and an auto reboot in safe mode with a generic overclock recovery screen (that doesn't even have the decency to be blue).

I haven't tried the massive voltages to NB and RAM you mention folks used. RAM because it's underclocked from spec anyway, and NB because it doesn't make sense to need such a huge jump for another 1 mzh clock.

But, my NB is on water so maybe I'll try cranking up the IOH a lot more.

While I don't want my chip to fry under my nose, my longevity expectations are limited to the next significant processor release.

[Extreme Newbie]

Originally Posted by stargate125645

I have no problem starting a new thread.

I have added the QPI information you provided (I presume that is from one of the tables you linked to before)? Any other voltages I should add? Also from that table, is the V_DDQ the memory voltage? I have it listed separately right now in my summary.

Regarding the memory voltage being within 0.5V of the core voltage, I took that from this:

Source: http://www.bit-tech.net/hardware/cpu...tecture-dive/3
Did I read it wrong? That second paragraph is what makes me believe they mean the core voltage.

It looks like you read that review correct as they do state its .5 from CPU voltage. However, this review states that is is .5 from QPI voltage.
http://i4memory.com/wp/article/327
I will do some more searching and see what I can find.

[stargate125645]

Originally Posted by Extreme Newbie

It looks like you read that review correct as they do state its .5 from CPU voltage. However, this review states that is is .5 from QPI voltage.
http://i4memory.com/wp/article/327
I will do some more searching and see what I can find.

They reference the same report as I, and suggest it is the uncore voltage.

DRAM Voltage - 1.66V: This is the closest to the 1.65V the Corsair Dominator DIMMs wanted and it’s within the 0.5V Uncore difference.

I take it uncore voltage is the QPI voltage? I have adjusted my summary accordingly.

Also, what is the V_DDQ from the Intel manual? I don't see it referenced in the BIOS or in any overclocking guides.

[nebuchanezzar]

Originally Posted by Patch

Hmm...gives me food for thought.

At Bclock 220 and turbo on/HT off, I can get into Windows and surf Firefox (but nothing CPU intensive). That's using Vc 1.45, VTT +220, PLL 1.87, 1.65 to ram, IOH 1.3. Since my 1600 mushkins tend to give me constant C1 beeps (they ring in my head now) when clocked over 1730 despite lots of power and loose timings, for higher CPU clocks I run the RAM at a lower multi (end up in low 1300's mhz) with tighter timings of 6-6-6-16.

Sooo....when trying to get bclock to 222 I've stepwise adjusted parameters individually, in different combinations, and all of them together up to the following levels:
Vc 1.55
VTT +320
PLL 1.95
IOH 1.34
RAM 1.66

I've upped the SB volts some and upped the IOH subcategories
I've adjusted CPU and IOH skew in different combinations up to 200ps.

Taken Turbo on and off
Tried different combinations of PPM, EIST, CXE on and off (keeping CXE off helps all my higher clock combos)
Tried disabling QPI fast link

With 222 bclock I still can't even get to a beep or a blue screen to direct me to a problem to address. I just get a brief power on/shutdown and an auto reboot in safe mode with a generic overclock recovery screen (that doesn't even have the decency to be blue).

I haven't tried the massive voltages to NB and RAM you mention folks used. RAM because it's underclocked from spec anyway, and NB because it doesn't make sense to need such a huge jump for another 1 mzh clock.

But, my NB is on water so maybe I'll try cranking up the IOH a lot more.

While I don't want my chip to fry under my nose, my longevity expectations are limited to the next significant processor release.

Correct me if I am wrong but, at +.320 with a default vtt of 1.20 you would be at @ 1.52 vtt voltage. That would somewhat make sense with such a high BCLK but that is a bit over 1.35. If your default is 1.10v then it would be about 1.42 which is back within my personal comfort zone. Can you check that with a multimeter or even in your bios? I'm curious what it is actually at.

220BCLK is actually quite impressive no matter what multis you are running. AFAIK the people who post higher BCLKs are not booting with them. They boot at a lower clock and then raise it with things like SetFSB or some overclocking utility as they cannot boot with such a high setting. They dial in just enough voltage to complete a certain benchmark and then hope it doesn't crash before it finishes and they get a screenie .

Good luck on working past 220 but really, anything over 200 is great and anything over 210 is stupendous.


@stargate125645
I can offer some links for things like the pdf for the i7 and quite a few good articles and reviews/overclocks that I have been reading if you want to consolidate some info.
Also, the .5 differential is vtt to vdimm specifically set by Intel. I have come across many users that also state a relationship between vtt voltage and vcore voltage recommending to keep vtt at least within .1v of vC and many recommending even closer but not over. That does somewhat conflict with tests by those pushing their memory to the 2000 speeds who often have vtt much higher than vC so take that as you will. I personally feel these chips are pretty darn tough and will take a lot more abuse than people think but that's just an opinion, they haven't been around long enough to extrapolate any true long-term results/effects yet.


edit
v_ddq is vdimm voltage. According to Intels own whitepapers(found here) page 22 lists recommended voltages. It lists v_ddq(VDimm) at maximum of 1.85. This is if you raise your vtt to 1.35 however as you must keep vtt ~ vD within .5v difference and no greater. However, like I mentioned, many OCers have pushed vtt well above 1.5v in order to run memory at voltages over 2.0v vDimm. For some good technical information(besides the whitepaper pdf I listed already) check out Lost Circuits articles on i7 overall and another focusing on the power draw/voltage/actual tdp found here.
Hope that helps a bit.

one last edit, if you check the forum at i4memory you will see Eva2000's results with the big brother to my mobo running very high ram speeds with quite a bump to vtt and vdimm voltages. He also includes some great general info on i7 overclocking. A recommended read even if you are not using a DFI X58 mobo.

[stargate125645]

Originally Posted by nebuchanezzar

Correct me if I am wrong but, at +.320 with a default vtt of 1.20 you would be at @ 1.52 vtt voltage. That would somewhat make sense with such a high BCLK but that is a bit over 1.35. If your default is 1.10v then it would be about 1.42 which is back within my personal comfort zone. Can you check that with a multimeter or even in your bios? I'm curious what it is actually at.

220BCLK is actually quite impressive no matter what multis you are running. AFAIK the people who post higher BCLKs are not booting with them. They boot at a lower clock and then raise it with things like SetFSB or some overclocking utility as they cannot boot with such a high setting. They dial in just enough voltage to complete a certain benchmark and then hope it doesn't crash before it finishes and they get a screenie .

Good luck on working past 220 but really, anything over 200 is great and anything over 210 is stupendous.


@stargate125645
I can offer some links for things like the pdf for the i7 and quite a few good articles and reviews/overclocks that I have been reading if you want to consolidate some info.
Also, the .5 differential is vtt to vdimm specifically set by Intel. I have come across many users that also state a relationship between vtt voltage and vcore voltage recommending to keep vtt at least within .1v of vC and many recommending even closer but not over. That does somewhat conflict with tests by those pushing their memory to the 2000 speeds who often have vtt much higher than vC so take that as you will. I personally feel these chips are pretty darn tough and will take a lot more abuse than people think but that's just an opinion, they haven't been around long enough to extrapolate any true long-term results/effects yet.


edit
v_ddq is vdimm voltage. According to Intels own whitepapers(found here) page 22 lists recommended voltages. It lists v_ddq(VDimm) at maximum of 1.85. This is if you raise your vtt to 1.35 however as you must keep vtt ~ vD within .5v difference and no greater. However, like I mentioned, many OCers have pushed vtt well above 1.5v in order to run memory at voltages over 2.0v vDimm. For some good technical information(besides the whitepaper pdf I listed already) check out Lost Circuits articles on i7 overall and another focusing on the power draw/voltage/actual tdp found here.
Hope that helps a bit.

one last edit, if you check the forum at i4memory you will see Eva2000's results with the big brother to my mobo running very high ram speeds with quite a bump to vtt and vdimm voltages. He also includes some great general info on i7 overclocking. A recommended read even if you are not using a DFI X58 mobo.

Very nice! I'll update the information today when I'm done with work. So the V_DDQ max of 1.85V listed by Intel assumes that the QPI/uncore voltage has been increased to the max of 1.35V?