Overclock.net › Forums › Intel › Intel CPUs › Haswell Overclocking Guide [With Statistics]
New Posts  All Forums:Forum Nav:

Haswell Overclocking Guide [With Statistics] - Page 591

post #5901 of 19540
Quote:
Originally Posted by creos7 View Post

ouch frown.gif i'm sorry, that is a bit on the poor-peformer side ya. I need 1.39 to hit 4.7, i'm broken-in solid 4.6 @1.28.
This is with your RAM and Uncore/CacheRatio/Ring at stock (i.e. manually set to x34)? Have you tried lowering RAM to 1333? (or 1300 whatever that standard is) to see if it allows for lower vcore?


hmmmm actually it was with it left to auto (the ring ratio clock and voltage) which I believe has 3.8 (the max turbo clock speed default) as the setting in gray, not x34. I could try manually setting the ring ratio multiplyer to x34... I assume then I would leave it's voltage on override mode @ 1.2V? Instead of auto where I had it. Then re-attempt to find my 4.4GHz lowest voltage...I wonder if I should also just toss the input voltage up to 2.1V to start as well.

The memory is set to 1600 & 1.5V which is the standard clock (1333MHz is the default but the memory is 1600 not 1333) with XMP off. These settings were memtest86+ verified for 3 passes so I feel confident enough memory wise.

so
1) Set ring ratio to 3.4Ghz Manual @ 1.2V
2)Set input volatage to 2.1V
3)Set memory to standard 1600MHz (not the mobos default, but this is the rated clock speed on the cards)
4)Find the stable lowest voltage for 4.4GHz clock, if low enough voltage, maybe push to 4.5, or 4.6.
5)Push ring/cache voltage up to the clock voltage
6))Find closest to 1:1 cache clock possible that is stable
7)begin lowering input voltage, checking for stability
8)begin lowering ring voltage, checking for stability.


Sound right?
post #5902 of 19540
Quote:
Originally Posted by KillerDJK View Post

hmmmm actually it was with it left to auto (the ring ratio clock and voltage) which I believe has 3.8 (the max turbo clock speed default) as the setting in gray, not x34. I could try manually setting the ring ratio multiplyer to x34... I assume then I would leave it's voltage on override mode @ 1.2V? Instead of auto where I had it. Then re-attempt to find my 4.4GHz lowest voltage...I wonder if I should also just toss the input voltage up to 2.1V to start as well.

The memory is set to 1600 & 1.5V which is the standard clock (1333MHz is the default but the memory is 1600 not 1333) with XMP off. These settings were memtest86+ verified for 3 passes so I feel confident enough memory wise.

so
1) Set ring ratio to 3.4Ghz Manual @ 1.2V
2)Set input volatage to 2.1V
3)Set memory to standard 1600MHz (not the mobos default, but this is the rated clock speed on the cards)
4)Find the stable lowest voltage for 4.4GHz clock, if low enough voltage, maybe push to 4.5, or 4.6.
5)Push ring/cache voltage up to the clock voltage
6))Find closest to 1:1 cache clock possible that is stable
7)begin lowering input voltage, checking for stability
8)begin lowering ring voltage, checking for stability.


Sound right?

1) Yes. Even though for some the default is 3.5Ghz, in your case I would try 3.4Ghz just for extra room. 1.2V for that low level is probably an overkill, but still better to rule that out uncore as a souce of instaiblity, so I agree - keep it at 1.2V for the initial testing.
2) IMHO this is way too high. I would say 1.9V should be plenty, maybe 1.95V, but no more than that. Why did you choose 2.1V?
3) Sometimes when a CPU is OCed it struggles to deal with higher speed memory even if it is rated. It may potentially need more voltage (not just the memory but the vcore itself). Chances are small but given that you seem to have a poor OC-er, I would recommend setting your memory to whatever the MoBo considers the default. If that is 1333Mhz, I would set it that way. It may (possibly) allow you lower stable voltages for your vcore. What is your mobo btw?
4, 5,... etc.... yes. That is basically following Darkwizzie's guide.

Please take what I said, particularly on (3) with a grain of salt. It is just what I would do.
post #5903 of 19540
Quote:
Originally Posted by creos7 View Post

1) Yes. Even though for some the default is 3.5Ghz, in your case I would try 3.4Ghz just for extra room. 1.2V for that low level is probably an overkill, but still better to rule that out uncore as a souce of instaiblity, so I agree - keep it at 1.2V for the initial testing.
2) IMHO this is way too high. I would say 1.9V should be plenty, maybe 1.95V, but no more than that. Why did you choose 2.1V?
3) Sometimes when a CPU is OCed it struggles to deal with higher speed memory even if it is rated. It may potentially need more voltage (not just the memory but the vcore itself). Chances are small but given that you seem to have a poor OC-er, I would recommend setting your memory to whatever the MoBo considers the default. If that is 1333Mhz, I would set it that way. It may (possibly) allow you lower stable voltages for your vcore. What is your mobo btw?
4, 5,... etc.... yes. That is basically following Darkwizzie's guide.

Please take what I said, particularly on (3) with a grain of salt. It is just what I would do.

I was just throwing 2.1V out there since I believe I've read that as the max for input voltage, the standard is 1.9V so I figure once I start to push the ring/cache clock up, 1.9V is just not going to work, and that starting is high (not necessarily 2.1V) might be easier. Especially since I have yet to enter a scenario where temperature, not my stability is the issue. In my 3 hours (possibly more if I let it run) XTU stable OC setup, I had it at 2V input memory, however with changes to memory etc. that may not be needed.

(Does a higher input voltage help in finding the lowest stable vcore voltage with the desired clock?)

I have a MSI-Z87-G45 motherboard (BIOS flashed to be the most up to date, all drivers). Think the 1333MHz hit from down from 1600Mhz memory clock is worth an extra 100MHz clock speed or a .5V lower core voltage? (I figure if it happens to help more than that It would indeed be worth it....I'm just trying to ask harder question smile.gif )

Actually though my h100i corsair link software is having issue, so I may have to halt any OC attempts until corsair properly supports W8.1 ....but that's another issue entirely. My fans were set to max during any previous OC attempts.
Edited by KillerDJK - 11/25/13 at 12:36pm
post #5904 of 19540
Quote:
Originally Posted by KillerDJK View Post


so
1) Set ring ratio to 3.4Ghz Manual @ 1.2V
2)Set input volatage to 2.1V
3)Set memory to standard 1600MHz (not the mobos default, but this is the rated clock speed on the cards)
4)Find the stable lowest voltage for 4.4GHz clock, if low enough voltage, maybe push to 4.5, or 4.6.
5)Push ring/cache voltage up to the clock voltage
6))Find closest to 1:1 cache clock possible that is stable
7)begin lowering input voltage, checking for stability
8)begin lowering ring voltage, checking for stability.


Sound right?

Not entirely.

1)1.2v is probably the very maximum you would need for x34.
2)Start low, not high. Better to increase to find stability than decrease to find instability. Start off at 1.8v and just gradually increase it in small increments. Keep core constant, adjust vrin within a 'window'. If still unstable by the time you hit the maximum you've set that window at drop it back down, increase vcore slightly then try again. Keep doing this and you should find minimum vrin/core for that multi.
3)1600 is probably a non issue but some of us (myself included) drop it down to 1333mhz running at the kit's recommended voltage. XMP off, timings on auto.
4)It's a good idea to start from the beginning, i.e, low multi's and work your way up. The main benefit is you end up with a good idea of how the chip scales, i.e, it'll give you a rough idea of what vcore the next multi up will need. Lower OC's are also a lot easier to achieve and in doing so you get good experience OC'ing that particular chip and get a 'feel' for what it's like.
5)More uncore is good but there's not much point spending ages trying to raise it. See DW's write up in the guide for some more on this.
6)Again, have a read through the guide.
7)As mentioned above, better to increase rather than decrease. It takes longer to decrease to find instability and more importantly you'll be running more voltage than required, especially with vrin which is constant.
8)Same as 7.

I gather most of this is from the Linus guide?
Edited by Doug2507 - 11/25/13 at 1:09pm
post #5905 of 19540
Quote:
Originally Posted by Doug2507 View Post

Not entirely.

1)1.2v is probably more the very maximum you would need for x34.
2)Start low, not high. Better to increase to find stability than decrease to find instability. Start off at 1.8v and just gradually increase it in small increments. Keep core constant, adjust vrin within a 'window'. If still unstable by the time you hit the maximum you've set that window at drop it back down, increase vcore slightly then try again. Keep doing this and you should find minimum vrin/core for that multi.
3)1600 is probably a non issue but some of us (myself included) drop it down to 1333mhz running at the kit's recommended voltage. XMP off, timings on auto.
4)It's a good idea to start from the beginning, i.e, low multi's and work your way up. The main benefit is you end up with a good idea of how the chip scales, i.e, it'll give you a rough idea of what vcore the next multi up will need. Lower OC's are also a lot easier to achieve and in doing so you get good experience OC'ing that particular chip and get a 'feel' for what it's like.
5)More uncore is good but there's not much point spending ages trying to raise it. See DW's write up in the guide for some more on this.
6)Again, have a read through the guide.
7)As mentioned above, better to increase rather than decrease. It takes longer to decrease to find instability and more importantly you'll be running more voltage than required, especially with vrin which is constant.
8)Same as 7.

I gather most of this is from the Linus guide?

Somewhat, I mean, a lot of my initial searching around took me by his haswell youtube video, which was an okay view I guess, but now I'm thinking was only half useful, especially since I probably do have a poor OCer, his initital recommendation of just starting at 4.6GHz @ 1.2V wasted much of my time.
post #5906 of 19540
Quote:
Originally Posted by KillerDJK View Post

I was just throwing 2.1V out there since I believe I've read that as the max for input voltage, the standard is 1.9V so I figure once I start to push the ring/cache clock up, 1.9V is just not going to work, and that starting is high (not necessarily 2.1V) might be easier. Especially since I have yet to enter a scenario where temperature, not my stability is the issue. In my 3 hours (possibly more if I let it run) XTU stable OC setup, I had it at 2V input memory, however with changes to memory etc. that may not be needed.

(Does a higher input voltage help in finding the lowest stable vcore voltage with the desired clock?)

I have a MSI-Z87-G45 motherboard (BIOS flashed to be the most up to date, all drivers). Think the 1333MHz hit from down from 1600Mhz memory clock is worth an extra 100MHz clock speed or a .5V lower core voltage? (I figure if it happens to help more than that It would indeed be worth it....I'm just trying to ask harder question smile.gif )

Actually though my h100i corsair link software is having issue, so I may have to halt any OC attempts until corsair properly supports W8.1 ....but that's another issue entirely. My fans were set to max during any previous OC attempts.

Yes, VCCIN/VRIN aka input voltage aka eventual voltage CAN cause instability but I think 2.1V is kind of pushing it. I think the default is 1.8V, not 1.9V.

Raising it should not cause higher temps, but it can be unsafe if too high depending on the quality of your mobo, your unique chip, etc. Conversely, if too low, it can cause instability and you can erroneously keep bumping up Vcore when in fact that will only exacerbate the issue. I just think 2.1V is a bit high. If you are comfortable with 2.1V - go for it smile.gif. You can check Dark's table, see what people are running (i think Dark himself is running it up there) so may be fine. Probably worth starting at 1.95V or above as your comfort level dictates.

As for whether the RAM downclock is worth it, only you can decide smile.gif. I thought the goal of this exercise is to get you a more reasonable Vcore at 4.4Ghz. So probably best to find the lowest stable and then try to bump RAM back up to 1600 and see what needs tweaking .At least that's what I'd do, but totally up to you.

I have the same issue with H100i -- TBH i can't believe Corsair has left Win 8.1 unsupported for over 5 weeks now but nothing we can do. It won't affect you other than your ears/noise tongue.gif. By default the fans use a "standard" profile which is pretty aggressive, but they will spin appropriately. At least they do for me.

Let us know how it goes, i.e. if you manage to find a lower stable Vcore for 4.4 (and hopefully with RAM back to 1600 eventually). GL
post #5907 of 19540
oh btw, do you have anything set for SA, DIO, AIO?
post #5908 of 19540
anyone have any info on i7-4470K's that say L317**** MALAY SR***

just got it and looking to overclock!
post #5909 of 19540
Quote:
Originally Posted by creos7 View Post

oh btw, do you have anything set for SA, DIO, AIO?


Those are RAM settings right? I did not touch their defaults, why? (those terms are not familiar to me)

KHX1600C10D3B1K2
http://www.kingston.com/datasheets/KHX1600C10D3B1K2_16G.pdf

This is my ram by the way, when only time the CL when to 10 was using the XMP... without it I believe it was set to 11. I will try when I redo my overclocking more carefully this time to use the settings in this sheet for 1333, then attempt the 1600 just in case that is an issue.
Edited by KillerDJK - 11/25/13 at 3:40pm
post #5910 of 19540
Quote:
Originally Posted by szeged View Post

under water at full load in florida's awful summer heat the max i saw on the hottest core was 91c lol, that was with intel burn test though, normal every day use at full load max i saw was roughyl in the mid 60s

It's gotten a little chilly since last two days, but the heat does kick in randomly on some days.....
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Intel CPUs
Overclock.net › Forums › Intel › Intel CPUs › Haswell Overclocking Guide [With Statistics]