Intel Haswell Overclocking Clubhouse.

[Frizz]

Tested for 4 hours ages ago in like 30+ Deg Cel ambient with pump settings on #2 which is barely audible. Not delidded, not sure if its worth it based on these results.

 

[LiveOrDie]

Tested for 4 hours ages ago in like 30+ Deg Cel ambient with pump settings on #2 which is barely audible. Not delidded, not sure if its worth it based on these results.

Don't use prime as it doesn't stress the FPU try ether AIDA64 using just the FPU test or IntelBurnTest V2, You will see your temps will be higher.

 

[TheHunter]

Ok about my 2400mhz ram OC adventures, I knew i should do better then just 2200mhz.
I bought it specifically because I saw it can do up to 2690mhz
http://uk.hardware.info/reviews/333...benchmarkoc-results-11-13-13-timings--17-volt

Anyway its not DRAM phase clock fault, or RTL IOR?.. I left both at auto

the main issue is/was in these 6-7 adv timings
tWR 16
tRFC 313 (214 is ok)
tWTR 9 (or 8, but didnt try yet)
tRRD 7 (or 6)
tRTP 9 (or 8)
tFAW 27 (or 26)
tWCL 7



I spotted this here - G.Skill adv timings
http://www.anandtech.com/show/6374/geil-evo-veloce-review-2x8gb-at-ddr32400-c111212-165-v

and now i get this with -tRFC 214 and 100% stable,

Usually the system shutdown by now with my old tweak DRAM phase clock = 9..


I also ran geekbench3 and saw a minor ram boost (although its 4.5ghz vs 4.7ghz), very small though lol, but hey at least i made it 2400mhz. Finally





EDIT: and 1T can lower L2 cache ns latency.

With 2T its 4.5 - 4.8ns, now max 3.3ns just like by 2133mhz using 1T.

I lost read and copy a bit, but its still over 35gb/s.




the fix was CL10-12-11-31, instead of CL10-11-11-31

 

[TheHunter]

Im still @ 2400mhz OC since my last post and no problems yet, I even lowered them a bit to CL10-12-11-28-1T



was at adaptive voltage..

 

[Kursah]

Well I just flashed to the newest BIOS which detailed that it "improves stability". I have gone back to using the newest OCCT and it's PSU test so can test overall system stability. I cannot seem to pass with my current settings. I do use linpack x64 + all logical cores which pushes another 10-12C load on each core (pushing upper-80s and low-90's with that enabled!). I have had some BSOD's which I haven't traced yet (Windows 8 BSOD sad face is lame! lol). I want to keep it at 4.5Ghz but if I can't pass my preferred stress test I may need to back off. I've preferred OCCT because of it's PSU test and it's ability to really heat systems up and push their limits. Apparently my limits have been found. This rig has been solid in gaming and everything I've thrown at it though...which makes it hard for me to justify that there may be an issue...but if it's waiting to crop up I'd rather be able to pass this test. I'm thinking of backing off the speed because I don't really want to push voltages too much further.

I will note that the current version of OCCT does NOT push voltages past the set specs if you're using adaptive mode unless you choose the CPU only test with AVX. Then my 1.24v turns into 1.35v+! Yikes! I will say nothing else I've done in normal use gets my GPU, MB or CPU at these temps so that's a plus.

 

[LiveOrDie]

Its not the best temps this is with a ambient of 30c this is a full loop with a 240mm rad, What do you guys think i think i could get better temps maybe trying to remount the block but i really cant be bothered you can see why below lol.

@ 4.4Ghz on 1.2v

 

[Durvelle27]

Guess I'll be joining. Have a i7-4770. Waiting for board to arrive though

 

[Kursah]

Well I finally got around to getting OCCT to pass a 6-hour PSU test with Linpack, all logical cores at 4.4Ghz. I tried 4.5Ghz up to 1.245v, I may try 1.25v, but that's gonna push my temps into the 90s with my air cooling. I did however back off and just happened to try 4.4 @ 1.21v and it went without issue and my load temps dropped a little too. I'm going to try 1.20 for 1-hour next. I'd be willing to give up 100Mhz for cooler temps and lower voltage, I'm hoping to hit the hi teens.

Edit: Well I tried 1.19 for 1-hour, good, 1.18 for 1-hour locked up, trying 1.185v now. I'll take that for 4.4ghz vs 1.24+v for 4.5ghz. Maybe if I had water I could push further.

 

[PolRoger]

It has been a while since I've posted here...

My chip is always "crunching" 24/7 and I had previously settled on a 4.6GHz clock but I recently updated my BIOS and thought I try again to get a stable 4.7GHz long term "crunching" overclock. I've managed to pass 64 hours @ 100% load but I'm still expecting that I'm going to run into a random BSOD sometime over the next few days.

 

[LiveOrDie]

It has been a while since I've posted here...

My chip is always "crunching" 24/7 and I had previously settled on a 4.6GHz clock but I recently updated my BIOS and thought I try again to get a stable 4.7GHz long term "crunching" overclock. I've managed to pass 64 hours @ 100% load but I'm still expecting that I'm going to run into a random BSOD sometime over the next few days.

View attachment 54458

What stress test are you running those temps seem to low?

 

[TheHunter]

Do you need that Vccsa @ 0.896v?


I tested this once and it looked like higher vccsa lowered cpuv a bit.. Is it the same by you?

 

[PolRoger]

What stress test are you running those temps seem to low?

I'm not stressing with Prime, Linpac or AIDA64... I'm just running BOINC Rosetta@home which is kind of like WCG or Folding@home. So far... I've now passed 87 hrs. with 8 threads @ 100% load. Cooling is custom water on an open bench with a triple 140 rad. Ambient temps right now ~18/19 C.

Do you need that Vccsa @ 0.896v?


I tested this once and it looked like higher vccsa lowered cpuv a bit.. Is it the same by you?

I've been pretty much running most of my overclocks with a slight bump to VccSA... running now with +.075? Also running CPU Digital I/O @ +.100 and CPU Cache @ 1.136v

 

[Durvelle27]

 

[...PACMAN...]

Getting a 4770k real soon. Any proof that the i7s clock higher with lower voltage?

Previously had a 4670k but needed 1.27v for a stable 4.5ghz.

Gonna be updating my cooling as well. Seriously considering H105, should be plenty right?

 

[Womper]

It has been a while since I've posted here...

My chip is always "crunching" 24/7 and I had previously settled on a 4.6GHz clock but I recently updated my BIOS and thought I try again to get a stable 4.7GHz long term "crunching" overclock. I've managed to pass 64 hours @ 100% load but I'm still expecting that I'm going to run into a random BSOD sometime over the next few days.

Heh, been a while for me too. I'm about to catch up on the latest BIOS and drivers, decided to check-in here.

Well I just flashed to the newest BIOS which detailed that it "improves stability". I have gone back to using the newest OCCT and it's PSU test so can test overall system stability. I cannot seem to pass with my current settings.

I'm still on the 1405 BIOS. Any verdict on the new 1707 BIOS?

I ran into a couple more random BSOD over the past 4 months, and each time bumped up my core and cache voltages. I'm at 1.285v adaptive on both core and cache now, still with a 4.7GHz core, 4.5GHz cache, and DDR @ 2133. I'm pretty happy with my chip still, I expected a few more BSOD to flush out.

Getting a 4770k real soon. Any proof that the i7s clock higher with lower voltage?

Previously had a 4670k but needed 1.27v for a stable 4.5ghz.

Gonna be updating my cooling as well. Seriously considering H105, should be plenty right?

I swapped my Swiftech H220 out when its pump seized. Swiftech fixed it, but I just don't need that much cooling now that I'm done stress testing my chip. PM me if you're interested in it.

I'm using my CNPS10x HSF again. I did a real nice job mounting it this time, my idle temps are now the same as ambient (~19C).

 

[Kursah]

Heh, been a while for me too. I'm about to catch up on the latest BIOS and drivers, decided to check-in here.



I'm still on the 1405 BIOS. Any verdict on the new 1707 BIOS?

I ran into a couple more random BSOD over the past 4 months, and each time bumped up my core and cache voltages. I'm at 1.285v adaptive on both core and cache now, still with a 4.7GHz core, 4.5GHz cache, and DDR @ 2133. I'm pretty happy with my chip still, I expected a few more BSOD to flush out.



I swapped my Swiftech H220 out when its pump seized. Swiftech fixed it, but I just don't need that much cooling now that I'm done stress testing my chip. PM me if you're interested in it.

I'm using my CNPS10x HSF again. I did a real nice job mounting it this time, my idle temps are now the same as ambient (~19C).

So far I can't complain. I can pass OCCT now back at my 1.241v @ 4.5Ghz and 4.0core on 1.06v (iirc...it might be 1.16v...whatever the stock variable was...), but I can only pass on the CPU tests, if I do the PSU combined test something happens so I'm still looking into that a little bit...though I don't crash in games so it's more of a personal challenge.

I can't say 1707 did much of anything...other than improve the response of the mouse cursor in the UEFI interface.

 

[Womper]

So far I can't complain. I can pass OCCT now back at my 1.241v @ 4.5Ghz and 4.0core on 1.06v (iirc...it might be 1.16v...whatever the stock variable was...), but I can only pass on the CPU tests, if I do the PSU combined test something happens so I'm still looking into that a little bit...though I don't crash in games so it's more of a personal challenge.

I can't say 1707 did much of anything...other than improve the response of the mouse cursor in the UEFI interface.

I did notice that the CPU microcode patch changed from 0x9 to 0x12 when I moved from 1405 to 1707. Everything seems normal, I plugged in my settings from before and off I went.

I tried to install the new chipset drivers, but the installer doesn't seem to find anything to upgrade. Strange, but I haven't had any problems anyways so whatever.

 

[TheHunter]

I've been pretty much running most of my overclocks with a slight bump to VccSA... running now with +.075? Also running CPU Digital I/O @ +.100 and CPU Cache @ 1.136v

I see your mobo over volts cpu or hwinfo64 reads it wrong. You set to 1.274v but per core its 1.296v? Or is that higher VCCSA and digital IO fault?

This is how mine looks like @ 1.283v

see that core2 @ 1.285v,
if i set main to 1.282v then core2 goes to 1.280v and it can be kinda unstable unless i raise other parameters lol, funny stuff how main voltage scales per cores.

1.280v makes most ~ 1.278-1.280v and its kinda weird, its stable but then later its not again..

 

[Solaris17]

my bus will not stabilize at 100 on my UD3H is their a setting im missing? I have taken pretty much everything off of auto but i flux between 95/96

 

[cadaveca]

my bus will not stabilize at 100 on my UD3H is their a setting im missing? I have taken pretty much everything off of auto but i flux between 95/96

set 100.3 manually. FSB has "holes", although every 0.1 MHz is selectable. Try the values between too, perhaps, but it's usually 100.3 that works on boards with 100.0 MHz "hole".

 

[Solaris17]

set 100.3 manually.

ill give that a shot thanks dave i read the review you did on the msi (which was the contender to this board) and while im not expecting miracles i have to have even numbers. this haswell overclocking is crazy diff though. i have alot to catch up on since i went straight nehalem to this.

 

[cadaveca]

ill give that a shot thanks dave i read the review you did on the msi (which was the contender to this board) and while im not expecting miracles i have to have even numbers. this haswell overclocking is crazy diff though. i have alot to catch up on since i went straight nehalem to this.

It's more complex, but actually simpler.

vCore is obvious, vccsa and Digi-IO/AnalogIO are memory controller, cache is cache. Keep that in mind, then everything from the past still applies.

Cache and CPU each have multi, do one at a time. No need to keep cache in sync with CPU speed, since it's not a bottleneck to connect to rest of system after about 3900 MHz. Faster is faster, of course, but cache is obviously the largest source of heat next to FIVR. Cache wasn't much of an issue with SNB/IVB, and Haswell has +25% cache performance it seems(makes sense given process size reduction, plus library enhancements), so keeping cache @ 75% of CPU speed is still pretty efficient for most workloads.

Memory controller is very different, if you had good sticks, they'll fly sky-high on 1150, might be worth keeping depending on size, etc.

 

[PolRoger]

I see your mobo over volts cpu or hwinfo64 reads it wrong. You set to 1.274v but per core its 1.296v? Or is that higher VCCSA and digital IO fault?

This is how mine looks like @ 1.283v

You have your "vcore" sensors for core 0/1/2/3 hidden... While showing all the core VID readings? I have mine set to show just the Core 0 VID and all four of my vcore sensors. My 4.7 o.c. is still not fully dialed in... I got up to a max of ~160 hrs load before BSOD with my previous settings but additional testing has proved not stable so I've now bumped up vcore trying to find long term stability.

 

[Solaris17]

Hm bumping to 100.3 didnt stick. im not sure what to do its almost their is some automatic setting getting in the way.

 

[TheHunter]

Try cpu strap @ 100mhz fixed and its source clock filtering to 60ohm. This will make a fixed blck 100.29mhz

*one older screen

like in my pic below when it shows 4701mhz

You have your "vcore" sensors for core 0/1/2/3 hidden... While showing all the core VID readings? I have mine set to show just the Core 0 VID and all four of my vcore sensors. My 4.7 o.c. is still not fully dialed in... I got up to a max of ~160 hrs load before BSOD with my previous settings but additional testing has proved not stable so I've now bumped up vcore trying to find long term stability.

Ah right, just re-enabled those and it shows exactly the same volts like by you..