Intel Haswell Overclocking Clubhouse.

[MxPhenom 216]

Do you guys mess with both inputs or just the initial? I have Initial and Eventual options.

 

[cadaveca]

You have to enable manual mode, then just set the main and leave initial and eventual alone.

(Like The hunter said he did).

 

[prescient]

cpu 1.008
cpu cash 1.16

reply to this


http://www.techpowerup.com/forums/threads/my-new-rig.202088/

i dont have flash usb

 

[cadaveca]

Should be good then. Aim for 4600 or 4700 @ 1.25V or so. Pay attention to DIO and AIO voltages too though. Also be sure that CMOS is reset when checking those voltages, since adjusting other settings in BIOS can adjust them as well.

 

[prescient]

ohh great news . when i install the cooler i will come here . and step by step . if i can do 5 ghz with 1.3 or would be ok

 

[cadaveca]

ohh great news . when i install the cooler i will come here . and step by step . if i can do 5 ghz with 1.3 or would be ok

4700 seems to be the breaking point. Many chips need boost for that. Every single multi above that possible on air is truly a blessing.

Anyway, mine is 1.040, gets 4600 no problem, even with memory clocked to 2666+. But voltages, vCPU, vCCSA, vDIO, vAIO, vINPUT need to be low too, sometimes this voltage can just be a tease!

5 GHz on air is a pretty lofty goal.

 

[MxPhenom 216]

You have to enable manual mode, then just set the main and leave initial and eventual alone.

(Like The hunter said he did).

Okay, so I disabled it and set it to 1.75v and disabled the iVR Fault management, and temps dropped 5-6 degrees on average during Test 2 of Prime95.

4700 seems to be the breaking point. Many chips need boost for that. Every single multi above that possible on air is truly a blessing.

Anyway, mine is 1.040, gets 4600 no problem, even with memory clocked to 2666+. But voltages, vCPU, vCCSA, vDIO, vAIO, vINPUT need to be low too, sometimes this voltage can just be a tease!

5 GHz on air is a pretty lofty goal.

Has 5GHZ on air even been done with Haswell yet?

@TheHunter

I am going to try that Asus ROG Realbench. I have been reading up on it and sounds like a cool application. Drills the system as if it would in real life. I also really like Using Battlefield 3/4 multiplayer for a round or 2 to determine stability.

Also wPrime as well.

 

[MxPhenom 216]

Ran RealBench in a infinite loop for 30 minutes and max temp I hit was 69c at 4.4GHZ 1.312v and memory at 2400 1.65v. No issues. Then ran 3DMark Skydiver and Firestrike with no issues.

EDIT: Just crashed after about 20 minutes of BF4. Wasn't of BSOD, just the app crash. Not entirely sure what caused it.

 

[MxPhenom 216]

Okay, so BF4 crashed, and instead of just crashing to desktop. System rebooted. No bugcheck in event viewer, nothing in there to give more information. Thinking i might need to tweak the VCCSA or vinput a bit more. Maybe vDIO and vAIO?

 

[cadaveca]

random crash/reboot most likely memory. try 1.7V there. Also, try boost v Input to 1.8 or so.

Myself, I'd just keep 2133. Make note of stock voltage values on all areas, then set each manually. Increase only what you adjust.

VCCSA should needs no or only slight boost from stock.


BTW, found two more kits of your memory, for less than what you paid for yours. Still want another one?

 

[MxPhenom 216]

random crash/reboot most likely memory. try 1.7V there. Also, try boost v Input to 1.8 or so.

Myself, I'd just keep 2133. Make note of stock voltage values on all areas, then set each manually. Increase only what you adjust.

VCCSA should needs no or only slight boost from stock.


BTW, found two more kits of your memory, for less than what you paid for yours. Still want another one?

I really want another one, but I am really broke man!

I want to have my CPU and Memory overclocked. I will stop at 2400 on the memory, 2666 like we tried before wasn't worth it IMO, though with what I know now, I think I could get it stable.

I have been reading a lot on the vinput. I set it to 1.79v for now, and am about to try BF4 again and see if I can finish a match. It was at 1.75v before. Which from one I am reading, higher vcore, means itll likely need higher vinput.

EDIT: So far so good with vinput at 1.79v. Played about 2 full rounds of BF4 will no crashes/reboots. Max temp in game 67c.

Might end up trying 4.5 4.6ghz at this point.

 

[cadaveca]

Might end up trying 4.5 4.6ghz at this point.

Yeah, it's just a matter of managing the temps. Most chips on most cooling won't be able to handle 1.9V. + CPU clock, + ram clock + cache clock. You have to choose which ones are best for your workload, and adjust accordingly. That's why we have all these different voltages in the first place, and by managing it inside the chip, they can keep it very concise...just the iVR takes the heat hit. Intel showed demos of this tech YEARS ago now, with many iVR modules, like 4 I think, outside the chip.


So, anyway, I think this is in part why they have changed the PCB for Deveil's Canyon, and adjusted capacitors on the package. If done right, then vINPUT can be less, and heat can be less. You just need to find out how you chip scales in that relationship, since you want to keep vINPUT as low as possible for temperature control.

 

[MxPhenom 216]

Yeah, it's just a matter of managing the temps. Most chips on most cooling won't be able to handle 1.9V. + CPU clock, + ram clock + cache clock. You have to choose which ones are best for your workload, and adjust accordingly. That's why we have all these different voltages in the first place, and by managing it inside the chip, they can keep it very concise...just the iVR takes the heat hit. Intel showed demos of this tech YEARS ago now, with many iVR modules, like 4 I think, outside the chip.


So, anyway, I think this is in part why they have changed the PCB for Deveil's Canyon, and adjusted capacitors on the package. If done right, then vINPUT can be less, and heat can be less. You just need to find out how you chip scales in that relationship, since you want to keep vINPUT as low as possible for temperature control.

yeah. Ive been reading even more on the input voltage stuff, and have learned a lot from the Haswell OC Guide at OCN.

As you know my board has the Initial Input Voltage, and Eventual Input Voltage option. The guide at OCN mentioned that changing the Initial Input Voltage on Asus boards hasn't yielded any difference, and said to try and tweak the Eventual Input Voltage, but I changed the Initial Input Voltage option, and it seemed to have fixed my crashes/restarts in BF4.

Maybe as a test, ill set Initial back to auto, and change the Eventual to what I have my Initial set too and see what it does. Might be interesting......

 

[cadaveca]

Might be interesting......

OR not.

Also, keep in mind the OCN guide is just a rehash of the ASUS' guide (obviously, since BIOS options apply to your ASUS board).

 

[MxPhenom 216]

OR not.

Also, keep in mind the OCN guide is just a rehash of the ASUS' guide (obviously, since BIOS options apply to your ASUS board).

Yeah, I have been reading the Asus ROG guide as well.

 

[TheHunter]

btw MXPhenom 216


I think its also Internal PLL overvoltage that can heat extra, default auto..

I was @ 130% cpu vrm current and I kept getting some higher temps. in something like LP2 or RE5 benchmarks, both not really temp. demanding, ok LP2 is a bit but still.. I had that disabled before but now with 2004 i forgot about it and left it at auto and now I was getting max 77C on core0 vs 71C after I disabled it (LP2 - test B, fixed bench, with jobthread=8).


And yeah I saw the same thing with SVID, 1.75v is a bit too low and can cause such reboots in BF3/4, sometimes not enough cpu current too, but 130% is enough.. Also too low VCCSA or too low cpu cache or ram current % that's if you OC ram..

*VCCSA 0.050 - 0.055v is more then enough - total 0.856-0.864v (stock 0.808 -0.816v), i keep IO analog an digital both at auto, I tried to test those once but it made everything more unstable, only Vccsa made a difference.

 

[MxPhenom 216]

On my way to having 4.5GHZ stable. Was able to run through about an hour of looping x264 stress test.

45x multi 1.34v (set in BIOS 1.35v LOAD) Initial Input Voltage 1.80, Eventual Input voltage 1.85.

I will test in BF4 later, as I am anticipating that I might need to bump the Eventual vinput a bit more.

 

[TheHunter]

I was losing hope with OC stability with these last two most recent BIOS updates. Then I tried something I hadn't since I first started OC-ing my setup a year ago. I did +0.001 for Adaptive and +.011 for additional voltage. This gives me 1.46v under load. Which so far is stable (so far...).

The kicker? I was running between +.005-+.010 on Adaptive for 1.151-1.156 and it was NOT stable. Both OCCT's Linpack solo and PSU test would crash/lockup/fail within 15-20 minutes. Temps were higher.

Running this lower voltage seems to be very good so far....OCCT PSU test is a very good load...especially with my GTX 770 pumping hot air into my case....between that and my CPU none reach more than the hi-70's at this point. I'm pretty stoked about it. I'll be super excited if I am running lower voltage than ever for the same OC!

I was testing this offset (additional voltage) a moment ago and then remembered your post.

It passed both mini stress tests (RE5 - variable test, lower reso -cpu bound & LP2 - variable, same settings) and 3dmark physics 3-4 times, 1-2 days earlier it would fail in RE5- 2nd or 3rd part really quick, if I had additional - auto and turbo 1.275v.


I remember I use to pass 4.7ghz @ 1.258v, I posted it here once, but later I found out its not really it.. Although I didnt use more then 0.010 additional voltage.

Now, I set additional 0.018v and turbo adaptive 1.250v making total ~1.268v, coreV max 1.280v.
And it passed all mentioned above , but in idle its now @ 0.737v instead of 0.712v


Funny stuff, I think this offset additional voltage acts as a buffer for each turbo multi & its voltage and so if a chip is programmed to run idk 4.6ghz between 1.20 -1.24v (upper values - stabler) you raise this gap a little so now you can use less main.
But I get higher temps now, *see bellow use to get max 62-65C on core0 in RE5, 3dmark physics temps are around the same - usually near70C threshold..

I also have cpu current 130% vs 120% its just that 4-5C gap, will test that again or maybe go for even lower cpuv , but I think 1.25v is kinda the limit for sure

 

[prescient]

should i disable paging files with 16 gb ram ?

 

[cadaveca]

Nope. But I'd set to 4096 MB min-max.

 

[MxPhenom 216]

Feeling pretty stable. My previos settings for 4.5 weren't quite right. I got worst score in Firestrike than at 4.4. So I jumped the vcore to 1.35v and Eventual Input Voltage to 1.91 and score jump up higher then 4.4GHZ. Then dropped vcore to 1.34 and left Eventual vinput at 1.91 and got even a few more points. So I think my final settings for 4.5ghz are 1.34v vcore and 1.80v Initial Input voltage, and 1.91v for Eventual Input voltage. Everything else same as before.

 

[prescient]

whoooo

with my old system 3d mark

P3616

http://www.3dmark.com/3dm11/286898;jsessionid=144E20EBF71F99D02ADE7AEAB53B2DA4?show_ads=true&page=%2F3dm11%2F286898%3Fkey%3DVNfkSSmkkdjmS8dhCtFCcbsBW4rb4X

the new system

http://www.3dmark.com/3dm11/8429601


now let me know one thing . we have the same system here if that in your sig is 3d mark 11 score
and you are so far ocied how come my system scores 12300? yours score 10800
is it the 16 gb ram ?

 

[MxPhenom 216]

whoooo

with my old system 3d mark

P3616

http://www.3dmark.com/3dm11/286898;jsessionid=144E20EBF71F99D02ADE7AEAB53B2DA4?show_ads=true&page=%2F3dm11%2F286898%3Fkey%3DVNfkSSmkkdjmS8dhCtFCcbsBW4rb4X

the new system

http://www.3dmark.com/3dm11/8429601


now let me know one thing . we have the same system here if that in your sig is 3d mark 11 score
and you are so far ocied how come my system scores 12300? yours score 10800
is it the 16 gb ram ?

That score isn't from 3DMark 11. It is 3DMark "13" Firestrike bench.

 

[prescient]

sure i should have figured out the evga is surly stronger . and your oc physics score is way higher
i will get that much increase if i oced my cpu ? i dont think od want to stretch my gpu so far at this point .

 

[MxPhenom 216]

sure i should have figured out the evga is surly stronger . and your oc physics score is way higher
i will get that much increase if i oced my cpu ? i dont think od want to stretch my gpu so far at this point .

Well I have a launch 780 (Launch products always tend to be the thing to get if you want to overclock). Bought it the morning the 780s launched after reading the review on TPU. It was probably the most expensive impulse buy I have ever done, but I don't regret it at all. And yes your Physics socre will be right around mine if you overclock to 4.5GHZ.