Official Haswell-E thread

[DooKey]

Im using the F8C aswell but on a Gaming 5.

My pc suffers from the crash loop bug if i plug a ext HDD into a USB3 slot.

The new Bios fixed it for the USB front panel but the back usb slots still causes an infinite crash loop.

Apart from that its been rock solid.

I think my i7 5820K is a dud tho. 1.265V just to get 4.2GHz

Dang, that's def a dud 5820K, sorry for the bad luck.

 

[dmoney1980]

I had to ship my corsair ddr4 kit back to amazon today. Started getting blue screens so I did a Memtest, sure enough errors right off the bat. Weird this is it worked fine for a week even with a few bf 4 sessions. Sucks that my 980's are coming Tuesday but now I have this problem to deal with

 

[EvilNodZ]

Dang, that's def a dud 5820K, sorry for the bad luck.

Yeah defo a dud, i cant get 4.3GHz without hitting TJmax gave up at 1.3V But its still a big upgrade over my i7 860 So im happy and i have had fun playing around overclocking it.

 

[taserbro]

I'm using F8c. Didn't have any problems with it. In fact, I think it's the best BIOS for the board so far.

Are you getting stuck at A9 when attempting to get into the BIOS?

Actually no, the bios flashes the splash screen briefly and then displays A9 before going into perma-black screen whether I solicitate to enter the bios or not, which effectively bricks the board. I first thought it might have been my bios being corrupted but the thing is it worked perfectly for a while, fixing all my issues with the F4 revision and then after setting it up, I booted into windows, installed the drivers, rebooted for sound card drivers and bam! nothing worked anymore and there was no way to fix it.

I've tried resetting the bios via both ways, with the switch or by powering down and removing the battery for a minute, reflashing (which yields the same result), every permutation of bios options on the board... It's probably just a 1 in a 1000, cosmic ray flipped the wrong bit sorta deal.

The other backup bios is still on the F4 revision which is outright unusable due to not letting me use any cpu options at all. I've already contacted gigabyte about it but realistically, I'm going to be boned out of my good rig for 5-7 business days.

 

[AdamK47]

Try disconnecting all USB and SATA devices and see if you can get further.

 

[biostud]

Yeah defo a dud, i cant get 4.3GHz without hitting TJmax gave up at 1.3V But its still a big upgrade over my i7 860 So im happy and i have had fun playing around overclocking it.

Same here. I haven't done a lot of testing but going over 42x seems to render it unstable.

 

[Mydog]

I've been lucky with my 5960X I think, 4.8 GHz stable with a bit high vcore but my water cooling can handle it.

 

[Ajay]

I've been lucky with my 5960X I think, 4.8 GHz stable with a bit high vcore but my water cooling can handle it.

Wow, that's sweet! I wonder what the max recommended vCore is for HW-E?

 

[Mydog]

Wow, that's sweet! I wonder what the max recommended vCore is for HW-E?

I believe it to ba 1.35V but with good cooling you can go higher at your own risk.

 

[taserbro]

Try disconnecting all USB and SATA devices and see if you can get further.

I actually tried that too, reducing the rig to the bare minimum before trying anything else but it wouldn't do anything but throw a black screen at me.

I ended up sending the board back and grabbing a asus x99 deluxe locally.
I'm not too used to the way this one works yet but 4.5ghz at 1.3v passes all my benches and rog stress testing so I'm not complaining, especially since I ended up with a notoriously poor batch number.

Happy ending I guess.

 

[biostud]

Had some problems with low BCLK in CPU-z (96-98Mhz) turns out Hyper-V influences the correct reading in CPU-z and HWinfo, so after I disabled Hyper-V I now get 99.98Mhz.

 

[Abwx]

I've been lucky with my 5960X I think, 4.8 GHz stable with a bit high vcore but my water cooling can handle it.

According to the temperatures on the right, and knowing that the CPU disspation with your settings and CB is about 220W, the thermal resistance between the die and ambiant is at most 0.16°C/Watt, using a core2 quad die/case thermal resistance of 0.113°C/watt as basis imply that your cooling apparatus has a thermal resistance of 0.06°C/watt, this is good enough to dissipate 500W with a CPU case that would be heated only 30°C over ambiant.

 

[Mydog]

According to the temperatures on the right, and knowing that the CPU disspation with your settings and CB is about 220W, the thermal resistance between the die and ambiant is at most 0.16°C/Watt, using a core2 quad die/case thermal resistance of 0.113°C/watt as basis imply that your cooling apparatus has a thermal resistance of 0.06°C/watt, this is good enough to dissipate 500W with a CPU case that would be heated only 30°C over ambiant.

Sounds about right, one 360x60 and on 280x60 radd with 2k rpm fans in push pull config. ambient is about 20-22C. However I got a Hailea water cooler hooked into the loop and it helps me keep the water temps in check.

As I use a lot of QDC's I can use the cooler in the CPU or GPU loop or combine them all into one big loop with four radds(2x 360x60 + 2x 280x60). All radds has fans in push pull but the two 360's has the push fans with shrouds. Fans are ofc controlled by software controller and are set up per radd.

 

[EvilNodZ]

I've been lucky with my 5960X I think, 4.8 GHz stable with a bit high vcore but my water cooling can handle it.

Interesting your TJ Max is 105DegC, Mines is 94Deg C but im using core temp 1.0RC7

 

[AdamK47]

Yeah, how is it at 105°? Is this an ES CPU? Mine is the usual 89° Tj. Max for a 5960X.

 

[AdamK47]

According to the temperatures on the right, and knowing that the CPU disspation with your settings and CB is about 220W, the thermal resistance between the die and ambiant is at most 0.16°C/Watt, using a core2 quad die/case thermal resistance of 0.113°C/watt as basis imply that your cooling apparatus has a thermal resistance of 0.06°C/watt, this is good enough to dissipate 500W with a CPU case that would be heated only 30°C over ambiant.

Running Cinebench isn't a very good indicator for thermals. It's pretty lightweight actually.

 

[biostud]

Yeah, how is it at 105°? Is this an ES CPU? Mine is the usual 89° Tj. Max for a 5960X.

My 5820K says 105 C as well.

 

[EvilNodZ]

Im sure my TJ max is higher when i remove thermal throttling in the bios. Do you have it switched off in the bios?

 

[AdamK47]

Im sure my TJ max is higher when i remove thermal throttling in the bios. Do you have it switched off in the bios?

Oh, that's probably what it is.

 

[Dylanharvey999]

Wish I could afford this setup!

 

[groovieknave]

Alright, I want to get into this OC game, I'm using the Asus X99 Deluxe with the 5820k, I've only used the auto tuning but briefly played with the multipliers but haven't had any luck higher than 4.1ghz.

Maybe I can get some tips for settings from people to see what pushes this properly?

I'm using the H110 corsair water cooling and it seems to cool the cpu at 30-40c or less no matter what I do.

 

[Idontcare]

Im sure my TJ max is higher when i remove thermal throttling in the bios. Do you have it switched off in the bios?

Within Realtemp or Coretemp, TJmax is not a value read from your CPU or your BIOS, it is set value that is programmed into the temperature monitoring software and is adjustable by the user.

If your realtemp software is reporting to you (via the GUI interface) that TJmax is "105C" then that means it is either falsely assuming that value or you have manually set it to that value.

TJmax doesn't change for your CPU, all that you as the user can do is instruct the mobo to ignore the CPU when the CPU instructs the mobo to throttle the CPU because it has gotten to hot. (you can setup the BIOS to ignore the CPU's thermal protection requests, but you can't change when the CPU issues those requests)

I recommend you look into adjusting the TJmax value in your specific instance of Realtemp so that it is correct, Realtemp needs to know the correct value because the Realtemp can only read the temperature offset reported by the CPU (CPU doesn't report temperature, it reports distance from TJMax, the offset to TJmax, have the wrong TJmax in Realtemp and you have Realtemp reporting the wrong "temperature")

 

[Grooveriding]

Alright, I want to get into this OC game, I'm using the Asus X99 Deluxe with the 5820k, I've only used the auto tuning but briefly played with the multipliers but haven't had any luck higher than 4.1ghz.

Maybe I can get some tips for settings from people to see what pushes this properly?

I'm using the H110 corsair water cooling and it seems to cool the cpu at 30-40c or less no matter what I do.

First update to the latest BIOS: http://www.overclock.net/t/1510328/...upport-thread-north-america/990#post_22893334

Then leave everything at stock, use manual voltage control, change your CPU input voltage to 1.9, CPU voltage to 1.3v and set your CPU ratio to 45.

Try to boot into Windows, if successful run a system stability stress test like the one included in the latest Aida64 Extreme beta build. If you pass go back to BIOS and either start reducing CPU voltage and running the stress test until you fail the test in order to get your lowest voltage needed for 4.5ghz, or try raising the ratio to 46 and seeing if you still pass. If you fail at 1.3v and 4.5 then reduce the CPU ratio to 44 and try again. I would not go over 1.3V on the CPU with your cooling setup. Make sure you watch your CPU temperatures as well throughout all of this.

You can also try adjusting the CPU current and LLC settings under the Digi+ power settings in BIOS if you are having a hard time getting your overclock stable. A lot of people have found LLC to 6 or 7 and cpu current at 120% can help. This will raise temperatures but can help to get an unstable overclock stable.

Once you find a stable CPU clock you can try raising the Max cache ratio to something like 40, you'll also need to set the CPU cache voltage to near 1.2 to get that stable. This is only if you want to play with the cache speed, it doesn't make that much difference.

Once you've found your stable overclock and necessary voltages using manual control you can switch over to Offset or Adaptive voltage controls so your voltage and clocks will dial down when the system is under a light load or at idle. Adjust offsets as necessary to get the needed voltage you determined in manual mode delivered to your CPU when it's under loads. Make sure Speedstep is enabled and CPU C states are also enabled, you need at least C1 to be set to enabled for voltages and downclocking to work with adaptive and manual mode.

This is a good thread to read : http://www.overclock.net/t/1510388/haswell-e-overclock-leaderboard-owners-club A lot of HW-E owners there overclocking the chips. And this PDF gives a little more detail on how to overclock on the platform in the context of what Asus saw binning a pile of 5960X chips : https://docs.google.com/file/d/0Bz2VRRbLPrZnYmJVVHM2UGxVS00/edit?pli=1

 

[Abwx]

Running Cinebench isn't a very good indicator for thermals. It's pretty lightweight actually.

About 90W at stock and thus 220W for said settings.

90 x (1.37/1.1)^2 x (4.8/3.1) = 216W

As an indication Hardware.fr got 151.2W at the CPU 12V rail , wich is about 135W for the CPU, at stock frequency and voltage using Prime 95, with the settings posted above the comsumption under Prime 95 would skyrocket to 320W.

http://www.hardware.fr/articles/924-6/overclocking.html

All three CPU tested, 5960, 5930 and 5820 were instable at 1.3V, max frequency was 4.1 for the 5960, 5820 and 4.2 for the 5930, theses are samples delivered by Intel.

 

[groovieknave]

First update to the latest BIOS: http://www.overclock.net/t/1510328/...upport-thread-north-america/990#post_22893334

Then leave everything at stock, use manual voltage control, change your CPU input voltage to 1.9, CPU voltage to 1.3v and set your CPU ratio to 45.

Try to boot into Windows, if successful run a system stability stress test like the one included in the latest Aida64 Extreme beta build. If you pass go back to BIOS and either start reducing CPU voltage and running the stress test until you fail the test in order to get your lowest voltage needed for 4.5ghz, or try raising the ratio to 46 and seeing if you still pass. If you fail at 1.3v and 4.5 then reduce the CPU ratio to 44 and try again. I would not go over 1.3V on the CPU with your cooling setup. Make sure you watch your CPU temperatures as well throughout all of this.

You can also try adjusting the CPU current and LLC settings under the Digi+ power settings in BIOS if you are having a hard time getting your overclock stable. A lot of people have found LLC to 6 or 7 and cpu current at 120% can help. This will raise temperatures but can help to get an unstable overclock stable.

Once you find a stable CPU clock you can try raising the Max cache ratio to something like 40, you'll also need to set the CPU cache voltage to near 1.2 to get that stable. This is only if you want to play with the cache speed, it doesn't make that much difference.

Once you've found your stable overclock and necessary voltages using manual control you can switch over to Offset or Adaptive voltage controls so your voltage and clocks will dial down when the system is under a light load or at idle. Adjust offsets as necessary to get the needed voltage you determined in manual mode delivered to your CPU when it's under loads. Make sure Speedstep is enabled and CPU C states are also enabled, you need at least C1 to be set to enabled for voltages and downclocking to work with adaptive and manual mode.

This is a good thread to read : http://www.overclock.net/t/1510388/haswell-e-overclock-leaderboard-owners-club A lot of HW-E owners there overclocking the chips. And this PDF gives a little more detail on how to overclock on the platform in the context of what Asus saw binning a pile of 5960X chips : https://docs.google.com/file/d/0Bz2VRRbLPrZnYmJVVHM2UGxVS00/edit?pli=1

Thanks so much... that's the best info I've got!