FRONTPAGE Intel i7 4790K Devil's Canyon CPU Review

EarthDog · Jul 18, 2014

Lol, yeah... It certainly helps to know how someone is using the term since it suddenly seems interchangeable. Like I said, to me, VID is stock voltage... Nothing more. Stock voltage is a static value for each cpu... It's idle voltage in Windows regardless if it comes from a mm or cpuz(if accurate, lol).

VID to me is NOT a dynamic value as it seems to be used here.

For example I ask, what is your stock vcore(VID)? I do NOT say what is your current VID. Vcore is the correct term there (again to me, lol)

RGE · Jul 18, 2014

yeah, vid has an enthusiast definition (ie = stock vid) which is different from intels, which is why I try to avoid vid discussions, but failed to here

my only issue with cpuz is I would rather it read actual/supplied voltage sensor like hardware monitor, so I can see the LLC like modification of vid (called for voltage), instead of cpuz reading bit for vid. Along with cpuz fixing mhz/bclk bug. But yeah .02v or the 99.89 bclk bug isnt going to bug everyone....maybe just me

EarthDog · Jul 21, 2014

LOL, yeah, you use it(VID) as a dynamic value versus me, VID = stock. If I want to know what your voltage is for the core, I ask what is your vcore, not VID.

Anyway......

PolRoger · Jul 22, 2014

Since Haswell was first released the VID/VCC for CPUZ has been switched/reversed.

Two posts by "Blameless" over on OCN discussing VID:

http://www.overclock.net/t/1238019/what-is-cpu-vid#post_16875895

http://www.overclock.net/t/1238019/what-is-cpu-vid#post_16876166

Maybe it has something to do with the inclusion of on die VRM for 4th gen chips? I suppose for the sake of clarity... When people now discuss how much voltage they are running for a given overclock some confusion can occur since CPUZ isn't actually reporting vcore (VCC).

Interestingly if you go back one generation and monitor an IB cpu... CPUZ reports correctly:

RGE · Jul 22, 2014

PolRoger said:
Since Haswell was first released the VID/VCC for CPUZ has been switched/reversed.

Interestingly if you go back one generation and monitor an IB cpu... CPUZ reports correctly:

Yep, which is why having to use HWM for vcore now, otherwise miss the up to .02V increase with built in LLC logic.

I was wondering same with fivr...maybe cpuz reading vid from cpu registers because either coder doesnt know register or access to info from fivr for vcore or whether HWM is reading vcore from mobo, like my fluke, and cpuz coder didnt want trouble of scan info for all mobos,....because otherwise dont understand why cpuz switched from vcore to vid. But then again, dont understand the persistent mhz/bclk bug either, when realtemp and older versions of cpuz 1.67 reads mhz just fine.

And on another note, finally got 5.1ghz delidded stable enough for CB R15, never came close to this with ihs on.

Spawn-Inc · Jul 22, 2014

rge said:
And on another note, finally got 5.1ghz delidded stable enough for CB R15, never came close to this with ihs on.

does that mean your running without the IHS on, or you swapped the tim between the 2?

what was your temp change before and after?

RGE · Jul 22, 2014

no, I didnt word that well, i meant to say prior to delidding. I am using CLP for tim1 (IHS to die) and pk1 for tim2 (wb to ihs).

My temps dropped 23C after delidding and using CLP for tim1.... got about half that temp drop with pk1 for tim1 (while waiting for liquid metal to arrive). It didnt help my 24/7 settings, but definitely made benching 5ghz to 5.15ghz more stable with less vcore.

outhouse · Jul 24, 2014

There is a guy over at evga that is running direct die CLU on water that is topped out 60C something, on the latest primi95 28. whatever

he is running low voltage though to get to 4.8 1.25 I believe

I may go naked myself.

RGE · Jul 24, 2014

Idontcare at anandtech got 4C lower temps at 4.7ghz with bare die vs IHS, if you try it post the temp decrease, you have probably seen his testing. I have only done bare die for short times testing temps, pain removing bracket and back when I did it the PCB was thinner and difficult to get even pressure so it wouldlnt boot correctly everytime. I lost my taste for bare die then. Thought about doing it just to check the temp difference on mine, and if it werent for bracket removal would have already done so, but wont be running bare die 24/7.

ATMINSIDE · Jul 24, 2014

outhouse said:
I may go naked myself.

On topic though, direct to die is always the coolest. If you have the capability with your cooler/block.

outhouse · Jul 24, 2014

rge said:
Idontcare at anandtech got 4C lower temps at 4.7ghz with bare die vs IHS, if you try it post the temp decrease, you have probably seen his testing. I have only done bare die for short times testing temps, pain removing bracket and back when I did it the PCB was thinner and difficult to get even pressure so it wouldlnt boot correctly everytime. I lost my taste for bare die then. Thought about doing it just to check the temp difference on mine, and if it werent for bracket removal would have already done so, but wont be running bare die 24/7.

this guy put the mounting bolts back in to keep the backplate.

seemed like a good idea.

I could imagine it raises the poassibility for pin damage. you know that CPU will pull off when removing the waterblock.

RGE · Jul 25, 2014

well delidding added 100 mhz to my suicide shot as well. 5.25 was instant freeze up to 1.5 vcore prior to delidding.

5.3ghz with 1.48v (base clock set to 100.05 to avoid cpuz bug).
http://valid.canardpc.com/mxtq3m

5.2 sp1m/sp32m 1.48v

outhouse · Aug 4, 2014

OK finally got my CLU

15C drop in temps using small's

7C drop using blend.

My rig was getting unstable due to two 780ti's pumping a lot of heat in my loop gaming. This stablilized it at 4.8

Woomack · Aug 12, 2014

Finally better than average haswell.
So far I made this:

Almost all settings at auto. Vcore in BIOS at 1.228V. Memory settings from XMP @ 2933 C12 1.65V. Other voltages and LLC at auto.
It also passed XTU @4.6GHz 1.17V which is stock VID.

Edit:
Little update - 4.8GHz 1.267V passed 1h prime95 and some other tests. I can't make 4.9GHz work stable. All the time error ( bs ) 101. I can make it boot @5.1GHz ~1.45V.

RGE · Aug 12, 2014

yep 4.7 with 1.228v and 4.8 with 1.267v is a nice chip, temps arent too bad either. You can probably run most benches at 5+ as well....and dang I need your ram for my benching...lol....i probably should have gotten 8mb instead of 16mb for overclocking, since I dont use 16mb anyways.

Woomack · Aug 12, 2014

Your Dominators will be faster in every benchmark. These high clocked kits have low performance, just look good. 2400 CL10 kit will be as fast if not faster than 3000 CL12 in this case and you can set your memory to 2600+ CL10.

I think that my CPU won't be so great afterall as I used prime95 28.4. In 28.5 it's crashing at the same settings with 124 error. I have to check it later if temp is too high or it needs higher voltage. Right now it's running above 90*C with only ~1.27V. My water cooling isn't great but it's enough to keep [email protected] 1.4V at 15*C lower temps.

Edit:

4.5GHz is running now for over 30 mins in prime95 28.5 at 1.15V so slightly undervolted ( stock is 1.17V ). Temps max ~80*C.
I don't really care about stable work at 4.7GHz+. I will probably run it on cold for some benchmarks and later undervolted at 4.2-4.5GHz. It would be still nice to set it stable @4.9GHz+.

RGE · Aug 13, 2014

yeah not your watercooling, just tim1 issue again. Be interesting to see what one of the good clockers do on cold, post your results if you do it. And even though you used 28.4 without fma3, wont take much more vcore to make up difference.

Funny, I have to change (release/renew) my ip address or use a proxy every time I try to visit this site...lol...the ip range of my carrier must be in this sites spam blocker.

Woomack · Aug 13, 2014

I already checked that my CPU doesn't need higher voltage in 28.5 as long as I keep low temps. 95*C+ causes 124 error after 1-2 mins of full load.
Right now it looks like:
4.5GHz 1.15V
4.6GHz 1.17V
4.7GHz 1.228V but in Prime95 28.5 it already reaches ~95*C and is crashing after some time.
4.8GHz 1.267V but the same issue with overheating, passed 1h+ P95 28.4
4.9GHz+ 101 blue screen and for now I'm not sure how to make it stable, have to test some additional voltages

Blueleader · Aug 15, 2014

I cannot wait to put my new Devils Canyon under stress. I'm upgrading from a 5 year old CPU with 800 million flat transistors to a 1.4 BILLION 3D Trigate transistor monster...WOW!

dicko555 · Aug 16, 2014

This is looking pretty good.

I also ran AIDA64 for a couple of hours and temps were higher, mid to high 80's. but averaged in the high 60's and low 70's.
Only really touched the high 80's occasionally.

Did NOT hit 90+

FRONTPAGE Intel i7 4790K Devil's Canyon CPU Review

Gulper Nozzle Co-Owner

Senior Member

Gulper Nozzle Co-Owner

Member

Attachments

Senior Member

Member

Senior Member

Senior Member

Senior Member

Sim Racing Aficionado Co-Owner

Senior Member

Senior Member

Senior Member

Benching Team Leader

Senior Member

Benching Team Leader

Senior Member

Benching Team Leader

New Member

Member

Attachments

Similar threads