cancel
Showing results for 
Search instead for 
Did you mean: 

3970X running too hot. RIVE AUTO Vcore too high? How low can you go?

jdbtwo
Level 7
I have a 3970X with an ASUS Rampage IV Extreme ( latest BIOS ), 64GB of 2400MHz G.Skill DDR3 F3-19200CL10Q2-64GBZHD and a Corsair H100i ( In a quad 2x-2x push-pull config ).

I've tried two thermal interface materials :

1) Indigo Extreme
2) Arctic Silver 5 ( what I'm currently running with )


( Note that I used Realtemp GT for measuring temperatures and prime95 x64 at the "maximum heat" setting for load testing below )

With the Indigo Xtreme, even after reflow, my load temps at stock were upwards of 75C+ but only about 29C idle. They did get high enough about 4 times to shut off my computer due to thermal protection ( 90C? ).

Now with the Arctic Silver 5, my load temps are "better" around 50-60C but idle is a craptastic 35C. That's pretty sad considering I tinted the CPU heatspeader and the H100i water block with Arctic Silver 5 and used the 1mm "vertical line" method of application on the CPU as well as rotating the H100i water block several times before bolting down.

Note that I've disabled Intel Enhanced Speedstep so I'm running at 4.0GHz most of the time. I've also disabled C1E .

Just FYI my batch number is 3238B264 .

I should be seeing idle temps of around 25C and load temps of around 40-45C.

CPU-Z says my load Vcore is around 1.384V ( after Vdroop ) and at 1.4V when idle when the Vcore is set to "AUTO" in the BIOS.

This seems quite high. The default "AUTO" Vcore is 1.256V. I've been running prime95 with small ffts for about 7 hours straight without errors at a manual 1.265 Vcore so it seems that I haven't reached the lower limit yet on Vcore and this is 4.0GHz at load. My temps have also improved remarkably, with the coolest core coming in at 49C .

So, I guess my question is, how low can you push the RIVE-3970X Vcore voltage with a load frequency of 4.0GHz before expecting instability?

Thanks,

jdbtwo
Definiti facultate nostro potis in impossibilem ultra videre modo sumus

[/HR]
There are only 11̅ types of people in the world : Those that understand balanced ternary and those that don't.
18,036 Views
13 REPLIES 13

BOLTS4BREKFAST
Level 12
My 3960x is at 4.7Ghz with H100i (stock thermal paste), 4 fan (push-pull), VCORE @ 1.375... Speedstep is off... All C-states OFF... So my CPU runs at 4.7GHZ 24/7 with 1.385vcore.

As far as pushing that low voltage with a 4.0Ghz clock.. All CPU's are different.. so yours might handle lower voltage on a higher clock.. All i can say is slowly try lowering the voltage and running benchmarks and games..

I dont really care for prime95 and other benchmarks like that.. because they are not real world usage.. but to each his own.
NZXT Phantom 820
Asus Rampage IV Extreme
NZXT HALE90v2 1200w PSU
Intel i7 3960X OC'd @ 4.7Ghz
Corsair H100i (Push/Pull)
MSI GAMING 4G GTX 980
16GB Corsair Dominator GT 2133Mhz
2 Samsung 840 Pro 256GB SSD's (RAID-0)
2 Samsung 850 Pro 256GB SSD's (RAID-0)
SoundBlaster Z PCIe Sound Card
Logitech Z906 5.1 Surround Sound
Sennheiser PC360 Headset
LG Blu-Ray Burner
Windows 10 Professional 64-bit

Melting_Point
Level 10
jdbtwo wrote:
That's pretty sad considering I tinted the CPU heatspeader and the H100i water block with Arctic Silver 5 and used the 1mm "vertical line" method of application on the CPU as well as rotating the H100i water block several times before bolting down.


Perhaps you should try the pea-shaped ball method, which has always worked fine for me.
Motherboard: RIVE (3602 bios)
CPU: Intel 3930K @4646MHz
OS Drive: 2 X Samsung 840 PRO (Raid 0)
Storage Drive: 2 X 1.5TB WD Caviar Black RAID 0, 2 X 3TB WD Caviar Red, Kingston V100 256GB SSD
Memory: 64GB G.SKILL Ripjaws Z (F3-12800CL10Q2-64GBZL)
GPU: Gigabyte GeForce GTX580 @795MHz - 1536MB GDDR5
PSU: OCZ ZX1250
Cooling: Phantek PH-TC14PE
OS: Windows 7 Ultimate x64. (EUFI)

Zka17
Level 16
Welcome, jdbtwo, to the ROG Forum!

Something went really bad there... since both the Indigo Extreme and Arctic Silver 5 are showing way too high temps, there left only two posibilies: wrong mounting of the cooler and your H100i are not working correctly... - I would try to "fix" the first one... are you sure that the block is correctly/evenly is touching the CPU's surface?

HalloweenWeed
Level 12
jdbtwo welcome to the ROG forums.
If you have set this:
Extreme Tweaker menu:
CPU Vcore V: Offset mode
Offset mode sign: + or -
CPU Vcore offset V: auto

Your Vcore will be fixed, and not reduce as it should at idle. This explains your idle temps. IDK what happens when set on CPU Vcore V: auto; but it may work the same way. Instead set:
Extreme Tweaker menu:
CPU Vcore V: Offset mode
Offset mode sign: +
CPU Vcore offset V: 0.005

About your shutdowns, I'm not positive about SB-E but W/X58 the CPU just downclocks at TJmax - which is 91*C BTW. I think SB-E is the same, meaning something else is causing a shutdown. Even the VRM ("Vreg," MOSFET) overheat causes only a downclock, so it's not that either. More likely it's this BIOS setting you need to set:
Monitor menu:
Anti surge support: Disabled

and/or

Digi+ power control submenu:
CPU current capability: 140%
VCCSA current capability: 130%
DRAM-AB current capability: 150%
DRAM-CD current capability: 150%


As for your core temp issues, first let me say did you try tightening the thumbscrews W/pliers? And you must do that slowly and evenly, not too much on one particular corner first. The first time I tightened mine I found that one was still not tight, and had a bad seat.

16709

Here are some of my TIM applications for your reference:

16708

16710

16711
i7-3930K; Asus RIVE; G.SKILL Ripjaws Z 4x4GB DDR3 1866; MSI 7870 2GD5/OC; Crucial M4 SSD 256GB;
Corsair 1000HX; Corsair H100, 4x Excalibur 120mm PWM CPU Fan p-p, AS5; SB X-Fi Titanium Fata1ity Pro;
Dell U2412m IPS 1920x1200; Cooler Master HAF 932 case; Tripp-Lite OMNIVS1500 UPS fully Line-interactive.
(EVGA site: ) And I have a second (wife's) computer, Eve.

Overclocking is useless to me if it is not rock stable.

Retired
Not applicable
And the fact that tighten too hard, result in higher temps.. and higher noise level..

Pliers, is NO NO

i think its like a zillions threads in this forum, regarding the H100, and how 2 mount it..

ok, not a zillion, thats the lap-top issues threads

But many

HalloweenWeed
Level 12
You see that pic W/my calipers in depth micrometer position? That was how I discovered my H100 mount was crooked, one corner was not down. And I had tightened them what I considered tight with my fingers. I used to be a machinist, I know what tight is, I am no wimp with my fingers. Three corners measured ~0.615" and the fourth measured ~0.675". That's almost a 1/16" difference. Upon remounting I used the calipers to tighten all my thumb nuts evenly, and discovered that the mount is designed to bottom out against the standoffs meaning you can use pliers, it does not hurt anything.

This is what I have done ever since, and I am at 4.7GHz stable now.

When you tighten all nuts to the tight against the standoffs position, it is about 0.605". Before tightening, and before applying TIM, the measurement is 0.620-0.625", meaning it's only 15-20 thousandths difference in movement. There is nothing wrong with that.
i7-3930K; Asus RIVE; G.SKILL Ripjaws Z 4x4GB DDR3 1866; MSI 7870 2GD5/OC; Crucial M4 SSD 256GB;
Corsair 1000HX; Corsair H100, 4x Excalibur 120mm PWM CPU Fan p-p, AS5; SB X-Fi Titanium Fata1ity Pro;
Dell U2412m IPS 1920x1200; Cooler Master HAF 932 case; Tripp-Lite OMNIVS1500 UPS fully Line-interactive.
(EVGA site: ) And I have a second (wife's) computer, Eve.

Overclocking is useless to me if it is not rock stable.

Retired
Not applicable
The major downside with the h100 as i see it, is the fact that the block is flat grounded really, really poor. its like a ashtray, and the pea sized paste is most likely to fill that void.. because there hardly any contact, in the center of the block, with the cpu. if any??, and its there the cores are located..

Only by grind/lapping the block in a proper way, -5c idle temps is gained..

If 2 surfaces is very flat, very little paste is needed, and no major force. just need to be tighten evenly, and with a equal force.. law of physics

springs are very good, at least thats my opinion, for use in cpu block mountings..

And with all do respect, weed. i dont mean 2 run you over..

If the block is seated slightly off in the plastic, or brackets not fully straight, i think you actually could make it worse, by measuring the corners, and tighten.. sometimes an other sense is the best tool..

The difference between a "super paste" and a "budget one", is at the most 2-3c gain MAX, with a few exceptions.. If the gain is higher, i think the mounting of the block is 2 blame.. max load can can have a slightly higher difference.. But not any major numbers.. max 5c at 70c

I guess i went some off-topic,, sorry..

maximiza
Level 11
it was 5 ibs of torque for my Kühler block, 2011 socket mounting screws, you need a torque driver set though. I think it is always 5 ilbs of torque for the socket 2011 cpu mounting block screws regardless of what block you use i think it is an intel spec.

Hillguy
Level 9
Take a look at this video on thermal paste application , at the end he does what he calls the " Cross" method , it works awesome :cool:

Another good part is how he demonstrates the old fashioned " Even Spread " method is probably the Worst way to do it :eek: