X6 1045T, Asrock 990FX Extreme4 overclocking

[VirtualLarry]

So far, I'm doing OK overclocking I think, but temps are pretty darn bad.

I'm at 3.510Ghz (260FSB, 13.5 multi), LLC on AUTO, vcore on 1.35v (with LLC, gets to 1.4v in OCCT).

OCCT 4:64-bit linpack (90%) gets to 65.5C on CPUTIN. CoreTemp reports 42C.

Heatsink is OCZ Vendetta. Case is Rosewill Blackhawk (5 fans stock).

Heatsink is configured with FAN at top, with mobo mounted, pulling air through heatsink, and then the top exhaust fan is right there next to the CPU fan.

Edit: OK, OCCT finished sucessfully, max temp was 67C on CPUTIN. CoreTemp max was 42C.

 

[sequoia464]

Those temps do seem high, I haven't used a six core Thuban yet so I don't know if this comparison means anything - but I'm running a FX6100 at 4200 with Turbo enabled at 4700 - voltage set to 1.38 and I don't exceed 59° running prime.

That is just with a Corsair H60 cooler - not the best cooling from what I have been reading. I think your OCZ Vendetta should cool close to the same as the H60 - fan dependent.

 

[Rhoxed]

are you lowering north bridge speeds while bumping the FSB? A high OC on the NB will result in much higher processor temps.

 

[VirtualLarry]

are you lowering north bridge speeds while bumping the FSB? A high OC on the NB will result in much higher processor temps.

I've been keeping NB and HTT speeds around 2000, which is I think the standard speed, right?

I was interested in overclocking the NB in the future, once I get a handle on the rest, but if the CPU is already running hot, I don't think I will.

Edit: Wow, even with LLC set to "DISABLED", my vcore under load is increasing from 1.35v (BIOS set) to 1.37v in OCCT's monitoring. (Edit: 1.39v) (Edit: 1.4v !)

Edit: Trying 282 FSB / 3.8Ghz. First try was at 1.3v BIOS, which got as high as 1.34v in OCCT. But Linpack crashed.

NB and HTT is 2000 or lower.

DRAM is 9-9-9-24, DDR3-1500, 1.5v. (Rated for 1600)

Edit: Stable so far at 3.8Ghz / 282 FSB, 1.375v BIOS (1.42v under load OCCT), temp CPUTIN is 73C, core temps are 48C.
Nevermind, BSOD.

Decided not to go for 3.8, now trying FSB 260 / 3.51Ghz, 1.3v BIOS, LLC Enabled.

Edit: It passed an hour of OCCT 4:64-bit linpack (90%) at 1.3v, LLC enabled.
Interesting that it will do that, but won't pass 3.8Ghz even at 1.375v.

Edit: That was with DRAM set to 13xx. Now trying with DRAM set to 17xx, 1.5v.
BSOD. Now trying 1.55v for DRAM.

Edit: Thubans are specced for 1866 DRAM speeds, aren't they?

Edit: BSOD, at 17xx and 1.55v.

 

[Rhoxed]

I've been keeping NB and HTT speeds around 2000, which is I think the standard speed, right?

I was interested in overclocking the NB in the future, once I get a handle on the rest, but if the CPU is already running hot, I don't think I will.

Edit: Wow, even with LLC set to "DISABLED", my vcore under load is increasing from 1.35v (BIOS set) to 1.37v in OCCT's monitoring. (Edit: 1.39v) (Edit: 1.4v !)

Edit: Trying 282 FSB / 3.8Ghz. First try was at 1.3v BIOS, which got as high as 1.34v in OCCT. But Linpack crashed.

NB and HTT is 2000 or lower.

DRAM is 9-9-9-24, DDR3-1500, 1.5v. (Rated for 1600)

Edit: Stable so far at 3.8Ghz / 282 FSB, 1.375v BIOS (1.42v under load OCCT), temp CPUTIN is 73C, core temps are 48C.
Nevermind, BSOD.

Decided not to go for 3.8, now trying FSB 260 / 3.51Ghz, 1.3v BIOS, LLC Enabled.

Edit: It passed an hour of OCCT 4:64-bit linpack (90%) at 1.3v, LLC enabled.
Interesting that it will do that, but won't pass 3.8Ghz even at 1.375v.

Edit: That was with DRAM set to 13xx. Now trying with DRAM set to 17xx, 1.5v.
BSOD. Now trying 1.55v for DRAM.

Edit: Thubans are specced for 1866 DRAM speeds, aren't they?

Edit: BSOD, at 17xx and 1.55v.

FWIW, I need 1.35v for anything above 3.4ghz, 1.375v for 3.6, 1.4v for 3.8, 1.45v for 4ghz, and 1.475+ for 4.2

I'm using a xigmatek thors hammer with push-pull 110cfm fans - max temp is 65C @ 4.2ghz settings

 

[Magic Carpet]

Edit: Thubans are specced for 1866 DRAM speeds, aren't they?

Natively? No, it depends on motherboard otherwise.

Edit: BSOD, at 17xx and 1.55v.

You're not doing too badly, overclocking-wise, you just need to fine-tune it and find the perfect settings. In my experience, they are unique to every motherboard. I am pretty sure, your chip can do 4.0 if you get the temps down. In my experience, the higher the temp... the less stable overclock is. I wouldn't go above 1.5v, unless you can tame your temps.

2) Make sure your ram is capable. I had sticks that wouldn't do anything 1666+ stably. Bump some voltage there as well. HT @ 2.4 is perfect for me. Again, it greatly varies from one mobo to the next, so take everything you read with a grain of salt.

3) Use good TIM, e.g. : MX-2, MX-4, NT-H1 or metal.

4) At 4.0 Thuban consumes heaps of power so not sure how efficient the cooler of yours is really is. Too much TIM, or lack of it.. can add up to the higher temps as well.

Ah hell, you know this yourself anyway, just more practice you need with your new toy 😀

 

[VirtualLarry]

Thanks for the suggestions, you guys. I don't think I'm going to go for 4.0, unless I get a different, better cooler.

I'm still debating if I screwed things up, by setting the CPU fan up for pull, rather than push.
My buddy who does AC/heating stuff, says that I should change it.

I dunno, when I though I had the TIM screwed up, because I didn't get it on all of the heatpipes, well, when I fixed that, it seemed equal or worse.
http://forums.anandtech.com/showthread.php?t=2223081

It would be nice to hit 3.8, with 1700 DRAM, and 2.4 NB, but that seems out of reach too, unless I bump the volts quite a bit. 73C at 3.8, with 1.375v, was kinda scary temp-wise, esp. since it didn't have enough volts to be stable.

I'm still a bit in the dark about which temp is the real temp. OCCT displays "CPUTIN" which I assume is BIOS temp, and that's about 23C higher than the reported core temps.

Normally, I go with the core temps not the BIOS temp, but in this case, my core temps are 6C at idle, which is like 18C lower than ambient temps, so that's impossible.

So if I add 20-odd C to the core temps, that gets me something around what CPUTIN is reporting. So I've been going with the CPUTIN temps.

 

[LOL_Wut_Axel]

Thanks for the suggestions, you guys. I don't think I'm going to go for 4.0, unless I get a different, better cooler.

I'm still debating if I screwed things up, by setting the CPU fan up for pull, rather than push.
My buddy who does AC/heating stuff, says that I should change it.

I dunno, when I though I had the TIM screwed up, because I didn't get it on all of the heatpipes, well, when I fixed that, it seemed equal or worse.
http://forums.anandtech.com/showthread.php?t=2223081

It would be nice to hit 3.8, with 1700 DRAM, and 2.4 NB, but that seems out of reach too, unless I bump the volts quite a bit. 73C at 3.8, with 1.375v, was kinda scary temp-wise, esp. since it didn't have enough volts to be stable.

I'm still a bit in the dark about which temp is the real temp. OCCT displays "CPUTIN" which I assume is BIOS temp, and that's about 23C higher than the reported core temps.

Normally, I go with the core temps not the BIOS temp, but in this case, my core temps are 6C at idle, which is like 18C lower than ambient temps, so that's impossible.

So if I add 20-odd C to the core temps, that gets me something around what CPUTIN is reporting. So I've been going with the CPUTIN temps.

When you can afford it, get a Corsair A50 or a CM Hyper 212+. You can easily get each for $25 on sale, but the A50 is a bit better when it comes to temps, though either of them will be vastly superior to your current OCZ Vendetta and will bring down temps by around 10C. Since you didn't do it correctly before, remember to do proper TIM application. Proper seating of the heatsink is important, too.

 

[VirtualLarry]

I bought a 212+ for $30 at MC when I picked up the new mobo and CPU.

I just didn't want to remove the factory heatsink backplate. I have fears about scratching the PCB around the CPU heatsink, and shorting something out from the back side.

Plus, although the CM 212+ is a 120mm fan cooler, the overall heft is about the same between the vendetta and the CM. The Vendetta has probably more surface area, since the plates on that stick out farther. Although that also means that you cannot configure it for push/pull.

 

[LOL_Wut_Axel]

Err, no, the 212 Plus has a much larger surface area. Could you listen for once? Replace the Vendetta with the 212 Plus and you'll see much improved results.

 

[Sunny129]

Larry, sorry i couldn't chime in sooner...

unfortunately i have no idea what CPUTIN is, but i've been using Core Temp for quite some time and its never done me wrong...and your Core Temp readings look just fine to me (42°C @ 3.5GHz & 48°C @ 3.8GHz, under load i presume) for a Thuban chip. to give you some perspective, in one of my DC rigs, the push/pull fans on my CM 212 EVO spin @ ~1000rpm, and core temps never exceed 50°C under 100% load 24/7 (1090T @ 3.7GHz). in my other DC rig, the push/pull fans on my CM 212+ spin @ ~800rpm, and core temps never exceed 44°C under 100% load 24/7 (1090T @ factory 3.2GHz). could i do better temp-wise on the OC'ed rig? sure, if i cranked up my CPU fans...but i like my silence (its in the master bedroom), and even under full load the temps are doing just fine.

if you insist on pushing 4GHz and beyond, indeed switch to the CM 212+ and use 2 fans in push/pull, still exhausting out the top...or, if you're not opposed to the sound of leaf blowers, just crank the RPM on your Vendetta's fan (or replace it with a high CFM fan). but honestly, you really should just use the CM 212+ that's already in your possession. as Axel said, the improvements will be tangible. plus, 2 fans at a lower RPM will move the same air a single fan will move at a higher RPM - you just might find that the RPMs needed to keep your 1045T cool @ 4GHz results in a noise level that is perfectly tolerable...just remember to invest in some quality fans. i'm using Scythe SY1225SL12LM-P 120mm fans on my 212's.

 

[VirtualLarry]

The thing is, I feel like I cannot trust CoreTemp.

I have two 1045T hex-cores.

One is in a K9A2 Platinum AM2+ board, along with four 9600GSO graphics cards.

I stopped BOINC on that machine, and I'm going to see how low CoreTemp reads on that rig.

On my new rig, with the 990FX Extreme4, CoreTemp reads 4C at idle, at 3.51Ghz / 1.3v LLC on. (It just dropped to 2C, actually.)

2C is way below ambients. It's 69-70F in the room here.

So if CoreTemp is 20C too low, then that means that in the new machine, core temps get to 62C, with the OCZ Vendetta.

In the K9A2 Plat. board, I'm using a stock AMD 4-heatpipe heatsink, and it gets as hot as 45C, in an Antec 300 full of TriCool fans on high.

Edit: Ok, after some minutes to cool down, the K9A2 board is showing 18C for idle temps. But for some reason, even without much running (CoreTemp, Precision), TurboCore is still running, some cores are showing at 800Mhz, some at 2700, and VID is fluctuating between 1.35v and 1.425v, even at idle.

So that's a 15C gradient between idle temps on both rigs.

Edit: I just checked, 18C is 64F. So even that idle temp is wrong.

 

[LOL_Wut_Axel]

The thing is, I feel like I cannot trust CoreTemp.

I have two 1045T hex-cores.

One is in a K9A2 Platinum AM2+ board, along with four 9600GSO graphics cards.

I stopped BOINC on that machine, and I'm going to see how low CoreTemp reads on that rig.

On my new rig, with the 990FX Extreme4, CoreTemp reads 4C at idle, at 3.51Ghz / 1.3v LLC on. (It just dropped to 2C, actually.)

2C is way below ambients. It's 69-70F in the room here.

So if CoreTemp is 20C too low, then that means that in the new machine, core temps get to 62C, with the OCZ Vendetta.

In the K9A2 Plat. board, I'm using a stock AMD 4-heatpipe heatsink, and it gets as hot as 45C, in an Antec 300 full of TriCool fans on high.

Edit: Ok, after some minutes to cool down, the K9A2 board is showing 18C for idle temps. But for some reason, even without much running (CoreTemp, Precision), TurboCore is still running, some cores are showing at 800Mhz, some at 2700, and VID is fluctuating between 1.35v and 1.425v, even at idle.

So that's a 15C gradient between idle temps on both rigs.

What about HWMonitor?

 

[VirtualLarry]

HWMonitor on the 990FX rig, at idle:

Temp
SYSTIN 28C
CPUTIN 26C
AUXTIN 128C

Fans
CPUFANIN 2660 RPM

Core Temps
(six listings) 3-4C

Power
PACKAGE 60.00W

HD WDxxxx
Temp 26C

HWMonitor on the K9A2 Plat. rig, at idle:

Temp
TEMPIN0 20C
TEMPIN1 22C

Core Temps
(six) 18-19C

Power
PACKAGE 111.10W

HD WDxxx
Temp 22C

 

[pelov]

The Core temps is an issue with a lot of the 45nm AMD chips where they give the wrong reading. My 955 reads 0C and has been that way out of the box on day one. The Thubans have the same issue.

The CPUTIN gives you an estimate of your core temps. Your idle temps seem fine.

Also want to chime in and say that the Hyper212+ is an awesome cooler especially in a dual fan setup. Just make sure you get the fans in the right orientation, because I made the mistake of being lazy and they were cooling themselves instead of having a nice breeze sweep across the CPU 😛 I wondered for a week why my temps never went down...

 

[daveybrat]

Also have to chime in and say that my Hyper 212+ is absolutely fantastic. One of the best coolers i've used in a long time. And don't worry about scratching the back of the motherboard, it won't happen. There is a protective plastic surface on the mount that touches the back of the motherboard. It was actually a fairly easy install and one that i should have done a long time ago.