Good temps for OC on i5 3570k

[Zardnok]

okay, so i've currently decided to push my CPU even further....

Now im running at 5.2gHz and 1.37v

Idle temp 30c

Load temp using IBT and Prime 95 is around 75-80c

Does this sound decent guys ?

Can you post your bios settings and a CPUz/CoreTemp/Stress Test screenshot? You are far beyond what I was able to get and am curious what I need to change to get closer.

 

[jellliston]

Can you post your bios settings and a CPUz/CoreTemp/Stress Test screenshot? You are far beyond what I was able to get and am curious what I need to change to get closer.

I'm Currently at work. but when i get some spare time, ofcourse id be delighted to post them

 

[jellliston]

Ambient? Fanspeed? Can you do 5 runs Linx 0.6.4 AVX enabled and check GFlops too?

So far seems like a very good chip to me, actually first time I saw 3570K pass 5Ghz I think. Mine hit 98c at 4.6 GHz 1.3V using HR-02 with TY-140 1300rpm fan at 20c ambient.

Anyway, for stresstesting anything under 105c is fine I guess.

Edit: noticed you got 16GB ram. I used 6144MB for Linx testing.

Ambient is 16c Fan speeds are 2500rpm (on both Corsair fans, performance mode on the H100) the two motherboard fans are running at 6200rpm (surprisingly quiet) when i get time, i'll post all the results

 

[coffeejunkee]

You live in a cool place...

The reason I ask about Linx GFlops is to see if there isn't any throttling going on.

 

[jellliston]

only ran an hour test as thats all i had time for but there ya go BIOS n stuff when i get round to it

 

[jellliston]

You live in a cool place...

The reason I ask about Linx GFlops is to see if there isn't any throttling going on.

What would be expected in the Gflops ?

Never used Linx before and im too tired to work it out

 

[BonzaiDuck]

What would be expected in the Gflops ?

Never used Linx before and im too tired to work it out

It need to be done with "deliberation."

You would download and install (if it doesn't just run from an .exe file) the latest LinX with the AVX extensions.

In Windows, open Task Manager. Go to Performance and open Resource Monitor, and select CPU. Note which theads ("CPU0" to "CPU7" with HyperThreading) are active, and which are parked.

Start LinX, set problem size to ~41,000 and Memory to "All." In "Settings," change "Threads" to 4. In Task Manager, select "Processes" and find "LinX32." Right-click, select "Affinity," unclick "all" and select four threads which are unparked, but ideally -- CPU0, CPU2, CPU4, and CPU6.

If the parking arrangement changes while then running LinX, in Task Man -> Processes right-clicke the 64-bit extension of the program and change the select so that four threads are active. You might do the same to LinX32, for the same four threads.

Run 10 iterations or so and watch the GFLOPS. Go to the "sticky" on stress-testing and performance-tuning, and read the exchange in the last few posts between IDontCare and BonzaiDuck.

 

[MustangSVT]

wow maybe you have one where they soldered the cpu instead of thermal paste

 

[bl00tdi]

I would use real temp to monitor your temperatures. Thermal radar is a decent ballpark estimate of where they are but is far from accurate.

As far the Intel PTPP, there's no guarantee that the RMA cpu will be as good of an overclocker. Golden chip is gold.

 

[DigitalWolf]

I would use real temp to monitor your temperatures. Thermal radar is a decent ballpark estimate of where they are but is far from accurate.

As far the Intel PTPP, there's no guarantee that the RMA cpu will be as good of an overclocker. Golden chip is gold.

I would imagine if they look at those temps with Real Temp it will display quite a bit higher. Tho I honestly haven't been able to understand why a piece of free software is so much more "accurate", than the software a company marketing to a performance crowd gives their customers.


I know with my 3570K what Asus's monitoring software displays under load.. is much lower than Real Temp. Perhaps this is a ymmv thing tho.

 

[coffeejunkee]

What would be expected in the Gflops ?

Never used Linx before and im too tired to work it out

Well, I think something like 140 GFlops at 5.2GHz. But you should compare with other frequencies so you can see a gradual increase and notice any lower than expected GFlops.

But I assume you have all power limits at max so might not be too relevant. Still I'm curious.

I would use real temp to monitor your temperatures. Thermal radar is a decent ballpark estimate of where they are but is far from accurate.

As far the Intel PTPP, there's no guarantee that the RMA cpu will be as good of an overclocker. Golden chip is gold.

I would imagine if they look at those temps with Real Temp it will display quite a bit higher. Tho I honestly haven't been able to understand why a piece of free software is so much more "accurate", than the software a company marketing to a performance crowd gives their customers.


I know with my 3570K what Asus's monitoring software displays under load.. is much lower than Real Temp. Perhaps this is a ymmv thing tho.

I'm thinking the Asus software displays overall cpu temp and not core temp. Also, Linx stressing gives higher temps than Prime. Nonetheless it's remarkable to see this cpu working at 5.2GHz.

 

[BonzaiDuck]

I'm thinking the Asus software displays overall cpu temp and not core temp. Also, Linx stressing gives higher temps than Prime. Nonetheless it's remarkable to see this cpu working at 5.2GHz.

There've been some "skirmishes" about this -- and I was somehow in the middle of it. Almost makes me jittery to mention the old "spec" at issue.

The Intel TCASE temperature spec -- meant to be measured at top-center of the IHS -- addressed case ventilation and cooling most obviously. But it is always hinged to the core (TJunction) sensor temperatures, and over a few generations of Intel processors (or at least three generations back) had been pegged at ~10C below the core average.

ASUS revised their Suite monitoring software, or so it appeared when I updated. It looks very much as though ASUS has attempted to provide an approximation or estimation of a "TCASE." Otherwise, I can't explain how the monitor CPU temperature value is always about ~10C less than what HWMonitor, RealTemp, CoreTemp or Aida64 show as my core temperatures.

Of course, instead of telling me outright that I'm nuts, we can just call it a "coincidence."

 

[jellliston]

Bit of a rant...

posted my spec on the WoW forums.

the number of people that are basically saying i will break my CPU within a week, and that even attempting 5.2ghz was a stupid idea really P**ses me off.. /rant

anyhow, I used RealTemp and Speedfan..

the temps for 5.2Ghz were at 83C Average, so the Thermal radar wasn't too far off

 

[TJCS]

Bit of a rant...

posted my spec on the WoW forums.

the number of people that are basically saying i will break my CPU within a week, and that even attempting 5.2ghz was a stupid idea really P**ses me off.. /rant

anyhow, I used RealTemp and Speedfan..

the temps for 5.2Ghz were at 83C Average, so the Thermal radar wasn't too far off

If you acknowledge the risks of going beyond intel's limit, then why give a rats ass about what they think. It's your money and you are the one with the 5.2ghz chip. Go enjoy the chip, while you still can... LOL

Here:

BTW, ever got around to doing the 10-pass LINX or IBT? Want to see what your GFLOPS looks like.