3570k - 4.4ghz Temps
3570k - 4.4ghz Temps
CompuTronix: geofelt: Prime95 isn't a true benchmark, nor should I rely on IBT. Consider trying OCCT instead, which employs more typical instructions. MeanMachine41: Prime95 is a demanding synthetic tool that pushes your system to its limits. Download the AIDA64 trial version and pair it with HWMonitor. Arrange them side by side on your desktop. Guys, Concerning Prime95, general statements don't hold up. The real concern lies in the VERSION specifics. It's accurate to say that running later versions than 26.6 isn't advised because they overload the CPU with AVX instructions. For accuracy, note that pre-AVX version 26.6 is perfectly fine. Version 26.6 Small FFTs is ideal for thermal testing as it provides a consistent workload and closely matches Intel's Thermal Design Power (TDP) values. You can find more details in this Sticky... you might want to read it: Intel Temperature Guide - CT Thank you for the info on P95. I get what you mean about VERSION specifics, but I know Prime95 is mainly a synthetic tester. Without constant temperature checks, it can cause issues. Through experience, I've used P95 many times. Still, it doesn't evaluate all subsystems well and I wish it tested GPUs too. Please clarify if my view is incorrect—I've found real-world tools like Realbench more suitable and less intense. On numerous stress tests, passing the benchmark means stability across all areas. I also combine P95 with AIDA64 and HWMonitor for better system insights during OC. For initial stability, I used P95 for 10 minutes on a small FFT and longer sessions, but my system crashed quickly when running demanding games, even though P95 said it was stable. So in short, it may not be as reliable as some claim. It seems opinion-based, and I'll try to review the link's content eventually. At the moment, it's a bit TLDR. I hope we can reach a common understanding.
I just finished reading the reasons behind why any version post 26.6 causes my temps to spike really high, thanks to you. Now using 26.6 I’m hitting 70-75°C max, down from about 85°C before. That’s a 10°C drop overall, but it’s still around 72°C which is bad for 1.24v. Also, it’s odd that my OC has no WHEA errors and feels more stable with LLC 2 (75%) instead of LLC 3 (50%), which tends to have those errors.
cameronmc88
,
You haven't mentioned, and no one has thought to ask; what is your ambient temperature? All computer temperatures rise and fall with ambient, which plays a major role in determining thermal performance.
What was you ambient temperature when you ran Prime95 Version 26.6 Small FFT's?
Guys
,
For those who haven't read the Temp Guide just yet, here's a few details:
There's a difference between
stability
testing and
thermal
testing. For
thermal
testing, the goal is to replicate Intel's TDP values as closely as possible, so as to establish a
valid
thermal
baseline
. This is explained in Section 13.
While I agree there are several utilities superior to Prime95 for system
stability
testing such as Asus RealBench, there are none more ideal for CPU
thermal
testing than Prime95 Version 26.6 Small FFT's, because it's a
steady-state
100% workload which yields
steady-state
Core temperatures, and is as close as you can get to replicating Intel's Thermal Design Power (TDP) values within 3%.
This means that after your CPU warms up and stabilizes about 4 to 6 minutes into the test, the Core temperatures are
very steady
, which is critical to analyzing thermal performance. How can you determine thermal performance if you run a utility with fluctuating workloads which causes fluctuating Core temperatures that look like a bad day on the stock market?
Intel's TDP values are a calculation AND a measurement. In order to achieve accurate results which agree between calculated and measured values, Intel must use a
steady-state
100% workload. Further, 100%
usage
doesn't necessarily equate to 100%
workload
.
AIDA64 is OK for
stability
testing, but it's not
steady-state
, and is confusing due to having too many test combinations. AIDA64's only
steady-state
test is "Stress CPU" when selected individually. Even though it shows 100%
usage
, it's far below TDP and 100%
workload
, which gives the user a false sense of security because of misleadingly low temperatures. Conversely, the "Stress FPU" test when selected individually, may instantly slam Core temperatures to Throttle at TjMax, way above TDP, so that doesn't work for
thermal
testing either.
OCCT? Which test? CPU: Linpack? With or without AVX? Neither is
steady-state
, nor are they close to TDP. CPU: OCCT? Notice the selection for "Small Data Set"? Sound familiar? Maybe like Prime95 "Small FFT's"? Core temperatures similar to AVX Version 27.7?
Steady-state
but too far above TDP. And OCCT already includes HWMonitor.
Note
: It's never a good idea to run more than one monitoring utility concurrently, because they may interfere with the accuracy of one another.
If you're going to talk about testing and specifications, then you need to be
very
specific. I've tested every utility, including several most of you have never heard of, and I can tell you without any doubts, that Prime95 Version 26.6 is the best utility for
thermal
testing. If you want to get a
valid thermal baseline
, then that's the way to go. Run it with Core Temp or Real Temp, which was developed specifically for Intel processors.
CT
72 at 1.24v isn't bad at all. My 3570k runs at 4.3GHz and with p95 26.6 small fft at 1/2hr on my h55 I reach 70°C at 1.16v. What catches my attention is that you actually increased the voltage. The 3570k begins with a solid stock vcore, which usually handles 4.3GHz just fine with just a boost multiplier. Once you turn off certain features and add an LLC filter, removing auto settings, you should be able to lower the vcore to 4.4GHz. I set mine at 1.19v without using any offset. For some reason, 4.3GHz appears to be the upper limit for auto settings on the 3570k.
Not sure about the interior temperature, but outside it's 24°C.
@Karadjgne I'm setting 1.24v since lower voltages cause WHEA errors; I need full stability. Also, I've been using Fixed instead of offset.
What are you suggesting to disable auto? I'm using manual voltages for vcore and ram.
Your temperatures are a bit elevated at 82C, Cameron. To keep the system running smoothly at 3.4GHz with a need of 1.248V, you’ll need to upgrade your cooling solution. A Corsair H100 series or similar all-in-one water cooler would be suitable if your case accommodates one. The only option to lower temps without compromising performance is to reduce core voltage, which might mean lowering your overclocking. Optimal temperatures should be 10-15°C above ambient during idle and 60-65°C under load. The maximum junction temperature on your chip is 105°C before throttling occurs, so staying away from that is crucial. Sustained temperatures around 80°C under load could harm the CPU over time.
cameronmc88 :
In your screenshot were you running Prime95 V26.6 Small FFT's?
Here's the normal operating range for Intel Core temperatures:
80C
Hot
(100% Load)
75C
Warm
70C
Warm
(Heavy Load)
60C
Norm
50C
Norm
(Medium Load)
40C
Norm
30C
Cool
(Idle)
Core temperatures in the mid 70's are safe
, so just keep it under 80.
Except for a hot peak on Core #2, it looks like you're OK, but a little marginal on your Phantek PH-TC12DX, which is a mid-range air cooler.
As
MeanMachine41
has pointed out, if you want to maintain 4.4 and still have some thermal ceiling to accommodate higher ambient temperatures, then you might want to consider upgrading your cooling.
However, keep in mind that unless you run apps that use AVX for CPU intensive tasks such as rendering and transcoding, it's unlikely you'll ever see these temperatures, nor will you when gaming, even with your current configuration.
CT
While playing games, the CPU isn't reaching even 60c despite using the heater, so it's probably because of running Prime95 V26.6 Small FFT. Should I replace the thermal paste?
"Maintaining temperatures around 80c during heavy use would eventually harm your CPU."
I'm planning to upgrade to a new Intel line next year, so if it fails within six months, it's not a big deal.
If you're not going beyond around 60, don't interfere with the cooler. The main goal of testing with p95 is to test the CPU up to its maximum. Because your limit is in the low 70s, you're fine. Since games usually don't hit 100%, any gaming you do will stay below your true capacity.