Arksun :
Yeah but here's the thing, in an earlier post dusktilldawn48 wrote: "so i reset all my bios to default and primed it, it turbos to 4.4 and after about 16 mins it jumps to 90C. " At stock speed/voltage/wattage, the 4790K should NOT be getting anywhere near 90C. If he's experience these kind of temps at stock speed/voltage with both his and his friends chip, to me that says the chip is not the problem. My 4790K with a Noctua NH-U14S cooler (single fan with voltage reduction adaptor to reduce fan speed) at stock speed with prime95 jumpes from about 33C to 54C with all cores 100%, then once prime95 starts pushing the wattage up by itself (a known issue with the current version of prime95) I find that once the wattage has gone from 88W to about 130W my core temps go up to 68C.
This is all still well below what dusktilldawn48 is seeing.
Hi Guys,
Interesting thread! I've been carefully reading through this, and unless I've missed it, I think something has been overlooked concerning Prime95, and the sudden increase in temperatures described above.
As I've seen in 95% of these temperature threads, nowhere over the course of all these posts has anyone asked or mentioned
which Prime95 test was being run. I think that "Blend" was run, and has caused a certain degree of confusion in this thread. If there's a problem with the current version, 28.5, I'm not aware of it.
Prime95's default test, Blend, is a memory stability test which has a cyclic workload with fluctuating temperatures. Large FFT's is a composite of Blend and Small FFT's. Both tests do indeed ramp up temperatures at approximately 16 minutes, when the test increments to less memory intensive iteration types. At 32 minutes it ramps temperatures back down, so Blend and Large FFT's both have a cycle time of about 32 minutes.
The first 16 minute segment of Blend is more memory intensive, so processor temperatures are lower. The second 16 minute segment is less memory intensive, so wattage and temperatures are higher. Shown below in the center of the Chart is the transition which occurs at approximately 16 minutes. The Chart is 13 minutes wide.
Shown below is Small FFT's, Blend, Linpack and Intel Burn Test.
Small FFT's is a steady-state 100% workload that does not fluctuate, nor is there a transition. Shown below is Small FFT's. The center of the Chart is approximately 16 minutes into the test.
SpeedFan Charts are auto scaling, so on the chart above you're looking at 1C increments. The dip in the middle of the chart is the cores following my ambient temperature down before the A/C compressor cycled off toward the right side of the chart.
If the chart above was scaled similar to the second chart, then the temperature would appear as a relatively smooth line. Here's what that looks like:
I just ran this on my 4770K which is an L3. It's de-lidded and lapped, so the temperatures are very close between cores compared to most of these processors.
Although this explanation doesn't solve
dusktildawn48's problem, I think you can appreciate that it's important to be specific when discussing Prime95 testing, so that we don't introduce another set of variables into the original problem.
Also, you guys might want to give this a read:
Intel Temperature Guide -
http://www.tomshardware.com/forum/id-1800828/intel-temperature-guide.html
Hope this helps to clear up at least part of the puzzle,
CT
Facebook
Twitter
Instagram