high cpu temperature during 100% load causing in game stutter ? Help required

[drumatticjim]

Hi guys, I am experiencing in game stutter during all games on my new build. Have done some testing and discovered the following (all software, os, bios and drivers are updated. Using monitoring tools memtest86 (4 passes no errors), prime95, cpu-z, realtemp, msi afterburner, furmark, LatencyMon). The i7 4790k is set to stock clock of 4ghz and the ram is set to xmp profile of 1866mhz.....

FURMARK @ 2560x1440 (2x MSAA)
note: It was mostly butter smooth with the odd micro stutter for a minute or so during the ten min test.
FPS: 39 (min 31, max 41)
GPU1: 69 c (min 47 c - max 70 c)
core: 1379 MHz / 70 c / 99%
mem: 3505 MHz / 10%
GPU Power: 74.8% TDP,fan: 51%

-------------------------------------------
CoreTemp standalone BEFORE using Prime95

frequency 4400.00 MHz
processor #0
Core #0 = 24 c (Min 22 max 56) load 0%
Core #1 = 22 c (Min 19 max 59) load 0%
Core #2 = 20 c (Min 19 max 57) load 0%
Core #3 = 25 c (Min 23 max 54) load 0%

------------------------------------------
CoreTemp with Prime95

frequency 4400.00 MHz
processor #0
power: 170 watts
Core #0 = 95 c (min 22 max 98) load 100%
Core #1 = 99 c (min 19 max 100) load 100%
Core #2 = 98 c (min 19 max 100) load 100%
Core #3 = 90 c (min 23 max 96) load 100%

-----------------------------------------

The GPU temp from the test seems ok? They also seem ok when monitoring via msi afterburner. However, the CPU result with prime95 and coretemp seem really high? Could this be the cause of the problem? What are the possible solutions? Do I need to reapply thermal paste between the cpu and h100i cooler? Also when using RealTemp under no load the thermal status for the cores reads: LOG LOG LOG OK
however when running alongside prime95 it would often read: LOG HOT HOT OK.

Many thanks in advance for the replies.

system:
i7 4790k
16gb 1866mhz cas 9 ram
120gb samsung evo 850 ssd
seagate 2tb 7200rpm hdd
msi gaming 5 z97 mobo
gigabyte gtx 980 g1 gaming windforce edition
corsair h100i
evga g2 750w psu
corsair k70 rgb keyboard
corsair h2100 headphones
trust gxt 152
asus rog swift pg278q

 

[drumatticjim]

12.88 is far too high. BIOS may also be wrong.

The expected tolerance for a decent +12v rail should be +/- 0.10 or 11.9 to 12.1, or more preferably +/- 0.05 or 11.95 to 12.05.

My +12v rail is 11.98 under load, and 12.01 at idle, which is normal and expected, and is well within +/- 0.05 tolerance.

The only way to know for certain is to check it with a meter ASAP.

Most meters have a selection to measure inside a 20VDC range. Make sure that the meter you purchase has enough decimal places, such as 20.00, or better yet, 20.000. You need to read the specs to be sure.

In the world of multimeters and electronics test equipment, just as with most other gadgets, you get pretty much what you pay for.

oh my bad, I meant to type 12.288 not 12.88. Still, this falls outside of the 11.9 - 12.1 range you mentioned in the previous post so I'm going to get a meter anyway to try it. I'm assuming the bios readout is more accurate than the software one?

 

[CompuTronix]

You posted just before I finished editing. Please re-read my above post.

 

[drumatticjim]

You posted just before I finished editing. Please re-read my above post.

Ah I see, so I have! Many thanks, I will keep you posted with the results

 

[drumatticjim]

12.88 is far too high. BIOS may also be wrong.

The expected tolerance for a decent +12v rail should be +/- 0.10 or 11.9 to 12.1, or more preferably +/- 0.05 or 11.95 to 12.05.

My +12v rail is 11.98 under load, and 12.01 at idle, which is normal and expected, and is well within +/- 0.05 tolerance.

The only way to know for certain is to check it with a meter ASAP.

Most meters have a selection to measure inside a 20VDC range. Make sure that the meter you purchase has enough decimal places, such as 20.00, or better yet, 20.000. You need to read the specs to be sure.

In the world of multimeters and electronics test equipment, just as with most other gadgets, you get pretty much what you pay for.

Hi CT,

So I have purchased a digital multimeter and tested the 12v from the 24pin mobo connector and from the 8pin cpu connector. The readout is much closer to what it should be than the readout from both the bios and OCCT software.
The readout is 12.16v (flickering between that and 12.17v) on both connectors.

Any thoughts? Is that still too far away as I have been reading a few posts and this seems to fall within the 11.6 to 12.4 limits that a few people seem to be quoting?

Cheers, James

 

[CompuTronix]

Well done! 😀

Your measurement is within a 1.5% tolerance, which is OK, however, is that at idle, or under heavy load running the Power Supply test in OCCT?

While you're at it, you might also want to measure the +5v and +3.3v rails, but at least now you know what your +12v rail is actually supplying to your system.

You must feel at least a little bit relieved.

It's a shame that MSI's Gaming 5 motherboard is so inaccurate measuring and reporting the +12v rail. Owners shouldn't have to purchase a multimeter to verify the voltages.

So to summarize:

(1) Your Core temperature benchmark (Prime95 v26.6 Small FFT's) is borderline but OK at 80C with ambient temperatures of 20 to 22C.

(2) As tested with a multimeter, your +12v rail is within a 1.5% tolerance, but tests must include idle and load (OCCT Power Supply Test) values.

(3) Frame stuttering problem remains, which questions nVidia drivers. 900 series owners report the issue as GPU usage spikes below 90%, but also occurs on your rig with a GTX 670.

Accurate so far?

CT

 

[drumatticjim]

Well done! 😀

Your measurement is within a 1.5% tolerance, which is OK, however, is that at idle, or under heavy load running the Power Supply test in OCCT?

While you're at it, you might also want to measure the +5v and +3.3v rails, but at least now you know what your +12v rail is actually supplying to your system.

You must feel at least a little bit relieved.

It's a shame that MSI's Gaming 5 motherboard is so inaccurate measuring and reporting the +12v rail. Owners shouldn't have to purchase a multimeter to verify the voltages.

So to summarize:

(1) Your Core temperature benchmark (Prime95 v26.6 Small FFT's) is borderline but OK at 80C with ambient temperatures of 20 to 22C.

(2) As tested with a multimeter, your +12v rail is within a 1.5% tolerance, but tests must include idle and load (OCCT Power Supply Test) values.

(3) Frame stuttering problem remains, which questions nVidia drivers. 900 series owners report the issue as GPU usage spikes below 90%, but also occurs on your rig with a GTX 670.

Accurate so far?

CT

Hi, thanks for the swift response

The measurement was at idle with only the 24pin connector plugged into the psu. I can check at load but I have a question, which 12v connector do I need to check using the multimeter? As surely the 24pin needs to be connected into the motherboard so that I can get into windows to run OCCT? Or am I ok to test it on one of the 'perif' cables which wouldn't be so vital to booting into OS?

I also checked the +5v and +3.3v and I think they were 5.05v and 3.31v (I say 'think' as I was more interested in the 12v reading at the time, so I can always go back and re-test these if needed).

I also swapped over the x2 8pin power connectors to the GPU with the spares that came with the evga psu but still the stuttering remains.

In regards to the summary, yes, points 1, 2 and 3 are indeed correct
In relation to the stutter, it still puzzles me how when cut scenes are playing, or if i'm running a benchmark test the graphics run super smooth, yet the stuttering only occurs when I'm actually trying to play the game.

Cheers,
James

 

[CompuTronix]

Since your PSU has a single +12v rail, it really doesn't matter which connector you probe, as they're electrically the same test point.

Concerning the frame stutter problem, here's a few thoughts:

(A) Temporarily downclock the CPU to it's base clock of 4.0 on all cores with Turbo Boost disabled, and restore all other settings to default / auto values.

(B) Temporarily downclock the memory to 1600MHz, which is the maximum memory clock supported by your processor's onboard memory controller.

(C) Uninstall the current nVidia driver, then install the driver which came with your GTX 980 from Gigabyte on their Driver DVD.

CT

 

[drumatticjim]

Haven't tried the above just yet, however on a side note, I have removed the H100i which was using the TIM already supplied with it. I noticed that only around 2/3 of the TIM looked like it was making proper contact with the CPU (left third looked like it hadn't made proper contact in comparison to the rest of the TIM). So I rechecked the standoffs which connect the H100i to the mobo and discovered the top left standoff was nowhere near as tight as the others (which wasn't apparent during installation). So I tightened this and applied some Arctic Cooling MX-4 TIM using the line method and slightly off centre of the CPU (to theoretically compensate for any slight imperfections in the flatness of the cpu and/or h100i block).

I have now turned the PC back on, run Prime95 v26.6 for 10mins on small ffts and I seemed to have improved the temps by almost 20c! Instead of the max core temp hitting 85c it is now topping out at 68c so I'm much happier about that! Now, onto fixing this stuttering..... lol!

UPDATE: Tested Prime95 with Turbo disable (cpu @ 4ghz) & RAM @ 1333mhz (Stock - XMP disabled) and temps max out between 53c and 57c across the four cores.

 

[drumatticjim]

Since your PSU has a single +12v rail, it really doesn't matter which connector you probe, as they're electrically the same test point.

Concerning the frame stutter problem, here's a few thoughts:

(A) Temporarily downclock the CPU to it's base clock of 4.0 on all cores with Turbo Boost disabled, and restore all other settings to default / auto values.

(B) Temporarily downclock the memory to 1600MHz, which is the maximum memory clock supported by your processor's onboard memory controller.

(C) Uninstall the current nVidia driver, then install the driver which came with your GTX 980 from Gigabyte on their Driver DVD.

CT

Hi CT,

I have tried the three things above and the problem still persists. Tried the base clock on the cpu, the ram, OC Genie profile and XMP profiles and a combination of each. Have also tried the graphics from the cd (after having used Display Driver Uninstaller to get a clean install).

Is there anything in the polling rates of the mouse? I've read somewhere that windows 8.1 has a low dpi, could this be causing conflict seeing as on quite a few games the game is choppy when I move with a mouse yet is super smooth when moving with the keyboard?

Cheers,
James

 

[CompuTronix]

I apologize, but I have no further recommendations, as I'm fresh out of ideas for this particular problem. Perhaps if you post a new thread in the Graphics & Displays Forum, someone familiar with this problem might be better able to help you.

CT

 

[drumatticjim]

Cheers, I'll give that a try. Many thanks for all your help Although it didn't solve my problem, I'll pick the solution as that article was a great read and helped solve the incorrect cpu temp readings and rule out certain potential problems from this troubleshooting I'm having to do