GTX970/1070 not boosting to full clock speeds.

MrHatty

Honorable
Aug 16, 2013
19
0
10,510
This has been an issue I've been trying to figure out myself for the past 4-6 months. Its strange how it comes and goes.

What's happening is my GPU(s) are not reaching their full clock speeds (on or off boost) while under load (in this case the Fire Strike benchmark). This started roughly back in August/September with my GTX970.

I bought this card when it was new and gave it a modest overclock out of the box. When issues started arising I noticed chugging frame rates in render intensive situations. I started monitoring clock speeds (first with ASUS GPU Tweak then MSI Afterburner) and noticed that my clock speeds were WELL below what I had put them at and had been stable with up to this point. Below is what they were running at compared to what they were supposed to be running at.

GPU Clock: 630Mhz (1390Mhz)
Memory Clock: 4200Mhz (7750Mhz)

I went straight into cleaning out the video drivers with DDRU and re-installing them fresh. This would temporarily solve the problem, but it would arise again within a weeks time. The 2nd time it happened, I re-seated the card along with removing the drivers. Same situation; after about a week the problem arose again. The 3rd time I moved the card to a different PCI-E slot, cleaned out the drivers, dusted the entire system out (earlier than I usually do as I dust my rig out every 2 weeks) and re-installed the drivers. Same scenario. GPU performance dropped after roughly a week.

At this point I figured that the GPU was crapping out. So around Christmas time I picked up an Asus 1070. Up until last night, this card has given me no issues. Last night however as I attempted to game, I noticed that games that normally run with no issues what so ever suddenly started chugging. I brought up GPU Tweak and saw once more that clock speeds were not hitting what they were supposed to be hitting. This time, only the GPU clock wasn't getting up. The memory clock performed exactly as intended.

GPU Clock: 734Mhz (~2000Mhz)
Memory Clock: 8000Mhz (8000Mhz)

The 1070 is NOT overclocked. At this point I'm eyeballing either power delivery or the mobo is on its way out. Below is my system specs:

Mainboard: MSI Z77 Mpower
CPU: Intel i7 3770k (overclocked to 4.2ghz)
GPU: Asus GTX 1070 Strix
Memory: Corsair Vengeance 16gb @ 1600mhz
PSU: Thermaltake Smart Series 750

What I did last night to try and solve/find the issue. None of which had an effect.

Cleaned out/reinstalled graphics drivers.
Dusted entire system.
Swapped PCI-E slots
Swapped PSU Cabling
Swapped/Reseated Memory Modules
Removed ALL CPU overclocks.

Right now I cannot remove individual compnents to test them in other systems. At this stage I'm considering the nuclear option and replacing the mainboard, cpu, memory, and psu.
 
Well, I have you tried to flash the BIOS of the motherboard? Considering its age, even relative to the 970, this could a reason behind the problems - although it wouldn't really explain why the card would work shortly after installing a driver.

You PSU should be delivering all needed, and unless you've received any obvious problems, then I doubt it's having too much trouble. Just a question, though, how old is it?

Lastly, do you have another, newer motherboard to test out the card on?
 
Most of the time this is caused by a driver error. Whenever a driver error occurs the card will downclock to a set speed. It will not increase it's speed again until you reboot the system.
Since both cards are using the same driver, both cards get downclocked.
Remove the overclock on the 970, if it is factory overclocked you may need to downclock it to default speeds.
 


Just a question for myself: Considering this, and the fact that the GTX 1070 isn't overclocked, how would he fix this considering that using DDU didn't work at first? (as mentioned, this information is just for me to learn ;))
 


If you read his post again, he stated that it would work for about a week, then happen again.
Both cards use the same driver, so if an error occurs ,both cards will get downclocked, until a reboot or reinstall of the drivers.
 

MrHatty

Honorable
Aug 16, 2013
19
0
10,510


Other than the GPU obviously, the components in my system are just over 5 years old. Along with the usual parts, the PSU is supplying power to an h100i CPU cooler, two solid state and one mechanical drive, 6 USB 3.0 devices and 2 USB 2.0.
 

MrHatty

Honorable
Aug 16, 2013
19
0
10,510


If the 1070 was throwing artifacts while under load, I would be more concerned with the GPU. My 970 started throwing artifacts (with and without its overclock) when these problems first arose which is what drove me to replace it with the 1070 to begin with.

Last night I let it rip on both the Firestrike stress test and Fur Mark. During the runs, while frame rates chugged, there was no artifacting or graphical glitches to speak of.

I currently have OCCT running and I'm seeing numbers that I'm not sure if I should be concerned with. Just a quick googling kind of tells me I should be.

The numbers in question are the outputs for the +3.3v and +12v rails. A couple places I'm reading indicate that those reading should be within half a volt or so of those numbers. The 3.3 reading holds between 3.24-3.28, but the +12v reading is down at 8.18. Again, just some quick googling (and common sense) tells me that should be pushing at or close to the 12v area.
 

MrHatty

Honorable
Aug 16, 2013
19
0
10,510


Took a look while running Overwatch and Firestrike. The two most common reasons where power limit and reliability voltage.
 

Automative

Prominent
Feb 14, 2017
1
0
510


Hi.
Just the same issue.
I've got ASUS STRIX GTX1070 OC for a few months.
Since the 378.49 driver was released I encounter a GPU locking at 734MHz. I'm monitoring with ASUS GPU Tweak 1.4.0.8.
Today I tried to install the 378.66 driver but it behaves quite similar.
Rollback to the 376.33 driver version solves this issue to me. The GPU clock demonstrates speeds from 278MHz to 2015MHz.

Btw while the GPU clock is locked at 734 MHz (378.49 and 378.66 driver versions) I still can't see any frame rate drops in GTA5. I haven't done any synthetic test to be sure but according to the GPU temperature it's loaded up as normal. So my theory is that the trouble belongs to representation of the GPU clock but not to GPU clocking directly.

That's my rig:
Mainboard: ASUS P8Z77-V Deluxe
CPU: Intel i7 3770k (3.99GHz under load)
GPU: ASUS STRIX-GTX1070-O8G
Memory: DDR3 Kingston HyperX 16Gb @ 2400MHz
PSU: Corsair HX850
Yes, old but still mighty.
 

MrHatty

Honorable
Aug 16, 2013
19
0
10,510


I've seen similar behavior. I'v eseen issues where most people seem to not like the GPU Tweak software (I enjoy it for the custom fan curve but I can get that with any software).

I did end up taking the nuclear option (because why not, lets cover all bases) and as it turns out, I'm getting this exact behavior from GPU tweak. However, with GPU-Z, it reports all normal in GPU heavy loads from Firetrike to general gaming. Still more testing needs to be done as the new components are still in the burn-in period.

The new build is as such:

Asus Prime Z270A-R
Intel i7 7700k (stock clock)
Corsair Dominator DDR4 (16gb @ 3200mhz)
Thermaltake Toughpower Gold 750

The 1070 remains. I'm half tempted to grab the 970 and see if I can replicate behavior.
 

Rascal1066

Prominent
May 10, 2017
1
0
510
i noticed this problem last night. during a game GTX970 pulling way too low watts and running too cool for a reference blower. quick check in firestorm showed really low core and memory clocks. i have my pc conected to a tv via HDMI and a 1080p monitor via DVI . this morning through trial and error i discovered that by having both monitor and tv on at the same time the graphics card ran on these loe clock speeds . cured by turning off the tv. this is a new problem i havent seen maybe caused by the latest windows 10 updates.
anyhoo this may be of help to some out there.