GTX 1070 8GB - Constant frametime spikes (stutters) in every game - Hardware & Software

Here is the problem: with every game I try playing, I get absolutely MASSIVE spikes in Frametime pretty much all the time in irregular intervals. The game can be locked in 60 or 120 FPS yet it'll still keep stuttering due to them, making games unplayable and I've been dealing with this ever since I bought the card some months ago. Some details that I've found while troubleshooting with a friend: GPU utilization hovers at 40-70% No one CPU core ever maxes out Voltage limit is being hit almost constantly Power % hovers at about 50% No problems with audio ever No stutter in video players Firefox visibly stutters while scrolling but I'm not certain if that's related Here are my specs: Mobo: ASUS PRIME B350M-A (AM4) CPU: Ryzen 5 1600 GPU: GIGABYTE NVIDIA GTX 1070 G1 Gaming 8G (rev. 1.0) RAM: 8GB DDR4 Running Windows 10 64bit (ver 1709) What I tried: Removing drivers with DDU, reinstalling the latest Turning off HPET in Windows (mobo does not have the option in the BIOS) Disabling Win10's fullscreen optimizations Disabling Superfetch Disabling Game DVR Geforce Experience isn't installed Setting absolutely everything to max performance on NVIDIA CP and Windows (checked the P-State on CPU-Z, it's always at P0) Messing with Power/Temp limits and Core Voltage in Afterburner Limiting card to a voltage of 825 mV as a test Changing global Maximum Pre-Rendered Frames to 1 Reseating the power cable Changing from an HDMI cable to DP Checking DPC Latency, results lower in this post Drawing a pentagram in blood on the ground and sacrificing a goat Some stuff I probably forgot Some pics: What the usual Rivatuner frametime graph looks like: https://files.facepunch.com/forum/upload/109692/5ce9a4f2-b54c-4af4-8d95-88f21c09b686/image.png Screenshots of Afterburner monitors: HITMAN: https://files.facepunch.com/forum/upload/109692/28060479-a95a-4b1b-882b-fe460a668df1/unknown.png Warhammer: Vermintide 2 https://files.facepunch.com/forum/upload/109692/b351319d-d734-4649-b06b-371f8036d832/unknown-1.png (First mark is after game spawns me on the hub after loading, second mark is start of intro on map load, third is closing the game after playing up to a bit) CPU: https://files.facepunch.com/forum/upload/109692/8a48b096-c820-4ab1-8ad6-da82f14e0773/unknown.png RUINER: https://files.facepunch.com/forum/upload/109692/37c227b5-fc67-45a0-a653-5f697270861a/unknown-1.png https://files.facepunch.com/forum/upload/109692/572ccef6-ae25-4b5d-b4f6-ed5e20bc0649/unknown-1.png LatencyMon results after running it for a while with Vermintide 2 open (idle): https://files.facepunch.com/forum/upload/109692/addda0cf-f71a-43f7-a355-d8d7fc36923b/unknown.png I'm at the end of my rope here. Any ideas?

Try and cut down the possibilities, sounds like its either memory related or the GPU is kill. How full/old are your hard drives and have you run chkdsk on them? Have you tried running on just 1 stick of ram/other ram not from your current build? If not memory then do you have another PC you can stick the GPU into to test it separately? What happens to firefox if you remove the GPU, onboard should be able to handle the desktop and a web browser. Also what is your power supply?

Added some information to OP. I got two 1TB hard drives; the one Vermintide 2 runs on (not-OS) is quite full indeed but not the one that has HITMAN and RUINER in them (OS). I'll have to try running chkdsk. This PC already only has one single stick of RAM and sadly I don't have access to any others I could use to test, neither another PC. This model of mobo also doesn't have an onboard GPU. Thanks for replying.

I had this issue with this exact card and for me the resolution to the problem was just that the VRMs were getting too hot even after doing the EVGA pad mod My PC had no exhaust fans and the only two fans active were the cpu / intake fans (which were set to not running / low speed) Adding a single exhaust fan fixed the issue and I might at some point just print a fan shroud so the stupid thing exhausts hot air instead of milling it all over inside the case

Given that you were playing around with limiting the GPU's power draw, just max its available draw out and see what happens. I edited my original post but if that volt limit is refering to the card then it shouldn't be hitting max voltage only utilising like 80% of the cards processing power. Im not aware of if that volt limit shown in the images is specific to a piece of hardware or not so feel free to fill in the blanks there

WRT the graph indicating voltage limit, it means that the graphics card is hitting the maximum amount of volts for the given core clock. You can thank nvidia for the somewhat complicated featureset of boost 3.0, but in general each temperature has an associated node on a voltage/frequency curve. https://files.facepunch.com/forum/upload/109677/e9fe88dc-28b6-43ad-b5df-a87f952d2556/1521412941.png For example, in the vermintide 2 set of graphs, the first marker is at 62 degrees, which corresponds to a node located at 1987MHz/1081mV. You're right in that usage % of as low as 50% shouldn't be triggering voltage limit, especially since power% is only at 85~50%.

To clarify, I haven't done any overclocks, this is all stock settings. Also fixed the HITMAN image, it was the same as Ruiner.

Well when the card is hitting the limit its use is a little over what the stock voltage is said to be (1.062v from google for a 1070) so I don't really see why there should be any issue with it running out of power if thats what the factory settings are anyway. Also Kovac I doubt its going to be a temperature issue, that chart shows the card being well within normal temperatures for a 1070

You can stick a spare fan right in front of it and see if it helps any

Tried CHKDSK, it found some errors on both drives and fixed them all, however that didn't seem to change anything. Here are some charts of the last Vermintide 2 reading, from the beginning of The Screaming Bell level up to some point after the fire corridor, including a horde on the way: https://files.facepunch.com/forum/upload/109692/3878b252-1cc7-439e-9702-41bfa8a4f449/image.png https://files.facepunch.com/forum/upload/109692/5fb9377d-8b2e-4fd9-9fb4-8b7162087b20/image.png Is it really normal for the game to be dropping frames this hard while uncapped, only ever reaching at max 72% usage at the beginning? This is puzzling.

Your best bet for diagnosing this is GPUView

This might sound stupid, but are you using the "Ryzen Balanced" power plan in Windows? Might help, it changes more than just timers. I also see you're running single-channel DDR4, that's a performance killer, dual-channel is a must on Ryzen.

I'm betting on either the HDD or Ram. Gpu is fine and I want it xoxo

I'm not running that plan, today was the first I've heard of it while googling earlier. I'll have to update my chipset drivers and try it. Speaking of which, why is my Ryzen running at 3.5ghz? Isn't the stock speed 3.2? Did I accidentally overclock it somehow? About the RAM, I've realized I'm stuck between a rock and a hard place: this stick was all I could afford at the time, but now if I want to upgrade I'll have to either get 2x8GB sticks at 3200Mhz so I can dual-channel, or another 8GB stick at 2400Mhz which I don't think it's worth it at this point. Been having nothing but grief since I put this PC together.

It's ok if the RAM is "low speed" like 2400, but single-channel operation is slow as shit on ryzen. Windows clock speed reporting is inaccurate, so if that's where you got 3.5 from just disregard. My old 2500k at 4.6GHz showed over 6ghz all the time. Definitely get the latest chipset drivers and BIOS, lots of critical fixes in them.

So, update: I've updated the chipset drivers, uninstalled the latest NVIDIA drivers (heard people were having problems with it) and installed the older 388.71 alone with nothing else and set the power plan to Ryzen Balanced. Here are the results on Vermintide 2: https://files.facepunch.com/forum/upload/109692/da218724-1e49-4fbd-8ad6-8f094aa7267b/unknown.png It seriously feels much smoother than before, visible frametime spikes were rather rare, barring the low GPU usage (CPU bottleneck?) and low frames during hordes it's OK I suppose. However, look at what happens when I tried running Half-Life 1: https://files.facepunch.com/forum/upload/109692/2eaacf39-19cf-4e18-919d-d310898c1eab/unknown-1.png What the absolute fuck?

Do you by any chance have ms edge running? For whatever reason I get stutters when using ms edge and hulu. Even if they're in the background

These problems appeared since you got the new video card? I also had some weird shit go down when I upgraded from GT 730 2GB to GTX 1060 3GB. I thought I had cancer when I booted up my PC and everything ran slow, stuttered, etc. Luckly installing the latest drivers fixed everything for me at least. Sometimes the games themselves have performance optimization issues.

In my experience chkdsk generally wont do much in terms of fixing HDD issues, its just a good way to see if there are a lot of faults which usually indicated a bad drive. Just to be sure it isn't an issue with the HDD that has been lessened by the updated drivers, can you move HL1 onto the other drive and see what happens? Given the drivers made a difference its kind of doubtful its an issue with the HDD or sata controller so I get the feeling this may be purely a software issue.

So that set of graphs from half life was fucking weird. Frametime spikes of that magnitude when the gpu sits at 5-8% and the processor doesn't go beyond 90% is a little too weird. Cue some collaborative testing on discord, including downloading GPUview and capturing a spike from half-life. https://cdn.discordapp.com/attachments/115879850392158210/425485429823897602/unknown.png Zoom, enhance, etc https://cdn.discordapp.com/attachments/327383471444066305/425537322998693899/unknown.png https://cdn.discordapp.com/attachments/327383471444066305/425538523077148672/unknown.png https://cdn.discordapp.com/attachments/327383471444066305/425537549277200394/unknown.png Looks like an interrupt that doesn't get resumed?

Done that, results as seen below. The major spikes seem to have gone, but on this save I've got one that I can consistently reproduce. Load my save, walk up some stairs and launch an MP5 nade towards two HECU, the game always stutters at this point even after multiple attempts. It's marked below, the plateaus are me loading the save: https://files.facepunch.com/forum/upload/109692/9c1dba0b-d93b-43e6-8ee2-19cd3151f013/image.png https://files.facepunch.com/forum/upload/109692/5f2788e8-b92e-43cb-b974-0f3227bfea6a/image.png Not sure what to think of this.

Whats your fps_max set to in HL2? Source can't handle going over 1000fps well at all, and will stutter out the ass.

According to the HL1 one, seems to be capped at 101.

Shit I read it as HL2 not HL1, my bad. GoldSRC is basically locked at 100fps. I'd edit my post but the save button does straight up nothing.

Well given it performs alot better on one HDD than the other this is starting to sound like it really is a memory issue (and outdated drivers intensifying it), but is it only in hl1 that that single stutter occurs? It may just be an issue with HL1 if it only ever happens in that one spot.

I used to also get an easily reproducible stutter by hitting the dummy at the hub (only the first hit though) in Vermintide 2, but my last test after doing all that with the drivers didn't have it happen. Though I'm not sure I understand, doesn't it performing better in one drive means it's an HDD issue? If memory really is the problem, how would I go around solving it? Get another stick for dual-channel? How do I make sure it isn't my CPU having a bottleneck? This low GPU usage is really suspect. Also considering just resizing my C: partition and installing a Linux distro to test games in it to see what happens, that would rule out software surely.

woops I meant a hard drive issue, sorry, wrote that response when I hadn't had any coffee yet. While AMD CPU's tend to run a bit shittier compared to Intel, if it was a cpu bottleneck then you would be seeing the CPU cores sitting at something like 80-90% while the GPU sitting lower. Looking at those charts even before you did the driver updates your CPU cores were averaging 50%-60% while your GPU is hitting 80% (ruiner is a good example of that as its constant) which is a good sign its not that, if the GPU was starved by the CPU then the CPU should be sitting very high. I would like to hear others opinions on this though. Its possible the SATA controller or RAM is slowing the CPU down (Had a dying SATA controller a bit ago and it tanked my system hard) but its most likely just the hard drive having issues, can't even begin to count how many I've gone through over the years. If you don't mind, how old are the drives? Though without being able to move parts to a different system to test individually there's really no way to conclusively prove which part is the fault.