Frametime spikes and stuttering after switching to AMD CPU?

[Pets4Ever]

Lol, it's a hotspot issue, as I said.

We'll see, he had that card in his pc before and wasn't having these issues. Maybe uninstalling the card caused it to flex and worsened the contact of the heatsink with the chip

 

[3x0]

Lol, it's a hotspot issue, as I said.

He tried a different GPU with the same results, read the thread

 

[Montelo]

running furmark puts the hotspot at 92

 

[dgianstefani]

We'll see, he had that card in his pc before and wasn't having these issues. Maybe uninstalling the card caused it to flex and worsened the contact of the heatsink with the chip

It's above 90 C. As soon as that hotspot reaches 100 it throttles hard. Plus, if the core is hitting those temps you can bet the rest of the card needs cleaning and maybe new thermal pads too. VRAM hitting overheat gives stuttering.

Any temp above 65 C from memory, NVIDIA GPUs steadily downclock, so you're only hitting full boost at 65 and below.

Chipset/RAM/SSD/VRM overheating from waste component heat etc or dirty heatsinks is also a big one. Any of those reach certain temp they will throttle hard and it will be noticeable.

A single core in the CPU reaching a certain temp, even if averages OK - instant throttle.

That's assuming there isn't simply a worn out part somewhere that can't deliver consistent perfomance. Dropping to lower performance state periodically.

He tried a different GPU with the same results, read the thread

So? Overheating can only occur in one card? Stuttering is classic borderline temperatures problem. RAM overheating for instance won't instantly crash, but dies get steadily more unstable as temps rise. Same with VRAM and cores, hence aggressive throttling which often manifests as brief stuttering.

Any PC temp above 90 C is instant red flag and fixing so nothing goes above 70 C should be first priority.

 

[Montelo]

So? Overheating can only occur in one card?

would be a weird coincidence though if both cards overheat no?

 

[Pets4Ever]

It's above 90 C. As soon as that hotspot reaches 100 it throttles hard.

Is that true? When I google max hotspot temp for the 3070 what I get is 110 C

 

[Montelo]

would be a weird coincidence though if both cards overheat no?

ive seen worse though a friend of mind had his cpu running at 90 and he had no stutter at all just his pc sometimes shut off if it was too hot

 

[Pets4Ever]

ive seen worse though a friend of mind had his cpu running at 90 and he had no stutter at all just his pc sometimes shut off if it was too hot

I hope it wasn't hitting 90 in a game. It really shouldn't hit 90 in a game...

 

[dgianstefani]

Is that true? When I google max hotspot temp for the 3070 what I get is 110 C

105 is max rated point it will downclock to point it essentially shuts down, that's also the temp VRAM will be rated for typically, as well as most components in power delivery etc. Even lower if not top tier components. It's Japanese caps rated for 105/110 C, chinesium stuff is lower. Sometimes it's a hard crash, sometimes it's just severe throttling. Regardless, at 90 C a GPU will downclock so bad I wouldn't be surprised to see major stuttering. Like I said, for every step above 65 C it downclocks from rated boost.

ive seen worse though a friend of mind had his cpu running at 90 and he had no stutter at all just his pc sometimes shut off if it was too hot

No stutter at all, just pc turning itself off to prevent damage. Nothing to see here.

Lol. Lmao.

 

[Montelo]

105 is max rated point it will downclock to point it essentially shuts down, that's also the temp VRAM will be rated for typically, as well as most components in power delivery etc. Even lower if not top tier components. It's Japanese caps rated for 105/110 C, chinesium stuff is lower. Sometimes it's a hard crash, sometimes it's just severe throttling. Regardless, at 90 C a GPU will downclock so bad I wouldn't be surprised to see major stuttering.


No stutter at all, just pc turning itself off to prevent damage. Nothing to see here.

Lol. Lmao.

haha i get your point , but overall it ran thats what i mean it was not ideal ofc haha

so can i just get a different gpu to test this theory out if its my gpu?

 

[dgianstefani]

haha i get your point , but overall it ran thats what i mean it was not ideal ofc haha

Anything above 65 C is not ideal for GPU, for CPU you generally want it to never go above 80 for similar reasons. For RAM it's about 40-45 C before you start getting errors, for VRAM I wouldn't be comfortable with anything above 75 C, same for SSDs.

Even if you don't get noticable performance issues from instability/downclocking, you get massively accelerated wear and tear above these temps.

 

[3x0]

Regardless at 90 C a GPU will downclock so bad I wouldn't be surprised to see major stuttering.

Nope, it just wont boost as much. The clocks will be higher than what GPUz Boost clock specification reads even with those temps.

 

[dgianstefani]

Nope, it just wont boost as much.

That's what downclocking is lol. If it's not at max boost under 3D load, it's downclocking. Unless you're arguing a part rated to turn itself off to prevent damage at 105 C should be regularly hitting above 90 C.

This is literally classic old paste dried up pads, time for repaste/repad, enjoy 60s temps.

 

[oxrufiioxo]

haha i get your point , but overall it ran thats what i mean it was not ideal ofc haha

so can i just get a different gpu to test this theory out if its my gpu?

Yes if you test a brand new gpu without overheating issues and have the same problems it isn't the gpu but you really should repaste the 3070 either way it's near the danger zone.

 

[Montelo]

Anything above 65 C is not ideal for GPU, for CPU you generally want it to never go above 80 for similar reasons. For RAM it's about 40-45 C before you start getting errors, for VRAM I wouldn't be comfortable with anything above 75 C, same for SSDs.

Even if you don't get noticable performance issues from instability/downclocking, you get massively accelerated wear and tear above these temps.

my ssds never go above 45 at all, gpu max is like 65, CPU never goes above 70

 

[dgianstefani]

my ssds never go above 45 at all, gpu max is like 65, CPU never goes above 70

Your GPU max isn't 65, that's average temp. Your max is 92 as you've already said, and we don't know what VRAM is.

I'm not going to bother writing more. You have a serious GPU temperature issue from being a five year old card that likely shipped with bottom of the barrel cheapo paste, and you need to fix it, even if it doesn't entirely fix the stuttering problem it's a priority.

 

[Montelo]

Yes if you test a brand new gpu without overheating issues and have the same problems it isn't the gpu but you really should repaste the 3070 either way it's near the danger zone.

its crazy that after only 3 years its already that bad, i had some gpus for over 6-7 years and it wasnt even half that bad

 

[3x0]

That's what downclocking is lol

Nope, cards have base clock, advertised boost clock and real clock that can often be higher than the advertised boost clock. It has been like that Since Kepler/Maxwell. Downclocking is when the card is below base clock.

@Montelo Open GPUz, let it run in the background, play a game for 10-15 minutes and post a screenshot of the main GPUz screen and the Sensors tab. Make sure to widen the GPUz window so we can se a larger portion of the sensors and hover over the GPU Clock to see what's the frequency like along with hotspot temps. You can move the mouse on the red bar to see specific values.

 

[Montelo]

Your GPU max isn't 65, that's average temp. Your max is 92 as you've already said, and we don't know what VRAM is.

didnt mean hotspot i meant the other temp

 

[Pets4Ever]

Like I said, for every step above 65 C it downclocks from rated boost.

65C seems overly conservative for hotspot. Where are you getting that number from?

 

[dgianstefani]

didnt mean hotspot i meant the other tempView attachment 405192

The other temp isn't relevant. The GPU sets clocks based on hotspot and average.

65C seems overly conservative for hotspot. Where are you getting that number from?

He's not. That's the average and is essentially meaningless unless it's above 65. Hotspot will cause throttling if it's near dangerzone for damage.

 

[Pets4Ever]

He's not.

?

 

[dgianstefani]

Nope, cards have base clock, advertised boost clock and real clock that can often be higher than the advertised boost clock. It has been like that Since Kepler/Maxwell. Downclocking is when the card is below base clock.

@Montelo Open GPUz, let it run in the background, play a game for 10-15 minutes and post a screenshot of the main GPUz screen and the Sensors tab. Make sure to widen the GPUz window so we can se a larger portion of the sensors and hover over the GPU Clock to see what's the frequency like along with hotspot temps. You can move the mouse on the red bar to see specific values.

You can focus on specific interpretations and may even be technically correct in a way, but modern GPUs can have 500-1000 MHz difference between base and boost clocks, or more. So I when I say it's downclocking due to temperature, you know exactly what I mean, because someone not being pedantic can understand how an overheating GPU will "downclock" from a previously consistent max boost, and also exactly how significant a 90 C+ reading is for real clocks.

In fact downclocking refers to reducing clocks, pretty obviously, whether that's from boost or other states doesn't really matter, it's just reducing clocks. Throttling is the specific term for reducing clocks below rated speeds, generally due to overheating.

 

[oxrufiioxo]

its crazy that after only 3 years its already that bad, i had some gpus for over 6-7 years and it wasnt even half that bad

Brand new cards can have this issue, just Google gigabyte 50 series issues lol.

Manufacturers especially during the pandemic really cut corners and even now you see issues on brand new cards. I would imagine 3 years isn't abnormal and most people probably repaste their gpus right after the warranty is up. Some cards have lasted me 5+ years without it being necessary but there really isn't a firm time table for when you will need to do it.

 

[dgianstefani]

?

His real hotspot temp is 25-30 C higher than that. Scroll up.

Brand new cards can have this issue, just Google gigabyte 50 series issues lol.

Manufacturers especially during the pandemic really cut corners and even now you see issues on brand new cards. I would imagine 3 years isn't abnormal and most people probably repaste their gpus right after the warranty is up. Some cards have lasted 5+ years without it beinf necessary but there really isn't a firm time table for when you will need to do it.

Most people don't ever repaste, they just go and buy a new card. Most people don't even know about repasting. If the card doesn't have phase change paste or liquid metal it will need repasting eventually, and often is a short time, even from new. Manufacturers do bare minimum stuff most of time, and old dirty heatsinks and dried out TIMs don't make things better over time.

Personally I repaste cards with good stuff from BNIB. The factory application is generally crap, or with bad paste, or both. There's some exceptions, like ASUS using PTM/Liquid metal on new cards. Modern consumer laws also state you can open your card and the warranty void if broken stickers are illegal etc.