RTX3070 won't OC the same it did on my old Mainboard

[PTG]

So recently I pretty much upgraded everything but my GPU (New Hardware: MSI Z490 Gaming Plus, I7-10700K, Lumen S28, new PSU...; OLD Hardware: Gigabyte Aorus B450, R5 2600X, NZXT X62 ...).
Back then on my old pc I could easily clock my 3070 FE to 2070mhz on 1.018V without any problems, after I upgraded the system the GPU wont be stable on this value. It will always drop really hard and won't keep the frequency I set for it on MSI Afterburner. Only with a higher power limit I was able to get it kind of stable on this clock. Ive been testing a lot and still don't know what's wrong. I changed windows, I checked for thermalthrottling and I also checked several NVIDIA and Windows settings. Nothing helped.
Any ideas?

 

[lawood]

Im kinda confused on your upgrade.

You had a b450 that can take all zen 3 processors but you decide to buy an I7-10700K?

One 5900x is cheap atm, you could buy that and use your OLD motherboard... any zen 3 processer smokes the 10700k in performance. Very puzzling decision.

 

[PTG]

Im kinda confused on your upgrade.

You had a b450 that can take all zen 3 processors but you decide to buy an I7-10700K?

One 5900x is cheap atm, you could buy that and use your OLD motherboard... any zen 3 processer smokes the 10700k in performance. Very puzzling decision.

I got it pretty cheap from a friend, almost not used. So why not

 

[pavle]

One idea would be that with the different CPU (i7 10700K) the card would be utilised more and therefore not stable with the old overclock/power limit. Have you ever checked what was GPU utilisation on your R5 2600X processor?

 

[PTG]

One idea would be that with the different CPU (i7 10700K) the card would be utilised more and therefore not stable with the old overclock/power limit. Have you ever checked what was GPU utilisation on your R5 2600X processor?

percentagewise it shouldve been the same. But I was thinking about the same. Since the old CPU was bottlenecking the GPU it now can reach its peak performance even tho the clock is the same

 

[lawood]

I got it pretty cheap from a friend, almost not used. So why not

You got the motherboard + cpu for cheap from a friend then?

Ask him if he had any problems with the mb because you should be getting much more performance from the 10700k, its much more powerfull than the 2600x.

 

[PTG]

You got the motherboard + cpu for cheap from a friend then?

Ask him if he had any problems with the mb because you should be getting much more performance from the 10700k, its much more powerfull than the 2600x.

He used his 3070TI on it and it worked totally fine. No problems at all

 

[lawood]

Have you tried doing a bios reset?

Not loading optimized results, shorting the usual buttons or taking off the cmos battery.

You could also try updating the bios the latest version if you haven't done that.

 

[PTG]

Have you tried doing a bios reset?

Not loading optimized results, shorting the usual buttons or taking off the cmos battery.

You could also try updating the bios the latest version if you haven't done that.

All I have done so far is checking different drivers, but didnt do a bios reset. Im also not quite sure how it works though

 

[lawood]

You should at least reset the bios and see if you see any improvements.


How to do that:

https://download.msi.com/archive/mnu_exe/mb/E7C79v1.1.pdf

Page 39

Resetting BIOS to default values

1. Power off the computer and unplug the power cord
2. Use a jumper cap to short JBAT1 for about 5-10 seconds.
3. Remove the jumper cap from JBAT1.
4. Plug the power cord and power on the computer

If you dont have a jumper (most people lose them) just use a screwdriver.
You could also remove the cmos battery and put it back in but I find that very annoying.

 

[GerKNG]

you have a stronger CPU and you are less often cpu limited/bottlenecked.

 

[PTG]

You should at least reset the bios and see if you see any improvements.


How to do that:

https://download.msi.com/archive/mnu_exe/mb/E7C79v1.1.pdf

Page 39

Resetting BIOS to default values

1. Power off the computer and unplug the power cord
2. Use a jumper cap to short JBAT1 for about 5-10 seconds.
3. Remove the jumper cap from JBAT1.
4. Plug the power cord and power on the computer

If you dont have a jumper (most people lose them) just use a screwdriver.
You could also remove the cmos battery and put it back in but I find that very annoying.

Oh you were referring to a CMOS clear, yea did that before few times. Dont think its the issue

you have a stronger CPU and you are less often cpu limited/bottlenecked.

I know the cpu is most likely not being limited/bottlenecked, but what are you trying to say?

 

[lawood]

btw, its normal a card not maintaining the target frequency, temps come into play.

 

[PTG]

btw, its normal a card not maintaining the target frequency, temps come into play.

Its not thermalthrottling. Card is barely reaching 75C anywhere

 

[lawood]

That doesnt matter. It will downclock even at 75, if you want full clocks always you need sub 70 temps, ideally sub 60.

You will HARDTHROTTLE at 85c, but your clocks with be lowering along the way.

You can see this very well with hwinfo "effective core clock". Monitors like msi afterburner are garbage for this and lead people into error because they only report the desired core clock, not the REAL core clock.

Because the card is not only worrying about the core temp, its also worrying about hotstop temp and mem temps.

 

[PTG]

That doesnt matter. It will downclock even at 75, if you want full clocks always you need sub 70 temps, ideally sub 60.

You will HARDTHROTTLE at 85c, but your clocks with be lowering along the way.

You can see this very well with hwinfo "effective core clock". Monitors like msi afterburner are garbage for this and lead people into error because they only report the desired core clock, not the REAL core clock.

Because the card is not only worrying about the core temp, its also worrying about hotstop temp and mem temps.

So I just tested it on max cooling and it showed me in GPU-Z max temp of 60 and max hotspot of 70 and it still downclocked by a lot

 

[lawood]

What is by a lot?

What is your target core clock, and what's the downclock?

Also, use hwinfo and see mem temps and EFFECTIVE core clock.

 

[PTG]

So the clock is dropping to around 1700mhz max
And in HWINFO I could see the average of the effective clock being at like 1783mhz while GPU clock was at 1883
Also the gpu temp limit shown by HWINFO is 83C and the clock was already dropping without reaching it

 

[evernessince]

As other users have said, reset the BIOS to default settings then enable XMP if applicable. If you upgraded from an AMD platform to an Intel Platform without reinstalling windows that could be causing you issues. You might have to go through and delete drivers on your system or re-install windows. Sometimes you can do a drop in place upgrade like that and other times there will be a billion small issues that will linger and cause you trouble. I'm going to be honest, I don't think it was worth upgrading from the AMD platform given the massive hassle it is when you switch your entire platform from installing the new hardware to re-installing windows and getting everything right. Especially considering how easy it would have been to drop in a faster AM4 CPU. If we are talking 8 hours of extra work, no savings are going to make up for that time difference because I could have worked that much time and gotten paid far more.

Run Cinebench R23 multi-core and single core in addition to Unigine Superposition and report the scores here.

You can run a program like HWInfo to monitor thermals during the benchmarks / while gaming. Report back CPU / GPU thermals and clocks during load.

 

[PTG]

As other users have said, reset the BIOS to default settings then enable XMP if applicable. If you upgraded from an AMD platform to an Intel Platform without reinstalling windows that could be causing you issues. You might have to go through and delete drivers on your system or re-install windows. Sometimes you can do a drop in place upgrade like that and other times there will be a billion small issues that will linger and cause you trouble. I'm going to be honest, I don't think it was worth upgrading from the AMD platform given the massive hassle it is when you switch your entire platform from installing the new hardware to re-installing windows and getting everything right. Especially considering how easy it would have been to drop in a faster AM4 CPU. If we are talking 8 hours of extra work, no savings are going to make up for that time difference because I could have worked that much time and gotten paid far more.

Run Cinebench R23 multi-core and single core in addition to Unigine Superposition and report the scores here.

You can run a program like HWInfo to monitor thermals during the benchmarks / while gaming. Report back CPU / GPU thermals and clocks during load.

So first of all I upgraded because I got the stuff REALLY cheap, otherwise I would have gotten another AM4 for sure.
Also I always reinstall Windows if I get new parts or so. First I installed Win10 and the issue appeared and since it was new hardware and my PSU was just 550w I thought it might have been the PSU so I upgraded that one. With the new PSU I installed win11 (clean install not upgrade) and the issue was still there, so I downgraded to win10 (again clean install) and the issue was still there. I can try to CMOS clear AGAIN but don't think it will change ANYTHING. XMP and all the stuff is on, but XMP shouldnt make a difference GPU wise, same for running cinebench. I tested it on several Benchmarks and games and it was everywhere the same (3DMark, Heaven, CyberPunk2077). Everywhere it was dropping the clock without any reason.
The thermals are also totally fine, the thermal limit for the gpu is set to 83C and it is dropping without even reaching it (neither hotspot nor general temp)
I hope this makes it a lot clearer

 

[evernessince]

So first of all I upgraded because I got the stuff REALLY cheap, otherwise I would have gotten another AM4 for sure.
Also I always reinstall Windows if I get new parts or so. First I installed Win10 and the issue appeared and since it was new hardware and my PSU was just 550w I thought it might have been the PSU so I upgraded that one. With the new PSU I installed win11 (clean install not upgrade) and the issue was still there, so I downgraded to win10 (again clean install) and the issue was still there. I can try to CMOS clear AGAIN but don't think it will change ANYTHING. XMP and all the stuff is on, but XMP shouldnt make a difference GPU wise, same for running cinebench. I tested it on several Benchmarks and games and it was everywhere the same (3DMark, Heaven, CyberPunk2077). Everywhere it was dropping the clock without any reason.
The thermals are also totally fine, the thermal limit for the gpu is set to 83C and it is dropping without even reaching it (neither hotspot nor general temp)
I hope this makes it a lot clearer

What was your scores in those benchmarks? Saying you ran them doesn't help us help you. We need to know the scores so we can see the level of performance you are getting.

Also need your CPU temps at max load and in what situation. You say you get 83c on the GPU but in what benchmark?

What model is your new PSU? You say you had a 550w, which is woefully under-powered, but your new PSU could still be bad depending on the quality.

 

[PTG]

3DMark with the 2070clock at around 1300. The average by 3DMark is 13220. With a different clock speed (1980) I can reach up to 13700.
My previous PSU was a Corsair VS550 and my current one is a Gigabyte Aorus P750W.
My GPU temp limit is 83C, after that it is thermalthrotteling (maybe). CPU temps are usually below 70 unless its a specific CPU benchmark (then high 70Cs)

​

​

 

[sneekypeet]

Are you using GPUZ?
I ask, as the throttle reasoning will show in the sensors tab. It should help to direct you as to what the issue is versus chasing your tail all day.

 

[PTG]

I do use it and its not showing any throttle

 

[sneekypeet]

I do use it and its not showing any throttle

Sorry, but I trust GPUZ above all other software for what is going on with the GPU.
In the sensors tab, watch what is going on while benching and see if its the software used or a real issue before continuing to chase things.

My initial guess was the PSU before i continued reading, but who knows from outside your room.