Asus 2070 RTX Overclocking - Help

[katulen]

Hey guys.

I wanted to get some feedback and guidance on my overclock for my Asus Strix 2070 RTX, and hope that some of you can help me

I messed around with it alot the last few days, and ended up getting these results which seem stable so far (4 days in & no crashes nor artifacts)

Voltage Core: 0%
Power Limit: 125%
Temp. Limit: 88c (max)
Core Clock: +140
Memory Clock: +600
Max Boost Clock: 2085
Max Voltage: 1.050v
Max Temp. Under Load: 65c
Max Fan Speed: 75%

Under max load or 99-100% GPU usage it runs 2085MHz & 7600MHz stable in games, 3D Mark Fire Strike & TimeSpy, MSI Kombustor & Heaven.

First of all, what is your thoughts on this overclock of my Asus Strix 2070 RTX ??

Secondly, i tryid to get some more juice out of it. But as soon as i increase clock or memory it will start crashing. I then thought about changing the Core Voltage % slider, but when i did that, it started to crash.
I then tryid to mess around with the voltage curve, without any success.
I did some research on the Voltages with the RTX series, and came to the conclusion, that most people don't have any success, getting any extra juice by changing the Core Voltage % slider.
What is your thoughts / opinions on this subject?

It is very much appreciated.

Thank you in advance.

Best regards,
Katulen - OC Newbie

 

[robot zombie]

Afaik core voltage slider does nothing with 20-series... same as pascal. Its locked down to a fixed curve. All you can really do is raise the ceiling via power limit. Card will always decide what voltage to run at what clocks otherwise.

Youre one stop lower than my Strix 2060 maxes at. Anything above 2ghz is pretty good. Cant ask for more from those cards really. Youre power limited and there is no way around that without modding that hasnt and may never be done on an RTX card. The Strix line doesn't have a lot of OC headroom because theyre sort of built to max the gpu on thier own.

No worries. Let it do its thing and enjoy that. One of the fastest, coolest-running, nicest cards you can get. Most people will never see numbers quite that good.

Only thing I might try is lowering the power limit to something like 120. If it isn't stable then thats that, but sometimes you can get a little better temperatures that way.

Id probably bring fan curves down too... thats just me. If 65C is what youre getting during sustained loads, such as long gaming sessions, then you have a lot of headroom. The GPU wouldnt care if it was hitting 75C, but at lower speeds the fans may last longer. Not to mention you may be able to make an already pretty quiet card a little quieter.

Thats getting into preference though. If you want something more to refine, thats probably it at this point.

 

[katulen]

Only thing I might try is lowering the power limit to something like 20. If it isn't stable then thats that, but sometimes you can get a little better temperatures that way.

So you mean moving the Power Limit slider from 125% to 120%, to see if its stable there?

Id probably bring fan curves down too... thats just me. If 65C is what youre getting during sustained loads, such as long gaming sessions, then you have a lot of headroom. The GPU wouldnt care if it was hitting 75C, but at lower speeds the fans may last longer. Not to mention you may be able to make an already pretty quiet card a little quieter.

Well tbh i dont even hear the fans, when its running at 75% fan speed, while having headset on. But i see what you mean about the longevity of the fans it self.
I might try it out and see what happens.

By the way, i've noticed while idling and not doing anything except being on the web and such, my Core Clock & Memory Clock keep spiking. Going from 300 core clock to 1410 core clock and back down to 300 core clock in a matter of a seconds. The same goes for the Memory Clock. Goes from 405 to 7600 and back to 405. And this keep happening.
Is this caused because of the GPU Power Settings being set to balanced or something else? And is it suppose to do this?

 

[robot zombie]

So you mean moving the Power Limit slider from 125% to 120%, to see if its stable there?

Yeah, sorry. In my head I think increases from base instead of total

But yeah... just kibd of a standard refinement. Sometimes you need less power than it seems. That 5% could be a significant overstep at the tippy top.

By the way, i've noticed while idling and not doing anything except being on the web and such, my Core Clock & Memory Clock keep spiking. Going from 300 core clock to 1410 core clock and back down to 300 core clock in a matter of a seconds. The same goes for the Memory Clock. Goes from 405 to 7600 and back to 405. And this keep happening.
Is this caused because of the GPU Power Settings being set to balanced or something else? And is it suppose to do this?

Yeah, thats normal idling for balanced. And then when it goes active it should rise to base clock until it's done.

If you switch to performance youll see it stick to base clocks the very first time it goes there. Only way it goes below that is if the driver restarts or something crazy like that. Most people run performance for overclocks. Though ime with 20-series I dont think it matters much.

 

[katulen]

Alrite, i will try mess with the PL and see how low i can get it before it becomes unstable with the current OC setup i've got. Cause it seems moving core clock or memory any more will make it unstable, and dont want that obv .

Yeah, thats normal idling for balanced. And then when it goes active it should rise to base clock until it's done.

Ahh okay. I just thought it looked weird, but if its normal i wont worry about it.

Another thing.
Do you know why Afterburner hits Power Limit "1", when its between 90-110% current Power % ? I find that very odd.

 

[John Naylor]

I'. lazy... I like to benefit from thoise who came before me and take advantage of the sseat off their backs ...

https://www.guru3d.com/articles_pages/asus_rog_geforce_rtx_2070_strix_review,28.html

We ended up with the following settings:

Temp Target 88 Degrees C

• Power Limiter: 125%
• CPU clock + 115 MHz
• Mem clock +100 MHz
• Voltage +100%
• FAN RPM default

Remember ... you will almost never get the highest fps at the highest memory and core OCs.

 

[xkm1948]

Use MSI Afterburner and do the OC scanner without increasing power limit, that will give you most accurate general overclocking headroom for your GPU

 

[katulen]

Remember ... you will almost never get the highest fps at the highest memory and core OCs.

What do you mean by that exactly?
That a 2000mhz core clock & 7300mhz memory clock will run same fps or perhaps even better?

Use MSI Afterburner and do the OC scanner without increasing power limit, that will give you most accurate general overclocking headroom for your GPU

I did try the OC scanner, but ended up with a lower overclock state. Can't remember exactly where it landed.
Thanks tho'.

 

[xkm1948]

What do you mean by that exactly?
That a 2000mhz core clock & 7300mhz memory clock will run same fps or perhaps even better?



I did try the OC scanner, but ended up with a lower overclock state. Can't remember exactly where it landed.
Thanks tho'.

OC scanner is a lot more accurate for daily usage.

Alternatively, you can set your OC manually and then use the "test" function and check for stability. Anything below 90% of confidence interval means your OC is not stable

 

[katulen]

OC scanner is a lot more accurate for daily usage.

Alternatively, you can set your OC manually and then use the "test" function and check for stability. Anything below 90% of confidence interval means your OC is not stable

Alrite, i will try that just for the sake of it.
However, wouldn't you claim the oc stable, when it passes all the stability tests / benchmark listed above, and also can play various games without crashing ?

 

[xkm1948]

Alrite, i will try that just for the sake of it.
However, wouldn't you claim the oc stable, when it passes all the stability tests / benchmark listed above, and also can play various games without crashing ?

I would say no. The OC scanner implementation is really good for the Turing GPUs.

At the end of the day it is your own definition of "stable". Do you want 7*24*365 stable? Or do you want 1 or 2 crashes a month? It is all relative. Also your OC today that is deemed "stable" WILL become unstable overtime as semiconductor undergoes aging process and OC will only accelerate the electron migration.

 

[katulen]

I would say no. The OC scanner implementation is really good for the Turing GPUs.

At the end of the day it is your own definition of "stable". Do you want 7*24*365 stable? Or do you want 1 or 2 crashes a month? It is all relative. Also your OC today that is deemed "stable" WILL become unstable overtime as semiconductor undergoes aging process and OC will only accelerate the electron migration.

I ran the test with my current OC, and it says 80% confidence.

I then did a scan again, cause i forgot what it came up with, and it showed +120 core clock, which was 2025 mhz core clock under load.
It said the limited factor was "Power", however, it never hit Power Limit (100%), and max voltage was 1.000v, so dont know why it said that.

And running 2025 mhz or 2085 mhz under load, isn't that the same when it comes to longevity of the gpu regardless?

 

[kastriot]

Well i wish you happy overclocking experience.

 

[John Naylor]

What do you mean by that exactly?
That a 2000mhz core clock & 7300mhz memory clock will run same fps or perhaps even better?

If you read TPU reviews, you will see that the card which gets the highest OC is as often as not, the one with the highest core or memory clocks. Here's a summary of TPUs test results on 2080 Tis

The card with the highest OC on the core is the Zotac Amp with 2145 came in 5th place in fps achieved.

The card with the highest OC on the memory is the Lightning but lets ignore that as it is the only one tested with Samsung K4Z80325BC-HC14. All the others used Micron MT61K256M32JE-14:A. The cared with highest memory OC not counting the Lightning was the Asus Strix w/ 2065 and it only managed 4th place in FPS.

The card, again excluding the Lightning cause of the unfair advantage, which hits the highest fps is the Gaming X Trio and it was in 7th place in Core OC and 3rd place in Memory OC

With the 2080, we have an instance where highest core = highest fps. The 2080 Gaming Trio hit 180.2 fps (5.2 fps faster than 2nd place) with a 1st place finish in core (2115) and a 5th place finish in memory.

I have seen this in my own testing all the time. I typically make a spreadsheet with columns listing core test speeds of say 1800, 1825. 1850 ..... and so on and then have rows doing the same incremental steps with memory. The "best OC fps wise" is never in the last column or row. Core 2050 / Memory 2025 may give higher fps than 2050 / 2050 or 2075 / 2025. I suspect it has to do with the way Boost 3 works since the numbers you set are not the same as what we see during the test ... instead the bounce around which is why 'average' is reported in the reviews.

 

[katulen]

If you read TPU reviews, you will see that the card which gets the highest OC is as often as not, the one with the highest core or memory clocks. Here's a summary of TPUs test results on 2080 Tis

Super interesting table. Thank you for that.
Would you be able to provide a link to where you found that table from? Would be much appreciated.

But this table you listed, is that done within 1 certain benchmark, or over a various of benchmarks/games ? Cause that matters alot if its actually stable or not.

The only "bouncing" i see on my GPU is when i start the load (benchmark, game or whatever) it clocks to 2130 and then downclocks within 5-15 seconds to 2085 and is stable there for the rest of the load duration - Memory sits on 7600 from start till the end. I don't know why this is happening though. It happens even if i put +50 on core clock, then the MHz is just lower obviously. I assume it is something to do with Boost 3.0.

Im not so clever on this subject, so might be a stupid question
Howwever, is GPU benchmarks purely based on min, avg & max fps ? Or is it also based on the stability of the current core clock & memory ?.

I might try make a table myself, and mess around with the core/memory a bit, to see what differences i get in fps throughout the different benchmarks.
Would be interesting to see.