[SOLVED] Help with undervolting an RTX 2070 Super ?

[Wolf96xz]

Hello everyone,

I'm trying to do an undervolt to my zotac rtx 2070 super twin fan with MSI Afterburner. The values I've put in so far are as follows:

- Power limit (%): 90

- Core clock (Mhz): +100

- Memory clock (Mhz): +300

According to the manufacturer's page, the memory clock of this graphics card is 1770 Mhz. So, I'd like to know how much I can reduce the power limit without losing performance. Is there any way to estimate it or do I just have to try it?

Official website of the manufacturer.

Thank you.

Regards.

 

[Newtonius]

Due to the varying nature of electrical components stability on a per card basis usually relies on what that specific card can handle. What undervolt or overvolt works for you may not work with me, so yes you'll just have to try it and see where your performance dips. It also varies per game as some games rely on fast GPU clocks speeds while others torture the VRAM with loads of texture data that need to be called quickly. Just feel around for that sweetspot but if you're already stable at 90% power and 100 on CC and 300 on MC then good for you man, that's impressive enough already.

 

[kerberos_20]

hi, most power draw comes from shaders (ie computing), i never had RTX so no idea how much power draws tensor cores, but going with just ROP/TMU, they have overally low power draw, which means resolution/antialiasing and texture mapping/anisotrophic filtering, if u reduce game shaders (most of effects such as shadows and other eye candy stuffs), gpu will put decently high fps at low power draw (thus normal gpu boost would be mostly maxed without throttling), once u add some heavy shaders and gpu starts computing heavily, power draw and temperature kicks in. if u block power draw, then u may see degraded performance on eye candy stuffs

 

[Wolf96xz]

I have done two tests with Battlefront 2, on the stage of Felucia with everything in ultra, minus the HBAO and others that I have lowered them a little.

First test results with 100% power limit and without touching anything:
- Temperature: 64 Cº max.

Second test with power limit at 90%, 100 cc and 350 mc.

- Exactly the same temperatures.

I just don't understand why this happens, in theory there should be a reduction in temperatures but it doesn't look like it. I did the two tests in a row, closing and opening the game when I was finishing making one, letting the graphics card cool down a bit. What can be going on so that temperatures don't drop even with the reduced power limit?

Thank you

 

[Karadjgne]

Most likely because your gpu, like everyone else's, is not hitting power limits to start with.

Power limit is the amount of power the card can possibly use when maxed out. It's not the amount of power the card uses always.

So let's say your gpu uses 1.2v and 50A (60w) when playing BF2. Your card limits might be 1.4v and 80A (112w). You lower card limits to 90% (101w), but that's not going to affect the 60w used by the game.

What it will affect is anything that pushes the limits, max boosts, rendering, or anything else that requires high amounts of gpu processing, and you just chopped it off at the knees, possibly creating instability as it tries to work at 100% of its clocks and memory speeds and cannot.

Gpu vendors have almost universally removed the ability to undervolt cards for a reason.

 

[Wolf96xz]

Most likely because your gpu, like everyone else's, is not hitting power limits to start with.

Power limit is the amount of power the card can possibly use when maxed out. It's not the amount of power the card uses always.

So let's say your gpu uses 1.2v and 50A (60w) when playing BF2. Your card limits might be 1.4v and 80A (112w). You lower card limits to 90% (101w), but that's not going to affect the 60w used by the game.

What it will affect is anything that pushes the limits, max boosts, rendering, or anything else that requires high amounts of gpu processing, and you just chopped it off at the knees, possibly creating instability as it tries to work at 100% of its clocks and memory speeds and cannot.

Gpu vendors have almost universally removed the ability to undervolt cards for a reason.

So would you recommend that I give up on undervolting the graphics card? If so, how could you lower the temperatures without affecting its performance? Especially in gaming, which is what I use this pc for. I am aware that temperatures are not worrisome right now, but if there is a possibility of reducing them, and incidentally also the noise of the card itself, I would like to know. To extend its life of use without losing power.

Thanks.

 

[Karadjgne]

Temps are more than just internal processes, they are also external.

Inside your case you have airflow characteristics. The airflow is going to be dictated by its variables. For instance, a high airflow fan moves a lot of air, but a high static pressure fan moves a lot of air. Significant difference. Almost always better to have airflow as exhaust, sp as intake. That way intake air actually reaches the gpu. Higher airflow fans don't have the force to move the air to the gpu very well, so it sits in a hot zone.

Not having decent exhaust means you don't create enough low pressure area in front of the fan, so case air doesn't fill the void. If there's an open vent above the rear exhaust or next to it, that's going to feed the fan before case air does, meaning gpu exhaust doesn't get sucked away, starts raising the case temp at a lower zone, and the gpu sits in a hot zone again.

Case airflow has a large impact on not only cpu temps, but also gpu temps, as a decent sized gpu is 2x-3x more power hungry than most cpus. It's the major heat source of any gaming pc.

64°C is a very decent gpu temp under gaming conditions, many usually sit closer to 80°C when pushed, so I'm not entirely sure why you feel the need to take drastic measures to lower temps.

What you want, and what's physically possible are often 2 different things.