NVIDIA GeForce RTX 3080 FE Overclocking

[Brent_Justice]

Introduction



The NVIDIA GeForce RTX 3080 Founders Edition was launched on September 16th, 2020 as part of the RTX 30 series.  The Ampere architecture-based GeForce RTX 3080 FE was released with an MSRP of $699 and replaces the GeForce RTX 2080 and GeForce RTX 2080 SUPER which were also launched at this same price point.  It also offers a nice upgrade from the GeForce GTX 10 series.In our review, we found the GeForce RTX 3080 Founders Edition actually performs faster than the GeForce RTX 2080 Ti, which is in a completely higher price bracket.  With the GeForce RTX 3080 FE you can now get faster than GeForce RTX 2080...

Continue reading...

 

[Daniel_Doty]

Awesome article Brent!!!!!

As you stated, it really isn't all that beneficial to overclock the card for the gain that is achieved.

It would also seem that the 3080FE does it's job quite well at stock settings and outperforms the 2080Ti.

 

[hubaduba]

9% faster does put it in 3090 territory for performance, and +14% power also puts it in 3090 territory. Ampere seems to be prophetically named, as the performance hinges on how much juice you can push through the chip.

 

[Stoly]

When it was first announced, I hoped Ampere could be the next Pascal when overclocking. Getting high clocks and still being more power efficient than the past generation. Or maybe pull a Maxwell, having even higher clocks in spite of powerdraw but much better performance.

Unfortunately its quite the opposite, kind of reminds me of Kepler; very little clocks, less performance and much more power.

So don't waste your time with OC on RTX3080.

BTW apparently some people are getting great results with undervolting the RTX3080, you should give that a try.

 

[Stoly]

BTW I think people should just skip OC versions of 3rd party cards, there's nothing to be gained.

 

[Brian_B]

When it was first announced, I hoped Ampere could be the next Pascal when overclocking. Getting high clocks and still being more power efficient than the past generation. Or maybe pull a Maxwell, having even higher clocks in spite of powerdraw but much better performance.

I definitely feel like Turing was this vs Pascal - especially since apart from the Raytracing, there wasn't really much difference in rasterization / efficiency. Turing really was "Add Raytracing, throw in more cores, boost the power envelope, push these new cards out there at huge margins".

But it does appear Ampere at least has some performance gains over Turing in that department.

With respect to overclocking - I don't see why anyone would be disappointed. Turbo algorithms are getting better, your getting more STOCK clocks out of these chips, it's less of a lottery. Sure, it's fun to sit there and tweak on clocks, and you can still do that if you really want - but better monitoring and good algorithms are taking a lot of the guesswork out and giving that to you out of the box. That's even more true with CPUs, where overclocking is all but gone, and the boost algorithms are starting to do a better job than manual overclocks.

 

[Stoly]

I definitely feel like Turing was this vs Pascal - especially since apart from the Raytracing, there wasn't really much difference in rasterization / efficiency. Turing really was "Add Raytracing, throw in more cores, boost the power envelope, push these new cards out there at huge margins".

But it does appear Ampere at least has some performance gains over Turing in that department.

With respect to overclocking - I don't see why anyone would be disappointed. Turbo algorithms are getting better, your getting more STOCK clocks out of these chips, it's less of a lottery. Sure, it's fun to sit there and tweak on clocks, and you can still do that if you really want - but better monitoring and good algorithms are taking a lot of the guesswork out and giving that to you out of the box. That's even more true with CPUs, where overclocking is all but gone, and the boost algorithms are starting to do a better job than manual overclocks.

I agree that we do get better stock performance thanks to boost algorithms, but nvidia did market its cards as highly overclockable for quite some time even with boost clocks. IMO pascal was the best example because it was both highly overclockable and still required less power than Maxwell and anything AMD.

To the very least, nvidia seems to have switched from that policy, not to mention the power efficiency. I recall that from maxwell on, Jensen promised ever increasing power efficiency. That 1.9x power efficiency increase is nowhere to be seen...

 

[Dan_D]

9% faster does put it in 3090 territory for performance, and +14% power also puts it in 3090 territory. Ampere seems to be prophetically named, as the performance hinges on how much juice you can push through the chip.

Minus the memory difference, yes.

 

[Armenius]

When it was first announced, I hoped Ampere could be the next Pascal when overclocking. Getting high clocks and still being more power efficient than the past generation. Or maybe pull a Maxwell, having even higher clocks in spite of powerdraw but much better performance.

Unfortunately its quite the opposite, kind of reminds me of Kepler; very little clocks, less performance and much more power.

So don't waste your time with OC on RTX3080.

BTW apparently some people are getting great results with undervolting the RTX3080, you should give that a try.

I honestly don't care about overclocking headroom. If the video card is already close to its limit from the manufacturer then OC headroom shouldn't matter. I'm buying the card for OOB performance. Any overclocking potential on top is just a bonus.