Cooling, Noise, and Power

My last set of benchmarks are some of the most important for people like me who do a lot of small form factor builds. In a large case you will most likely have all of the cooling you need and getting a high wattage power supply isn’t hard. Even noise is an issue with small form factor builds because typically they will be running warmer than other builds, spooling up the fans much higher than in other builds. So I always make sure to take a look at all three areas.

My first set of tests were focused on power usage. I did one using 3DMark Fire Strike curing the combined benchmark that puts a load on the CPU and GPU with in game realistic loads. Here our test system pulled 393 watts. This was right in between the original GTX 1080 Founders Edition and the overclocked GTX 1080. This was also 37 watts lower than the RTX 2080 FE. In my second test, I use AIDA64 to load the GPU up completely without any other CPU load. This is a higher load on the GPU than 3DMark but without the CPU load, you can see that our entire test system only pulled 270 watts. Again falling in between the two GTX 1080’s and above the overclocked GTX 1070 Ti as well.

graph34

graph35

My next tests were noise focused. For this, I used our decibel meter and tested the dual fan RTX 2070 FE with the fans turned up to 100% fan speed and again at 50% fan speed. Getting idle results in the office are inconsistent, but these two fan speeds show a good idea of the noise range you can expect from a card. A case with good cooling and a decent GPU cooler will most likely never go above 50% but a small form factor build or someone overclocking will be more likely to hear 100% fan speed or close to it. Now the RTX cooler design with two fans was a touch louder than pas Founders Edition cards at full speed and the 2070 was slightly higher than the 2080’s. I think this is due to the fans being closer together. This was still better than the GTX 1080 Ti. You can see in the RPM graph why the 1080 Ti was louder. The three RTX cards all spun at the same speed and the cards with aftermarket coolers ran much lower speeds and were much quieter as well.

graph36

graph37

My last tests were temperature focused. Here I loaded up the RTX 2070 FE using the AIDA64 stress test and documented the temperature once it leveled off. This is checked at 30 minutes or longer if it hasn’t leveled off by then. I test at both stock fan speeds and with the fans turned all the way up to get a look at how much performance is left in the cooler. The RTX 2070 and the RTX 2080 both came in at 66 degrees, the stock fan profile must target this temperature. With the fans turned all the way up things ran much cooler at 49 degrees. Interestingly enough the RTX 2080 did run two degrees cooler. I think that is because of the larger size of the card. Both results were in the bottom half of our charts, not bad at all for a non-aftermarket card.

graph38

graph39

When doing temperature testing (with the stock fan profile) I did also grab a few pictures of the card with our thermal imager. I was curious to see if the backplate would show a hotspot like both RTX 2080’s did and I really didn’t see that at all. In fact, the only warm areas on the RTX 2070 were at the top and bottom of the card where the heatsink is visible and where air is blown out. I did note the temperature of the motherboard below the card. With air blowing out the bottom on to the motherboard, I would highly recommend not putting an M.2 drive below any of these Founders Edition cards.

other 2

other 3

other 1

 

Log in to comment

VaporX's Avatar
VaporX replied the topic: #38570 16 Oct 2018 15:57
So based on your testing compared to my own in house testing with 3D Mark (I use 3D Mark as it gives a direct horse power to horse power comparative without brand specific optimizations getting in the way) the Sapphire Vega 56 is only about 3% slower but costs 20% less :-) NICE....
garfi3ld's Avatar
garfi3ld replied the topic: #38571 16 Oct 2018 19:29
Just saying, you could send a Sapphire Vega 56 over and we can put those numbers down on the same system. We have been lacking the 56 for all of this testing :(

We have 500 guests and one member online

supportus