Date: Sunday , September 23, 2018
We did a quick and dirty GeForce RTX 2080 Overclocking Preview with Scanner a couple a days ago when we first took delivery of the RTX 2080 cards we purchased from NVIDIA. (Our RTX 2080 TI card purchase has been delayed.) In that review we talked about using NVIDIA's new OC Scanner technology, which did work, and did work fairly well. OC Scanner however does not touch the memory clocks, so of course we wanted to find out what the RAM was capable of.
There was a bit of confusion as to warranty and using OC Scanner. After being told by official NVIDIA support on its website that OC Scanner is not recommended and does void your warranty, NVIDIA PR reached out to us and told us that OC Scanner does NOT void your RTX card warranty. We are glad to get a solid answer on that for sure.
Surprisingly, we used OC Scanner on two different systems, one AMD, and one Intel, and got very different results. We also got very different results than a website that signed NVIDIA's NDA and were sampled directly. Keep in mind that we are using privately purchased retail cards from NVIDIA, and these cards were not given to us for review. We will discuss that a bit later.
For all of our overclocking below, you are looking at a fully heat-loaded system. However we are testing on an open testbench, with an ambient temperature of ~75F/24C, so surely we are still seeing overclocking levels shown in its best light. We use the Heaven "benchmark" to load these cards. While being a bit old, Heaven is still very capable of pushing a GPU to ~100% load, and it loops endlessly so it is easy to set up and tweak while the card is loaded. This all means that none of our results are shown with a "cold card." So in that vein, we are showing off the 2080's overclocking abilities at its "worst." Heaven is running at 1440p, with Ultra quality settings, Extreme tessellation, and 8X MSAA.
We have set out base line at 1800MHz because that is the stated boost clock for the RTX 2080 Founders Edition. As you can see, as is usual for NVIDIA GPUs since boost has come into existence, we are getting in-game clocks well above our base line, averaging 1878MHz at stock.
Our Scanner clocks did well in our opinion averaging 1966MHz. Again, keep in mind, OC Scanner does not touch the RAM clocks.
We also decided to simply increase our fan speed to maximum which is 3700RPM. Previously, while letting the card control its own fan speed, the maximum we had seen was ~1980RPM. With our XSPC water cooled system, the decibel reading at 4 feet is ~47.7dB. With the RTX 2080 FE fans at 100%, after spin-up, the decibel reading calms down to ~54.8dB. While certainly this is "loud," I did spend a many hours working around this sitting feet from me on an open test bench, and I did not find the sound profile annoying. Surely, you can hear the RTX 2080 FE fans, but it did not grate on my nerves. YMMV, but if you are gaming with headphones, this should be a non-issue. That said, the tradeoff that you get in terms of overclock, is probably not worth the tradeoff being that it is only about a ~40MHz increase.
For our Hand OC, we are using fans at 100%. In fact, we have pushed all of our exposed tweaks to their maximums besides GPU Clock and Memory Clock. Target Power and GPU Temp are at +124/88C. The RTX 2080 GPU Voltage is also pushed to +100.
The screenshots above show our hand overclock on the left and our stock clocks and settings on the right after the 2080 FE has been heat-loaded.
The only two variables left to dial in are GPU Clock and Memory Clock. Our value used for GPU Clock was +155. Once you start playing around with the overclock on this card you will find that it moves its GPU clocks in increments of 15MHz. If you dial these in at +145 you will likely immediately get a 2055MHz clock, with +150 you will get 2070MHz, and with +165 you will likely see 2085MHz. These values were seen with a "warm" card. If you apply these clocks on a "cold" card, you will see higher values. However, once you run these settings on a fully heat-soaked card you see different clocks, so these are very dynamic in how the card reacts to heat and GPU load.
Our final clock setting of +155 resulted in a static GPU clock of 2055MHz. We were also able to push the Memory Clock to a +1300 setting resulting in a static RAM clock of 8300MHz. Stock is 7000MHz. All overclocks above gave us artifact-free runs. Our GPU ran at 61C with an exhaust temperature of 101F/38C. Certainly our fans at 100% are moving a lot more air than those are at stock.
We do not look to Heaven to be a good tool for gameplay evaluation, but this Hand Overclock gave us an 11.72% increase in framerate over stock clocks.
We did have multiple crash issues running EVGA Precision X1 Scanner on our AMD Ryzen system we were testing on. I was only able to get Scanner to complete fully twice on our Ryzen system. I did move over to an Intel based system to see if we had the same issues there. Scanner ran much better on our Intel system with an 8600K, but it was not without difficulties as well. The process was much more successful on the whole however. Interestingly, OC Scanner gives your system a "Score" at its completion. Our best AMD score was "93." Our best Intel score was "41." The OC Scanner overclock was slower on the Intel system as well by about ~30MHz under load. Remember we see dynamic clocks while using the Scanner overclock. One thing I did find interesting, was a HardForum user posting a link to the Hardware Canucks review in which they used OC Scanner on a 2080 FE card sampled by NVIDIA. They got a OC Scanner score of "138."
If you like our content, please support HardOCP on Patreon.