This ASUS GTX 1080 Strix OC deal seems unreal!
This ASUS GTX 1080 Strix OC deal seems unreal!
Althought OC is not new to me I never dig into it because I felt the margin gains were not worth the "risk". Thing is, as long as you keep inside that given interval (considering the OC playing margins are already capped by the manufacturer) the risk is pratically inexistente. So straight to my question:
For the last 2 days I´ve been tweaking my Strix 1080 with GPU Tweak II (most recent version with proper clock numbers
). In Watchdogs 2 so far, I´ve managed to hit 2050 MHz without a slight problem - card stays at 60ºc, fan at 40%, voltage at 1050 mV, GPU usage 98-100% and power target between 98-106%. By default, GPU Tweak II only allows a certain degree of increase in both core clock and memory clock but there´s an option that removes that limit and allows you to go further as your tests permit. So I uncapped the frequency ceiling and added another 10 MHz (wich should be the incremental step I will use from now on, instead of 20). Anyway I do not intend to go over 2100 mHz and will probably stop at 2080 MHz and start tinckering with memory clock instead (I´ve seen some pretty crazy OCs going from 10k to 11k on memory frequency).
Regarding memory tweaking, everywhere I read I see that memory clocks OC more than core clocks. Mine is yet at 10010 - I´m somewhat reluctant about memory tweaking because usually coolers are more focused on the GPU chip than memory chips and there is no temperature reading from the memory chips so…
My concern: can I trust those numbers from GPU Tweak II considering last two versions were misreporting clocks? From where I´m seating it´s actually quiet - my case is nicely ventilated with two exhausting fans on top and the air coming out isn´t even warm.
What do you guys think? Should I trust GPU Tweak II numbers? Are these results inside the OC pattern others are getting? Should I push further on core clock or start pushing memory clocks and see what happens with temperature?
Sorry for long post.
Be well
I favor Msi Afterburner or EVGA Precision X over ASUS GPU tweak, both perform well with an ASUS card. I’ve also noticed several problems with GPU tweaks. Personally, I use Msi Afterburner for my GPU overclocking and have never encountered any issues.
The cooler on the Strix 1080 is sufficient for any overclocking or the other programs I mentioned will prevent you from raising the power limit or voltage excessively, which could cause significant heat. If I were you, I’d push the core to its limits until it becomes unstable, then reduce it slightly—starting around 2080MHz and increasing in 20MHz steps until stability returns. For most 1080 models, this would place the max core overclock between 2100MHz and 2250MHz.
Regarding memory, the Strix model features active cooling for all memory chips, ensuring they stay in contact with heatsinks and are actively cooled. You’re correct that it lacks a temperature sensor, so you’re taking some risk when overclocking memory. However, I haven’t heard any reports of memory failures in Strix cards, and the cooler design appears sufficient to maintain chip temperatures even during a modest overclock. If you choose to overclock memory, do so after reaching your stable core clock. Keep in mind that memory overclocks can sometimes interfere with core overclocks, so it’s best to apply memory overclocks once your core is fully stable. I’d suggest increasing memory frequency in 50MHz steps and checking stability before proceeding.
I favor msi afterburner or evga precision x more than ASUS GPU tweak, both should function well with an ASUS card. I've also heard about several problems with GPU tweaks. Personally, I use msi afterburner for my GPU overclocking and never faced any issues.
The cooler on the Strix 1080 is sufficient for any overclocking or the other programs I mentioned will prevent you from raising the power limit or voltage excessively, which could cause significant heat. If I were you, I'd push the core to instability, then gradually reduce it until stability returns—typically around 2100mhz to 2250mhz for most 1080s.
Regarding memory, the Strix model features active cooling for all chips, ensuring they stay in contact with heatsinks and are actively cooled. You're correct that it lacks a temperature sensor, so you're taking some risk with overclocking memory. However, I haven't heard any reports of failing memory chips in Strix cards, and the cooler design seems adequate to maintain temperatures during a bit of overclocking. If you decide to overclock memory, I recommend doing it after reaching your stable core clock. Memory overclocks can sometimes interfere with core overclocks, so it's best to apply memory overclocks once your core is stable. You could increase the frequency in 50mhz steps and check for stability before repeating.
Exactly what I think.
I've been thinking about keeping my OC at 2080/2100 and memory at 10500/10800, if it proves stable. With Watchdogs2 offering a broad range of graphics settings—from detailed textures to lighting effects—it could serve as a solid benchmark reference.
Thanks for your response.
From what I've learned, Watch Dogs 2 can be quite demanding on the CPU, which might make it less ideal for testing GPU performance. It's worth checking if the game consistently uses 100% CPU during playtime; if so, it should serve as a solid benchmark. If not, alternatives like Unigene Heaven or games such as Witcher 3 could be better suited.
GPU usage stays mostly at 100%, just a slight hover without dropping below 98%. Before overclocking it would stay between 96-98% and the temperature would never exceed 50ºC. That’s why I chose to boost my card, thinking it was clearly lagging. I almost decided on core clock adjustments—yesterday I increased it by 20 MHz, now it runs at 2063 MHz and if I calculate correctly, that’s about 1063 mV.
I’ve observed something interesting—the card maintained around 1050 mV during a nearly 70 MHz range. Yesterday it jumped to 1063 mV and I was sure I saw an improvement in FPS. Today my goal is to hit 2080 MHz and stop there. After that, I’ll focus on memory tuning.