14:20 makes sense, LM's great thermal conductivity helps for hot spots. i used LM on my i5-4690k and if u think about it, the TIM needs to conduct ~100W thru half a small stamp's worth of die size compared to thermal paste which tends to conduct heat from the IHS, which is about the size of 2 stamps, into a heatsink, which is about the same size, thus needing at least 3x as much thermal conductivity to not be the bottleneck. would apply for a gpu with hot spots too. my 3080 xc3 rn is 87C core 100C hotspot (the future gave better sensors!) and it's a bargain mining gpu. i'm getting LM in since i need to reapply for my aio (i think the copper sucked it all up, my temps kinda suck now) so i'll have LM anyways and i'm already performing surgery. the 3080 xc3 is a copper coldplate so it's possible, but i may have to come back in half a year and see if it needs a touchup before i seal it up forever.
I'm glad I found this video, because my build plan calls for LM on a liquid cooled 3080, and I'd only seen videos with people using LM with air coolers. It boggles my mind because the air cooling bottleneck just won't be able to take advantage of the heat transfer properties of LM. The reason it looks so great on CPUs is that it is typically used with a delid to remove the middling to bad TIM between the die and IHS. GPUs on the other hand are always direct die, so you don't have that extra bottleneck of the IHS. Feeling a lot better about it. Hopefully EK rolls out FE blocks soon.
yup once you go liquid metal on a GPU you cant go back, im going to be doing it on my 3080 and 3090 as well once the blocks come out. When you order your waterblock just make sure the coldplate is nickel plated so the LM doesnt get absorbed into bare copper
Frame Chasers yep, already did that for my CPU block-nickel plated velocity block. If you check out Optimum Tech’s dual radiator Ncase M1 you’ll see what I’m working on.
10degrees can make a big difference on clock due to the card declockong at set temps or percents. Ive noticed mine declocks at 50c and 60c. Because of this you will always clock higher at lower temps even if it's just a few degrees that happens to keep you under the declock threshold. Thanks for sharing the results, I'm about to do mine tomorrow. Had great results on cpu. I7700k from 4.2 now 5.2
lol im going thru ur old videos. ur a true enthusiast. i got a 3080 xc3 recently and considered LM but decided against it since the thermal conductivity increase probably doesn't matter with a low temp diff (air cooler) anyways. but it is fun to do. used it on my 980ti ftw with nail polish for the caps. i gained 1-2C lmao but it was an experience, something i think you can def relate to
Most tests I've seen were done with stock air coolers. Their results were lack luster due to the limits of air cooling. Other tests I've seen were done on old GPU's with 28nm process. The transistor density was not high enough to create hot spots, unlike CPU's even at 45nm would show. I have a 5700XT with a liquid loop on metal. I had similar results going from conventional grease to liquid metal. (8 - 10C drop in hot spot and edge temps)
10 degrees definitely accounts for the extra 60 MHz. Boost speeds have been temperature dependent since Pascal, keeping the core under 50C produces the highest boosts, by 60C you have lost your highest boost bins, by 70C you are getting down to "normal" boost clocks (~1900s), by 80C you are down to base clocks. Most people just don't get to experience this because normally you need liquid cooling to go down that low.
I mean like I had to actually increase the mhz slider to go that high, it doesn’t boost passed the number I set in afterburner. Before I had the slider at 2160, and then the clocks would settle at 2145 under load at 50c. After the liquid metal I increased the slider to 2205, and it locked there. I’d say the temperature difference allowed the card to remain locked at the highest boosted bin which was 2205, but the liquid metal and transference of hotspot heat is what allowed it to boost that other 40mhz. I think that makes sense, haha
@@FrameChasers hi I have a question, I have a Founders 2080ti, I have a system with 4 radiators, and double pump in serial, I have a problem if you can say so, practically my video card in full load reaches 34 degrees, at 2070 mhz the problem is that I can not go beyond 2070 mhz with the coreclock !!!! I do not understand why, when it was 42 degrees with a single pump and 3 radiators it always went at 2070mhz, now that I pulled down another 8 degrees not about OC anyway !!! memories are stable at 8350mhz .. why? how can i fix to go on coreclock? I have an ek waterblock .. thanks
I really love those kinda videos. Your statement @ 3:04 shows that you know your stuff. Many people are actually having this issue with their 1080 Ti's and the default response in many forums is "dying card". Also your music choice really surprised me because I used to listen to this :D Sub earned!
2200 is an amazing OC dude, congrats! Going to put liquid metal on mine 2080ti next teardown. Is that a Bitspower block? Mine used different heights for the memory pads - some were .5mm and others were 2.5mm. I ended up selling my Bitspower block and replacing it with a Heatkiller 4, which resulted in massive drop in temps (11c less under load). the HK block is 36% heavier than the bitspower block, so more surface area, and thinner thermal pads. I used the 17w/mk thermal pads fujipoly ultra extreme. You should do a video on your shunt mod, I don't see it on the card...
Do you mean the waterblock? Its just an EK reference one, the cheapest one I could find at the time. I've never actually tried any other brand of waterblock except EK. All the memory pads are the same though, they seem to be .5mm. 17w/mk thermal pads?? now this I want to look into, I feel like my memory could overclock further if I had better pads, I was thinking of trying thermal paste on the memory IC's as well. I can do that, I figured people already knew about the resistor shunt mod. Good Idea il do that!
more mhz means more heat so you gain in theory more than than 10 degress. You always have to calculate than in normal stock condition you would generat even more heat with this more mhz and from this level you go down. So its more than 10 degress.
14:20 makes sense, LM's great thermal conductivity helps for hot spots. i used LM on my i5-4690k and if u think about it, the TIM needs to conduct ~100W thru half a small stamp's worth of die size compared to thermal paste which tends to conduct heat from the IHS, which is about the size of 2 stamps, into a heatsink, which is about the same size, thus needing at least 3x as much thermal conductivity to not be the bottleneck. would apply for a gpu with hot spots too. my 3080 xc3 rn is 87C core 100C hotspot (the future gave better sensors!) and it's a bargain mining gpu. i'm getting LM in since i need to reapply for my aio (i think the copper sucked it all up, my temps kinda suck now) so i'll have LM anyways and i'm already performing surgery. the 3080 xc3 is a copper coldplate so it's possible, but i may have to come back in half a year and see if it needs a touchup before i seal it up forever.
I'm glad I found this video, because my build plan calls for LM on a liquid cooled 3080, and I'd only seen videos with people using LM with air coolers. It boggles my mind because the air cooling bottleneck just won't be able to take advantage of the heat transfer properties of LM. The reason it looks so great on CPUs is that it is typically used with a delid to remove the middling to bad TIM between the die and IHS. GPUs on the other hand are always direct die, so you don't have that extra bottleneck of the IHS.
Feeling a lot better about it. Hopefully EK rolls out FE blocks soon.
yup once you go liquid metal on a GPU you cant go back, im going to be doing it on my 3080 and 3090 as well once the blocks come out. When you order your waterblock just make sure the coldplate is nickel plated so the LM doesnt get absorbed into bare copper
Frame Chasers yep, already did that for my CPU block-nickel plated velocity block. If you check out Optimum Tech’s dual radiator Ncase M1 you’ll see what I’m working on.
10degrees can make a big difference on clock due to the card declockong at set temps or percents. Ive noticed mine declocks at 50c and 60c. Because of this you will always clock higher at lower temps even if it's just a few degrees that happens to keep you under the declock threshold. Thanks for sharing the results, I'm about to do mine tomorrow. Had great results on cpu. I7700k from 4.2 now 5.2
I appreciate your shunt mod videos and your 2080ti mod videos and overclocking. They have been much help.👍
lol im going thru ur old videos. ur a true enthusiast. i got a 3080 xc3 recently and considered LM but decided against it since the thermal conductivity increase probably doesn't matter with a low temp diff (air cooler) anyways. but it is fun to do. used it on my 980ti ftw with nail polish for the caps. i gained 1-2C lmao but it was an experience, something i think you can def relate to
Most tests I've seen were done with stock air coolers. Their results were lack luster due to the limits of air cooling. Other tests I've seen were done on old GPU's with 28nm process. The transistor density was not high enough to create hot spots, unlike CPU's even at 45nm would show. I have a 5700XT with a liquid loop on metal. I had similar results going from conventional grease to liquid metal. (8 - 10C drop in hot spot and edge temps)
10 degrees definitely accounts for the extra 60 MHz. Boost speeds have been temperature dependent since Pascal, keeping the core under 50C produces the highest boosts, by 60C you have lost your highest boost bins, by 70C you are getting down to "normal" boost clocks (~1900s), by 80C you are down to base clocks.
Most people just don't get to experience this because normally you need liquid cooling to go down that low.
I mean like I had to actually increase the mhz slider to go that high, it doesn’t boost passed the number I set in afterburner. Before I had the slider at 2160, and then the clocks would settle at 2145 under load at 50c. After the liquid metal I increased the slider to 2205, and it locked there. I’d say the temperature difference allowed the card to remain locked at the highest boosted bin which was 2205, but the liquid metal and transference of hotspot heat is what allowed it to boost that other 40mhz. I think that makes sense, haha
@@FrameChasers hi I have a question, I have a Founders 2080ti, I have a system with 4 radiators, and double pump in serial, I have a problem if you can say so, practically my video card in full load reaches 34 degrees, at 2070 mhz the problem is that I can not go beyond 2070 mhz with the coreclock !!!! I do not understand why, when it was 42 degrees with a single pump and 3 radiators it always went at 2070mhz, now that I pulled down another 8 degrees not about OC anyway !!! memories are stable at 8350mhz .. why? how can i fix to go on coreclock? I have an ek waterblock .. thanks
I really love those kinda videos. Your statement @ 3:04 shows that you know your stuff. Many people are actually having this issue with their 1080 Ti's and the default response in many forums is "dying card". Also your music choice really surprised me because I used to listen to this :D
Sub earned!
Definitely want to try with my gpu.
great video and amazing results~
i gues ima throw now some liquid metal on my 1080 aswell~ you just gained an new sub there nuuuumber 93!
Yusss thanks man! We’re almost at 100 I’m so excited, welcome!
2200 is an amazing OC dude, congrats! Going to put liquid metal on mine 2080ti next teardown. Is that a Bitspower block? Mine used different heights for the memory pads - some were .5mm and others were 2.5mm. I ended up selling my Bitspower block and replacing it with a Heatkiller 4, which resulted in massive drop in temps (11c less under load). the HK block is 36% heavier than the bitspower block, so more surface area, and thinner thermal pads. I used the 17w/mk thermal pads fujipoly ultra extreme. You should do a video on your shunt mod, I don't see it on the card...
Do you mean the waterblock? Its just an EK reference one, the cheapest one I could find at the time. I've never actually tried any other brand of waterblock except EK. All the memory pads are the same though, they seem to be .5mm. 17w/mk thermal pads?? now this I want to look into, I feel like my memory could overclock further if I had better pads, I was thinking of trying thermal paste on the memory IC's as well.
I can do that, I figured people already knew about the resistor shunt mod. Good Idea il do that!
more mhz means more heat so you gain in theory more than than 10 degress. You always have to calculate than in normal stock condition you would generat even more heat with this more mhz and from this level you go down. So its more than 10 degress.
Awesome results.
I thought the old saying was measure twice fill the loop once?
I end up usually filling and measuring 5 times 😑
What is the liquid you put around the chip before applying the liquid metal?
Flitz polish, use it on both surfaces before liquid metal application so it can really dig in