Ask GN 113: Liquid Metal Pads? GPU Transplant? Side Intake Fans? Ft. Bearded Hardware
video description
Date: 2020-05-06
Related videos
Comments and reviews: 10
Techkey1
Very good and informative video. Thank You. I use to call the stuff I'ed mod Frankensteins when I was finished with them lol. I haven't done a crazy one in a long time. What I use right now for my GPU. XFX AMD Radeon VII My Cooling method = NZXT Kraken X62 280mm NZXT G12 (Personally Modded) does not fit with the G12 natively. Noctua NF-A14 iPPC-3000 PWM 4-PIN (X2) Noctua NF-A9x14 PWM (X1) Innovation Cooling IC Graphite Thermal Pad + NT-H1 Thermal Paste (yes that's right I said it, I combine, I have always done that and have had no problems lol I always get confused looks about it) MSI AFTERBERNER v4. 6. 2. 15743 Beta 4 IDLE TEMP=27C UNDERLOAD TEMPS=43C (With 8 hour gaming fluctuations included) 47C (with 100% stress testing and no fluctuations) Fan 35/50% (Idle/Full Load) 1. 150V (1150mV) Power Limit % +99 Core Clock 1900(MHz) Memory Clock 1176(MHz No Crashes) - 1200(MHz doable however for me screen goes blank after about 6 hours) I've seen better clocks from others. All tested under 100% load 4K MAX 2 hours Using Kombustor v4. 1. 5. 0. I do everything in 4K unless that option is not available. I only use Noctua and/or Artic Silver for my Pastes' and Innovative Cooling for Pads and if Liquid Metal, I use Thermal Grizzly (Just my personal preferences. I remove pre paste from AIO and replace, add thermal pad, then paste, then GPU. (Sooo. AIO/paste/pad/paste/GPU) Just in case anyone was wondering. - Yes it sounds crazy but I have been fine for years doing this on GPU's and CPU's and it works for me (especially with these results, do what works for you. - My Current Build NZXT H510 Bgears b-PowerMiner 1200W 80Plus Plat PSU ASRock X570 Taichi G. Skill Ripjaws V Series 32GB DDR4 3200 (X1) AMD Ryzen 7 3700X OC'ed to 4. 0GHz (on air with stock cooler and paste until I get my AIO) 43C Idle/68C Full Load. XFX AMD Radeon VII + NZXT G12 NZXT Kraken X62 280mm Noctua NF-A14 iPPC-3000 PWM 4-PIN (X3) Noctua NF-F12 iPPC 3000 PWM 4-PIN (X1) Noctua NF-A9x14 PWM (X1) Intel 660p Series M. 2 2280 2TB (Main drive) Samsung 860 QVO Series 2. 5 4TB SSD (Secondary drive) WD 10TB Surveillance (X1) WD 5TB Black (X2) WD 2TB Green (X1)
reply
Very good and informative video. Thank You. I use to call the stuff I'ed mod Frankensteins when I was finished with them lol. I haven't done a crazy one in a long time. What I use right now for my GPU. XFX AMD Radeon VII My Cooling method = NZXT Kraken X62 280mm NZXT G12 (Personally Modded) does not fit with the G12 natively. Noctua NF-A14 iPPC-3000 PWM 4-PIN (X2) Noctua NF-A9x14 PWM (X1) Innovation Cooling IC Graphite Thermal Pad + NT-H1 Thermal Paste (yes that's right I said it, I combine, I have always done that and have had no problems lol I always get confused looks about it) MSI AFTERBERNER v4. 6. 2. 15743 Beta 4 IDLE TEMP=27C UNDERLOAD TEMPS=43C (With 8 hour gaming fluctuations included) 47C (with 100% stress testing and no fluctuations) Fan 35/50% (Idle/Full Load) 1. 150V (1150mV) Power Limit % +99 Core Clock 1900(MHz) Memory Clock 1176(MHz No Crashes) - 1200(MHz doable however for me screen goes blank after about 6 hours) I've seen better clocks from others. All tested under 100% load 4K MAX 2 hours Using Kombustor v4. 1. 5. 0. I do everything in 4K unless that option is not available. I only use Noctua and/or Artic Silver for my Pastes' and Innovative Cooling for Pads and if Liquid Metal, I use Thermal Grizzly (Just my personal preferences. I remove pre paste from AIO and replace, add thermal pad, then paste, then GPU. (Sooo. AIO/paste/pad/paste/GPU) Just in case anyone was wondering. - Yes it sounds crazy but I have been fine for years doing this on GPU's and CPU's and it works for me (especially with these results, do what works for you. - My Current Build NZXT H510 Bgears b-PowerMiner 1200W 80Plus Plat PSU ASRock X570 Taichi G. Skill Ripjaws V Series 32GB DDR4 3200 (X1) AMD Ryzen 7 3700X OC'ed to 4. 0GHz (on air with stock cooler and paste until I get my AIO) 43C Idle/68C Full Load. XFX AMD Radeon VII + NZXT G12 NZXT Kraken X62 280mm Noctua NF-A14 iPPC-3000 PWM 4-PIN (X3) Noctua NF-F12 iPPC 3000 PWM 4-PIN (X1) Noctua NF-A9x14 PWM (X1) Intel 660p Series M. 2 2280 2TB (Main drive) Samsung 860 QVO Series 2. 5 4TB SSD (Secondary drive) WD 10TB Surveillance (X1) WD 5TB Black (X2) WD 2TB Green (X1)
reply
UltraMind
#ASKGN I have this very interesting question! Why reviewers never mention the single and double precision GFLOPS performance of computational devices like CPUs and GPUs? The GFLOPS performance is in fact very revealing characteristic of the device! GFLOPS allow to accurately compare devices that are very similar in performance, it shows the advantage of binned parts, GPUs with better VRMs etc. For example I am still running i7 2600k and I was wondering how much better are new processors like 9900k or 3900x. I measured my CPU with Aida64 showing 240 GFLOPS single precision and there is my answer, 9900k is at 1226GFLOPS (5 times better) and 3900x is at 1546GFLOPS (6 times better. Here we can also catch the single core advantage of the 9900k dividing 1226 GFLOPS / 6 cores = 204. 3 GFLOPS per core, while for 3900x 1546 GFLOPS/12 cores = 128. 8 GFLOPS per core. GFLOPS data is from lanoc. org. they measure cache speeds, cache latency, GFLOPS, GIOPS. It would be nice if GN presents at least GFLOPS charts for CPUs and GPUs. Thanks.
reply
#ASKGN I have this very interesting question! Why reviewers never mention the single and double precision GFLOPS performance of computational devices like CPUs and GPUs? The GFLOPS performance is in fact very revealing characteristic of the device! GFLOPS allow to accurately compare devices that are very similar in performance, it shows the advantage of binned parts, GPUs with better VRMs etc. For example I am still running i7 2600k and I was wondering how much better are new processors like 9900k or 3900x. I measured my CPU with Aida64 showing 240 GFLOPS single precision and there is my answer, 9900k is at 1226GFLOPS (5 times better) and 3900x is at 1546GFLOPS (6 times better. Here we can also catch the single core advantage of the 9900k dividing 1226 GFLOPS / 6 cores = 204. 3 GFLOPS per core, while for 3900x 1546 GFLOPS/12 cores = 128. 8 GFLOPS per core. GFLOPS data is from lanoc. org. they measure cache speeds, cache latency, GFLOPS, GIOPS. It would be nice if GN presents at least GFLOPS charts for CPUs and GPUs. Thanks.
reply
QuantumBraced
AskGN: Steve, could you talk about the slowdown of Moore's Law? It's not discussed nearly enough in the tech youtuber space. It seems to me progress has slowed down to a crawl. I just compared Geekbench 5 scores between the 3950X and the Q9650 quad-core from 2008. The 3950X is less than 3 times faster in single-core and 10 times faster in multicore. That is pathetic improvement over 12 years. In the 80s, 90s and first half of 2000s, you could expect at least 100x improvement over a decade and it was all single-core so perfectly scalable. RAM prices have not budged in 5 years, 16GB of DDR4 2666 costs roughly the same now as it did in 2015. Again, unthinkable in previous decades. We know generally the reasons why it's happening, exponential growth is not sustainable, but could you please give your take on it. Thanks!
reply
AskGN: Steve, could you talk about the slowdown of Moore's Law? It's not discussed nearly enough in the tech youtuber space. It seems to me progress has slowed down to a crawl. I just compared Geekbench 5 scores between the 3950X and the Q9650 quad-core from 2008. The 3950X is less than 3 times faster in single-core and 10 times faster in multicore. That is pathetic improvement over 12 years. In the 80s, 90s and first half of 2000s, you could expect at least 100x improvement over a decade and it was all single-core so perfectly scalable. RAM prices have not budged in 5 years, 16GB of DDR4 2666 costs roughly the same now as it did in 2015. Again, unthinkable in previous decades. We know generally the reasons why it's happening, exponential growth is not sustainable, but could you please give your take on it. Thanks!
reply
Fletcher
Indium solutions can be as good as 80 W/mK, but there are numerous issues with using them as thermal pads in DIY computer applications. You mentioned the issues with compression and flow at low pressure and temperature, but there are also issues with oxide formation, especially with high In content solutions. Mostly I've seen them in power semiconductor and photonic applications, as well as cryogenic applications. It might be interesting to use In for cryo cooling for extreme OCing, but applying it would be more like soldering the die to the pot than applying thermal grease. Another potential use case would be in high power modular systems, like those found in military and aerospace environments. Ultimately, as cool as the idea is I can't see it being practical outside of extreme applications.
reply
Indium solutions can be as good as 80 W/mK, but there are numerous issues with using them as thermal pads in DIY computer applications. You mentioned the issues with compression and flow at low pressure and temperature, but there are also issues with oxide formation, especially with high In content solutions. Mostly I've seen them in power semiconductor and photonic applications, as well as cryogenic applications. It might be interesting to use In for cryo cooling for extreme OCing, but applying it would be more like soldering the die to the pot than applying thermal grease. Another potential use case would be in high power modular systems, like those found in military and aerospace environments. Ultimately, as cool as the idea is I can't see it being practical outside of extreme applications.
reply
Phu
The GPU can be done, I am a Rework engineer who worked to install and do research on the first BGA machine. It is NOT for the average hobbyist. It takes a stereo microscope a set of hot micro tweezers a ton of patience and a bunch of time. In the defense and aerospace industry I have done them to increase memory capacity of obsolete systems on the B-52 avionics, I did it in the early 2000s to update the encoder boards in the Hubble telescope (It was cool because I built the originals at Sperry univac when I first started out) I also did them to update the U2 for NASA as part of a avionics update to allow it to operate on civilian networks. BGA processors are not easy but it can be done as a once in a lifetime experience. Btw, if you overheat the board, its toast forever.
reply
The GPU can be done, I am a Rework engineer who worked to install and do research on the first BGA machine. It is NOT for the average hobbyist. It takes a stereo microscope a set of hot micro tweezers a ton of patience and a bunch of time. In the defense and aerospace industry I have done them to increase memory capacity of obsolete systems on the B-52 avionics, I did it in the early 2000s to update the encoder boards in the Hubble telescope (It was cool because I built the originals at Sperry univac when I first started out) I also did them to update the U2 for NASA as part of a avionics update to allow it to operate on civilian networks. BGA processors are not easy but it can be done as a once in a lifetime experience. Btw, if you overheat the board, its toast forever.
reply
Ben
kind of a lazy answer from buildzoid -- 0. 3ft of 22g wire is 0. 005ohms. 0. 5ft for 0. 008ohms. so yeah, wire length and gauge affects it, but it's not going to defeat the mod if you account for it. and having a bread board opens up ways to work around it. the resistance of your wires could be overcome by boosting the DC on the breadboard from the 12v source. So it wouldn't matter if you have extra resistance going in (say 11. 75v); cause you could boost to 12. 05v on your breadboard then the wire back in takes it down to 11. 95. Add a couple transistors and a cap and you could add a pot to fine tune it. sounds like a really cool project to me. I might give it a go especially since it can be tested with a bench supply / not hooked up to the card.
reply
kind of a lazy answer from buildzoid -- 0. 3ft of 22g wire is 0. 005ohms. 0. 5ft for 0. 008ohms. so yeah, wire length and gauge affects it, but it's not going to defeat the mod if you account for it. and having a bread board opens up ways to work around it. the resistance of your wires could be overcome by boosting the DC on the breadboard from the 12v source. So it wouldn't matter if you have extra resistance going in (say 11. 75v); cause you could boost to 12. 05v on your breadboard then the wire back in takes it down to 11. 95. Add a couple transistors and a cap and you could add a pot to fine tune it. sounds like a really cool project to me. I might give it a go especially since it can be tested with a bench supply / not hooked up to the card.
reply
Dale
I once decided I wanted to change cases from my old HAX-X. I bought the HAF-X when it came out to upgrade and migrate an existing build from my most favorite case of all time. I decided some RGB and better aesthetics were needed. I tried a fractal meshifiy was disappointed in OC stability and loss of frames due to temps. I tried a master case h500 which was better but also disappointed. All I/O QOL inmpovments missing on my old cases. Both my 10+ year old HAF-X and Antec 1200 Remain my all time favorites and they cool better too. To this day both run my primary machines. Both cases provide better temps and higher more stable overckocks than the cases I returned. Thanks Newegg! TL; DR - BRING BACK SIDE INTAKES AND OPEN FRONT ENDS!
reply
I once decided I wanted to change cases from my old HAX-X. I bought the HAF-X when it came out to upgrade and migrate an existing build from my most favorite case of all time. I decided some RGB and better aesthetics were needed. I tried a fractal meshifiy was disappointed in OC stability and loss of frames due to temps. I tried a master case h500 which was better but also disappointed. All I/O QOL inmpovments missing on my old cases. Both my 10+ year old HAF-X and Antec 1200 Remain my all time favorites and they cool better too. To this day both run my primary machines. Both cases provide better temps and higher more stable overckocks than the cases I returned. Thanks Newegg! TL; DR - BRING BACK SIDE INTAKES AND OPEN FRONT ENDS!
reply
Carboxylated
ASUS is doing liquid metal on all their ROG PC's going forward into 2020. They have an automated robot arm that applies the LM to the CPU/GPU with extreme precission and also uses a cradle system that is held above the entire MOBO to protect or shield it from LM splatter upon application. I want to see if ASUS will do this for customers who call and send in their laptops for a fee. I want to liquid metal paste all my computers and macbook's. I use all my systems rigorously and everyone of them run really hot under the demanding settings I sport. LM would solve a lot of my heat issues and CPU throttle. To much of a p%ssy to apply LM on my own, I would ruin some hardware for sure!
reply
ASUS is doing liquid metal on all their ROG PC's going forward into 2020. They have an automated robot arm that applies the LM to the CPU/GPU with extreme precission and also uses a cradle system that is held above the entire MOBO to protect or shield it from LM splatter upon application. I want to see if ASUS will do this for customers who call and send in their laptops for a fee. I want to liquid metal paste all my computers and macbook's. I use all my systems rigorously and everyone of them run really hot under the demanding settings I sport. LM would solve a lot of my heat issues and CPU throttle. To much of a p%ssy to apply LM on my own, I would ruin some hardware for sure!
reply
pagaiba
Hey Steve, in case you want some more info on the Indium sheets, it's nothing specifically made for CPU's but usually used to cool down lasers. It's used as a TIM and it's bassically just a very thing sheet of Indium that gets crushed in between. Because indium has thermal conductivity of 86W/mk and is very soft, with enough preassure it will plastically deform and fill those microscopic gaps. The perforated sheets exist because since it's 2/3 or 1/2 of the material without the perforations, it takes that much less force to deform them so it might be more doable under a CPU block. You can get them on ebay and aliexpress, and there are some specialized companies making them.
reply
Hey Steve, in case you want some more info on the Indium sheets, it's nothing specifically made for CPU's but usually used to cool down lasers. It's used as a TIM and it's bassically just a very thing sheet of Indium that gets crushed in between. Because indium has thermal conductivity of 86W/mk and is very soft, with enough preassure it will plastically deform and fill those microscopic gaps. The perforated sheets exist because since it's 2/3 or 1/2 of the material without the perforations, it takes that much less force to deform them so it might be more doable under a CPU block. You can get them on ebay and aliexpress, and there are some specialized companies making them.
reply
salut
Why doesn't AiO coolers just use standard car coolant? I've used some extremely old new stuff (from Woolco to help date it) to restore a cooler that was completely gunked up. I've left that sitting for around 2 years because I didn't have a use for it/didn't have the proper mounting hardware (only had the LGA 2011 screw plate) but I recently took it and ziptied it to a 3770 and it worked perfectly. Absolutely 0 issues except maybe I underfilled it as it was noisy in some orientations, but that didn't affect performance. Also tried some experiments by cooling the radiator in a bucket of dry iced coolant, which also worked fine.
reply
Why doesn't AiO coolers just use standard car coolant? I've used some extremely old new stuff (from Woolco to help date it) to restore a cooler that was completely gunked up. I've left that sitting for around 2 years because I didn't have a use for it/didn't have the proper mounting hardware (only had the LGA 2011 screw plate) but I recently took it and ziptied it to a 3770 and it worked perfectly. Absolutely 0 issues except maybe I underfilled it as it was noisy in some orientations, but that didn't affect performance. Also tried some experiments by cooling the radiator in a bucket of dry iced coolant, which also worked fine.
reply
Add a review, comment
Other channel videos
