|
SwissArmyDruid posted:You'd think that Steve would be pitchforking Nvidia over this poo poo by now. He might lose his hair over it, and nobody wants that.
|
#
?
Oct 22, 2018 20:52
|
|
Have you seen my apex seals? I seem to have lost them.
|
|
| # ? Jun 11, 2024 05:08 |
|
|
VelociBacon posted:Besides the heatsink on the VRMs, that goon should probably also rig up some 120mm fan to blow across them/the other pcb elements. A little airflow goes a long way. The G10/G12 mount actually has a smaller (80mm?) fan mounted towards the front (power connector-side) of the card for exactly that reason: to put some airflow over the VRMs. The VRAM chips don't tend to get a whole lot of airflow, but they usually run pretty cool anyhow, though throwing some heatsinks on those if you've got 'em left over from the VRMs won't hurt anything.
|
|
#
?
Oct 22, 2018 21:23
|
|
|
craig588 posted:I put a H100i on my 680 way back and forgot to plug in the fans. It got to the 60s with just the silent case airflow. GPUs just don't get hot like CPUs. I'm skeptical that GPUs run cooler than CPUs under gaming loads. Even the 680 dissipates sustained ~175W under full load, which is roughly equivalent to a 8700K@5GHz (minus a little bit for VRMs, VRAM, etc.). A 2080 Ti is dissipating closer to sustained 275W.
|
|
#
?
Oct 22, 2018 21:48
|
|
|
The big difference is that GPUs usually run bare-die while CPUs have an IHS. If we went back to bare-die packaging on CPUs, temps would come down a fair bit especially on blast-furnace parts like the 9900K.
|
|
#
?
Oct 22, 2018 21:57
|
|
|
Stickman posted:I'm skeptical that GPUs run cooler than CPUs under gaming loads. Even the 680 dissipates sustained ~175W under full load, which is roughly equivalent to a 8700K@5GHz (minus a little bit for VRMs, VRAM, etc.). A 2080 Ti is dissipating closer to sustained 275W. No, they do. The amount of heat (energy) generated doesn't really matter as far as temperature goes - what matters is how quickly you are able to transfer that heat away from the source. Thermodynamics tells us that heat transfer goes quicker the sharper the temperature gradient is. It doesn't matter if you're generating 15W or 150W, if you can transfer it away fast enough you stay cool, if you can't you heat up until you reach equilibrium. The reason GPU's stay so much cooler is precisely because they're better at heat transfer, and that pretty much comes down to two things. First, Pascal GPU's have gigantic silicon dies compared to an 8th gen Intel CPU (and Turing even more so). A 1080Ti has a 471mm² die, more than three times bigger than the 8700K's 149mm², but it doesn't generate anywhere near three times the heat (and the 2080Ti is even bigger at 754mm²). That means the energy density is much lower, and the heat transfer occurs over a bigger area. Second, the GPU's have no heatspreader - the die is in direct contact with the coldplate via a single layer of thermal paste, compared to on a CPU where it goes silicon -> thermal paste (or liquid metal, if you've delidded) -> thick copper heatspreader -> more thermal paste -> coldplate. So, even when delidded, a 8700K is severely handicapped in the heat transfer department when compared to a 1080Ti. TheFluff fucked around with this message at 22:14 on Oct 22, 2018 |
|
#
?
Oct 22, 2018 22:07
|
|
|
Makes sense, thanks! I was thinking of delidded CPUs, but I didn't consider the die size.
|
|
#
?
Oct 22, 2018 22:17
|
|
|
Stickman posted:Makes sense, thanks! I was thinking of delidded CPUs, but I didn't consider the die size. Yup! That's why I can cool a 1080Ti with a 120mm radiator and a single fan just north of "off completely," while my moderately OC'ed 5820k has a 280mm rad on it and still can get dicey when run full tilt with fans at 100%. Physics is cool!
|
|
#
?
Oct 22, 2018 22:24
|
|
|
Even liquid metal is a poor thermal conductor compared to copper - we're talking like 70 W/mK to somewhere in the 3-400 W/mK range. It's great compared to thermal paste though, which tends to be somewhere around 10 W/mK.
|
|
#
?
Oct 22, 2018 22:25
|
|
|
TheFluff posted:Even liquid metal is a poor thermal conductor compared to copper - we're talking like 70 W/mK to somewhere in the 3-400 W/mK range. It's great compared to thermal paste though, which tends to be somewhere around 10 W/mK. More like 5W/mK, the really good stuff can push to 12W/mK. I have both my (delidded) CPU and my GPU in the same custom cooler loop and it is not uncommon for the CPU to be 20C warmer than the GPU at load. The GPU consumes nearly 2x the power of the CPU, but it is dramatically closer to the coolant than the CPU could ever be. Also to the person wanting a silent PC, perhaps think about a custom loop cooler? A full cover GPU block requires no additional fans, plus effectively cools the memory/power delivery and you can also cool the CPU in the same loop with only a single superior copper radiator vs having to use multiple AIOs with their aluminum construction. And if you use compression fittings on flexible tubing it is really hard to screw up, if you are competent enough to install an AIO on a previously air cooled GPU going full custom loop is likely well within your skill level. Obviously the one big downside is cost, a custom loop can easily end up at 2x to 3x the cost of a couple AIOs.
|
|
#
?
Oct 22, 2018 23:00
|
|
|
Indiana_Krom posted:Obviously the one big downside is cost, a custom loop can easily end up at 2x to 3x the cost of a couple AIOs. The nice thing is that you can usually keep using it for several cpu/gpu generations and you can resell the parts down the road. Aint no one wants your 2 year old AIOs
|
|
#
?
Oct 23, 2018 04:27
|
|
|
What happened to needing to drain it every year or so? I looked into getting one and using a big heater core to passively cool the loop, but needing to mess with it every year scared me off.
|
|
#
?
Oct 23, 2018 04:32
|
|
|
some people need to drain twice a year other people go over a year no problem. Look at what additives people are using and what kind of life they get. I have noticed that the people who need flush often are using older loops that have probably never been cleaned properly and are just seeding the new fluid with old contaminants every cycle. You can make it extremely easy to do if you plan ahead with a fill port on the top and a drain port with a ball valve or quick disconnect. Its like $12 more in fittings and 10min extra that saves you so much time in the future. Its one of those thing where if you do it regularly before you have a problem its easy. Having to break it down and scrub out stuff is the real hassle. Its a hassle but you still have a working loop at the end unlike an AIO that gets tossed. Fauxtool fucked around with this message at 04:41 on Oct 23, 2018 |
|
#
?
Oct 23, 2018 04:38
|
|
|
Fauxtool posted:some people need to drain twice a year other people go over a year no problem. Look at what additives people are using and what kind of life they get. I have noticed that the people who need flush often are using older loops that have probably never been cleaned properly and are just seeding the new fluid with old contaminants every cycle. Yeah, part of the reason I went with the Evolv X even though I'm air cooling initially is that it has a fill and flush port built into the case.
|
|
#
?
Oct 23, 2018 05:41
|
|
|
For a budget compute card, does anyone have experience using the 1070ti or 1080 for small machine learning projects? There seem to be a lot of them used on Ebay, and the 1070ti new is pretty reasonable. Most information online focuses on equipment that is way out of my reach. For what I do, a higher core/thread CPU (vis-a-vis my venerable 4770k) would do just great, but I don't want to buy a new CPU/RAM/MB combo until 7nm shakes out next year. GPU compute is way, way easier than it used to be and I feel kind of silly not doing it.
|
|
#
?
Oct 23, 2018 05:42
|
|
|
BIG HEADLINE posted:Yeah, part of the reason I went with the Evolv X even though I'm air cooling initially is that it has a fill and flush port built into the case. Is that the chintzy plastic sliding door on the power supply shroud? That really bothered me in what is an otherwise super premium case
|
|
#
?
Oct 23, 2018 05:51
|
|
|
Fauxtool posted:Is that the chintzy plastic sliding door on the power supply shroud? That really bothered me in what is an otherwise super premium case No, the fill and flush ports are elsewhere. The sliding door...not even Steve could figure out what that was for. My guess is giving you the ability to route cables through if you don't give a poo poo about clean routing, since it wouldn't make much sense to put a reservoir down there. Only other reason I went with it is that only it and the updated Define R6 have front-panel USB-C 3.1 Type 2 at the moment.
|
|
#
?
Oct 23, 2018 05:53
|
|
|
Specifically, the fill and drain ports are little grommet holes on the radiator bracket at the top and at the front of the case. They're for pouring liquids into and/or out of and shouldn't be noticeable unless opening up the panels in question. Witness: https://imgur.com/gallery/D8gjcbE After spending the last week looking at case reviews I'm pretty sure I've settled on a Define S2 for getting my feet wet with water cooling. Space for two 360 mm radiators, lots of flexible mounting options for pumps and reservoirs, same front panel as the R6 update, no loving RGB LEDs. No drain port either, but I think I'll manage. If I use the top fill port I can always just turn it upside down.
|
|
#
?
Oct 23, 2018 06:05
|
|
|
Fauxtool posted:unlike an AIO that gets tossed. What is the life cycle of an AIO?
|
|
#
?
Oct 23, 2018 06:17
|
|
|
Aeka 2.0 posted:What is the life cycle of an AIO? You can always get a lemon and have them fail early, but generally 3-5 years, with some going longer. They're like PSUs - don't trust them past their warranty.
|
|
#
?
Oct 23, 2018 06:18
|
|
|
Aeka 2.0 posted:What is the life cycle of an AIO? When the warranty is up, throw them away.
|
|
#
?
Oct 23, 2018 06:19
|
|
|
Xerophyte posted:Specifically, the fill and drain ports are little grommet holes on the radiator bracket at the top and at the front of the case. They're for pouring liquids into and/or out of and shouldn't be noticeable unless opening up the panels in question. Witness: https://imgur.com/gallery/D8gjcbE do you really need space for 2 360s? Most people dont need more than a single 360 unless they are attempting to go full silent and just about every ATX case made in the last 4 years has space for a 360 in the front An integrated drain port is not a necessary thing at all. Just route a tube under the power supply shroud and cap it off. Your fill port is just unscrewing the top of the reservoir. Aeka 2.0 posted:What is the life cycle of an AIO? The nice thing about an AIO is that if they fail in a way that damages the system the manufacturer is generally very quick about writing a check to replace the parts. Now that Im in a position to have a computer just for gaming separate from my work computer it makes buying things that look nice but may break later an option instead of reliability and price/perf. I cant see myself going away from a 240-360 AIO and monster 3 slot GPU in the future.. So many RGBs and so easy to build. Fauxtool fucked around with this message at 06:32 on Oct 23, 2018 |
|
#
?
Oct 23, 2018 06:28
|
|
|
Fauxtool posted:do you really need space for 2 360s? Most people dont need more than a single 360 unless they are attempting to go full silent and just about every ATX case made in the last 4 years has space for a 360 in the front Well, I am sort of planning on going full idiot and getting a 9900K and 2x 2080 Ti to run renders on. That system potentially go up to something like 600W of heat for the loop to deal with, which is a lot for a single radiator. The sane thing would of course be to get no more than a single very expensive GPU that'll only last a year, but if I were doing the sane thing I wouldn't be looking at assembling a custom watercooling loop in the first place. I agree the fill/drain ports aren't a big deal, it's not an operation you do more than maybe once a year and the ports don't make much of a difference. The S2 still has one, though.
|
|
#
?
Oct 23, 2018 06:57
|
|
|
Finally got my 1070ti to replace the 1060 in my machine, and lol the difference between the Armor and Gaming X coolers is ridiculous. I think the card weighs twice as much. I can definitely see why that Armor cooler would be a bad move on a higher end card. The 1070ti seems to be driving the Dell AW3418 fairly well, averaging 60+fps on the Hitman benchmark which is what I'm playing at the moment so all good. G-Sync is brilliant.
|
|
#
?
Oct 23, 2018 10:55
|
|
|
Hey folks. Been having problems relating closely to my graphics card. Ordinarily I have two screens plugged in. Often when I plug in my Oculus Rift, the screen freezes and I have to hard shutdown. Often it just doesn't come back on, I do not even get the "beep". Sometimes it does come back on. One way I got it to do so was to remove my graphics card and plug into my mobo. The graphics card is a 1080 GTX with up-to-date drivers and no overclocking. The PSU is a Corsair RM 750x. One thing I am thinking of trying is updating to windows 10. I have been on 8.1 (a bit silly I know). Oculus recommends this generally. But does that actually have a chance of changing anything? And is there anything else I can try?
|
|
#
?
Oct 23, 2018 13:10
|
|
|
Xerophyte posted:Well, I am sort of planning on going full idiot and getting a 9900K and 2x 2080 Ti to run renders on. That system potentially go up to something like 600W of heat for the loop to deal with, which is a lot for a single radiator. Part of me still wants to do a CLC with an oversized pump (or secondary boost pump?) and then hang the radiator out the window to get that excellent cooling gradient. I mean, I won't. But I want to.
|
|
#
?
Oct 23, 2018 13:25
|
|
|
Yudo posted:For a budget compute card, does anyone have experience using the 1070ti or 1080 for small machine learning projects? There seem to be a lot of them used on Ebay, and the 1070ti new is pretty reasonable. Most information online focuses on equipment that is way out of my reach. For what I do, a higher core/thread CPU (vis-a-vis my venerable 4770k) would do just great, but I don't want to buy a new CPU/RAM/MB combo until 7nm shakes out next year. GPU compute is way, way easier than it used to be and I feel kind of silly not doing it. Yeah I've done a bit with a 1070Ti, I compared running some tensorflow stuff with the 1070Ti vs my 6600k and it was training about 8x faster. MaxxBot fucked around with this message at 18:57 on Oct 23, 2018 |
|
#
?
Oct 23, 2018 18:54
|
|
|
Aeka 2.0 posted:What is the life cycle of an AIO? fwiw I've been running a Nepton 140XL for just about 4 years with it in the wrong orientation (tubes on top) and it is starting to get short of fluid that's one of the two typical failure modes, the other being a pump that dies
|
|
#
?
Oct 23, 2018 21:11
|
|
|
The worst is the fittings crack out of warranty and let fluid leak on 1000 dollars of hardware.
|
|
#
?
Oct 23, 2018 21:19
|
|
|
craig588 posted:The worst is the fittings crack out of warranty and let fluid leak on 1000 dollars of hardware. How did you manage to lose two thirds of a gtx 2080ti
|
|
#
?
Oct 23, 2018 21:21
|
|
|
Paul MaudDib posted:fwiw I've been running a Nepton 140XL for just about 4 years with it in the wrong orientation (tubes on top) and it is starting to get short of fluid On this - I run a corsair H115i Pro with the rad on the top of my case. Is this bad for the pump? I don't want to put it anywhere else so I guess I don't really care but it'd be nice to know I'm not doing anything retarded. I'm wondering actually if I might benefit going to a 320mm rad from the 280mm since I'm going to a 9900k.
|
|
#
?
Oct 23, 2018 21:32
|
|
|
MaxxBot posted:Yeah I've done a bit with a 1070Ti, I compared running some tensorflow stuff with the 1070Ti vs my 6600k and it was training about 8x faster. Wow. That would make things much easier for me. I will try to pick up a 1070 ti or 1080 used or perhaps on Black Friday. Thanks for the feedback.
|
|
#
?
Oct 23, 2018 22:28
|
|
|
Statutory Ape posted:How did you manage to lose two thirds of a gtx 2080ti if useless tensor cores die, do they even exist in the first place?
|
|
#
?
Oct 24, 2018 00:07
|
|
|
VelociBacon posted:On this - I run a corsair H115i Pro with the rad on the top of my case. Is this bad for the pump? I don't want to put it anywhere else so I guess I don't really care but it'd be nice to know I'm not doing anything retarded. I'm wondering actually if I might benefit going to a 320mm rad from the 280mm since I'm going to a 9900k. Nah. The idea is that you want the tubes on the bottom of the rad for an AIO since the rad is also the reservoir. If you have the tubes on top, once it starts losing a little fluid and gets air bubbles, those bubbles will float to the top and eventually get sucked into the tubes, causing a reduction in performance and potentially stalling out the pump. If the tubes are on the bottom, it's much harder to get air in the lines. So with yours mounted on the top of the case (and presumably with the tubes coming out the bottom of the rad), the only thing you should consider doing is altering the screw depth slightly so that the tube-end of the rad sits lower than the non-tube-end, again to help keep any air pockets away. DrDork fucked around with this message at 00:44 on Oct 24, 2018 |
|
#
?
Oct 24, 2018 00:40
|
|
|
Does anyone know what the 2080TI's target load temperature is supposed to be? I have the FE and I set it to 83C like my 1080Ti but I couldn't find a source for the 2080Ti anywhere the Nvidia cooler/Mhz overhead is honestly better than I thought it would be, I can pretty much lock the thing at 2000Mhz and it keeps itself at the aforementioned 83 at like ~60% speed in my tiny rear end ITX case e: the nvidia specs page has the max thermal limit set at 89C and the 1080Ti's was 94C so I'm assuming I should shoot for 78C instead of 83C. phooey Scionix fucked around with this message at 06:19 on Oct 24, 2018 |
|
#
?
Oct 24, 2018 06:12
|
|
|
I believe 83 is still the default temperature limit, so it should be fine to stick with that.
|
|
#
?
Oct 24, 2018 07:46
|
|
|
Just slide the power slider all the way to the right and let the card thermal limit itself where it will IMO.
|
|
#
?
Oct 24, 2018 09:29
|
|
|
Turns out it doesn't matter because the 650w SFX PSU I bought to make my ITX life easier can't handle a 2700X and a 2080Ti  The good news is my old ATX PSU should fit but it's not gonna be comfy
|
|
#
?
Oct 24, 2018 13:47
|
|
|
Is it old? Have you confirmed the actual draw is higher than 650? 2080 ti's are pretty high powered I guess.
|
|
#
?
Oct 24, 2018 18:20
|
|
|
LRADIKAL posted:Is it old? Have you confirmed the actual draw is higher than 650? 2080 ti's are pretty high powered I guess. It's brand new, but I highly doubt I could pass GPU, then CPU stress tests individually, then have the computer just shut off repeatedly during mixed loads and it not be the PSU
|
|
#
?
Oct 24, 2018 18:31
|
|
|
|
| # ? Jun 11, 2024 05:08 |
|
|
Sounds about right. Yea not OC'ed it might be fine, but you may be using a good chunk when both are stressed OC'ed. Stock you should be fine though.
|
|
#
?
Oct 24, 2018 18:48
|
|
