|
I know times are tough and the end is near but if anyone is looking for a deal on a 980Ti I've finally gotten off my rear end and made a thread on SAMart. https://forums.somethingawful.com/showthread.php?threadid=3827623
|
#
?
Jul 20, 2017 23:07
|
|
|
|
| # ? May 31, 2024 06:52 |
|
|
Klyith posted:c. bigger fans require more watts to spin. I don't know what the fan power system of the average gpu is rated for, it's probably plenty, but as long as I'm throwing away the warranty to diy the fan I'd rather not burn out the controller. (if I'm doing this I'm gonna rewire the connector the original fans used, so I can keep the auto fan speed controlled by the gpu.) I've got two 120mms running off of my GPU so hopefully this isn't actually a thing.
|
|
#
?
Jul 21, 2017 00:57
|
|
|
Fruit Chewy posted:I've got two 120mms running off of my GPU so hopefully this isn't actually a thing. Just check on the fans themselves and figure out the wattage. Most 120mm fans are around 3.6W, with 80mm fans not much lower (2.5-3 ish), so unless the controller is specced just barely above the requirements for the 80mm fans, you probably should be fine (unless you got some strange 7W monster fan).
|
|
#
?
Jul 21, 2017 01:40
|
|
|
Fruit Chewy posted:I've got two 120mms running off of my GPU so hopefully this isn't actually a thing. it's probably not a thing, as I said they likely have plenty of headroom. most mobo headers can do 1 amp -- that's enough to run any two fans that aren't screaming 5000 rpm delta fuckers. GPUs probably do the same, they have to support centrifugal blowers and those take more juice than normal fans. and if yours didn't blow up the first time you ran the fans 100% then it'll probably keep working. but when doing that type of swap it's good to be on the safe side. check the old fans' spec if possible and use something the same or less.
|
|
#
?
Jul 21, 2017 02:02
|
|
|
I'm pretty sure a 1200 rpm 120mm fan only uses like 0.15A.
|
|
#
?
Jul 21, 2017 02:16
|
|
|
Paul MaudDib posted:NVIDIA can axe the GSync tax down to a fraction of its current level without hurting their per-monitor net revenue. Or they could make revenue from both things because the competing product is bad for all the reasons that you've mentioned. The only people buying it are people who make bad decisions and if it was closer to 50/50 or worse AMD had a bad product that people couldn't get enough of (think along the lines of the Xbox 360 where it established dominance during a period of infamously horseshit reliability), then you could see Nvidia do that. There's no good reason to cut margins aggressively against a product that sinks itself. Nvidia is in it for money, not for bloodsport. There's no good reason to burn money if the people you win over are budget shoppers and a tiny number of people who would buy a card that is objectively bad, and investors aren't moved at the thought of cutting their realized profits just to put some people at a competing company on the unemployment line. Craptacular! fucked around with this message at 03:43 on Jul 21, 2017 |
|
#
?
Jul 21, 2017 03:37
|
|
|
tldr get hosed boi
|
|
#
?
Jul 21, 2017 04:22
|
|
|
Craptacular! posted:Or they could make revenue from both things because the competing product is bad for all the reasons that you've mentioned. The only people buying it are people who make bad decisions and if it was closer to 50/50 or worse AMD had a bad product that people couldn't get enough of (think along the lines of the Xbox 360 where it established dominance during a period of infamously horseshit reliability), then you could see Nvidia do that. Majority of people I know were buying into the freesync ecosystem with 144hz 1080p monitors and RX480s. That all changed with miners though. Freesync and AMD is still attractive to the mid range buyer if prices go back to normal. Not everyone wants to spend a lot of cash on a super nice 1440p Gsync monitor to go along with their $400+ graphics card.
|
|
#
?
Jul 21, 2017 04:36
|
|
|
Craptacular! posted:Or they could make revenue from both things because the competing product is bad for all the reasons that you've mentioned. The only people buying it are people who make bad decisions and if it was closer to 50/50 or worse AMD had a bad product that people couldn't get enough of (think along the lines of the Xbox 360 where it established dominance during a period of infamously horseshit reliability), then you could see Nvidia do that. Yup, the top 20% spenders make up 80+% of the overall profits from iPhones to P2W gaming whales. As long as that 20% are buying into Gsync I don't see NV lowering prices one bit (lol as if rational companies cares about what the vocal misers think) and even then NV would rather figure out on how to differentiate their solution further than Freesync than lower prices to compete when the pressure is on.
|
|
#
?
Jul 21, 2017 05:20
|
|
|
Traditional thermal paste vs liquid metal based paste. TL:DR probably just stick with regular paste but liquid metal is not without merit sometimes There have been liquid metal repasting videos popping up on Youtube specifically about Pascal. I mentioned this in the buttcoin thread then decided to test it myself. I had plenty of liquid metal experiences with delidding and so on but never bothered to use it on a GPU, and I had some on hand... so I thought but it was too old for comfort and I bought some Thermal Grizzly Conductonaut. A little preface - I've repasted 20+ gpus in the last 3 years and am a pretty big proponent of it. Sometimes you can shed 10+ degrees off your GPU this way, which affects everything from max OC (in the past) and most importantly fan speed. These days you'd gain more boost clock if you dont OC and reduce fan speeds. On the other hand nothing might happen at all and there is no improvement. After having done so many I have discovered this is a case by case issue. It depends on the specific cooler and GPU you are using, not even the brand. I figured that the TIM was just one of a few parts that can be a thermal limitation, and if thats not the bottleneck then there will be no improvement. The other big components are how much the heatsink can dissipate, ambient temperature around the cooler, and of course how much heat the GPU itself puts out. I was testing this simply at first in my own personal real world scenario (max software OC). I added some different tests along the way and so I dont have some data I wish I had. I have one Gigabyte 1080 G1 and one new EVGA 1080ti SC2. The Gigabyte 1080 already had its stock paste changed a few times over a few months for various reasons and I did have Heaven 4.0 data on its stock paste luckily. This is a pic of my notes page, sorry its not super intuitive  As you can see the G1 1080 had virtually no effect. I expected this, the cooler is what I would call moderately adequate. I was glad I had this data complete because there are plenty of cards out there with the cooler itself being the absolute bottleneck and these results are in line with what I would have thought. If you go down to the first yellow 1080ti field you will see a fan speed improvement of 4% for regular thermal paste, which isnt insignificant. But then the liquid based paste netted a pretty whopping change. A five degree drop is awesome in itself but corresponding 8% fan speed drop makes the card effectively silent under full load + OC.  This lead me to the idea to start pointing fans at the GPU coolers to see if increasing their heat dissipation to perhaps move the bottleneck to the TIM and check if there was an improvement. Unfortunately I only got this idea after I used the liquid metal on the first card (1080ti) so there is plenty of missing data there. I used a big desk fan and a single 140mm Antec standard case fan. The test bench itself is open air with zero case fans sitting in a 22*C office in a better control environment than I'd be able to manage at home. And the results... varied even just between two different cards. The 1080 G1 responded to the desk fan with a 3 degree drop on the thermal paste and liquid metal respectively. It responded even better to a single case fan pointed directly at it just 1 inch away. I found one specific position gained the best on that card, covering the furthest two fans away from the GPU. That got 4 degrees on the paste and 6(!) degrees on liquid metal. The information I wish I had was the stock paste for these tests but I can probably safely assume it would do no better than the Xigamatek paste I was using. (case fan positioning like this)  The case fan test was particularly interesting here and I was looking forward to the results on the 1080ti so that I could attach a fan like that full time. Weirdly, though, the case fan actually increased temperatures across the board on the 1080ti - sometimes dramatically. The only way I could get an improved or neutral result was by positioning it far away. On the other hand the desk fan produced awesome results in respect to fan speed in partiulcar. 38% is hardly turning over and its looping a full OC load in Heaven over and over. Sad though because a desk fan mod is probably not feasible in my RVZ02 case lol. Test conditions: Heaven looping at 1440p for at least one entire run for heat soak, then a few runs in every test condition with short pauses between tests to reduce the fan and temps to roughly 50 degrees and 20%. Many tests ran for hours while I was working and the results for these cards in these conditions are very accurate. Open style test bench case with no additional case fans other than the ones noted. And there you go. In Conclusion: Changing the TIM helps. Unless it doesnt. Airflow helps. Unless you make it worse. On a serious note, I can see liquid metal being an advantage in some scenarios. Those with giant heat dissipation relative to the GPU heat output would stand the best chance to benefit. Those with watercoolers on the GPU would likely benefit the most (im most interested in this). But when all is said and done I will openly recommend regular good thermal paste for one huge reason: mounting pressure on GPU coolers varies greatly and coolers with exposed heatpipes on the plate have little grooves and valleys. True liquid metal TIM is basically like a layer of paint (when done "correctly") rather than a layer of paste. If your cooler mounting interface isnt exactly flat you will run into major problems if you dont have exactly as much liquid metal as you need. However, you simply cannot risk using too much to have it squeeze out the sides. It will come into contact with pins near the GPU and cause shorts even if you try to cover these up thats not a risk you want. The second problem is the mounting pressure can be pretty stiff to barely on there at all. This causes all kinds of issues with the TIM as you can imagine. However I can say with some certainly you can do it correctly if you want to in any normal scenario. The G1 1080 had a pretty badly ground heatpipe interface and low mounting pressure. I had to do it twice, but I got it done. And finally I will say that I've seen a trend starting in the 900 series cards where the TIM is either done correctly or the bottleneck to remove heat has shifted to the coolers much more than in the past. The 600 and 700 series cards would very consistently produce good to great results. Sometimes ridiculous results, once on an ASUS 660ti I dropped 15 degrees from 70 to 55 with paste alone. But those gains dropped quite a bit for Maxwell and on. But despite that they are still there to be had and its easy to do on most cards and I'd always recommend it if youre comfortable with such a thing. I did run into my first difficult to remove cooler ever, the EVGA 1080ti, so its *possible* it will seem like more trouble than its worth. 1gnoirents fucked around with this message at 22:47 on Jul 21, 2017 |
|
#
?
Jul 21, 2017 22:15
|
|
|
Sweet post, nice to see your test results.
|
|
#
?
Jul 21, 2017 22:26
|
|
|
The inconsistency between the case fan and desk fan could be turbulence in the wrong spot which can do bad things to effectiveness of heatsinks. I used to know someone who did aerodynamic design work for heat exchanger systems and he'd always say that the more you know about aerodynamics the more it might as well be magic. Edit: If you have the space for it, turning the 2 slot cooler into a 2.5 slot cooler by removing the shroud and stock fans to replace them with x25mm thick fans should give you good results. I generally repaste and also "refan" every card I own at the same time after owning them for about a month to ensure I won't have to warranty them. The combination of both changes at once gives much less noise and 20-30 degree lower temperatures, depending on how bad the stock fans were or how hot the stock card would get. As long as you're removing the cooler switching the fans isn't much more effort, as long as you have some extras laying around it's probably worth doing. craig588 fucked around with this message at 23:06 on Jul 21, 2017 |
|
#
?
Jul 21, 2017 22:57
|
|
|
craig588 posted:The inconsistency between the case fan and desk fan could be turbulence in the wrong spot which can do bad things to effectiveness of heatsinks. I used to know someone who did aerodynamic design work for heat exchanger systems and he'd always say that the more you know about aerodynamics the more it might as well be magic. I would agree that must be it. For whatever reason, the design of the EVGA cooler simply could not have a 140mm fan anywhere near it pointed at it. At best, it didnt do much, but at worst the temperatures spiked fast to 80+ degrees and 75%+ fan speed (like with 15 seconds). I tried from every position including both exposed sides of the card having the fan blow across the heatsink fins. The Gigabyte 1080 cooler was the opposite, the closer you could get it the better, even directly touching the cooler. But the EVGA loves the general airflow pointed in its direction from afar so... magic craig588 posted:The inconsistency between the case fan and desk fan could be turbulence in the wrong spot which can do bad things to effectiveness of heatsinks. I used to know someone who did aerodynamic design work for heat exchanger systems and he'd always say that the more you know about aerodynamics the more it might as well be magic. That is interesting and I want to try that on the 1080ti especially though Im skeptical theyre going to make that task easy. I technically have all the space in the world in an RVZ02 because I dont put the side on the GPU chamber. edit: Saw a good deal on nanoxia 140mm pwm fans and went for it, thanks for the idea I think it'll work awesome with this 1gnoirents fucked around with this message at 23:25 on Jul 21, 2017 |
|
#
?
Jul 21, 2017 23:01
|
|
|
I keep seeing it mentioned that nVidia sales are down year on year, which is confusing to me since nVidia doesn't release cards on a yearly schedule. If you assume that each sales cycle is 18 months, and that there's a glut of sales at the start of a cycle and a lull of sales towards the end, why would year on year sales figures be useful?
|
|
#
?
Jul 22, 2017 05:18
|
|
|
Well, i guess it depends on how wide a look you take at things. I would say Nvidia's gaming sales are doing juuuuuuuuuuuuuuuuuuuuust fine, and if RX Vega is what we think it is...
|
|
#
?
Jul 22, 2017 05:30
|
|
|
Cygni posted:Well, i guess it depends on how wide a look you take at things. I would say Nvidia's gaming sales are doing juuuuuuuuuuuuuuuuuuuuust fine, and if RX Vega is what we think it is... I'm the $500m revenue increase predicted for this quarter.
|
|
#
?
Jul 22, 2017 10:51
|
|
|
ConanTheLibrarian posted:I'm the $500m revenue increase predicted for this quarter. Mostly in gaming, no less. Come on Volta!
|
|
#
?
Jul 22, 2017 19:39
|
|
|
That's not a projection. Nvidia's Q1 FY18 is feb-april 2017. The spike is when Pascal was released.
|
|
#
?
Jul 22, 2017 19:55
|
|
|
K8.0 posted:That's not a projection. Nvidia's Q1 FY18 is feb-april 2017. The spike is when Pascal was released. Pascal was released in May 2016? These numbers are not making much sense to me.
|
|
#
?
Jul 22, 2017 19:59
|
|
|
FY = Fiscal year, which isnt a normal year. Oct->September is one year, even if that starts in 2016 its considered all FY 2017, and the quarters put the months in different places. I dont know much more
|
|
#
?
Jul 22, 2017 20:18
|
|
|
Palladium posted:I'm pretty sure a 1200 rpm 120mm fan only uses like 0.15A. Sometimes it's not the operating current draw but the start-up draw that can cause problems. When they get energized they'll likely pull peak current until UEFI/BIOS/software indicates to throttle down. It's not one of those things that usually causes a problem early on but long-term can lead to that line trace failing.
|
|
#
?
Jul 22, 2017 20:48
|
|
|
AT forums are saying someone at PDXLAN will win an RX Vega and specs will be announced. It'll all be over soon and then we can get hype for Navi and Volta will probably be a lovely GPU lineup.
|
|
#
?
Jul 22, 2017 22:19
|
|
|
1gnoirents posted:FY = Fiscal year, which isnt a normal year. Oct->September is one year, even if that starts in 2016 its considered all FY 2017, and the quarters put the months in different places. I dont know much more Nvidia's Fiscal Year 2018 actually started in February 2017, so that chart only shows past values, not forecasts. http://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-first-quarter-fiscal-2018
|
|
#
?
Jul 22, 2017 22:21
|
|
|
NewFatMike posted:AT forums are saying someone at PDXLAN will win an RX Vega and specs will be announced. I'm sure they might win one but I doubt they will get it before official release.
|
|
#
?
Jul 22, 2017 23:18
|
|
|
NewFatMike posted:AT forums are saying someone at PDXLAN will win an RX Vega and specs will be announced. Second prize: two RX Vegas.
|
|
#
?
Jul 22, 2017 23:30
|
|
|
Just finished swapping the TIM in my Asus 1060 Dual from the stock blob to some Arctic Silver 5 i had around. At 50% fanspeed and 24-25c ambient, only 1 80mm case fan. Base clocks had the core consistently at 1900mhz and the ram at 8Ghz. OC was 2000mhz and 9ghz. Stress test was Fire Strike on loop for ~30min.  Lil' drop, although the 1060 is already pretty cool and the Dual's fan is way overbuilt for the TDP. AS5 is dirt cheap and it didnt take long to do, so good stuff.
|
|
#
?
Jul 24, 2017 00:39
|
|
|
Dumb question, but does AS5 ever "expire"? I have a fairly large tube that's probably enough to last a lifetime in my desk and I'm pretty sure I got it when I was in high school and put my first PC together, so it's going on 15 years old.
|
|
#
?
Jul 24, 2017 03:58
|
|
|
It may harden but squeeze out some, throw it out and the rest that comes out should be good. Unless it separated in which case just mix it back up.
|
|
#
?
Jul 24, 2017 04:06
|
|
metallicaeg posted:Dumb question, but does AS5 ever "expire"? I have a fairly large tube that's probably enough to last a lifetime in my desk and I'm pretty sure I got it when I was in high school and put my first PC together, so it's going on 15 years old. It should be fine but if it's a conducive formula then you should avoid using it on GPUs.
|
|
|
#
?
Jul 24, 2017 04:17
|
|
|
metallicaeg posted:Dumb question, but does AS5 ever "expire"? I have a fairly large tube that's probably enough to last a lifetime in my desk and I'm pretty sure I got it when I was in high school and put my first PC together, so it's going on 15 years old. Officially most thermal pastes list a shelf life of 1-2 years. Unofficially they are usually good for a long time (and if they weren't you'd need to reapply on your PC every couple years anyway). The stuff in the tip of the tube is probably dried out, so squeeze some into a coffee filter. There's really no reason to bother trying to save it though, AS5 is no longer the silver bullet it used to be. Modern thermal pastes are just as good and also not conductive, which is a huge plus for usability.
|
|
#
?
Jul 24, 2017 04:23
|
|
|
AS3 was conductive, but AS5 is not. AS5 is very slightly capacitive (like most thermal pastes), but ive never had it interfere with anything in practice.
|
|
#
?
Jul 24, 2017 04:29
|
|
|
metallicaeg posted:Dumb question, but does AS5 ever "expire"? I have a fairly large tube that's probably enough to last a lifetime in my desk and I'm pretty sure I got it when I was in high school and put my first PC together, so it's going on 15 years old. Noctua NT-H1 is about $5 on NewEgg and Amazon, and gets a few degrees cooler than the AS5 in pretty much every test. Old AS5 will work in a pinch (every nerd on the planet seems to have a decade-old syringe of the stuff) but if you actually want a slight but real improvement, I'd say get some NT-H1 at a minimum.
|
|
#
?
Jul 24, 2017 05:19
|
|
|
I have some arctic silver 2 from like 2002, still perfectly fine. Still nice and runny, goes on thin with a razor blade just like it did in 2002 when men were men and overclockers used 6000 rpm delta fans. As long as the plunger side keeps a good seal thermal paste will last forever, it goes bad when the hydrocarbons evaporate off. The big innovation in thermal pastes is that the new ones are thick and stiff out of the tube, then thin down with heat cycles, so you use a lot more. Because at some point someone realized that selling a lifetime supply in each tube wasn't encouraging much repeat business.
|
|
#
?
Jul 24, 2017 06:14
|
|
|
I moved from using AS5 for the longest time to Arctic MX-4, since it does a slightly better job plus it's non-conductive. Actually still have half an 8g tube of AS5 around too, pretty sure I bought it back in like 2011...poo poo's gonna be around with Twinkies and cockroaches after armageddon.
|
|
#
?
Jul 24, 2017 07:15
|
|
|
I moved off of AS5 to Gelid GC-Extreme, which was the hot poo poo at the time that I bought it.... are we still plateaued with regards to thermal pastes? I even have some of that stupidly hard-to-use Shin-Etsu that you have to pre-warm before you squeeze it out that by all rights I should have thrown away years ago.
|
|
#
?
Jul 24, 2017 08:13
|
|
|
this is the best/most recent test I know of: http://overclocking.guide/thermal-paste-roundup-2015-47-products-tested-with-air-cooling-and-liquid-nitrogen-ln2/6/ the difference between any of the more common after-market pastes seems negligible
|
|
#
?
Jul 24, 2017 14:53
|
|
|
My rule of thumb is that you have to use Hellman's Real Mayonnaise, miracle whip won't work: http://www.hardwaresecrets.com/does-mayonnaise-last-as-a-thermal-compound/
|
|
#
?
Jul 24, 2017 15:01
|
|
|
eames posted:this is the best/most recent test I know of: Hm. Looks like I'm good, unless I want to step up to liquid metal for non-GPU applications, then.
|
|
#
?
Jul 24, 2017 19:50
|
|
|
Twerk from Home posted:My rule of thumb is that you have to use Hellman's Real Mayonnaise, miracle whip won't work: http://www.hardwaresecrets.com/does-mayonnaise-last-as-a-thermal-compound/ What about american cheese (https://www.hardocp.com/article/2009/04/07/thermal_paste_shootout_q209/2)?
|
|
#
?
Jul 24, 2017 21:50
|
|
|
|
| # ? May 31, 2024 06:52 |
|
|
Are video card prices just insane right now or am I not looking in the right places anymore? I'm trying to figure out how to replace an 8gb RX 480 and I. . . .can't. I'm starting to wonder if I'd be better off getting a new gysnc monitor instead of feeesync, and just undo what seems to be a year old mistake
|
|
#
?
Jul 24, 2017 21:52
|
|
