Heating Questions with the 30xx cards
NotAnArtist
Posts: 374
The 30xx cards are announced. Now it's time to ask about a new build. My old machine, which I built 8 years ago, never had its i5-2500K or GTX-1060 go above 50 degrees no matter what I did with it, Iray included. (Now it jumps to 70, actually no, I don't render anymore).
But the 30xx's are much hotter. The air blows from one fan directly over the RAM and largely over the CPU cooler. The other half of the card routes it out the back of the case. So only half the total heat of the GPU hits those components. I guess that solves half the heating problem!
My new build will use a 3080. I'll wait to see if a full blower card comes out, not the 'half-blown' version shown by Nvidia, but I'd like to get an opinion from an experienced builder on the rest of the plan:
-I bought an NH-D15S cooler because I don't want liquid cooling. I'll add a 2nd fan to it if needed.
-I will get a Ryzen 7 3700X, which will run cooler with its 65W TDP.
-A 1000W PSU, modular, would be sufficient. Maybe overkill. Is it true that a Gold rating isn't sufficient for this sort of system? My old machine uses a Corsair TX-750 Bronze - quite a step down!
-I also got a kit of 64 GB G.Skill RAM. Wondering how hot that'll get. I won't be overclocking anything, though. There'll be 3 fans blowing on them from the front. Oh, maybe I oughta add a fifth fan on top.
Heck, I think I've got this covered, but it'd be nice to hear from someone experienced... Don't want to mess this up. The build will eat most of my savings and my SS income is very low. But, I'm old and I worked hard for this, and have waited a long long time ;-)
If there's another thread covering these issues, I'm sorry I've missed it. Two days of studying articles, forums and videos have my eyes caking over.
Thanks,
-Ken
Daz 3D is part of 
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
Gold is more than sufficient. The differences between 80+ gold and above are very small. 80+ means the PSU is 80%, at least, efficient at 20% load, 50% and 80%. 80+ Bronze goes to 82% at 20% and 80% and 85% at 50%. 80+ Gold is 87%, 90%, 87%. Above that the gains are a couple of percent each. So in a year of 24/7 use a 1000W PSU platinum might save you a couple of dollars of electricty over the 80+ Gold. That matters to me running a datacenter. It might not matter to you.
Don't get blower cards! There is more than a lot of doubt that Nvidia's card design will work. None of the AIB's are using designs like that, and they have much more experienced engineers in regards to cooling GPU's. Get one of the 3 axial fan 3080's from a good brand, Asus or Gigabyte, and you should be fine. Ram gets warm but not hot. don't let it worry you.
Hello! I'm so glad to see your response, kenshaw!
PSUs: I worried that the small differences in percentages in the chart might have referred to an effect on the stability of the components that the PSU feeds. If small differences aren't critical to that extent, I really thank you for saving me quite a few dollars.
Re the GPU designs: OK, I missed reading of any doubts that Nvidia's mixed blower/open-air design would work. I guess it's not a new idea then after all?
But about that heating issue:
I was going to add a 3070 to the 3080 later on. It'd be more CUDAs than the 3090, at a cheaper cost - and something I could choose to include or not in the renders.
That's why I wanted blowers, to avoid the need for liquid cooling. You didn't mention why you don't like blowers, though. I have no memory of reading about problems with them.
Oh! Was there something about failure rates? I'll research it.
If this be the case, I'll consider the 3090 and 'settle' with that ;-)
So, bottom line, and I very much appreciate your help on this (and the case questions in my previous thread) >
Even a hot 350 watt card in the system I described above, open-air, 3-fan, would be fine?? Wow! I feel like having a drink, not to moan this time, but to celebrate!
Take care! I'd better let you go - there's probably a lot of folks in line behind me by now :-)
-Ken
Blowers just don't cool very well. There is some debate about how well they work in multi GPU setups compared to "standard" designs. The issues seems to come down to how much air flow is in the case otherwise.
So if you're getting a second card and your case doesn't have a lot of airflow you might want blowers. That's a decision you'll have to make.
As to these new Nvidia coolers. I'd very much wait til they come out and the reviewers get hold of them.
350 is hot and you want some decent cooling, and you definitely want to make sure there is plenty of airflow. I'd think long and hard before shoving too of these in tight against each other without watercooling. But one by itself is fine with 3 axial fans,assumingthe design is otherwise adequate.
@NotAnArtist
You're making an important build decision that should not be based on opinion, but rather evidence.
I can provide at least one real world data point showing that blowers are excellent at cooling, and they even don't need good air flow to do so. This is my system, 4 2080ti blowers that have been rendering full tilt all night, every night, for over a year and a half. You would not believe how hot the air vented out the back of my system is, hot air that is removed from the system and not being recirculated over other system components in the case, as you imagined. And note how closely the three motherboard mounted ones are together. This system renders RayDAnt's benchmark scene in 1:14, very nearly exactly 2x relative to some other 2 x 2080ti systems also listed there, so no throttling is occurring.
Also, I was also told by an engineer at System76 that for multi-GPU systems, blowers are essential, his words, presumably not because they don't cool very well.
But do your own research; everyone's got an opinion, but there's most likely nobody here who is truly qualified to help you.
Except you know data.
https://techbuyersguru.com/founders-edition-vs-open-air-video-card-coolers-definitive-analysis
You can search every, and I do mean every, PC tech forum on the web and you'll get the same data. Blowers are better in one scenario. Bad cases.
Did you read what they actually did to test? The whole point of a blower is to keep the air inside the case cooler and vent hot air outside the case, but yet they tested it with a non-blower inside the case alongside the blower. They didn't test two blowers. Do they truly not understand how blowers work? That's like saying your refrigerator performs poorly and you've got a space heater inside of it.
Even still, I did find it interesting how easy it would be to totally overwhelm the open air coolers so that they become the massively worst choice with a mere +190Mhz overclock:
"But when pushed to the limit, our dual open-air ACX cards start to show their limits. In this test, the top card is now hitting 84 °C, which leads to a massive thermal throttle. This is despite the fact that its fan is running at 100%. The card is maxed out, but it's not providing maximum performance. You can see that in the framerates for the dual ACX setup (again, follow the blue lines). It's the slowest, loudest setup, and it's not doing great things for our CPU either. In other words, choosing matched open-air ACX cards is actually the worst choice if you're running dual overclocked GTX 1080 GPUs!"
The blower was worse solo, in a dual on top and dual on bottom. By what logic could it be better with two?
BTW did you not look at the charts? The article text is wonky AF. Look at the DATA! I do not care at all what someone wrote. I care about THE DATA! LOOK AT THE DATA!
I buy the triple fanned ones personally. I got a 2080 super, above a 1070, the 2080 super still stays pretty cool. I do have a big case, with plenty of vents and a few case fans. Run the case fans at 100% and run aggressive fan profiles on my cards though.
By the logic that blowers acheive their benefit from not re-circulating hot air inside the case, and one might be expected to underperform next to a GPU that is re-circulating hot air inside the case, instead of a peer sibling that, like it, doesn't.
I am certainly not the only one who noticed that the authors are gentlemen and scholars when their chart seems to imply that you are correct. But the same people are "wonky AF" when their words seem to imply that the chart doesn't speak to your point at all.
Another thing I found interesting from the text:
"When we push our cards to a +190Mhz overclock (which amounts to around 12-13%), the situation barely changes. The cards are still running their fans at the same levels, and only the ACX card increases in temperature."
They are telling you, clear as day, that blowers have enough of a safety margin in their ability to cool your GPU that they are impervious to other changes in your system, while the open air cooler goes south very quickly, with a very small change. 84C and massive throttling from just +190MHz? What if you should have another heat generating component in there, say a third graphics card to drive your monitors?
Simply saying "Blowers just don't cool very well." as an absolute removes a lot of context and nuance. That's all I'm saying: like most things in life, it's complicated.
!
The blowers are never OC'd because they hit 80C, the temp at which the cards throttle, running at the base clock! The data is right in the charts! Why do you keep doing this?
(It takes me awhile to read, and re-read)
I have a Phantek P500A case. [H:510mm x W:240mm x D:505mm]. There'll be lots of fans.
My first card will be an open-air 3080. I'll study every source I can find before I commit to any form factor. If I get it & the cooling looks good, a 3070 may follow, but it will take a long time before there's enough evidence either way on how well the blowers function vs whatever else will be out there.
I just want to assure you guys that I'm not as quick to jump at buying things as it might seem from the way I write.
OK, I just ordered a 3090 blower ;-)
I think opinions are worth a lot. Most here are based on experience, and since each of you have different experiences, the opinions clash.
It's not pure science. If I were building a quantum computer, I'd definitely go somewhere else. But really, my only purpose here is to listen to you folks and go from there in my own research.
@kenshaw:
The techbuyersguru article, aside from both of your comments on it, further tells me to not even 'speculate' on what the environment with two cards, any format, both above 300 watts, would be like.
Look, one 3080, Iray-wise, is MORE than I had expected, at less cost. There's no reason to make it a blower by itself. I'm leaning toward an open-air format with 3 fans, because, I can't find anything wrong with it now.
The reason I questioned open-air at all is because blowing 300+ watts of heat onto the CPU and RAM seemed a bad idea. Now comes the next step: more research prior to and after the actual release of the cards. Argh.
PS: FWIW, here's a good short video for neo-builders like me, describing some very basic issues between blower & open-air GPUs:
Which GPU cooler is better? Blower vs axial/open air
Thanks for the fan speed suggestions! Lots to keep in mind here.
Rhetorical Q: If bad cases are a factor in how blowers perform relative to open-air, possibly leading to your disagreements above, would there be other factors to come? With significantly hotter cards, would every other factor follow equivalently? In weather, for example, there are threshholds that throw all sorts of predictions off.
Thanks everyone! I'm amazed at this community!
-Ken
Continue past the chart that seems to confirm your assertion, to the last chart where the GPUs are overclocked.
The last slide shows that in the overclocked scenario, even with an invalid dual GPU setup that compromises the blower, the open air coolers are equivalent, or worse. It is not difficult to imagine the results if they had tested a properly configured system with two blowers.
Again, your statement that blowers just do not cool very well needs context.
@NotAnArtist I'm impressed by your rational approach.
You bring up an interesting question re the environment. I can only offr a single data point based on my experience: The ambient temperature hasn't mattered much with my blower setup. I've left my system rendering during the day when it was uncomfortably warm in the room, probably over 100F. The idea, from the reference kenshaw posted, that blowers have more of a margin and don't fall off a cliff when conditions are less than ideal, like the open air GPUs did, really resonated with me for that reason. They seem "Murphy's Law proof" and "idiot proof" as well :) I think any good engineer would be interested in a healthy safety margin to account for things beyond his control.
In any case, enjoy your system when it comes together! And please update RayDAnt's benchmark page whenever Daz should support the new cards; that's what everyone will want to know.
Not seem to! DOES! In every single chart. In every single chart the blower underperforms! Why is this even still being debated? The 1080 axial fan card is a factory OC version and it still handily beats the blower in every single head to head test.
It matters. How much it matters is usually not that relevant. The temps we're talking about are degrees celsius not farenheit. So room temp varying between 68F (20C) and 72F (22C) is pretty minor. If your room temp gets a lot higher than that then yes it would be a factor. If your room is much cooler that could also be a factor, although at much cooler temps condensation can become an issue (that's why putting PC's in refridgerators doesn't work out so well).
The external factor that tends to matter most is dust. A dusty environment is really hard on a PC. You'll clog the dust filters, which inhibits air flow, which will need more frequent cleaning. And the inside will get dusty, no case is perfectly airtight, so you'll need to get inside and dust that out more frequently as well. Pet hair is a lot like dust so that can be a major PITA as well.
Probably because you just ignored what I pointed out and keep saying things that are objectively not true, expecting caps and exclamation points to convince the OP of something other than what they are intelligent enough to have figured out for themselves.
Again, I call your attention to the last chart. In the overclocked scenario, even their strange, mismatched configuration matches or beats the dual open air configuration. That chart, alone, as well as the text of your own reference suggest that blowers have the headroom the OP is interested in, while dual axial fans do not.
That last part was the single best argument for blowers I've ever read.
I know. It totally ruined my gouda.
Definitely. And everyone will be amazed I actually accomplished something.
! The axial fan card in the test is overclocked from the factory. It beats the non OC blower in every test. It was only when the axial fan card was further overclocked that it had any issues beating the blower. Since the blower was already thermally throttling at base clock even if they tried to set an overclock that did nothing. I get that you have this strange investment in "proving" that everything I write is wrong but this is just insane. The data is right there and you're arguing that up is down.
*facepalm* Blowers are actually infamous for getting fouled with dust and hair. The fans have terrible motors and no clearance. They're also very difficult to clean.
I really strongly suggest you get actually informed on the subject. There's an actual reason the entire Pc enthusiast community loathes the things.
You mean like from the experience of running 4 of them on an average 10 hours per day for the last year and a half?
I am most certainly a part of that community and, having owned and operated a multi GPU setup of both types, I have loved every aspect of my blowers and would recommend a set of blowers to anyone contemplating a multi GPU setup.
Last week, I just opened my case for the first time in a year and a half. I needn't have, because nothing internal even required cleaning and everything continues to operate just as well as on the day I installed them. There is nothing that I can complain about, and nothing akin to the problems you listed.
I am not sure exactly what the entire PC enthusiast community would loathe, certainly not the hassle-free operation, stable thermals even across a wide range of ambient temperatures, the reliability even rendering overnight ever night for the past 19 months, or the 1:14 performance.
I'm done with that one. The last slide is the most relevant one of all, but you won't acknowledge its existence or its implications.
The only thing that is strange is the way you seem to know everything about everything, and no one who disagrees with you can ever be in any way correct. But can we continue this discussion later? I'm watching a video about a 24G 3090 that exists.
Since the OP seems to have reached their decision this thread is being closed.