I don't usually ask for hardware advice, but...
joseft
Posts: 310
As in title, i am usually one to offer hardware advice rather than ask for it. But my situation is a bit unusual and im struggling to find specific information that may help indicate which of my two options is better. So im hoping some other hardware saavy peeps may be able to offer some informed opinions.
I run 3 GPU's. Two Titan X for gaming and renders, plus a third smaller GPU (in this case a GTX 770) to run my 4 displays when im rendering and not gaming. Due to the size of the GPUs and PCI layout of my mobo, the 770 sits in between the titants, obviously very close together which has always created cooling headaches.
The 770 is now quite old and the fans are on their way out, so i am looking at replacing it with a 2060. I am trying to decide what style 2060 to get for better overall system cooling under the circumstances (being the 'piggy in the middle' so to speak)
At the moment all 3 GPUs have big triple-fan open air cooling systems on them, which i know is not ideal for multi-GPU setups, particularly with 3+ GPUs and tiny gaps in between them. So at first i figured i would get a blower style 2060, as having that sitting in between the two titans would be a better solution, given it would dissipate its own heat out the back of the case, instead of throwing it into the other GPUs like another open-air setup would.
But i see that Gigabyte and MSI are making smaller mini ITX style 2060 cards, like this: https://www.gigabyte.com/Graphics-Card/GV-N2060IXOC-6GD#kf
That is obviously another open-air type cooler, but given the card itself is significantly smaller in length and also 4mm shorter in height/width, being in between the two titans would give them more space to breathe and air to flow. So i am struggling to determine which option would have the most positive impact to the overall cooling of the setup. Blower style, or the smaller but open-air type cooler.
This is all going in to a Thermaltake core X9 case, with what i think is a fairly good fan setup that should have good case airflow.
Performance/price between the two options is not a factor, this is purely about cooling.
Thoughts?
Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2025 Daz Productions Inc. All Rights Reserved.
Comments
Personally I don't like too much a 2060 in a compact size since it's a 160W card. And mixing a blower with open-hairs crush my personal aesthetic feelings. So I'd go with a dual-fan open-hair. Then you can add more fans to the case if needed since the X9 has practically no cooling limits.
I'd rather question the 2060 itself as a choice for the viewport. With your configuration it'll be a little monster between two old brothers. Unless you plan to update the titans too that is. But then three blowers will do better as a cooling solution.
I do intend to upgrade the titans eventually, just not yet. At the moment they are still more than enough for gaming and rendering performance is still very good also. I will not upgrade them until the RT cores on the 20 series are fully utilised by renderers, at the very earliest. More likely scenario is i will skip the 20xx generation and wait for the one after. Hopefully by then we will be seeing HBM2 and more than 12gb on non-titan/quadro cards.
So I guess the 2060 is to foreplay with rtx until it gets better with 30xx cards. Otherwise you could just keep the 770 and wait to upgrade them all together. I don't like to mix generations but it's not a bad plan it'll work fine in my opinion.
My choice would be depending on the air flow management. I assume the CPU is liquid cooled? If so, is the radiator used as exhaust or intake? How many exhaust fans? A third card that is Of the open air style will increase the temperature in the case and might effect the CPU cooling depending on setup and also cause all cards to run hottter. What are your current temps?
ciao
td
The CPU is liquid cooled yes, the radiator is currently setup as intake. I should have mentioned that it is all currently setup in a different case. The Thermaltake core X9 is new, and i have been waiting for some cables to arrive before i rebuilt my machine inside the new case. Knowing now that my 770 is on its way out, its just convenient to replace it when i rebuild inside the X9.
The current case is a corsair 780T, so its already supposed to be a good case for air flow, but with so much hardware jammed inside (i also have 6 internal hard drives in it, which inhibit the the airflow coming in the front intake fans), i think it still struggles. The X9 has far more space inside and is incredibly versatile in that you have so much ability to customize your fan setup and air flow needs, and it supports a rediculous amount of fans.
Currently my temps under load are along these lines. The 770 driving the displays is never under high load, so that temp is a result of heatsoak from the other two. I also have custom fan profiles, and often reduce the factory thermal throttling threshold in an attempt to keep in-case temps down (GPUs are listed top to bottom):
CPU: 50c
Titan1: 80c (my custom throttling threshold)
770: 60c
Titan2: 70c
in-case: 45c - this is with my case fans all running at 100%
I know having the space for the open air coolers to breathe makes a huge difference, the bottom titan always runs significantly cooler because the fans are on the bottom of the card, so the one on the bottom has plenty of open space underneath it and is not eating the heat from another GPU below it. The other two cards only have millimeters of space between them. To illustrate that further, i have been running a render on only the bottom titan for the past 25 minutes, and its temp is sitting on 57c, well below my custom throttling threshold and my custom fan curve still only has the fans running at 42% at this temp. In comparison, the top titan is idling at 47c.
I expect those temps to be considerabley better once it is all inside the X9 case. The fan setup i have planned for that is two 140mm high volume exhaust fans on the rear (top and bottom), a 200mm intake on the front, 200mm intake on the top, 2x 120mm intakes underneath at the front to help direct fresh air upwards (this will make sense if you have seen the x9 in detail), and two 140mm intakes on the side-bottom where my hard drive cages will be. I think i will change my CPU radiator fans to exhaust once it goes into the X9 too. This should still maintain positive air pressure inside the case. Definitely do not want negative air pressure on the X9, it is more open than most cases, so having negative air pressure is going to suck a lot of dust inside.
I just rebuild my system. It’s in a Corsair Osidian 800D. I use two GPU, a GTX1080 and a GTX1080ti. Currently the 1080 also drives the two displays. They are both open air style 3 fan cards from Asus. The cpu (i9 7920X) is liquid cooled with a 360mm radiator set up as exhaust at the top with three fans in push configuration. In addition there is a large exhaust fan in back. Intake is done via the bottom of the case with two intake fans behind a dust filter. Under full GPU rendering load my temps are:
cpu 35C
1080 Ti 71C
1080 55C
if I would add a third card, I would likely go for blower style to have less hot air in the box.
Ciao
td
I'm scratching my head about how mining rigs like the one in the link below can have 8 GPU's shoulder-to-shoulder, but there's so much concern here about having just two or three.
https://www.amazon.com/Hydra-Tower-Mining-Case-Ready/dp/B07B4QDWPK?ref_=fsclp_pl_dp_6
I've had a 1080ti and 1070 next to each other for years with no problems.
those are all blower style coolers, thats why it works. 90% of the heat they generate is vented outside the unit. There is also no other components inside the unit generating heat and contributing the complexity of the system cooling requirements.
are your two GPUs blower style or open air coolers? and are they literally right next to each other on your mobo with ~2-3mm of space between them? i dont think they would be. And do you actually monitor your temps to see if they start throttling?
PCI slot priority/layout on most mobos i have seen means if you run two GPUs, there is ample space in between them for air flow. e.g if using two cards, you use slot 1 and slot 3. So there is an open slot between them.
You talk about "just" 2 or 3. The difference between 2 and 3 is quite big when you consider what i said above, and the fact that open air type coolers do not do well in this kind of setup. If i was to take the 770 out of my system to open up all that space between the two titans, both my overall in-case temps and the individual temps of the two titans would be significantly better.
Not sure what you mean by "blower" and "open air", but these are just factory 1080ti (MSI) and 1070 (FE I think?). And yes, they're in adjacent slots (well, as adjacent as possible since they take up a couple slots as I recall).
And yes, I'm a monitoring fanatic. And no, I've never seen them throttle. As I've posted many times before, the 1080ti gets into the mid-high 70's C and flattens there, and the fans flatten at about 78% as I recall and they stay there as long as I render. So there's a ton of headroom for the fans to crank up. For the 1080ti I recall it might throttle up near 90C, and max temp allowed is 105C or more, but I've never been even close. Again, the engineers design them to crank up the fans so they stay below the danger or throttling zone and stay there. No need to worry unless you've done something dumb like blocked the cooling system or done overclocking or messed up the fan curves or drivers or something.
Honestly, I keep hearing this continuing paranoia about overheating due to Iray and I just don't get it. And people adding water cooling to CPU's, when most apps nowadays barely use the CPU's?
Pretty much precisely why i am finding it difficult to decide. Given the 2060 is generally not gonna be used for renders, its not going to be under a huge amount of load so that single fan should not cripple it, but it will still be venting its heat inside the case, as opposed to outside if it was a blower.
I looked at octanebench to get an indication of the 2060 vs the titan performance, and you are right in that the 2060 is a bit faster but the difference is minimal. Majority of what i do is going to need more than 6gb of of vram. So for the moment the titans will still be my render cards
By the way, and I wish someone could explain this to me, there seems to be a basic assumption in much of the tech community that when you do stuff like rendering with Iray and GPU's, the cooling system won't be able to handle it, and the whole thing is destined to go into some sort of thermal runaway. There seems to be no acceptance of the possibility of a "stable operating point" where the fans and the rest of the cooling system can match the heat generated and everything will just stay that way as long as you're rendering. A bit like saying when you drive a Lamborghini it can't maintain 60mph on the highway, it's destined to lose control and speed up to 250 mph and crash.
the difference between blower style and open air: https://www.howtogeek.com/365215/what’s-the-difference-between-a-blower-and-an-open-air-gpu-cooler/
two pictures halfway down that page show the difference perfectly. The 1080ti thermal throttling threshold is 84c, not sure about the 1070 but its probably going to be the same, i think that is generally fairly uniform for each generation. Thermal throttling thresholds and max temps are different, i know and most others know that having them run at 84 degrees or a bit higher is not going to damage them, but it is reducing performance.
You talk about engineers and all that, and you are right. But they cannot replicate every single environment and pc setup that anyone could possibly ever have, and then make their devices always stay below their thermal throttling thresholds in every possible situation. Well they probably could, but it would mean every GPU is water cooled and more expensive.
At the end of the day, i know full well that my hardware throttles due to temps. And i dont want that. Yours may be fine, and mine probably would too if i was only running two GPUs. But i am not, and i wont. Just because you dont have throttling problems, does not mean others dont.
And just to touch on the topic of CPU water cooling, it is used far more than you think in plenty of applications. And gaming. These days games are more and more reliant on CPU than they have been in the past. Not only that, another reason people go with water cooling is to keep the temp of the device down, which means less heat dissipated into the case that then affects the temp of other components. Say all you want about it not being necessary, and for you it may not be. But for some people and some setups it is.
So why not just get a case like the mining case I referenced, or at least use the same concept. If it can handle 8 GPUs right next to each other, I'd think it would work fine for your three. Though I'm still not sure how a big ol' full tower case doesn't afford even more airflow and cooling capacity.
If the 2060 is never going to be used for gaming, then blower. Because it's not likely to ever get near its 160w tdp outside gaming.
To answer the question above about how 8 card machines go on, have you never opened a sever room door and been suprised by the sudden noise?
I have a 980ti and a 1080ti, both 'blower' type (exhaust through the back of the card). If I don't run a custom fan profile they will both hit the designed thermal limit in something over 15 minutes, at which point they start to cut back on the core and memory clock speeds. With a custom fan profile that ramps up to top speed more aggressively there's no problem. The normal fan profiles are designed for gaming and don't cool soon enough.
The X9 is a very unusual case so a lot of typical advice is right out the window.
A huge amount depends on how close the 3 cards are going to be and how mant fans you're adding to the system and where. If your airflow is front to back through the case and you're using 3 fans or more as intake I'd go with a blower style card. What I'd do is not put the new card in the middle. I'd then undervolt the Titan that you never use for video out. That will cut down on your power draw significantly and therefore the amount of heat you generate.
Miners, knowledgeableones at least, undervoltage their mining cards. That saves them money oput also cuts down on the heat generated. Mining, like GPU rendering, is not greatly affected by GPU clock speed.
BTW that sort of case would also be ideal for a render rig if the same thing was done or you used all blower style cards.
You can't undervolt a card that much but how much you undervolt a card is up to you, it's like overclocking but in reverse. But at some point the card doesn't have enough power to be stable or even work at all. You should find a good tutorial on undervolting GPU's vefore you try it for anything too extreme.
It can make a very large difference.
Even a drop of 25 watts is signicant. First it reduces the amount of power being drawn by the PSU from the wall. The farther from 50% rated max that a PSU is drawing the less efficient it becomes. That ineffiency is waste heat. Second a card reducing its draw from 250 to 225 watts means the GPU generates only 90% roughly of the waste heat. That can be the difference between thermally throttling and not. That also means less heat in the rig which means less heat being absorbed by other components and less heat that has to be exhausted.
BTW its not I think undervolting has a significant effect on GPU cooling, I know. Both my own personal experience building and maintaining servers, where we normally undervolt every component we can to save on our absurd cooling costs and power bill, and the experience of many others in multiple GPU setups.
Am currently in the process of upgrading from a Core X5 (the slightly smaller, otherwise virtually identical sibling of the Core X9) so I have a good amount of familiarity with the issues you will be facing with that case. As to your 2060 dilemma, I'd suggest sticking with a decently performing stock pcb blower card design (since that will keep your initial costs down and aftermarket cooler upgrade options fully open - should that avenue present itself) and simply swapping it in for the 770 in the new case for the time being. If you end up not liking the results, my next suggestion would be buy a $20 PCI-E extension cable like this one and custom mounting the 2060 somewhere away from your rendering cards altogether (one of the major advantages of cases like the X5 or X9 is that you actually have enough room to do this sort of thing.) At least until the time comes for you to upgrade your Titan X's to something more beefy (at which time I'd highly suggest moving to a custom water cooling loop for your GPUs since you'll both have the space and doing so will mean never having to worry about GPU thermals again - just water leaks...)
You're only thinking of cost per watt in terms of electrical power usage - not also in terms of thermal heat dissipation (which, as others have already stated, is a much more complex and potentially costly issue in pc building.) GPUs don't operate in a vaccuum. Computers are an entire ecosystem of parts which interact both electrically and thermally in ways that can cause massive performance uplifts/downdrafts in one or more components as a direct result of seemingly miniscule changes in electrical/thermal performance of just a single other component. Eg. 25 additional watts of GPU produced heat floating around in a high performance build can very easily lead to significant thermal throttling of a system's CPU VRM. You have to consider the whole picture.
Just because something sounds intuitively obvious doesn't mean it's factually correct.
Of course, more heat seems intutively bad, but that doesn't mean it matters. The internet tech world is overflowing with non-specific worries like this that people seem to cling to, but I'm trying to get folks to focus on numbers and facts rather than general paranoia that ends up wasting their money.
Yes, 25 watts additional might matter to someone. Especially someone who either bought a junk computer or didn't clean the dust bunnies or overclocked without sufficient cooling or some other dumb thing. But raising the ambient temperature in the house a few degrees by turning up the heat or failing to turn off the aircon can also heat up your computer. Do people worry about turning the thermostat down 5 or 10 degrees when fretting about their computer temperatures? That has a DIRECT effect on computer temps.
So yes, you do have to take the whole picture, rather than focus on just one aspect. But it's important that we put numbers and facts to it rather than just vague paranoia.
So does 25 watts matter for most users? Well, I've got a reasonably high end computer with dual GPUs that generate a max of about 400watts when rendering, and a 25 watt increase is only 6% additional power/heat generation. And my computer has been humming along WAY below throttling temps for years. And with a 6% increase in heat my fans would crank up a bit more and bring the temps down again. Now if someone wanted to mount a 25watt light bulb inside their PC and see if it affected temps significantly and caused throttling, then that would be some useful info.
For some reason people love to hold on to this idea that their machines are going to melt, and don't seem to consider the fact that they are designed to protect themselves from overheating and maintain a normal temperature. And people are so convinced that a 6% increase in power will suddenly throw the whole ecosystem into thermal runaway? I don't get it.
If someone wants to worry about an additional $1 a year in additional power costs, or the vague likelihood that a few additional watts will throw your system into thermal throttling, then that's fine with me. I just encourage people to search the facts rather than blindly follow the never-ending list of non-specific concerns.
BTW, I just saw an interesting test where a guy took a computer and pulled of the CPU cooler/heatsink and all computer fans, so there was effectively zero cooling, and ran a CPU stress test to see if the the whole thing would melt. It didn't. In fact it self-throttled as it was designed to do and settled into a temperature in the 90's C. With ZERO cooling, just a bare CPU. Yes, it throttled performance, but the computer was fine, and the average user would never have known the difference. So even with no cooling whatsoever your PC will likely be fine.
Okay, so I mounted a 25 watt light bulb in my PC and checked whether it made any difference
Well, actually, I tried doing a video render that uses all of my Ryzen 7 1700 8 cores at the same time as I did an Iray render in DAZ. When doing just an Iray render (no video) my CPU uses about 30 watts, so I figured if I render a video that will increase quite a bit. And it did...it increased the CPU power by about 20 watts t0 50 watts (TDP of 65 watts). So now my PC case has 20 more watts to worry about while it's cranking the two GPU's doing a render.
Previously, as I posted above, when doing just a render my GPU's (1080ti + 1070) get into the mid-high 70C range and stay there (actually the 1070 runs a few degrees cooler), and the 1080ti fans flatten at around 78%. Well, if I do a video at the same time and generate 20 watts more in the case that should increase, right?
Well, it didn't. GPU stayed at 77C and the GPU fans didn't even increase, they stayed at 78%. So it seems at least one data point shows that a 20 watt increase doesn't noticeably affect the temperature of the components. If anyone has any other data points I'd love to hear about them.
Fyi none of the numbers you cite here are accurate because TDP isn't a measure of electrical power consumption. It's an amorphous (since it is calculated differently per manufacturer per generation of hardware) measure of thermal power output a component is said to put out when operating at a specific level of performance. If you want to know the true power consumption of eg. your CPU you need to use either a wall plug mounted current electricity usage monitor or a current clamp mounted on your CPU's 12v power supply cable(s.)
Modern Nvidia GPUs, by design, don't sit within a fixed range of clock speeds/performance level and then thermally throttle (severely reduce clock speeds/performance) like modern CPUs do when excess heat generation becomes an issue. They retroactively thermally scale clock speeds/performance upwards from an official guaranteed minimum to infinity based on preset limits on internal power consumption and GPU die temperature (usually around 75-80c.) Meaning that increasing cooling performance on your graphics card will always lead to higher clock speeds/better performance up to your card's preset internal power consumption limit (usually related to the number of power pins it has.) Furthermore, all modern default bios settings on Nvidia cards feature fan speed curves which effectively eliminate (for the sake of noise reduction) the top 30% or so of the usable range.
Meaning that if you happen to have a 1080ti on stock bios fan curves in a multi-GPU system that's registering around 77C with fan speeds at around 78% at full load, you are almost certainly looking at a card being held back from its actual max level of performance by excess heat. What you are describing here is almost certainly a thermally limited system.
More heat always matters in the specific case of modern Nvidia GPUs. With that one exception, more heat only matters if it pushes a system component beyond its thermal throttling safety point. At least in the short term; electrical component operational life is almost always tied to operating temperature. And usually in an exponential (rather than linear) fashion I'm afraid...
I'm not talking about thermal throttling. I'm talking about thermal limiting (a whole different mechanism from thermal throttling) which modern Nvidia cards use to intelligently regulate the max performance threshold of your card without ever violating its max thermal or power consumption limits.
I don't. I think it is thermally limited because that is what your previously stated steady-states of 77C (78C now?) and 78% max fan speed indicate based on the way thermal limiting works in the Nvidia creation that is GPU Boost. The whole point of thermal limiting in Nvidia GPUs is to keep the card at a safe temperature distance from the card's true thermal throttling point so that thermal throttling never even has the chance to happen. This is because the dramatic shift in clock speed caused by thermal throttling causes massive visual anomalies in live graphics rendering. And remember - these GPUs are primarily targeted at consumers for gaming. Same with fan speeds. A modern stock Nvidia card will never reach 100% because fan curves have been tweaked to avoid excessive amounts of noise - not because they are more than capable of dissipating the max heat put out by the card at full load.
As a rule, with modern Nvidia GPUs the better the heat dissipation is in your case/GPU cooler, the higher your GPU will be able to clock under full load (theoretically all the way up to infinity - unless your card's power consumption limit gets in the way first.) Unless you are seeing early 70s or lower under full load with a 100% fan speed, you are almost certainly looking at a performance capped thermally limited card. And none of this has to do with overclocking.
Current monitor = power monitor. Also, the term "power dissipation" has never appeared in any of my posts. Electrical outlets consume and dissipate power. Computer components like GPUs and CPUs consume power and dissipate heat.