Another 1080 ti? please help

kully_badman_129620027b

So I am an experienced Octane user and decided to switch to iRay a year ago and have noticed that it is immensly slower than octane. The reason why I decided to stick with it is becuase I think it produces truly stunning interior scenes whereas Octane (imo) excels at producing visually captivating exterior scenes.

My current setup is;

AMD ryzen 1700x CPU

Kingston HyperX 2400MHz RAM

EVGA supernova P2 850W PSU

GTX 1080Ti GPU

Will I see a considerable difference in render speeds if I add another GTX 1080Ti?

Please help me, Thank you for reading. XD

 

 

 

ebergerly

Theres a thread here somewhere where people post render times for the same scene. I think it said a second card might give 40% improvement with DAZ Iray. Or something like that. But you cant use SLI.

SpottedKitty

There might be something you can do about that "immensely slower" Iray rendering. Which versions of DAZ|Studio and the NVidia graphics card drivers are you using? The first couple of D|S versions using Iray had a few bugs, which NVidia fixed, and the new 10-series graphics cards need a recent-ish version of the NVidia driver; if you don't have this, Iray will ignore your graphics card and fall back to rendering on the CPU, which is much slower.

mjc1016

Also...define 'immensely slower'.

Remember when comparing renderers, you need to compare them as closely as possible, doing the same exact steps with the same assets and matching shaders as possible, too.

kyoto kid

mjc1016 said:

Also...define 'immensely slower'.

Remember when comparing renderers, you need to compare them as closely as possible, doing the same exact steps with the same assets and matching shaders as possible, too.

...that was what I did to compare 3DL and Iray CPU render times  I saved two versions of the same scene at the same pixel resolution with shaders optimised for each render engine. The only difference was the lighting I used, in the 3DL version it was AOA's advanced lights and in Iray it was the Sun/Sky setting.  In the end 3DL rocked, rendering in about 14 min, while the Iray version took around two hours tewnty five min. if I remember correctly.

ebergerly

Okay I found the thread with the benchmark info:

https://www.daz3d.com/forums/discussion/53771/iray-starter-scene-post-your-benchmarks#latest

You might want to download it and try a render with your existing 1080ti and see if you get similar results. That might tell you if there's something amiss with your present configuration.

I think they were getting less than 2 minutes with a single 1080ti, and 1 minute 20 sec with two of them.

BTW, one of the most recent posts (3rd from the last or something) is a chart I put together in an attempt to summarize many of the results. It might help you not have to scroll thru a ton of pages

Gator

ebergerly said:

Theres a thread here somewhere where people post render times for the same scene. I think it said a second card might give 40% improvement with DAZ Iray. Or something like that. But you cant use SLI.

Noo, you'll nearly halve the rendering times adding a second card.  There's one or two posts with the times of one card then two that show that.

 

As for speed against Octane, forgot exactly what version I used last - 2.x.  I really didn't notice all that much difference in speed really, biggest thing I noticed was that Iray seemed to handle darker scenes with less fireflies (that could be due to me not optimizing 100% correctly, but with some tweaking and forum help I never got there).

Leonides02

kully_badman_129620027b said:

So I am an experienced Octane user and decided to switch to iRay a year ago and have noticed that it is immensly slower than octane. The reason why I decided to stick with it is becuase I think it produces truly stunning interior scenes whereas Octane (imo) excels at producing visually captivating exterior scenes.

My current setup is;

AMD ryzen 1700x CPU

Kingston HyperX 2400MHz RAM

EVGA supernova P2 850W PSU

GTX 1080Ti GPU

Will I see a considerable difference in render speeds if I add another GTX 1080Ti?

Please help me, Thank you for reading. XD

 

 

 

Yep, you'll definitely see a good increase in your render speed. It's all about the CUDA cores, and you'll be doubling those. 

I'm curious - I have Octane, but to me the way it converts Iray shaders is crap.The scenes (exterior and interior) rarely look as good as they do in Iray since the shaders are native. Have you found a way around this?

 

ebergerly said:

Okay I found the thread with the benchmark info:

https://www.daz3d.com/forums/discussion/53771/iray-starter-scene-post-your-benchmarks#latest

You might want to download it and try a render with your existing 1080ti and see if you get similar results. That might tell you if there's something amiss with your present configuration.

I think they were getting less than 2 minutes with a single 1080ti, and 1 minute 20 sec with two of them.

BTW, one of the most recent posts (3rd from the last or something) is a chart I put together in an attempt to summarize many of the results. It might help you not have to scroll thru a ton of pages

Something to keep in mind: I have THREE 1080 ti's and it still took me 52 seconds with the GPU load. That's the annoying bottleneck right now, but it's not as big a factor if you're rendering a big scene.

ebergerly

ebergerly said:

Noo, you'll nearly halve the rendering times adding a second card.  There's one or two posts with the times of one card then two that show that.

Well, yeah, but after looking at a long list of render times in that thread and comparing, I came to the conclusion that I took all the results with a grain of salt. And that grain of salt amounted to about +/- 15 seconds or so. And that was due to stuff like some people first doing a render, then noting the time of a second render, which is usually significantly faster since everything is loaded and ready to go. Same with some folks setting their 3D view on Iray rather than something else. 

I think you really need to allow for variations and consider any benchmark results as ballpark numbers. And in a scene that renders on the order of a couple or a few minutes, that variation can be quite a bit. 

For the OP, I figure if he's getting around 2 minutes or less for a single 1080ti he's in the ballpark. My 1070 got a little over 3 minutes, and even those results varied. 

kyoto kid

...another issue could be scene file size though I doubt most scenes would exceed 11 GB (unless you are running under W10 which reserves about 1.9 GB of VRAM leaving you with 9.1 GB to run the displays and render with). Even so it would take a very "heavy" scene to cause it to dump to the CPU. The largest scene file I have is around 8.6 GB (without the Daz programme open) which has a fairly detailed set, 8 characters, several emissive lights, and a drizzly fog atmospheric effect.

ebergerly

Oh yeah good point....the OP has a Ryzen like me. And the older chipset software (like May or something) would dump to CPU when it shouldnt. Something about a new power mode for Ryzen.

Gator

ebergerly said:

ebergerly said:

Noo, you'll nearly halve the rendering times adding a second card.  There's one or two posts with the times of one card then two that show that.

Well, yeah, but after looking at a long list of render times in that thread and comparing, I came to the conclusion that I took all the results with a grain of salt. And that grain of salt amounted to about +/- 15 seconds or so. And that was due to stuff like some people first doing a render, then noting the time of a second render, which is usually significantly faster since everything is loaded and ready to go. Same with some folks setting their 3D view on Iray rather than something else. 

I think you really need to allow for variations and consider any benchmark results as ballpark numbers. And in a scene that renders on the order of a couple or a few minutes, that variation can be quite a bit. 

For the OP, I figure if he's getting around 2 minutes or less for a single 1080ti he's in the ballpark. My 1070 got a little over 3 minutes, and even those results varied. 

I don't know why you don't want to believe a second card greatly helps with render times, but you seem to be stuck in that belief.  From what I recall of that thread, most of the data supports the fact that multiple cards help greatly.  Sure, there may be a few exceptions I don't recall, but they are outliers.  Daz and Nvidia have stated the most important factor for Iray performance is the number of CUDA cores. 

 

I just ran through the Iray starter scene since I have some 1080 Tis now.  Here's my results.

 

2 Titan X Pascal cards in i7-6700K system 48 GB RAM

2 Titan X Pascal enabled with OptiX
Total Rendering Time: 1 minutes 8.59 seconds

Pre-loaded
Total Rendering Time: 58.75 seconds

 

1 Titan X Pascal enabled with OptiX
Total Rendering Time: 2 minutes 11.86 seconds

Pre-loaded
Total Rendering Time: 2 minutes 0.94 seconds

 

2 1080 Ti cards in AMD FX-8320 system with 32GB RAM

2 1080 Ti enabled with OptiX
Total Rendering Time: 1 minutes 19.37 seconds

Pre-loaded
Total Rendering Time: 59.72 seconds

1 1080 Ti enabled with OptiX
Total Rendering Time: 2 minutes 13.21 seconds

Pre-loaded
Total Rendering Time: 1 minutes 52.60 seconds

 

For both the Titan X Pascal and 1080 Ti systems my times reflect many other's experience.  Adding a second card nearly doubles performance.  This isn't gaming with SLI. 

OP, I recommend loading the scene linked above to compare your times, if vastly different then something may be wrong.

ebergerly

ebergerly said:

I don't know why you don't want to believe a second card greatly helps with render times, but you seem to be stuck in that belief. 

 You seem to think I care how much a second card helps. I'm only looking at the data that people post. If you have different data then post it. I just copied what people posted. I'm certainly not a "second cards don't help" fanboy (if there is such a thing..) or something. I don't care. 

And your data seems to prove my point...

• 2-1080ti render times can vary between 1 minute and 1.3 minutes.
• 1-1080ti can vary between almost 2 minutes and 2.25 minutes

So by your numbers a second card can make a 40-50% improvement. Which is about what the other numbers show. And same as when I said "about 2 minutes for a single 1080ti". Which is what you got.  

So what's your point? 

Havos

If the render time is short, ie 5 mins or less, then a lot of that time will be compiling the shaders and other start up work. This will require much the same amount of time regardless of how many cards you have. For longer renders, an hour or more, set up time is less relevant, and the speed up for having a second card would be closer to double.

Thus the actual percent increase varies a lot depending on time it takes to render, and why the performance test Sickleyield set up is not really ideal for these high performance cards.

Gator

ebergerly said:

ebergerly said:

I don't know why you don't want to believe a second card greatly helps with render times, but you seem to be stuck in that belief. 

 You seem to think I care how much a second card helps. I'm only looking at the data that people post. If you have different data then post it. I just copied what people posted. I'm certainly not a "second cards don't help" fanboy (if there is such a thing..) or something. I don't care. 

And your data seems to prove my point...

• 2-1080ti render times can vary between 1 minute and 1.3 minutes.
• 1-1080ti can vary between almost 2 minutes and 2.25 minutes

So by your numbers a second card can make a 40-50% improvement. Which is about what the other numbers show. And same as when I said "about 2 minutes for a single 1080ti". Which is what you got.  

So what's your point? 

I only mean to point out that you're simply spreading misinformation. 

There will be other system and software variables, as well as different vendor cards & driver differences, so what I get with a 1080 Ti may not be exactly the same as what someone else gets.

Your math is wrong.  Pre-loaded 1080 Ti 59.72 seconds for two cards vs. 1 minute 52.60 seconds for one card.  112.6 seconds is 189% of 59.72 seconds - just a little short of 200%.  As I stated, nearly double.

ebergerly

Scott, we're saying the same thing. You're quoting "how much worse with one card" and I'm quoting "percent improvment with two cards".

One minute with two cards versus two minutes with one card is either:

• 50% improvement by having two cards, or
• 100% worse by having one card, or
• One card takes twice as long (which is your 200%)

It's different ways of saying the exact same thing. And I'm not spreading misinformation, I'm copying what others have said. Geez guy relax. 

Gator

 

ebergerly said:

Scott, we're saying the same thing. You're quoting "how much worse with one card" and I'm quoting "percent improvment with two cards".

One minute with two cards versus two minutes with one card is either:

• 100% improvement by having two cards, or
• 100% worse by having one card, or
• One card takes twice as long (which is your 200%)

It's different ways of saying the exact same thing. And I'm not spreading misinformation, I'm copying what others have said. Geez guy relax. 

Almost...  Fixed. 

(Technically not 100%, but I assume you're using a rounded figure).

I'm not uppity about it, again I'm just pointing out that saying adding a second card will only offer 40% improvement (initially) to 40-50% improvement is wrong.

ebergerly

I think your math is a bit off....

% improvement is something like this:

• Starting speed = 2 minutes
• Ending speed = 1 minute
• % improvment = (2-1)/2 = 50%

You have improved by 50% of the original time, since 1 minute is 50% of 2 minutes. Yeah, it's also "100% faster", but typically that's not how it's reported. Those are two different things. 

Gator

ebergerly said:

I think your math is a bit off....

% improvement is something like this:

• Starting speed = 2 minutes
• Ending speed = 1 minute
• % improvment = (2-1)/2 = 50%

You have improved by 50% of the original time, since 1 minute is 50% of 2 minutes. Yeah, it's also "100% faster", but typically that's not how it's reported. Those are two different things. 

Reviewers typically do that with computing regarding times.  You're nearly doubling your scene processing power, which translates to render times being cut in half.

ebergerly

That's right. "Render times cut in half" is the same as saying "50% improvement in render times".

But it's not a "100% improvement in render times", because that would mean it takes 0 seconds to render. 

 

Gator

ebergerly said:

That's right. "Render times cut in half" is the same as saying "50% improvement in render times".

But it's not a "100% improvement in render times", because that would mean it takes 0 seconds to render. 

 

I think the disconnect is that the way you stated isn't common lingo.  At least not anywhere I've been or reviews I've read.  You wouldn't say 50% improvement in render time, you'd say a 50% decrease in render time, or render time cut in half.

GarrettDR

scott762_948aec318a said:

ebergerly said:

ebergerly said:

Noo, you'll nearly halve the rendering times adding a second card.  There's one or two posts with the times of one card then two that show that.

Well, yeah, but after looking at a long list of render times in that thread and comparing, I came to the conclusion that I took all the results with a grain of salt. And that grain of salt amounted to about +/- 15 seconds or so. And that was due to stuff like some people first doing a render, then noting the time of a second render, which is usually significantly faster since everything is loaded and ready to go. Same with some folks setting their 3D view on Iray rather than something else. 

I think you really need to allow for variations and consider any benchmark results as ballpark numbers. And in a scene that renders on the order of a couple or a few minutes, that variation can be quite a bit. 

For the OP, I figure if he's getting around 2 minutes or less for a single 1080ti he's in the ballpark. My 1070 got a little over 3 minutes, and even those results varied. 

I don't know why you don't want to believe a second card greatly helps with render times, but you seem to be stuck in that belief.  From what I recall of that thread, most of the data supports the fact that multiple cards help greatly.  Sure, there may be a few exceptions I don't recall, but they are outliers.  Daz and Nvidia have stated the most important factor for Iray performance is the number of CUDA cores. 

 

I just ran through the Iray starter scene since I have some 1080 Tis now.  Here's my results.

 

2 Titan X Pascal cards in i7-6700K system 48 GB RAM

2 Titan X Pascal enabled with OptiX
Total Rendering Time: 1 minutes 8.59 seconds

Pre-loaded
Total Rendering Time: 58.75 seconds

 

1 Titan X Pascal enabled with OptiX
Total Rendering Time: 2 minutes 11.86 seconds

Pre-loaded
Total Rendering Time: 2 minutes 0.94 seconds

 

2 1080 Ti cards in AMD FX-8320 system with 32GB RAM

2 1080 Ti enabled with OptiX
Total Rendering Time: 1 minutes 19.37 seconds

Pre-loaded
Total Rendering Time: 59.72 seconds

1 1080 Ti enabled with OptiX
Total Rendering Time: 2 minutes 13.21 seconds

Pre-loaded
Total Rendering Time: 1 minutes 52.60 seconds

 

For both the Titan X Pascal and 1080 Ti systems my times reflect many other's experience.  Adding a second card nearly doubles performance.  This isn't gaming with SLI. 

OP, I recommend loading the scene linked above to compare your times, if vastly different then something may be wrong.

Scott, what motherboard are you using? I am preparing to build me a new system and I am going with 2 nvidia 1080TI cards. I plan on getting the i7 7700K processor. Also, are you liquid cooling your cards? Whats temps and load percentage when rending large scenes?

ebergerly

Dude you're just too funny. 

Gator

ebergerly said:

Dude you're just too funny. 

Well, words do matter. 

Initially you just said "a second card might give 40% improvement" - an improvement to what?

I think most people, myself included, when hear the word improvement will assume you refer to performance.  When referring to times, increase or decrease would be commonly used.

Gator

Garrett R. said:

scott762_948aec318a said:

ebergerly said:

ebergerly said:

Noo, you'll nearly halve the rendering times adding a second card.  There's one or two posts with the times of one card then two that show that.

Well, yeah, but after looking at a long list of render times in that thread and comparing, I came to the conclusion that I took all the results with a grain of salt. And that grain of salt amounted to about +/- 15 seconds or so. And that was due to stuff like some people first doing a render, then noting the time of a second render, which is usually significantly faster since everything is loaded and ready to go. Same with some folks setting their 3D view on Iray rather than something else. 

I think you really need to allow for variations and consider any benchmark results as ballpark numbers. And in a scene that renders on the order of a couple or a few minutes, that variation can be quite a bit. 

For the OP, I figure if he's getting around 2 minutes or less for a single 1080ti he's in the ballpark. My 1070 got a little over 3 minutes, and even those results varied. 

I don't know why you don't want to believe a second card greatly helps with render times, but you seem to be stuck in that belief.  From what I recall of that thread, most of the data supports the fact that multiple cards help greatly.  Sure, there may be a few exceptions I don't recall, but they are outliers.  Daz and Nvidia have stated the most important factor for Iray performance is the number of CUDA cores. 

 

I just ran through the Iray starter scene since I have some 1080 Tis now.  Here's my results.

 

2 Titan X Pascal cards in i7-6700K system 48 GB RAM

2 Titan X Pascal enabled with OptiX
Total Rendering Time: 1 minutes 8.59 seconds

Pre-loaded
Total Rendering Time: 58.75 seconds

 

1 Titan X Pascal enabled with OptiX
Total Rendering Time: 2 minutes 11.86 seconds

Pre-loaded
Total Rendering Time: 2 minutes 0.94 seconds

 

2 1080 Ti cards in AMD FX-8320 system with 32GB RAM

2 1080 Ti enabled with OptiX
Total Rendering Time: 1 minutes 19.37 seconds

Pre-loaded
Total Rendering Time: 59.72 seconds

1 1080 Ti enabled with OptiX
Total Rendering Time: 2 minutes 13.21 seconds

Pre-loaded
Total Rendering Time: 1 minutes 52.60 seconds

 

For both the Titan X Pascal and 1080 Ti systems my times reflect many other's experience.  Adding a second card nearly doubles performance.  This isn't gaming with SLI. 

OP, I recommend loading the scene linked above to compare your times, if vastly different then something may be wrong.

Scott, what motherboard are you using? I am preparing to build me a new system and I am going with 2 nvidia 1080TI cards. I plan on getting the i7 7700K processor. Also, are you liquid cooling your cards? Whats temps and load percentage when rending large scenes?

It's an older board now, the Intel system is on an Asus ROG Maximus VIII Hero.  When rendering, I see the cards running 100%.  My 1080 Ti's are water cooled, they usually run about 49-50 C with EVGA's default turbo clock which is overclocked vs. Nvidia spec.  The Titan X's are air cooled, they'll be 80-84 C I think and I will only get base clock with the top card as there's too much heat.  Bottom card not much over base if at all.

My current Intel rig will get the water cooled 1080 Ti's.  Then I'll probably build a Threadripper system, and put the Titan-X Pascals in it and probably water cool them too.

GarrettDR

What will your new motherboard be? I was looking at ASRock - Fatal1ty Z270 Gaming K6 ATX LGA1151, and the Asus - MAXIMUS IX HERO ATX LGA1151. I want a mb that will have the space provided between both gpu cards. Did you install the water cooling boards on yours? Im thinking of getting dual  Asus - GeForce GTX 1080 Ti 11GB POSEIDON cards since they have both water and air. Your views on this will be appreciated. I am also thinking about installing my 1050 ti to use just for my three monitor setup.Thanks!

Gator

@Garrett R.

Well, most likely I'll build a Threadripper.  I'm waiting for the 1900x to be released, and see how that performs and hopefully there will be some Labor Day sales.  They use a different chipset, the x399.  I may bite the bullet and get the ASUS ROG Zenith Extreme for it.  Haven't really decided.  I have the EVGA 1080 Ti Hybrid cards, they are an AIO water cooler setup.  I chose those as I've never done water cooling so it was something simple for me to get my feet wet water cooling.  I like the idea of an AIO with EVGA's warranty.

GarrettDR

I forgot EVGA had those hybrids. I am in the same boat. I never did water cooling before. I appreciate the info!

kyoto kid

ebergerly said:

ebergerly said:

I don't know why you don't want to believe a second card greatly helps with render times, but you seem to be stuck in that belief. 

 You seem to think I care how much a second card helps. I'm only looking at the data that people post. If you have different data then post it. I just copied what people posted. I'm certainly not a "second cards don't help" fanboy (if there is such a thing..) or something. I don't care. 

And your data seems to prove my point...

• 2-1080ti render times can vary between 1 minute and 1.3 minutes.
• 1-1080ti can vary between almost 2 minutes and 2.25 minutes

So by your numbers a second card can make a 40-50% improvement. Which is about what the other numbers show. And same as when I said "about 2 minutes for a single 1080ti". Which is what you got.  

So what's your point? 

...considering I am still stuck in the Iray CPU rendering slow lane where times are measured in hours worrying about 20 to 30 seconds to even a full extra minute seems odd to me. I understand that for animating yes, it does add up when you are talking about say 2,880 frames (2 min) as each extra minute would equate to a total of an additional 48 hours for rendering that 2 min sequence. This is why pro studios rent time on, or have their own render farms Granted I only create and render single frame scenes as my old system is not up to the task of animation and I cannot afford a high menory GPU card like a 1080 Ti or even a 1070 (the latter due to the great cryptomining rush of 2017).

Gator

kyoto kid said:

ebergerly said:

ebergerly said:

I don't know why you don't want to believe a second card greatly helps with render times, but you seem to be stuck in that belief. 

 You seem to think I care how much a second card helps. I'm only looking at the data that people post. If you have different data then post it. I just copied what people posted. I'm certainly not a "second cards don't help" fanboy (if there is such a thing..) or something. I don't care. 

And your data seems to prove my point...

• 2-1080ti render times can vary between 1 minute and 1.3 minutes.
• 1-1080ti can vary between almost 2 minutes and 2.25 minutes

So by your numbers a second card can make a 40-50% improvement. Which is about what the other numbers show. And same as when I said "about 2 minutes for a single 1080ti". Which is what you got.  

So what's your point? 

 

...considering I am still stuck in the Iray CPU rendering slow lane where times are measured in hours worrying about 20 to 30 seconds to even a full extra minute seems odd to me. I understand that for animating yes, it does add up when you are talking about say 2,880 frames (2 min) as each extra minute would equate to a total of an additional 48 hours for rendering that 2 min sequence. This is why pro studios rent time on, or have their own render farms Granted I only create and render single frame scenes as my old system is not up to the task of animation and I cannot afford a high menory GPU card like a 1080 Ti or even a 1070 (the latter due to the great cryptomining rush of 2017).

Keep in mind that's a really simple scene at a low res.

Latest project I'm working on had a ton of specular materials in it, along with 8 Genesis 3 figures in some scenes.  Still images took 8-9 hours to render off the two cards...  double to 16-18 hours?  ACK!  Granted that's the worst.  I probably average 4-5 hours with the two cards for render times.

Halving a few minutes is no big deal, halving hours is!

dragotx

scott762_948aec318a said:

kyoto kid said:

ebergerly said:

ebergerly said:

I don't know why you don't want to believe a second card greatly helps with render times, but you seem to be stuck in that belief. 

 You seem to think I care how much a second card helps. I'm only looking at the data that people post. If you have different data then post it. I just copied what people posted. I'm certainly not a "second cards don't help" fanboy (if there is such a thing..) or something. I don't care. 

And your data seems to prove my point...

• 2-1080ti render times can vary between 1 minute and 1.3 minutes.
• 1-1080ti can vary between almost 2 minutes and 2.25 minutes

So by your numbers a second card can make a 40-50% improvement. Which is about what the other numbers show. And same as when I said "about 2 minutes for a single 1080ti". Which is what you got.  

So what's your point? 

 

...considering I am still stuck in the Iray CPU rendering slow lane where times are measured in hours worrying about 20 to 30 seconds to even a full extra minute seems odd to me. I understand that for animating yes, it does add up when you are talking about say 2,880 frames (2 min) as each extra minute would equate to a total of an additional 48 hours for rendering that 2 min sequence. This is why pro studios rent time on, or have their own render farms Granted I only create and render single frame scenes as my old system is not up to the task of animation and I cannot afford a high menory GPU card like a 1080 Ti or even a 1070 (the latter due to the great cryptomining rush of 2017).

Keep in mind that's a really simple scene at a low res.

Latest project I'm working on had a ton of specular materials in it, along with 8 Genesis 3 figures in some scenes.  Still images took 8-9 hours to render off the two cards...  double to 16-18 hours?  ACK!  Granted that's the worst.  I probably average 4-5 hours with the two cards for render times.

Halving a few minutes is no big deal, halving hours is!

This is the boat I'm in and why I'm wanting to add a 1080TI to my rig.  Right now most of my renders are running around 20 hours to get to 100%, but I render at stupidly large resolution, and have a bad habit of sticking 6 or more characters in the scene.  I'm running one 1070 in my machine right now, looking to eventually add a 1080ti.  I just haven't quite been able to justify the expense.  If I ever start making money off my work, that's the first place it's going