Quadro VS Geforce -- What's the truth?

[RobH2]

I've spent about 10 hours reading all I can find on GPU cards, Quadro, GeForce, RTX, etc. It gets very opinionated and confusing. 

There seems to be some "corporate" misinformation so that more Quadro cards are sold to professionals at a better profit margin. There are some vague references to Quadro cards having built-in features that are critical to 100% functionality in Max for modeling, rendering and viewport performance. However, I can't find any specific info that states what that is. 

What's your experience and feeling? I'm mainly asking pros, pros in large studios or users who do high-end work or entertainment work.  I use a Quadro K5200 that is fine but does not support real-time raytracing as well as the newer cards. I don't really want to buy a $6000 Quadro card. 

The specs of the GeForce RTX 2080 ti are very attractive. I could run two RTX 2080 (non-ti) cards for about the same price as one ti and get more VRAM.  

FUNDAMENTAL QUESTION: If I buy a GeForce RTX card, what "WON'T" work properly in Max? That's what I can't figure out from all of my reading. 

[RobH2]

I'm just circling back around to say my RTX is still thrilling me.

No regrets now in over 6-months. In fact, I received a private message from a user who asked of the RTX could eliminate their need for using a Render Farm. While the answer to that question is not quite that simple I did offer my opinion and some actual benchmark tests I've run using both the RTX 2080Ti and the K5200. I'll post that response below as well at the tests. Note: the tests file link will expire on 11/6/19. So anyone who finds this thread after that date won't have access to the link but I've attached that same file to this post. 

-------------------BEGIN PASTE-----------------

I have some data for you. I did a lot of tests before buying the GeForce card. I've always (for 23 years now) used the NVidia Professional or Quadro cards. I've never been really impressed with them and they've always been very expensive. So this was a scary switch for me. However, I'm glad I did it. 

I've attached some tests. I've always tested my systems over the years and I save the tests so I can compare them to see if I'm running fast. Sometimes my machines feel slow and by running a test I can tell if it's just me or if it really is slow. Usually, the test shows that the machine is fine and it's just me. So I ran tests before I bought the RTX. I 've included those. 

The tests with the most recent date are the RTX 2080Ti and the older test, Feb and March I believe, are the K5200. I tested with two different programs. I like a 2nd opinion I guess. You can go to the GPU or Graphics areas and see the significant speed increases. For some reason, DirectX10 got a lower score in one test but in all others, the RTX blows the K5200 away. If I had run the test again the DirectX10 would have probably improved as that low score is probably a momentary glitch. And, my overall machine scores improved. 

I can't tell you that you can eliminate a Render Farm or not. It's not that simple. The RTX only speeds up GPU rendering. A lot of my work needs CPU rendering because the GPU doesn't support every one of V-Ray's features. The RTX won't make your CPU renders any faster. I always use a Render Farm these days because if a frame takes 8-minutes to render and I have 1200 frames, that's 160 hours on my machine. That's about an hour on a farm using 100's of machines. 

What I can tell you is that you'll get faster IPR/GPU previews on your machine than you did. And, if you do render your work with GPUs, it will be about 2x-3x faster than it is with the K5200. Again, will that eliminate a Render Farm, it depends on how much time you have and how many frames? If the K5200 renders 900 frames overnight in 8 hours, then I'd guess the RTX will render those 900 frames in 3 hours. So it will save you time. But, no amount of GPU or RTX or whatever can eliminate a Render Farm when you have a large project. 

I have a Dual Xeon machine with E5-2650 V3 processors. So I have 40 CPU render threads. My machine is much faster than most Render Farm's CPU renders. But the power of the Render Farm is not "one" machine, it's that there are 1000s of them. That's the power of the Render Farm. So I say that no graphics card, RTX or otherwise, will eliminate a Render Farm. But what it will do is speed up what you are currently doing. Then you can decide. If your normal project with the K5200 takes 24 hours to render, then maybe the RTX will render it in half that time. There is probably a level where you can't have your machine tied up for so many hours rendering. Mine is 10 hours. If I can't render in 10 hours overnight or less, it goes to a farm. So on some level it can eliminate a Render Farm. Say the K5200 would take 20 hours to render. That would be Render Farm for me. And then consider that the RTX might render that same scene in 10 hours. Well then, the farm isn't necessary as it's now under my 10-hour cutoff, so in effect, I've eliminated the farm for that case. 

Anyway, I hope this helps. I have had the card over 6-months now and have never looked back. I still have my K5200 in a box knowing I could put it back in if the RTX ever let me down. To be honest, I don't think that expensive K5200 will ever go into a machine again. Want to buy it...lol....?

I hope this helps. 

Rob 

Here a link to the tests in a .zip file: (it will expire in 7 days)

https://drive.google.com/file/d/1i8wnHQBeuJUuORMk4tg4ECPbvJvILM_v/view?usp=sharing

[jibijib]

I've worked with Sony and Microsoft on AAA titles, we have always used Geforce graphics cards, not a single person there used a Quadro or anything else. I currently have Nvidia Geforce 2080 RTX in my work machine, it runs everything (Substance, Game Engine tools, etc) flawlessy. But I have a Geforce 980 with 16GB and a four year old processor at home and everything runs great!

Recently I worked on a project that needed something high performance so we had a single AMD (I think?) something or other card which I remember had outputs for 16 monitors - it was about £3000 at the time. The stuff we were doing was major performance intensive stuff - we all used GeForce cards in PC's that were updated fairly regularly for each release. 🙂

Companies will tell you you need anything to sell it, it's all marketing. If it makes you feel better, by all means buy it, but like driving a 200+ mph supercar, it's 99% unneccessary.

[senorpablo]

If something cost 4-5x as much, and it isn't absolutely clear what the practical benefits are, there probably aren't any. If there were any practical and tangible benefits, you can be sure that NVidia marketing, industry writers and users would be able to articulate them. 

Quadro's are a complete waste of money. 

[seanpattison]

Just wanted to post and say thank you for this thread. I pushed my company to get me a 2080ti when they kept trying to lean me towards Quadro. I do architectural visualization with Maya/Vray but I have been using Unreal / Twin Motion with VR more and more these days and the Quadro K5200 can't hold up anymore.

I figured a high-end gaming card will work perfectly fine with Maya, Photoshop, etc so I am glad to see that's the case. So many of these apps are moving toward real-time rendering so this is why I pushed for the 2080t over a newer Quadro. I don't need hyper-accurate wireframe lines or precise simulations. I need real-time viewing power. The card should be here in a week or so!

[whitebirchstudios]

I recently purchased a Titan RTX. When doing my research, it wasn't really a question between Quadtro VS GeForce, is was basically which GeForce. I am not a fanboy of either, I just wanted the biggest bang for the buck. Years ago I purchased a Quadro card, spent a ton of many, and it wasn't any better than the cheap gaming card that came with my machine.

[carrionfernandez]

RobH2, I'm late but I'm curiose about your benchmark. Can you answer me these questions in 3D Max?

* Which one can move more polygons?
* Wich one can move more vertices (in wireframe mode)
* Wich one can move faster the objects in playback animation?

* Wich one can move faster doing paint weight and moving vertices with Skin modifier?

* Wich one can move faster with CAT animations?

* Wich is faster doing a rendering (for example in Vray)

Of course the price of the 2080Ti is around 1200€uros, and the K5200 is 850€uros. So I guess that 2080Ti is faster than K5200. But I want to know how much faster.

Thanks!

[RobH2]

---

@carrionfernandez wrote:

RobH2, I'm late but I'm curiose about your benchmark. Can you answer me these questions in 3D Max?

Hi, 

I can't answer your questions because I'd have to reinstall the K5200 and run tests. I don't have time for that. What I can do is provide benchmark results and you can get an idea of the differences from them. The speed improvement of the 2080Ti is dramatic. The system is identical. I didn't alter anything else, same RAM, same Processors, etc. 

Regardless, I don't think I'd buy a K5200 because the card has been out for over 5 years. If you decide to use a Quadro card, look at a newer model and get the newer technologies. Plus, I was always disappointed with the K5200. It never seemed to perform as I expected. 

I'll answer the questions I can below. All of your questions involve a "comparison" that I can't do without reinstalling the K5200 but both cards performed all of the tasks you asked about without issue. 

* Which one can move more polygons?  Can't test
* Wich one can move more vertices (in wireframe mode)  Can't test
* Wich one can move faster the objects in playback animation?  Can't test

* Wich one can move faster doing paint weight and moving vertices with Skin modifier?  Can't test

* Wich one can move faster with CAT animations?   Can't test

* Wich is faster doing a rendering (for example in Vray)  If using GPU rendering, 2080 Ti is much, much faster. CPU renders are dependent on your, well, CPU. So your machine will differ from mine. 

I've had the card for a while now and I haven't found anything that doesn't function perfectly in Max or otherwise. The card has never done anything odd or strangely in Max. Have a look at the benchmarks and I hope it helps you evaluate the card. 

[Anonymous]

Rob,

I would be interested to know what settings you use in the nvidia control panel for your 2080ti.  Would you be able to post them?

I've bought a 2080ti system January 2019 and I didn't have any issues until I tired to do a little work from home.

I'm a cadcam engineer and I use powermill / hypermill.. during the calculation of a fine, complex toolpath my system just switched off and restarted without warning.. it happened a few times.

I use my home system for playing games, max settings usually, and I have not had any problems till now.

At work I use a Quadro P4000 and recalculating the toolpaths on this caused no issues what so ever.

I have yet to revert my system to the driver the cam company recommends, but I was wondering if it could be tweaked enough in the settings.

[RobH2]

I use the default NVidia settings and haven't found any need to modify them. If I used CAD and had tweaked something to get around issues my settings might be helpful. But unfortunately, I have nothing that will shed any light on your problem. 

If you do some reading about the Quadro vs the GForce cards you'll find a few obscure references that discuss a difference that may impact some functions in CAD software that requires, I believe I'm naming this correctly, "double digit precision." The Quadro cards have the ability to perform a few extra calculations that Max doesn't seem to care about. I don't use any high end CAD software so from my perspective, I don't see any downside to using the 2080 Ti. 

If you are using some CAD program that is dependent on this "one" feature difference, then you may, in fact, need a Quadro card. I'm not an expert in CAD so I just can't say. I'd do a little reading and also contact the vendor of that CAD software to discuss it. 

[CAMedeck]

I had to chuckle when I came across this discussion.  The same debate comes up every year or two.

However, I have a partial answer as to the difference between the two cards.

Quadro = Enterprise level hardware.  GeForce = Consumer level hardware.

So what does that mean?  Let's assume there are two new cards coming out; Quadro Meme and GeForce Yuyu.  If I go out and buy one of the first Quadro Meme cards, then buy another one five years later, they will be identical.  The RAM will be made by the same vendor, they will both be supported by the same drivers.  Every component is the same throughout the lifespan of that Quadro Meme.  

However, the much cheaper GeForce Yuyu will use whatever components are the least expensive at the time they are needed.  Drivers will need to change throughout the life of the Yuyu to accommodate the changes in component manufacturers.  

So what difference does it make, why make an enterprise card?  Big companies like to standardize their IT stuff as much as possible.  And being able to use a disk image for setting up new computers saves a lot of time.  They know the drivers on that image will work for as long as they use that Quadro card.  

Like I said, this is a partial answer.  It doesn't cover any performance differences, just the support aspect.  But it should help people understand the price difference at least.

[RobH2]

Sure. Thanks for confirming all of that but we sort of already know it. It comes up every few years because technology changes so quickly. You mentioned that as if it's a bad thing. I don't believe reviving the discussion every few years is moot. 

Likewise you kind of imply that "consumer" level is somehow inferior by saying they use the least expensive parts. Believe me, the "professional" level parts are being price shopped and negotiated. Sure, coming off the line they may test every single one, but I don't think inexpensive in computer component manufacturing always means inferior or defective.  😊

What we are trying to discern here is whether the cost of the "professional" cards, regardless of the components inside, are necessary for the vast majority of we "sole-proprietors" who don't need to 'Image' machines for large scale rollout. 

What we are trying to really figure out is not what's best, Professional or Consumer cards, but what computational features will and won't affect our everyday use of 3ds Max. There are some differences in the architecture that affect some CAD programs. But those issues can be discussed on the respective CAD vendor's forums. 

I've been using Max and Quadro cards for 25-years, well, maybe Quadro wasn't available that far back, but I did buy the first ones that came out whatever date it was. I've always been disappointed by the performance and now, just a year ago trying a GeForce RTS card, I have to say, I've wasted a lot of money over the years. This "consumer" level (I don't think it's a dirty word and I'm a 100% 3d visualization professional) card has been working like a Ferrari on steroids and way outperforms my Quado cards in every way. To be fair, I never invested $6K into the top of the line Enterprise Quadro cards, we little guys don't commonly have budgets for that, but I can't imagine they would provide a significant enough performance boost over the 2080 Ti, that I'm currently using, to justify the additional investment.  

Best part is if this "consumer" level 2080Ti craps out in 2 years, I'll get it's newest and hottest big brother at 1/4 the Quadro cost and keep on rolling. We all like to update our hardware every 2-4 years. I'd rather do that for $1500 instead of $6000. 

I for one have seen no downside by using the GeForce 2080Ti. I use Max on average 15-hours a day and I render hard and push the GPU to the max every day. This card has not skipped a beat. In fact, the past year with it has been more rewarding in Max than I've had in 25-years of using it. Plus, it's insane on the RTS games I play and in UE4 development. I can't be more pleased with the choice to "try" consumer NVidia cards to see what would happen. Nothing remotely bad has happened.  🙂 

[CAMedeck]

Rob, my apologies.  I had no intention of trivializing the discussion or making anything seem inferior or cheap.

Not only is technology constantly changing, but so is the way we use it.  Frequent discussions help us share knowledge when new and better ways to work come about.

Less expensive does not mean inferior, and I wasn't aware I'd made it sound that way.  But buying parts on the fly from the lowest bidder will always net a lower price than contracting with the lowest bidder to supply a part for the next several years.  

I only brought this up because it had been mentioned earlier in the thread and I wanted to help clear up the issue.  I know from many years of GeForce vs. Quadro talks that not many people understand the reasoning, and sometimes assume incorrectly that the parts are of a higher quality when you spend more money.  Which is not the case.

After using Quadro cards for most of my career, I now have a GeForce card in my workstation.  Most of my time these days is spent in Unreal Engine and I am currently using a GeForce RTX 2080 Ti which works well but for the small amount of RAM onboard.  

[seanpattison]

Hey I just wanted to chime in.

I actually just got the 2080ti a few days ago to replace my Quadro k5200 and I couldn't be happier. Maya is running like a champ. The viewport runs nice and smoothly, wireframes look really crisp, hyspershade is more stable/performs better and my Vray RT rendering is blazing fast. I also now have the ability to use renderers RTX features.

All this with the added bonus of being able to use programs like Twin Motion and the Unreal Engine for my real-time walkthroughs. Now I don't need to worry so much about my fps in walkthroughs in an out of VR.

[RobH2]

Hi Chris, no need to apologize. I didn't take offense. One thing that we don't get now that we all talk to each other with our fingers instead of our ears is tone-of-voice. Emojis help with that for sure. I realize your intent but it did, obviously...lol..., spur quite a diatribe from me. I type very quickly so I'm prone to long posts... 🙂 No harm, no foul. All dialog is good as long as people stay civil. I've long said that the Max Forum is one of the best on the planet. It rarely devolves into inane name-calling and insults like many forums out there that suffer immensely because of the lack of decorum.

It's all good... your input has a long history of being valued!

[RobH2]

@seanpattison 

Glad to see someone with the same experience as me so that I know mine is not a one-off. I too had most recently a K5200. I'm feeling the same bliss and now that I spend 1/4 of my time in Unreal Engine, I'm feeling that bliss there even more strongly. True platform-agnostic "real-time rendering" is coming like a freight train. I'm excited to be on board. 

Thanks for the feedback. 

[carrionfernandez]

Ok here I am going again.
I really don't understand that some of you were using Quadro for a long time and now you are very happy to change to GTX because is cheaper and equal faster (its difficult for me believe it), let me show you a few thing about 3DMax, a program that I know well. I will show you a few pictures, that you can see how big differences can be in your workflow in 3ds Max

The firs image 1) Standar+DefaultShading, as you can see my 980M GTX can move (orbit) very very easy 25millions of polygons, my GPU is working almos 90%. Thats perfect. In fact, I dont care if is Default Shading or Wireframe mode.

If i finish here my benchmark looks like 980m can be a good buy because nowadays is very cheap (so many "youtubers" just move around the object). But is not true, and now it will came the big problems.

The picture 2) Standar+DefaultShading+Playback.jpg, show you a playback animation at 0.201!!! Frames!!!
The GPU and the CPU is working at very low percentage and it's not taking the best performance of the computer (probably the drivers are not doing well his job)

In the last picture (3), show you if you try to move the vertices of a single object that contain around 6millions of polygons, you can move it at 0.130 frames per second!!!! That impossible work with that frame rate.

To sum up, move the viewport is the easiest thing for a graphic card, work with the polygons is another thing, and for animation is another another thing.

And I use 3DS Max for animation, and in some parts the card isuseless because 3DMax only use CPU. The question is...¿maybe a Quadro is design to give the maximun power in any situation of 3dMax like animation, moving vertices...etc?

You guys have been working for so many years in 3DMax, between professional studios and home edition you will know the answer... I hope so 🙂

P.D: About renders, the 980m is really fast with iRay, so render is not my problem. And also the last 3Dmax 2020, run much much faster that my old 3dMax 2016 (around 2x faster), great performance in 2020 version

[RobH2]

If you'll upload your teapot file so we are working with the same base I'll do some tests. I don't use a GTX card however. I'm using the more expensive RTX 2080 Ti. A test will be interesting. 

Also, what's the little program you are using to view the CPU and GPU usage? I'll use the same thing if you can tell me what it is. .

[carrionfernandez]

I have attached the files from 3DMax, and the Open Harware Monitor to view the CPU and GPU.

The file of Teapot Vertices, you must convert all the objects to "Editable Poly", and Attach all the objects into a single object. So, you will have one object, and then you can try to move the vertices.
I must do it like that beacuse the file can wieght around 800mb, and its too much for attached here.

I will wait for your benchmark.

Thanks for your time 🙂

[RobH2]

Here are screenshots of my tests. Test 1 and Test 3 are pretty similar to yours. Test 2 in moving the vertices I got a much better frame rate, about 2.1 (about 16x faster than your frame rate). It was a little slow but very usable.

I think the vertex count of 25mil is on the extreme side and may not be a fair commentary on Max’s abilities. With a file with that many polys, you’d be doing a lot of instances, proxies and other things to keep the count down. I’ve done some pretty large civil engineering projects that had a lot of plants and Forest Pack instances in them. As detailed as those were, they usually could be kept to 5mil polys.

I’ve also attached a comparison of the three tests as one image to make comparison easier. I also did an animation test as an alternate. I animated the teapots that were attached across the x-axis and got a really good frame rate moving those 25mil polys.

This was an interesting test. I’m not sure what it determines exactly but it does compare two machines doing the same tasks. I’m curious to know what “takeaway” you interpret from these tests.

[Note: The 'Attachments' feature is acting odd. I'm not sure what's going to happen when I hit 'Post.' I may have to come back and edit a few times to get all the attachments to upload.]

[RobH2]

Here are the other images. I could only attach 3, and only one-at-a-time by editing the post over and over.