Quadro vs GeForce for Creative Professionals
Imagine if you were a rally racer, but you bought a Formula 1 car, because it’s “professional.” Confusion around this topic is pretty common. Quadros are marketed towards professionals and GeForce cards are marketed towards gamers. Yet you’ll find many creative pros — including many with more than enough money to buy a Quadro — using GeForce cards. To add to the weirdness, Nvidia recently started a separate line of GeForce “Creator Ready Drivers.” What’s going on?
Quick recommendations
I strongly recommend reading the article, so you can make a more informed purchase, but if you just need some super-high-level suggestions…
PHOTOGRAPHY - GeForce RTX 2070 | Amazon
GRAPHIC DESIGN - Quadro RTX 5000 | Amazon
VIDEO EDITING (HD to 4k) - GeForce RTX 2080 | Amazon
VIDEO EDITING (above 4k) - Titan RTX | Amazon
COLOR GRADING - Quadro RTX 5000 | Amazon
MOTION GRAPHICS - GeForce RTX 2080 | Amazon
3D DCC & GPU RENDERING - GeForce RTX 2080 ti | Amazon
MULTI-PURPOSE - Quadro RTX 5000 | Amazon
Critical Knowledge: Basic Understanding of Precision & Teraflops (TFLOPS)
Let’s start with an analogy of a car transmission. Imagine that the transmission only has 3 gears.
1st gear = FP16, also called “half precision”
2nd gear = FP32, also called “single precision”
3rd gear = FP64, also called “double precision”
However, unlike a car transmission, we’re not constantly switching between all gears while driving. Instead, your software is using specific “gears” for specific jobs, no matter how you “drive.” The gears in this analogy are about the accuracy of your wheels in response to your steering command. Hence “precision” — each successive precision mode (gear) is more accurate than the last, but performs more slowly.
Creative software typically uses primarily (often exclusively) FP32/single precision. Conversely, if you’re involved with CAD, engineering, AI/machine learning, and other applications, the other two precision modes become way more relevant.
Now, what about Teraflops, or TFLOPS?
Where the precision modes are about accuracy, TFLOPS are about power, directly impacting speed. To be a smidge less basic, “theoretical speed.” Wait, so why is it theoretical? Imagine that we’re plotting out the performance of a car by its engine and tires. These are major factors in how the vehicle performs, but of course, not the only factors. So, TFLOP numbers don’t tell you everything, but quite a lot.
It’s similar in mindset to “miles per gallon” when shopping for a car. It’s not the whole story, but it’s enough to go by in a practical sense.
With TFLOPS and precision modes in mind, let’s look at a Quadro and a couple of GeForces.
Quadro RTX 5000
-$2,220
-FP16 - 22 TFLOPS
-FP32 - 11 TFLOPS
-FP64 - 0.35 TFLOPS
GeForce RTX 2080 ti
-$1,200
-FP16 - 27 TFLOPS
-FP32 - 13.5 TFLOPS
-FP64 - 0.42 TFLOPS
GeForce RTX 2070
-$500
-FP16 - 15 TFLOPS
-FP32 - 7.5 TFLOPS
-FP64 - 0.23 TFLOPS
So, if we’re talking strictly about brute horsepower, and if that’s a major concern for you (like for 3D artists), GeForce cards aren’t just more cost effective, you can shortcut to equal or better performance than a Quadro. If your software can use two video cards equally, you could buy two 2070’s and beat the Quadro, or two 2080 ti’s and absolutely slaughter the Quadro.
So, does a Quadro always lose like this? No way!
Basic Understanding of VRAM & ECC
Similar to the primary memory (RAM) for your computer, your video card has its own, dedicated memory: Video RAM, or VRAM. Any RAM is basically like a hard drive, but it only stores files temporarily (it literally “drains out” when you turn off your computer), with the benefit that it’s the absolute fastest way to read the files. This is used by all of your software, but useful in different ways, again, depending on exactly what you’re doing.
Quadros prioritize having higher amounts of VRAM. A lot of buyers target this specifically, along with other benefits discussed below.
For example, video editing software can put particular stress on your VRAM’s “framebuffer” when trying to play your footage. The framebuffer is the stage before the image is sent to your screen to view - it’s a collection of bitmap images with all of the necessary color and pixel information. In some scenarios, the information of your footage could overwhelm a GeForce’s framebuffer, and your playback performance will be affected. This is highly dependent on the format of your footage, resolution, framerate, color bit depth (which I’ll explain later), and other demands of your VRAM happening simultaneously.
There’s a catch, though. To build the image information that becomes your framebuffer, the brute horsepower of your card (in the TFLOPs we discussed earlier) is involved. Slower FP32 means a slower time building up to the framebuffer. It’s still a tricky call. Video editors are understandably split between Quadros and GeForces.
Another scenario is in GPU-based 3D rendering. GPU renderers like Octane and Redshift work by primarily storing your entire project in your VRAM. If your scene is too big to fit in the VRAM, the excess gets dumped into your computer’s regular RAM, and this slows down renders that are often already many hours long. So, if you often work with extremely large scenes, you’d benefit from the higher VRAM storage that Quadros offer. That being said, a lot of 3D artists don’t work with scenes big enough to make this relevant, so there’s no need to go beyond GeForce. Meanwhile, they can benefit from the same or better FP16 & FP32 performance at a fraction of the cost.
The other big factor is that the Quadro’s VRAM has Error Correction Code, or ECC. Basically, the information being processed by your VRAM is double-checked for errors. This is important when you must be 100% free of occasional glitches and inaccuracies at all times, without exception — a demand most notable in CAD/engineering applications. For creative professionals, again, it’s a bit of a toss. A lot of creatives are perfectly fine with the possibility of having a little glitch or an inaccurate result on rare occasion. For the most part, it’s not going to happen in any way that affects your work.
ECC isn’t saying that non-ECC is unreliable, it’s about mission-critical situations where you have zero tolerance for the possibility of a VRAM miscalculation. Considering CAD users are plotting out hyper-accurate, real-world measurements to forward for manufacturing/construction, you can see how this is extremely useful for them. However, video production companies may be handling very high-profile projects with a lot of money on the line, and even though they understand that a GeForce giving them VRAM miscalculations is “really unlikely,” they might have some peace of mind with “really, really unlikely.”
GeForce RTX 2080 ti
-$1,200
-8gb VRAM
Quadro RTX 5000
-$2,200
-16gb ECC VRAM
Titan RTX
-$2,400
-24gb VRAM
Quadro RTX 8000
-$5,500
-48gb ECC VRAM
That’s a situation where a Quadro “may” be the better choice, but what are situations where it is?
Bit Depth & 10-bit Output
GeForce cards offer 10-bit only for video games and some video players, but are still 8-bit for everything else, including your video editing, 3D design, or graphic design software. Quadros offer 10-bit for everything. This matters for people doing color-critical work. It’s not that you can’t work in 8-bit — a lot of people do — but you’re working with a handicap.
“10-bit” is a “bit depth,” and bit depth is about luminance — the range of brightness you’re able to perceive. Since color interacts with brightness (luminance), this means it impacts the range of colors you’re able to see as well. 10-bit is a big deal for certain users. HDR standards in newer TV’s and computer monitors also interact with 10-bit, so if you’re targeting HDR, but are working in 8-bit, you’re kind of working blind.
Before you jump on this point alone, bear in mind that you also need a 10-bit monitor to view a 10-bit image. I dive into that a bit more in this article.
Graphic designers in particular often don’t need to be in the highest range of FP32 power, though of course, if you can get up there, it’ll help when things get heavy. For many graphic design scenarios, you’re likely fine with a low-mid level Quadro, and you’ll get that juicy 10-bit.
“Top bin” components
Quadros are “top bin” hardware, meaning off the factory line, they’re double-checked for reliability. Similar to ECC, this doesn’t suddenly mean your GeForce choices are unreliable. But it does mean that if you have zero tolerance for the possibility of failure, a Quadro can give you some peace of mind. Quadros have a longer theoretical life span, and are better at putting up with constant, 24-hour stress.
Before you jump on this point alone, consider the price differential between Quadros and GeForces (including many other models not listed in this article). In a lot of situations, if your GeForce card happens to fail, you can run to the store and get another one — even with the cost of re-purchasing the GeForce, you’d still have spent less than on the Quadro.
Lower impact on your power bill
Quadros are manufactured in a more efficient manner, resulting in a lower power draw, and by extension, less impact on your power bills. However, if you’re only buying one or two video cards, the difference is typically negligible. Where this really starts to make a difference is if you have a facility with multiple computers that are working every day, stressing a lot of video cards at the same time. If you’re a major production company with dozens of computers going at once, Quadros can save you a good chunk of annual change.
Certified drivers
Quadro-exclusive “certified drivers” are made available for optimal performance & stability on specific software. The reported benefits of these drivers appear to be inconsistent. Some people report a “night and day” improvement, others report a negligible difference, and some report a worse experience. That being said, you may still consider it beneficial to at least have the option to explore certified drivers.
Other benefits
There are a host of other advantages with Quadros that largely aren’t related to creative software, or may only offer occasional benefit. For example, Quadros have optimizations for the simple-shader viewports in CAD software, which can also cross over to benefit some creative (DCC) 3D software, but not all.
More to come.
I’ve got more to add to this article, and will return with updates. Thanks for reading!