Quadro RTX 4000 vs RTX 2080 Ti

So,

I have been using the RTX 4000 for a few months and also got an RTX 2080 Ti for my new PC. However, the 2080 Ti arrived early, so I tested both in the old PC and here are the results from Holomark.

Quadro:

Holomark 2 R6 v2,61

Total Score: 232556
Total Runtime: 1054.67 sec

GPU scores: 217460
GPU_01 - 533.30 fps - Cube 4 tests
GPU_02 - 56.20 fps - UDT Shape
GPU_03 - 625.00 fps - Wireframe
GPU_04 - 526.30 fps - Shaded
GPU_05 - 109.90 fps - Rendered
GPU_06 - 91.70 fps - Block Rendered
GPU_07 - 939 units Nurbs @ 5 fps in Wireframe
GPU_08 - 343 units Nurbs @ 5 fps in Shaded
GPU_09 - 64 units Nurbs @ 5 fps in RenderSpeed
GPU_10 - 204.10 fps - RenderMesh Render
GPU_11 - 588.20 fps - RenderMesh RenderSpeed
GPU_12 - 500.00 fps - JoinedMesh Render
GPU_13 - 909.10 fps - JoinedMesh RenderSpeed
GPU_14 - 138 units mesh @ 15 fps in Shaded
GPU_15 - 402 units mesh @ 15 fps in Render
GPU_16 - 978 units mesh @ 15 fps in RenderSpeed
GPU_17 - 238.10 fps - mesh in Rendered Studio
GPU_18 - 30.60 fps - Nurbs in Rendered Studio
GPU_19 - 588.20 fps - Block Illustration
GPU_20 - 156.00 fps - 2D single
GPU_21 - 94.30 fps - 2D massive (20x)

CPU scores: 15096
CPU_01 - 8.89 sec - Booleans and Contours
CPU_02 - 2.77 sec - Twist and Taper (UDT)
CPU_03 - 4.96 sec - Meshing Mini
CPU_04 - 0.03 sec - Extract Render Mesh
CPU_05 - 0.03 sec - Join Render Mesh
CPU_06 - 12.03 sec - Reduce Mesh
CPU_07 - 0.58 sec - Calculating Technical display
CPU_08 - 3.83 sec - Making Silhouettes

System manufacturer
System Product Name

NVIDIA Quadro RTX 4000 - 4095.0 MB
DriverVersion: 26.21.14.4292

IntelĀ® Coreā„¢ i7-3930K CPU @ 3.20GHz
NumberOfCores: 6 NumberOfLogicalProcessors: 12
MaxClockSpeed: 3.2 GHz

TotalPhysicalMemory: 32.0 GB

Microsoft Windows 10 Pro

  • None - 64-bit

Rhino 6 sr 25 64 bit
AA level = 8

2080 Ti:

Holomark 2 R6 v2,61

Total Score: 268237
Total Runtime: 1427.1 sec

GPU scores: 255980
GPU_01 - 625.40 fps - Cube 4 tests
GPU_02 - 58.10 fps - UDT Shape
GPU_03 - 909.10 fps - Wireframe
GPU_04 - 625.00 fps - Shaded
GPU_05 - 119.00 fps - Rendered
GPU_06 - 99.00 fps - Block Rendered
GPU_07 - 1106 units Nurbs @ 5 fps in Wireframe
GPU_08 - 336 units Nurbs @ 5 fps in Shaded
GPU_09 - 56 units Nurbs @ 5 fps in RenderSpeed
GPU_10 - 238.10 fps - RenderMesh Render
GPU_11 - 588.20 fps - RenderMesh RenderSpeed
GPU_12 - 588.20 fps - JoinedMesh Render
GPU_13 - 1111.10 fps - JoinedMesh RenderSpeed
GPU_14 - 197 units mesh @ 15 fps in Shaded
GPU_15 - 519 units mesh @ 15 fps in Render
GPU_16 - 2079 units mesh @ 15 fps in RenderSpeed
GPU_17 - 263.20 fps - mesh in Rendered Studio
GPU_18 - 28.40 fps - Nurbs in Rendered Studio
GPU_19 - 35.20 fps - Block Illustration
GPU_20 - 158.50 fps - 2D single
GPU_21 - 77.50 fps - 2D massive (20x)

CPU scores: 12257
CPU_01 - 8.70 sec - Booleans and Contours
CPU_02 - 2.16 sec - Twist and Taper (UDT)
CPU_03 - 13.14 sec - Meshing Mini
CPU_04 - 0.03 sec - Extract Render Mesh
CPU_05 - 0.03 sec - Join Render Mesh
CPU_06 - 11.92 sec - Reduce Mesh
CPU_07 - 0.98 sec - Calculating Technical display
CPU_08 - 3.83 sec - Making Silhouettes

System manufacturer
System Product Name

NVIDIA GeForce RTX 2080 Ti - 4095.0 MB
DriverVersion: 26.21.14.4292

IntelĀ® Coreā„¢ i7-3930K CPU @ 3.20GHz
NumberOfCores: 6 NumberOfLogicalProcessors: 12
MaxClockSpeed: 3.2 GHz

TotalPhysicalMemory: 32.0 GB

Microsoft Windows 10 Pro

  • None - 64-bit

Rhino 6 sr 26 64 bit
AA level = 8

According to Holomark, the RTX 2080 Ti is better by some way, and even double up in some areas. What was strange is that the Quadro RTX 4000 performed slightly better with a Studio Driver - over a Quadro driver.

Anyway, I saw and read a lot about these two cards on here and thought Id chip in as I had a good benchmark platform.

Best

F

4 Likes

I have no idea why this is surprising? The 2080 ti has almost double the Cuda cores and Vram. Why do people keep buying these massively overpriced and under speced Quadros?

3 Likes

Itā€™s not, itā€™s just that people keep asking about it and I found myself in a position to weigh the two separately on the same system.

Personally I think it would be interesting to test the quadro 5000 against the 2080 ti.

I agree though, the quadros are stupidly over-priced. With the latest Nvidia Studio Drivers, the GeForce Cards are far superior in price for performance.

1 Like

Two quick questions:

Are the two sets of test results youā€™ve posted supposed to be identical?

How does the size and power consumption of the two cards compare? Iā€™m looking to upgrade and the Quadro RTX 4000 is top of my list because my PC case is small and a single slot, relatively low power GPU fits the bill nicely. I grew tired of graphics cards that wail like a banshee when they are under load several years ago. Iā€™m not going back there :smile:

Unless youā€™re running some other software that ā€œneedsā€ a Quadro, donā€™t bother with it unless you get the absolute top-of-the-line. Paying an insane amount of money for a minuscule performance benefit over the top GTX isā€¦well at least something, instead of paying more than a top-shelf GTX for less performance.

1 Like

hi,

Sorry about that, dumb mistake, its been editted now.

The Quadro is 160W and the 2080 Ti is 260W.

I think the quadro is a decent card and Iā€™ve enjoyed it, but the 2080 Ti blows it out of the water. That goes for Lumion, Twinmotion and Enscape as well, where the bechmarks of the Ti beat out even the Quadro 5000 by quite some margin.

I use Twinmotion a lot and I can tell you that performance difference is very noticealbe. It is louder than the than the Quadro, but i dont mind it at all.

I spend almost all of my time on 2D and 3D CAD in Rhino, with infrequent sessions of ā€˜renderingā€™ or visualisation. That is probably why Iā€™m still using a Quadro K4000, which has served me (very) well over the years. However, the promise that Cycles is showing within Rhino has made me start to look at render-capable GPU. The past 5 years of hardware development has seen my needs not only met, but probably surpassed. I no longer require bleeding-edge technology. That is why the RTX 4000 seems a good fit. Using the same Turing chip, what would be the equivalent mainstream graphics card?

It has 2304 cuda cores, the same as a RTX 2070, but a slim design, so probably at the same speed as an RTX 2060.

Theoretically a big upgrade from the k4000:
https://gpu.userbenchmark.com/Compare/Nvidia-Quadro-RTX-4000-vs-Nvidia-Quadro-K4000/m716215vsm7730

1 Like

Any graphics card with a blower cooler is a hard pass for me. They donā€™t cool anywhere near as well as a dual or triple fan cooler. Theyā€™re incredibly load by comparison also. Iā€™d recommend the RTX 2070 super or a used 1080 ti. The RTX 3000 series cards are out in Sept. Iā€™d keep that in mind as well.

2 Likes

If you have a small system, a blower is your best bet. The fan ones work better in a bigger system with lots of air flow. The quadro RTX 4000 is an amazing card and the Turing chips are cherry picked for performance; itā€™ll be like coming out of the dark from the K4000! Itā€™s the same swap I made. I just prefer the 2080 ti for complex rendering in twinmotion.

If your based in the UK let me know and we can make a deal for mine? Itā€™s two months old.

To be honest with you, I never understood these performance comparison threads. The thing with performance is, that

a.) the card doesnā€˜t matter at all, because the situations requires much less performance.

Or b.) The card doesnā€˜t really matter, because the performance lack is too big anyway.

To be within the performance range, where it matters is a very rare case. For high performance rendering, nowadays you can make use of cloud or local cloud computing, which outperforms any card by far.

What rather should be discussed is the advantage of Quadros compared to Geforce cards regarding anything else then performance, such as warranty,(hardware) reliability or official CAD support. Many CAD software, as already mentioned, optimize for a specific card and driver version, assuring that it reliable works.

The card performance matters a lot, especially if youā€™re doing anything that requires GPU rendering. I.e. rhino raytracing, vray, maxwell, twinmotion, lumion, enscape, etcā€¦

Yes cloud rendering can be faster, but it also costs money and some of them base the price off your GPU/CPU speed anyway.

You know people say this for a decade now, but renderings did not become better by any means. Still people believe to have the latest stuff to save 10 minutes for their rendering. But the funny thing is, then they add another set of polygons to their szene and still require the same time for rendering it. Its super realistic, but still images lack quality. This may becomes philosophical, but I personally rather disagree.

Edit: Of course Iā€˜m not saying take an Onboard card. But its about these 5 to 15% deltas, which also depend on so many other factors. Driver, Application, Szene Characteristics, Degree of (szene) optimisation. Amount of post processing, Type of Shaders, Amount of Light Sources, Type of Rendering, Resolution ā€¦

i can see the argument youā€™re making when it comes to creating ever more complex geometries and rendering them. However, if youā€™re just trying to render a piece of wood in maxwell render (for example) with a Quadro 4000 to SL 17 itll take 5 hours as opposed to 20 mins with the RTX 4000 on GPU render.

As it happens, I replaced the stock cooler on the K4000 with an after-market one by Scythe, which has two very low profile 100mm fans. It works very well and got rid of the whine of the stock cooler and the constant spooling up and down. Iā€™d definitely consider doing something similar to the new card, whichever one I end up with. Thereā€™s just about enough room in my itx case, which has a lot of perforations next the to the GPU slot.

I didnā€™t know about an upcoming RTX 3000. Thanks for the tip-off. I want to get a card with four monitor outputs, which the RTX 4000 has.

btw - whatā€™s with the ā€˜Meshing Miniā€™ times in those stats? Seems a bit odd. Does the GPU help out in some way, or is it likely just a glitch?

I put a morpheus cooler on my Titan xp with noctua nfa fans. I use msi afterburner to control the fans. I keep an eye on loads and temps with rainmeter desktop apps. It was absolutely worth the cost, time and effort.

Hahaā€¦ Must be a glitchā€¦ my system is a decade old now!

It seems strange that for the CPU stats, the RTX 4000 soundly beats the 2080Ti in both the Meshing Mini and Calculating Technical Display scores, while for the others the reutls are all but identical. Irrespective of whether or not your system is getting old (mine is too, as indeed am I :smile:), it would be nice to know if the difference can be attributed to the GPU.

Hi Matt,

Not sure what you mean old bean? Theres a 20,000 point difference in GPU scores. I think the CPU scores should be read independently of the GPU scores as they only test the i7 6 core processor and RAM. The glitches there are likely to do with the fact that my old CPU and RAM can be a bit tempermental.

Holomark can correct me if i am wrong, but I think the first test is on soley on the GPU and the second is solely on the CPU.

So, either there my CPU is glitchy, or there is roughly a 3000 point margin of error in holomarkā€™s benchmark.

Best,

F