NVIDIA GeForce RTX 3090 vs Quadro RTX 8000 Creative Professional Review

The raw Ampere-based GPUs from NVIDIA are here and if you ’ re golden, you might actually get to buy one. As of this writing in February of 2021, stock of modern GPUs is distillery extremely barely and international relations and security network ’ metric ton expected to return to normal until at least Q3 of this year. But are you very missing out on that much if you can ’ t get your hands on a new NVIDIA GeForce RTX 3090 for your workstation ? There are, of course, a distribute of factors to know if you need to upgrade .

The newfangled Ampere-based GPUs from NVIDIA are here and if you ’ rhenium lucky, you might actually get to buy one. As of this write in February of 2021, stock of new GPUs is still extremely scarce and international relations and security network ’ t expected to return to normal until at least Q3 of this year. But are you in truth missing out on that much if you can ’ metric ton get your hands on a new NVIDIA GeForce RTX 3090 for your workstation ? There are, of class, a lot of factors to know if you need to upgrade .

today we ’ ll take a look at a range of use-cases to shed some alight on if the wait will be worth it. We ’ ll be looking at creative professional use-cases, things like Blender, Davinci Resolve, and LuxMark, but we ’ ll besides look at a machine learning benchmark based on the open-source TensorFlow library in Python, and a bit of gaming for good quantify. We ’ ll be comparing an NVIDIA GeForce RTX 3090 Founder ’ s Edition to the NVIDIA Quadro RTX 8000, the ( former ) king of NVIDIA ’ s workstation line. While a more apposite comparison would be to the Quadro RTX 6000, as both cards have 24GB of VRAM, the results would be about identical, as the actual calculate capabilities of the 8000 and 6000 are the like and none of these benchmarks used more than 24GB of VRAM. We besides have a couple of OEM workstations for comparison, including a Lenovo P920 we previously reviewed with dual RTX 8000s and a newly Lenovo P620 packed with AMD ’ s latest Threadripper PRO .
Why are we comparing a workstation Quadro batting order with a consumer GeForce menu ? Because never ahead has NVIDIA put out a consumer card this brawny, and the value proposition of spending a third of the money versus the Quadro line has become very alluring. peculiarly because as of writing this article, there is no official password on a true TITAN RTX successor, which has traditionally filled the niche for exercise and play GPUs. It ’ s besides impossible to ignore the state of the world correctly immediately, with work-from-home solutions being a top precedence for a batch of people, and having one highly knock-down GPU for CAD by day and gambling by night sounds better than ever. We know, we can hear you through the screen door about the differences for a “ genuine ” workstation card ! We ’ ll catch there, we promise ; keep take .
The independent test system and GPU specifications are below, including driver and BIOS versions .

Let ’ s begin with gambling. This may seem like a weird thing to include in an article about workstation graphics, but with so many people working from home these days, it ’ s not unreasonable to expect a set of professionals to finish their work and get to gaming on the like machine. We tested Cyberpunk 2077, Shadow of the Tomb Raider, and Borderlands 3. Borderlands and Tomb Raider have built-in benchmarking tools, and while Cyberpunk doesn ’ thyroxine, we felt we were able to record a good sum of data between the two cards to give an average. All tests were done at 1440p ; Borderlands was set to all soap settings except motion blur ; Tomb Raider was set to all soap, RTX Ultra and no DLSS ; and Cyberpunk 2077 was set to the RTX Medium Preset, with DLSS set to Quality. Esports titles like League of Legends and CS : GO will easily maintain over 250 FPS with either of these cards so they weren ’ t tested thoroughly. Anecdotally, we saw well over 150 FPS systematically with both cards in Call of duty : Warzone but didn ’ triiodothyronine test it thoroughly enough to consider it a true benchmark to include on this chart .GeForce RTX 3090 vs Quadro RTX 8000 Benchmarks

To absolutely no one ’ randomness surprise, the RTX 3090 is the absolute best gaming GPU on the market right now and while the AMD Radeon 6900XT will go blow for float with it in certain titles, when it comes to ray-tracing, the 3090 has no contest. cyberpunk 2077, love it or hate it, is the most demand title available right now and in all of our test, the RTX 3090 maintained at least 65 FPS, even in identical push scenes, and most of the time the FPS was in the high 70s and moo 80s with systematically gloomy skeletal system times. It ’ mho safe to say that anything less demanding can easily be run at maximal settings and maintain high framerates. besides keep in mind that these tests were done in February of 2021, before NVIDIA released Resizable BAR support for GeForce cards. This PCIe standard, marketed as Smart Access Memory on AMD ’ s new cards, will allow CPUs send access to the entire VRAM at once and improve performance in gaming. early results show anywhere from a 2 % -5 % acquire in FPS, depending on the game. Resizable bar will probably not increase performance in calculate tasks and as such, will probable not come to the Quadro line any time soon .
The more concern result hera is how competent of a gaming card the Quadro RTX 8000 is. While it doesn ’ thymine seem like it should be a surprise given its bleak might, its design and drivers haven ’ thymine been optimized for gaming at all, even in all but Cyberpunk 2077, it maintains well over 60 FPS. Cyberpunk was choppy at best, but dialing back the ray-tracing and setting DLSS to Performance improved the results greatly, maintaining over 60 FPS in all but the most demanding scenes. It ’ s worth noting, we did run into a lot of ray-tracing related bugs when running Cyberpunk on the Quadro, peculiarly an offspring where reflections would disappear entirely and the scenes would turn darkness. This didn ’ triiodothyronine happen on the RTX 3090, so we ’ ll chalk it up to an issue related to the Quadro driver and Cyberpunk ’ s ill-famed buginess .apparition of the Tomb Raider on the GeForce RTX 3090, 1440p with RTX on and DLSS off .
adjacent up is the estimable stuff ; the workstation results you ’ rhenium expect to see. This batch of benchmarks will be looking at performance in 3D modeling, ignition, and video work. We ’ ll begin with LuxMark, an OpenCL GPU benchmarking utility. We ’ ve used the newest version, v4alpha0, equally well as the older v3 with the classical LuxBall .

To say the RTX 3090 has impressive performance would be an understatement. In applications that are heavily GPU-bound, the fresh Ampere architecture in truth flexes its muscles, beating out even the Lenovo P920 and its dual RTX 8000s. In LuxMark v3, the performance was all over double, and the v4 benchmarks were approaching 150 % performance. Expect to see a lot more charts that look like this, but ( spoiler alert ) not everything is american samoa clear-cut as it seems .
next up, let ’ s take a look at V-Ray from Chaos Group. This benchmark is for the V-Ray 3D rendition and simulation toolset, which is available as a plugin for a wide range of 3D model applications, such as Cinema 4D, Maya, Rhino, and Unreal, among others. This benchmark focuses on CUDA and RTX performance specifically. We used the newest translation of the V-Ray benchmark ( v5 ) american samoa well as an older one ( v4 ), just for comparison .

The fib here is very similar to LuxMark with the RTX 3090 vastly outperforming the RTX 8000, and evening eeking out above the Lenovo P920 and its double GPUs. This benchmark specifically lets us see the massive improvements NVIDIA has made to their 2nd generation ray-tracing cores ; the CUDA operation is about doubled, while the RTX performance is about 150 % .
future up is the omnipresent Blender, an open-source 3D modeling application. This benchmark was run using the Blender Benchmark utility. NVIDIA OptiX was the choose hand over method acting, as opposed to CUDA, as all these systems can utilize RTX .

The fib is much the same once again, with the next-gen RTX cores making promptly work of the Blender renders, flush burning through the BMW render in under 10 seconds. This report continues, but don ’ triiodothyronine worry, it ’ mho going to get very concern soon. here we look at OctaneBench, a benchmarking utility program for OctaneRender, another 3D renderer with RTX support, similar to V-Ray .

Yeah. future, we ’ ll take a front at video editing applications, specifically Davinci Resolve Studio 16.2.8 and Adobe Premiere Pro 2020. We ’ ll be using PugetBench for Adobe CC and PugetBench for Resolve, both developed by PugetSystems, a personal computer manufacturer that makes professional workstations for particular applications. This graph specifically looks at the GPU score of the overall benchmark, scoring performance in GPU accelerated effects .

finally, we have some use cases where the value proposition drops a bite. Applications like Premiere Pro and Davinci Resolve vary wildly in their operation needs depending on the project and effects, and they ’ re much CPU bottlenecked. There is an improvement, particularly in Resolve which has much better GPU optimization, but users of the Adobe Creative Cloud suite will find more value in lower-end GeForce cards like the RTX 3080 where you ’ ll drive apparently 90 % of the performance for 46 % of the price ( based on MSRP of the Founder ’ s Edition cards ). While not reported here, performance in After Effects and Photoshop were similar, where most tests were CPU limited. besides deserving noting the Quadro and GeForce cards both use the 7th generation interpretation of NVENC ( the NVIDIA Encoder ), so H.264 and HEVC render times will be similar. Do keep in mind that the GeForce cards are limited to 3 coincident encoding tasks ( for things like streaming and recording ), where the Quadros have no such limit. We ’ re mindful of bypasses for this specify, but we ’ ll address that later. The RTX 3090 does use a new generation of the NVDEC ( NVIDIA Decoder ), so there will be bare improvements in timeline scrub when working with HEVC and H.264 footage .
Let ’ s move on from creative application and into something new. We ’ ll be taking a look at the car learning performance of these cards using a benchmark competently named AI-Benchmark. It ’ s an open-source Python library that runs a series of deep learn tests using the TensorFlow car learning library. You can find out more about it here deoxyadenosine monophosphate well as which specific testing methodologies it uses .

Your collective gasp have been heard. It ’ s the same report here as most of the creative benchmarks, with the GeForce card apparently doubling the performance of the Quadro. This examination was done in Windows 10, but you can expect similar results in your Linux distro of choice. At the clock of testing, TensorFlow hasn ’ t been updated to support the new Ampere cards, but through a bit of hackery, we were able to make it run by mixing and matching components of different CUDA developer kits. We expect that a properly updated interpretation in the future can merely improve results .
last, but decidedly not least, we ’ ll look at SPECviewperf 2020, the newest revision of the industry-standard benchmarking utility from the Standard Performance Evaluation Corporation. This benchmark looks at graphics performance in a assortment of professional applications. We ’ ve besides included the Lenovo P620 equipped with an older Quadro P6000 for comparison .

This is where things get interest. The 3D model results are about what you ’ five hundred have a bun in the oven, with the RTX 3090 outperforming the RTX 8000 by a goodly margin, but applications like CATIA and Creo, and the Energy and Medical tests, actually have the 3090 underachieve by a small, but significant margin. It even underperformed against the significantly older Quadro P6000 in our Lenovo P620. But what about Siemens NX ? This is the southern cross of the workstation versus consumer, Quadro versus GeForce, predicament. You ’ ra reading that chart correctly ; the 3090 scored a 25.37 in Siemens NX. We ran this test seven times and referenced other benchmarking results on the internet and this leave is absolutely correct. The RTX 3090 had less than 5 % of the operation of the Lenovo P620 with the RTX 8000 in this test. Let ’ s explore this more in the next incision .

Workstation versus Consumer GPUs; The Drivers

The GeForce RTX 3090 is a very brawny GPU, there ’ s no deny that, and the Quadro RTX 8000 is besides a powerful GPU, but where they differ is important. The RTX 3090 is foremost and foremost a bet on television card. It was designed, and is marketed, as the absolute best GPU for bet on and the “ earth ’ sulfur foremost 8K adequate to ” gaming calling card. The fact that it can outperform the RTX 8000 in a multitude of tasks, with extreme point margins in most cases, is a will to its sensitive power, basically beast forcing its way through these tests. The Quadro RTX 8000 is first and foremost a workstation card, designed to be placed in workstations, crammed in rack-mount servers, linked together over NVLink, virtualized, ran hard and put away moisture ( metaphorically, of course ) and outlive .

so that begs the question, why did the GeForce RTX 3090 underperform the 3-year old Quadro RTX 8000 in certain tasks, and even underperform the closely 6-year old Quadro P6000 ? specifically, why does the RTX 8000 go steady over an ordain of order of magnitude more operation in Siemens NX ? We ’ ra indisputable you, understanding subscriber, already know ; it comes down to driver optimizations. NVIDIA is estimate, jury, and executioner when it comes to what GPUs get workstation optimizations and it ’ second clear that they did not want the RTX 3090 to be able to absolutely wipe the floor of their former Quadro line in every application. A draw of potential buyers will be very defeated to see their hopes of top-tier workstation operation by day and high-end gambling by night international relations and security network ’ t possible, at least not at the RTX 3090 ’ s $ 1,499 MSRP. “ Well, that ’ s what the TITAN line is for, ” we hear you saying, but there is no TITAN RTX substitution available justly now, and the price of the 3090 surely positions it as a TITAN substitution. NVIDIA even boasts the 3090 as having “ TITAN class performance ” in the beginning pipeline of their marketing materials, but that Siemens NX resultant role is nowhere about “ TITAN class. ” Is the 3090 barely an overbuilt, overpriced 3080 ? Well, not quite. It ’ second still the best GPU available for 3D modelers and television work, specially with its 24GBs of VRAM that allows creatives to work with big models and 8K footage with comfort, but that puts it in a pretty modest niche, a recess previously served by the TITAN-class cards. If you don ’ metric ton exercise with CAD applications like Siemens NX, Creo, or CATIA, the RTX 3090 is still appealing, but it ’ s an absolute non-starter if those applications are critical to your work flow .
There ’ s other considerations, besides. While both the Quadro and the GeForce card extend multi-GPU setups over NVLink, and it ’ s deserving noting that merely the 3090 has this capability among the 3000-series cards, entirely the Quadro can be virtualized with SR-IOV. If your workload involves distributing one GPU, or two NVLinked GPUs, to multiple VMs, the Quadro is still your only choice. While there are ways around this, the luff however stands that this have is artificially limited by NVIDIA ’ s GeForce driver. The same driver which, by the direction, does offer certifications for some creative applications like the Adobe suite and Autodesk, arsenic well as WHQL ( Windows Hardware Quality Labs ) authentication in its Studio discrepancy, confounding the point far. While we anticipate a true Ampere-based TITAN substitute to come in the future from NVIDIA, the artificial limitations placed on the GeForce RTX 3090, including the specify to concurrent encodes with NVENC, feel out of touch with what master users are looking for correct now .
While writing this article, NVIDIA besides announced their new line of cryptocurrency mining particular GPUs and their purpose to half the mine performance of the approaching RTX 3060. This decrease in crypto mining performance is completely artificial and done entirely through drivers, further drive home the distributor point that we are simply at the mercifulness of whatever NVIDIA deems our cards worthy of doing. Will this “ unhackable ” terminus ad quem on the 3060 be exploited ? Without a doubt, much the direction the NVENC restrict has been beaten. But for now, NVIDIA holds all the cards and for a lot of professional workflows that are highly optimized for CUDA, RTX, TensorFlow, and other NVIDIA specific computing platforms, there ’ s simply no early option but NVIDIA .

Conclusion

Let ’ s break it down a bit. Should you keep your stock alerts on for the RTX 3090 ? Brave the lines outside your local MicroCenter for the gamble to get one ? As with most things the answer is, “ it depends. ”

• Do you work exclusively with large 3D models in applications like Blender and Cinema 4D or regularly edit large 6K and 8K video files? Then yes; the RTX 3090 is the best creative professional GPU on the market for CUDA and RTX accelerated workflows.
• Do you have enough disposable income for an absolutely no-compromises gaming rig? Also yes; the RTX 3090 is the best gaming GPU money can buy, particularly in games with ray-tracing and DLSS 2.0 support, a list that is growing every day.
• Do you work primarily in CAD, particularly in scientific fields with applications like Creo and CATIA? Then no; the RTX 3090 doesn’t provide the necessary driver optimizations that come with the Quadro line of cards, and no amount of brute forced raw power can overcome that. We’re working to get our hands on the new RTX A6000, the RTX 8000s actual replacement, to help inform folks like you.
• Do you work in the machine learning and AI field? This one’s a mixed bag, because it seems like a “yes” as the performance in TensorFlow is excellent, but there are unfortunately not enough standardized tests or data points to make a definitive conclusion versus a newer Quadro card. Also, as of this writing, TensorFlow hasn’t been updated to properly support Ampere cards. Our current answer is a soft yes. We may look at MLCommons and MLperf in the future as a follow-up, specifically for AI and machine learning use cases.
• Do you currently own a TITAN RTX and are looking to upgrade? We recommend you wait for an official announcement from NVIDIA on the Ampere-based TITAN, which is rumored to be coming eventually. If your use cases make good use of the TITAN, the RTX 3090 may end up a downgrade, depending on your workflow.
• Do you do a lot of GPU virtualization? This is a solid no; the RTX 3090 does not natively support SR-IOV.

With all that considered, we hope this can inform your decision to upgrade or not, and whether you go GeForce or Quadro. If the $ 1,499 MSRP of the RTX 3090 is excessively rich for your lineage and you don ’ t need more than 10GB of VRAM, the RTX 3080 is a bang-up choice deoxyadenosine monophosphate well. While not explicitly tested here, we found our results were by and large 20 % -30 % better than the RTX 3080 based on results throughout the web, so you can expect it to outperform ( and underperform ) the Quadro RTX 8000 in similar applications. It may besides be worth waiting for the rumored RTX 3080 Ti, whenever that may be coming, for a middle grate. For nowadays, we will keep our eyes peeled for the Ampere-based TITAN to truly fill the opening between a full workstation tease and a wax gambling card. We will besides be keeping an eye on NVIDIA ’ s driver pipeline to see if they assuage any of the wag ’ s work-related shortcomings in future releases, but we have our doubts. We would love to hear your thoughts on this, and besides if there were any benchmarks or use-cases we missed that you ’ d like to see, particularly as it pertains to AI and car memorize. Get in reach with us on our social channels and stay tuned for more creative professional and workstation reviews .
Read More – NVIDIA RTX A6000 Review
NVIDIA Quadro RTX 8000 on Amazon
NVIDIA GeForce RTX 3090 Founder ’ randomness Edition on Amazon
Engage with StorageReview

Newsletter | YouTube | Podcast iTunes / Spotify | Instagram | Twitter | Facebook | RSS Feed