Just to put things in perspective, the GTX 1080 launched May 27th, 2016, that was the last consumer launch of a new series of cards and microarchitecture from Nvidia. That is nearly two and a half years ago. In that time the US elected a new president, we saw a total eclipse over the US, the Cubs broke their 108-year long curse and won the world series, we have even been through TWO Olympics, and not to mention the whole crypto craze! Even crazier, the PC market has completely changed as well, The Intel i7-6700K was the top dog in the consumer market with its 4 cores and the bleeding edge was the recently introduced i7-6950X with its 10 cores. Now 6 and 8 core CPUs are the norm in the mainstream lineups and AMD has the 2990WX with 32 cores on the high end! So even with the 1080Ti and later the 1070Ti, we are long overdue for big graphics launches and Nvidia is going all out. Today both the RTX 2080 Ti and RTX 2080 launch, I’m going to check out the 2080 Ti first and dive into the new Turing architecture, the use for RTX over GTX, the completely new cooler, and of course performance so hang on we have a lot to get through and a short time to get there (RIP Bandit).

Product Name: Nvidia RTX 2080 Ti Founders Edition

Review Sample Provided by: Nvidia

Written by: Wes Compton

Pictures by: Wes Compton

Amazon Affiliate Link: HERE

 

Turing, RTX, Ray Tracing, what is all of this stuff?

So if you haven’t been following things the last month or so you might have missed Nvidia’s announcement of the new cards and with that one of the most glaring changes. In fact, if that is the case you might even think I had a typo in the title of this article. Typos do happen, especially for me. But no the RTX before the 2080 Ti is real, they did change from GTX to RTX. The reason for this was with the new architecture Nvidia has been focusing on real-time Ray Tracing, the holy grail of gaming. For years now (it goes back to the 70’s even) ray tracing has been used in special effects rending to get the right lighting effects. But up until now, video games could only use multiple effects to try to fake the same look. Ray tracing is at its core, rendering out what each ray of light does including bouncing off of reflective surfaces. This takes a lot of power to figure out and in some cases pixels on the screen might be hit by multiple light rays, compounding the issue and the work needed to do that.

So how do the new cards handle Ray Tracing? Well, first I need to mention the architecture itself. We are finally moving away from Pascal, the architecture used in the 10 series cards. The new architecture is called Turing, a nice nod to Alan Turing who played huge roles in theoretical computer science and an early part in AI. Nvidia built-in ray tracing all the way down into each of their SM’s. That is their Streaming Multiprocessors, each GPU has a specific number of SMs, this determines the GPU size in addition to things like the L2 cache, memory controllers, and their GigaThread Engine. Below I have a picture of a Turing SM and you can see there is one RT core or ray tracing core for each SM. Along with that, each SM has four INT32 cores, four FP32 cores, and four Tensor cores.

nvidia 4

nvidia 5

The INT32 and FP32 cores are your normal shaders, but the Tensor Cores are new and interesting. These are focused on deep learning and to go with them Nvidia has also developed DLSS (Deep Leaning Super-Sampling). These “leverage a neural network to extract multidimensional features of the rendered scene and intelligently combine details from multiple frames to construct a high-quality final image”. This allows them to use half of the samples when rendering and to use the deep learning AI to fill in details, giving you similar quality at a much better performance.

They have also switched over from GDDR5X to the new GDDR6 and that memory is running at 14 Gbps, significantly higher than the 11Gbps of the 1080 Ti and later the upgraded 1080 as well. So what is interesting about this new architecture is it is going in a different direction from the last few from Nvidia. In the past, they have focused on cutting out what isn’t needed for gaming like the asynchronous compute performance. An area that AMD has dominated. But with DX12 taking advantage of asynchronous compute, we now have the INT32 and FP32 cores back together. This new direction has lead to an architecture that isn’t focused on just one thing, opening up the potential for big jumps in performance as game developers take advantage. You see, in the past everything basically ran through the shaders, but now with the Tensor Cores and the RT Cores games can actually run their AI independently of everything else on the Tensor Cores and have their real-time ray tracing as well with the RT cores. The image below shows what I mean.

nvidia 1

Okay, what about the new cards? Well, there are currently two cards out today, with the RTX 2070 coming out in October. What is interesting though is at least currently they don’t replace the previous models, in fact, there isn’t any indication yet that the GTX lineup is going away. Both the RTX 2080 Ti and the RTX 2080 are priced above the GTX 1080 Ti. Is Nvidia pushing to create a new level of high end? Maybe, I’m not really sure. But let's look at the specs. When just looking at CUDA cores we can see that the 2080 Ti is with its TU102 GPU is much larger than the 1080 Ti, same with the number of texture units though they have the same number of ROPs. Clock speeds for the 1080 Ti range from the core speed of 1350MHz and up to the boost clock of 1635Mhz, this is both lower and higher than the 1080 Ti. The RTX 2080, like the 10 series cards is clocked higher than the Ti model but the Ti makes up for the lower clock speeds with significantly more CUDA cores. The 2080 Ti also has more VRAM and a wider pipe with its 352-bit memory bus over the 256-bit memory bus of the 2080. This is similar to the 1080 Ti and the 1080, in fact, they both have the same amount of memory as well (11GB and 8GB).

The Turing GPUs are now made at 12nm using the same FinFET process as Pascal. But even with that smaller process, the die sizes are crazy, the TU102 is 754 mm squared, this is the second largest GPU die ever made and over twice the size of the GTX 1080. The 18.6B transistors take up a little space. The RTX 2080 is a low lower with 13.6B transistors but still above the 12B of the GTX 1080 Ti and nearly double the 7.2B of the original GTX 1080.

specs

All of that does need a lot of power to push, even with the smaller manufacturing process. The RTX 2080 comes in at 260 watts, this is higher than Nvidia’s single cards in the past. The GTX 690, their dual GPU card did have a TDP of 300. The RTX 2080 is much better with a 225-watt TDP. The price is the other big factor with these cards. As I mentioned they don’t really push the 10 Series cards out. The RTX 2080 Founders Edition comes in at $799, Nvidia says aftermarket cards may come in at $699, but as we have seen in the past normally having the Founders Edition or reference cards priced higher lead to other cards matching or going higher. The GTX 1080 Ti launched at $699 and that is where the price is as well, pricing is only now calming down from the crypto craze. The RTX 2080 Ti is even higher with its Founders Edition pricing set to $1199, blasting past the 1k price point. The starting price for aftermarket 2080 Ti’s is $999, but again we will have to wait to see if cards come in at those prices.

Also announced was the RTX NVLink, a replacement for the SLI bridge. With the HB SLI bridges just coming out with the 10 series cards it is disappointing to see them go away in favor of the NVLink. The NVLink isn’t really new, Nvidia has been using it on their server cards for a while now. But there are a few things to note about it. For starters, it will run you $79, so it is more expensive than the HB SLI bridges. It will also only work with the RTX 2080 and RTX 2080 Ti, the also announced RTX 2070 does not support it. So in addition to seeing multi-card configurations die for the x60 cards a while back, it has now also gone away for x70 cards. I’ve also heard that they will not be included with motherboards like HB SLI bridges have been. Aftermarket versions will be available, but they will be rebranded versions of the Nvidia bridge pictured below and at the same price.

nvidia 2

The new bridges do offer a LOT more bandwidth. Nvidia included this slide in their presentation showing the difference. Sadly there wasn’t any scale on the graph. Diving into the whitepaper shows that the numbers are in GB/second bidirectional bandwidth. Allowing for 4k144Hz and 8K surround bandwidths. Each link offers 25 GB/s each direction, the TU102 has two and the TU104 has one. All of this means you can now hook up more monitors to different GPUs and there is bandwidth to send the frames over to the video card.

nvidia 3

Nvidia has also brought on board new overclocking features. For starters, GPU Boost is now in its 4.0 variation. The biggest change here is now the profiles are user editable, allowing you to push the limits more as needed. But the biggest introduction in this area is Nvidia Scanner, a built-in overclocking option that spends up to 20 minutes testing and retesting to put together the ideal profile. Auto overclocking isn’t really new, but this is an improvement on what has been done in the past. It goes as far as testing different voltage points and then interpolating between them and testing over and over with a mathematical algorithm that checks for failures while testing to find the limits without hard crashes.

The new cards have also made big changes in their overall look as well. I’ll include these teaser pictures, but I will dive into the cooler in the card layout section.

image 1

image 2

image 3

 


Packaging

So the packaging for the RTX 2080 Ti managed to be completely different and nearly the same all at the same time. How is that possible you ask? Well, the shape of the box is exactly the same. It pulls apart by pulling the top off as well. But the printing on the box has a new angled line design all over the box with the GeForce RTX logo in the bottom left corner as well as along the side just like before with the GTX logo.

image 6

image 7

Inside of the box, the card itself came with a transparent plastic covering it. This was a big step up over the tinted static bags of the past when it comes to letting you see your nice new expensive video card right away. The downside, of course, is if you ever want to sell it later and you save the box, you won’t have a static bag to keep it safe. With the metal backplate and everything, I’m not sure how needed it is anymore anyhow though. The card sits in a foam tray cut to hold it perfectly and there is more foam up in the top to help hold the top as well. There is then a small thin box for all of the documentation and accessories tucked away behind the card into the foam, just like in the past. Inside you get a support guide and a quick start guide. The quick start guide has the 2080 Ti branding on it but the support guide is more generic. Then you also get a DisplayPort to DVI adapter. With the 10 series Founders, Editions Nvidia dropped the DVI port for more airflow and obviously, it's not coming back for the 20 Series!

image 8

image 9

image 10

 


Card Layout and Photos

So I always do a card rundown, but in the past, the Founders Edition and reference cards have been mostly the same with just the exterior design changing from time to time. With the 700 series, Nvidia introduced the all-metal construction that they first used with the dual GPU GTX 690 to their high-end reference cards. That same design stayed with the GTX 980 and GTX 970 only with the letters being filled in and later the all black Titan Black design. Now the 10 series of cards did bring in a new design with its tessellated shape but it was still basically the same design only with a larger heatsink. The GTX 1060 Founders Edition brought the metal shroud to the cheaper cards as well. But now with the 20 Series of cards, Nvidia has completely changed things up, dropping the blower design altogether for a more aftermarket design with dual fans.

The new design does still feature an all metal construction, and that alone is one of the main reasons I have been loving the reference and Founders Editions for a while now. They feel solid in your hands, nothing flexes and really they look like they were designed with the PCB where a lot of the aftermarket designs look like they just bolted the same cooler on another GPU and sometimes they do that. The new design is actually heavier than the GTX 1080 was and this might be the best looking card yet in my opinion. I love the clean look, it does remind me of a few of the older EVGA designs only with a metal shroud and it also reminds me a little of my all time favorite card design, the HD7970 from XFX.  The RTX 2080 Ti branding is in chrome just like before, but it is now in between the two fans, not at the end.

image 11

image 12

The back of the RTX 2080 Ti FE also has the RTX 2080 Ti branding and I like this decision. The 10 Series had this as well but you couldn’t see it at all and frankly this is the most visible part of the video card when installing in most cases. I’m a little unsure of the silver used though, but it does help the black branding stand out. These aren’t going to blend in with the dark all black builds like the past designs did with the black backplate. But pairing this up with one of the silver MSI motherboards would look killer.

image 13

So the two fans on the front of the card are more than just doubling up the fans. The old design had a thick fan that didn’t blow down, it would pull air in and blow it across the card. This new design has two fans blowing down through the heatsink, over the vapor chamber that covers the entire card and through a few small holes in that to get to the PCB itself. This design means that no longer is the Founders Edition sealed at the top and bottom. So from the bottom, you can see the heatsink that runs the length of the card. The end is now sealed and still has mounting points for case or server brackets. Then the top is open as well. So most of the air from the fans will come out the top and bottom, with a little making it out the rear I/O. This also means all of the heat from this card will end up inside your case, so make sure your case has the airflow to deal with that. The top and the bottom of the card do have the long V-shaped design that the other generations had, only now it connects the metal around both fans.

image 15

image 16

image 17

So I mentioned that some airflow will go out of the rear I/O, but as you can see there won’t be too much. For dropping the DVI connection they still didn’t really fill the space with much other than including the Nvidia logo now. The PCI bracket is now fully blacked out, not a black chrome and it looks awesome! Black makes sense it should better blend in with most cases now. For connections, you get three full-sized DisplayPort 1.4a with a max resolution of 8k at 60Hz. There is an HDMI connection but the whitepaper didn’t mention what generation. Then there is a Type-C connection. That is the one that should stand out the most to people, we finally have video cards with Type-C output! The Type-C connection also supports VirtualLink, a new open standard for VR headsets that can allow power, display, and data all through one cable. This is huge because currently you need all three connections and a hub to get everything to work, cutting it down to one plug could make VR more usable and portable.

image 14

This is a better look at the V-shaped design that I was talking about before that carried over from the past cards. Then there is also the logo up on top. It is a little different this time with it being Geforce RTX, not GTX. It is also not green with the lighting off anymore, they went with a clear transparent plastic not the semi-transparent rubber with green on top. Sadly at least currently the logo only lights up in green, I have heard that there are RGB lights under the logo though and a controller built in. I don’t know if Nvidia did this to get the perfect green or if they are leaving the option open for in the future to change the logo to other colors. I hope RGB is coming, I’ve been requesting it for the past few generations. Not because I don’t like the green, it does look good. But matching the logo to your build rather than just turning it off would be nice!

image 18

image 19

So with that 260-watt TDP dual power plugs wasn’t a surprise. The RTX 2080 Ti FE uses two 8-pin power connections where the GTX 1080 Ti only required an 8-pin and a 6-pin. The plugs are flipped around the traditional way, not the notched PCB way that aftermarket cards have been doing to get even more heatsink in around the plugs. In fact, you can see that the heatsink baseplate goes around them, leaving a nice gap.

image 20

So the new fans are a pushdown design, I already mentioned that. So they look different than in the past, but Nvidia did keep those metal center caps to keep things feeling like home. The fans are also much larger and with 13 blades they are a lot denser than some of the aftermarket card fans. They remind me a lot of the Asus Strix cooler fans though.

image 21

So then we have the new x8 Nvlink that replaced the HB SLI bridge. You can see that the pin density is significantly higher and the connection is a lot wider as well. I love that for the Founders Edition Nvidia did also include a matching metal cover that keeps the clean look when you aren’t running multiple cards.

image 22

image 23

I grabbed our GTX 1080 Ti for some comparison shots as well and you can see just how different the new cooler design is. They are the same length but I did notice the RTX 2080 Ti is a mm or two taller when they were next to each other.

image 24

image 25

 


Our Test Rig and Procedures

So before getting into testing, I did want to address our test platform. With the gap between the last card reviews, I decided to go ahead and upgrade our test platform from X99 to X299. I debated on Ryzen and Z370 as well and both would have been great as well. But this seemed like a good use for our i9-7960X and the Asus ROG Rampage VI Apex is a killer board that was going way underutilized for how good it was sitting in our storage racks. To go with that Kingston upgraded us to their A100 M.2 SSD for those NVMe speeds and the capacity needed to store all of our benchmarks and their 3000 MHz Predator DDR4 for memory. Everything then went in the same Primochill Wetbench as previously used, with the same Corsair power supply and Noctua cooler because all of those have been very stable.

image 4

image 5

Our Test Rig Configuration (with affiliate links)

CPU - Intel i9-7960X

Motherboard - Asus ROG Rampage VI Apex

Memory - HyperX DDR4 Predator 3000MHz

Storage - Kingston A1000 960GB M.2 SSD

Cooling - Noctua NH-U12S

Power Supply - Corsair AX1200w

Case - Primochill Wetbench

OS - Windows 10 Pro 64-bit

tests

 


Synthetic Benchmarks

To start off my testing I always like to take a look at a few different synthetic benchmarks. These help give us a first look at overall performance. I also find that they are some of the best tests at comparing card to card. They don’t really translate to a specific FPS for in game performance, but I have those tests later. So for performance tests, I really like the 3DMark tests. They used to be branded under the Futuremark name but they are now from UL. But in addition to that, I take a look at the performance in Unigine Superposition. I like that benchmark because it is based on an engine that is used for other games and the test has a wide range of resolution and detail tests.

Starting with the 3DMark testing. I tested using the three available Fire Strike benchmarks. All three are DX11 based and they focus on 1080p for performance, 1440p for the extreme test, and then 4k for the ultra test. In all three tests, the RTX 2080 Ti really stood out against the previous generation cards. I took a look through and percentage wise the RTX 2080 Ti went up 24.3% in the performance test over the GTZX 1080 Ti, 21.1% in the extreme test, and 20.1% in the ultra test. That is a nice jump, but in the Time Spy and Time Spy Extreme tests that are DX12 focused the increase was a LOT better. Time spy jumped 46.7% and time spy extreme 48.2%. It is clear that while the number of Cuda cores went up, adding that asynchronous compute back into the picture is really making a difference in DX12 titles. This is where AMD has normally outperformed Nvidia and with DX12 being the future it is good to see Nvidia turn their focus to it.

graph1

graph2

graph3

graph4

graph5

Now the Unigine Superposition results show right away that this is a DX11 title, it actually surprising that Unigine didn’t go with DX12 or Vulkan support for their newest test. Overall the performance jump from the GTX 1080 Ti to the RTX 2080 TI FE is good, but not out of this world good. That of course still puts the 2080 Ti at the top of the charts in every one of the tests.

graph6

 


VR Benchmarks

So VR benchmarks are still a little rare, but as someone who enjoys playing VR games, I couldn’t just skip them entirely. I did test and include the basic SteamVR benchmark that is already looking extremely dated. That test basically just gives you a good, bad, and okay result and with all of my testing being in the higher end none of the cards would be considered not good for VR. In fact, most of the cards all tested with the 11 that the test maxes out at.

graph7

Moving on, the main tested I used was the UL VRMark test (formally Futuremark) and I ran all three benchmarks. There are a few things to keep in mind here. First, each test has a goal FPS that they recommend. The orange and blue rooms are 109 FPS and the new Cyan test is 88.90. They have a target on this test where other UL tests don’t because VR headsets require smooth playback or it will make you sick. I skip the overall score and prefer the pure FPS results. So the orange room test is a current day performance focused test, similar to SteamVR. Cyan is a DX12 test that is more demanding than the orange room test but not as demanding as the future looking blue room test. So the results for the 2080 Ti ended up being interesting. It dominated in the DX12 cyan test and again in the future looking blue room. In fact, this is the closest I’ve seen a result get to that 1090 FPS target on that test. The future is almost here! But the orange room test, well the 2080 Ti flopped here for some reason. It performed as well as needed for good performance, but below some of the others including the 1080 Ti. This could be a driver bug but it was weird that it was only for that one result.

graph8

 


In-Game Benchmarks

Now we finally get into the in game performance and that is the main reason people pick up a new video card. To test things out I ran through our new benchmark suite that tests 11 games at three different resolutions (1080p, 1440p, and 4k). I also slipped in a few variations on the same games for comparisons like DX11 to DX12, OpenGL to Vulkan, and a couple of games are just tested at their highest setting and lower but still high detail options to show the performance difference when things are turned down slightly. In total, each video card is tested 51 times and that makes for a huge mess of graphs when you put them all together. To help with that I like to start off with these overall playability graphs that take all of the results and give an easier to read the result. I have one for each of the three resolutions and each is broken up into four FPS ranges. Under 30 FPS is considered unplayable, over 30 is playable but not ideal, over 60 is the sweet spot, and then over 120 FPS is for high refresh rate monitors. This covers all of the games tested except Final Fantasy XV that we have a score rather than a FPS, because they like to be different.

So how did the new RTX 2080 Ti perform? Well, let's just say you aren’t going to have trouble playing any games anytime soon. When looking at the 1080p results everything was above 60 FPS, but a majority were in the 120 FPS or higher range. Really this doesn’t show the whole picture, most of the 60+ results were actually in the 100-119 range. 1440p was the same, everything in the sweet spot of 60 FPS or higher with a lot in the 120+ FPS range. Then there is 4k, this is the range I was most curious about. Can we finally take advantage of the higher refresh 4k monitors? Well yes, 4k60 is basically where this card shines. There were a few results just a hair under 60 FPS (Deus Ex, Wildlands, and Shadow of the Tomb Raider). SotTR came in at 58, and all three only came in less than 60 FPS on their highest possible detail settings. Running high rather than ultra is about all you need to keep these averaging above 60 FPS as well.

graph27

graph28

graph29

Of course, I did include all of our normal graphs as well. I wouldn’t just leave you with our roundup numbers. I just think the 51 results per card across 6 cards takes a crazy man or woman to dig through (like me). So does anything stand out, good or bad? Well for starters, the RTX 2080 Ti FE tops the charts in every single game. There were a couple results like Hitman DX12 at 1080p where the 1080 Ti was faster, but the 2080 Ti did really well. Especially in the 4k results. Most of the games are DX11 and like in 3DMark the 2080 Ti is ahead but not in proportion to the increase in price.

Before all of the other charts, I did want to touch on the Final Fantasy tests at the top here. The test I ran all of the cards through showed about what you would expect. In line with all of the other games. But I also went back and tested the RTX 2080 Ti with a test that Nvidia sent out. I normally don’t do this, but right not it is the only look at DLSS. The test was set up entirely by them with bin files to run so I don’t even know what the TAA or DLSS settings are, so keep that in mind. They set it up to show the same level of detail and I can confirm that they did look the same when watching them run. The end result shows the performance gap you could potentially see by using DLSS that uses the new deep learning tensor cores. The overall jump was 39.4%, I will also be testing this on the RTX 2080 as well, but I am not including this in our benchmark suite. I may make our own test later when the FFXV test is updated to support DLSS.

graph26

graph25

Here are the rest of the results, enjoy!

graph9

graph10

graph11

graph12

graph13

graph14

graph15

graph16

graph17

graph18

graph19

graph20

graph21

graph22

graph23

graph24

 


Compute Benchmarks

So I expanded out our compute testing to better match our recently expanded CPU test suite. A lot of people, gamers included, use their PCs for video production, mining, and rendering. So it is important to look at every aspect. Sadly I ran into issues with a few of our tests with the RTX 2080 Ti, but rather than not include them I want to address them. We can update them later when they are fixed, but it is important for people to know if/when they might have issues.

So the first test I ran was the GPU Compute benchmark in Passmark’s Performance Test 9. I like this because they test in a few different ways and give you an overall score. The 2080 Ti did top the chart here, but I was a little surprised it was so close to the 1080 Ti.

graph30

Now the V-Ray benchmark showed what the RTX 2080 Ti is capable of. This might not be a surprise given the ray tracing focus of this card, but I don’t think V-Ray actually takes advantage of the new RT cores. This is just from raw performance. I was surprised to see the Vega 64 struggle in this test though.

graph31

Next, we have Cinebench using the OpenGL test, well frankly I thought this one was going to get tossed out. All of the Nvidia cards were right in the same range and the results didn’t really make much sense. I assumed we were CPU limited but testing the Vega 64 said differently.

graph32

Cinescore, lol well this one ended up being a mess. I really like this benchmark because they test encoding performance across every single format you can think of and then do it at every resolution all the way up to 8k. That is because Cinegy does really studio production software for the big boys, not just what people use at home for YouTube or Twitch. Sadly though the RTX 2080 Ti results look a little low, that is because a few of the tests would just fail and that dropped the overall score. Even worse, the Vega 64 didn’t complete the test at all because it failed every test in an Nvidia specific format.

cinescoreproblem

graph33

Then last but not least we have Blender, this was the test I was most excited about. Blender is used a lot and people use their CPU and or their GPU so it is good at looking across the platforms to see what works best. Sadly the RTX 2080 Ti crashed on this one as well. I spoke to Nvidia about this one and I’m told that Blender needs to be updated to CUDA 10 to fix this issue, so we will just have to wait for that before seeing numbers. Currently though the Vega 64 complete curb stomped the 10 Series Nvidia cards. Having asynchronous compute is important here and we know the RTX 2080 Ti does well with asynchronous compute, so it will be interesting to see how this turns out later.

graph34

 


Cooling, Noise, and Power

Performance in games or your work compute tasks are always the priority, but there are a few other aspects of video cards that are a big deal as well depending on your PC and situation. Those issues are cooling performance, overall noise, and power usage. Not everyone is going to care about all three, but I would lay money that only the most die-hard fans would put up with all three being extremely bad. So I always make sure to take a look at all of those other aspects as well. This becomes especially important later when I start taking a look at aftermarket cards as these areas are sometimes the main differences between different card designs and different companies.

Anyhow, I start off with power usage, really all of these revolve around it. Power usage turns into heat and more heat means bigger and faster fans. So going in I knew the RTX 2080 Ti FE was going to pull some power, Nvidia has it listed with a 260-watt TDP and while company to company they calculate that differently, this is the highest single card TDP I’ve seen from them. So to test power usage I got out our Kill-A-Watt and hooked it up in line between the test bench power and the wall and did a few tests. Now keep in mind these results are the entire system, not just the video card. But they do give a good look at expected usage and help you make sure you get a larger enough power supply. Depending on your motherboard and CPU (and any accessory in your case) you could use less or more, not to mention if you stressed your CPU along with the GPU you could see higher numbers as well.

Anyhow, I did two tests, one with the combined test in 3DMark Fire Strike, this gives us a look at every day in game power usage. Then my second test was using the AIDA64 stress test to load the GPU only. As you can see in the Fire Strike test the RTX 2080 Ti FE is up there, but both the 1080 Ti FE and the Vega 64 pulled more. This is partially due to the design of the Turing architecture, you can expect more power usage in games that are designed to use the RT and Tensor Cores as they go unused in tests like this. But even still, the card did pull a good amount of wattage. The Vega 64 liquid cooled card still pulls away, especially in the AIDA64 stress test where it pulled 475 at peak where the RTX 2080 Ti FE was at 373. The 2080 Ti looks high when compared to the Pascal cards, but Vega helps put things in perspective.

graph35

graph36

Moving on to noise I have two graphs below. One takes a look at the noise levels of the fans from 18 inches away with the fan speed set to 100% and again 50%. I do it at those speeds because “under load” fan speeds can vary depending on room temperature and the game. The new dual fan design wasn’t what I would consider to be quiet, but it was in line with the GTX 1080 FE and quieter than the GTX 1080 Ti with its higher fan speed. I think this is an area where aftermarket cards will really be able to do better. I also have a graph showing the 100% fan speed on each card tested, this helps put the noise levels into perspective as well.

graph37

graph38

So with power usage and noise out of the way, that just leaves cooling right? This is the biggest one, you can deal with power usage with a larger power supply and noise can be subjective. But if everything overheats then none of it matters. For cooling testing, I tested all of the cards twice. I put them under load using AIDA64’s stress test again, this time tracking the temperatures. I ran the load until the temperatures leveled off. I did this with the stock fan settings and again with the fans turned all the way up. This way I can see what you will get out of the box and then the best possible performance. The delta between the two gives us a look at potential for overclocking as well. So with the stock fan settings, even with the large die and higher power usage, the RTX 2080 Ti FE ran at 73 degrees. This I’m sure is a combination of the new dual fan cooler and lowered fan profiles due to the new cooler running better. The 100% fan speed results were similar, the RTX 2080 Ti FE was well below both 1080 FE options and not far above some of the aftermarket cards.

graph39

graph40

I also snapped a few thermal images while doing the stock fan profile tests just to see how warm things were and to see if there were any hot spots. No big surprise that the top edge with the exposed heatsink and one of the two areas where hot air can blow out would be hot. The fan side of the card, in general, was much cooler and the hottest spot was on the backplate

other 2

other 3

 


Overall and Final Verdict

So the gap between the last architecture launch and Turing has made me and everything else hungry. It is always exciting seeing the tech move forward and the CPU improvements this year and last year have really set the bar high, people want big improvements. That is why leading up to this launch there was backlash after backlash. Situations like the Toms article recommending to buy the 2080 Ti even before anyone had tested the card. Well, that isn’t a concern now right? Benchmarks are out. By this time (because I’ve been running late) you have already read multiple reviews and hopefully, you have at least peaked at the 78 different tests run across six different cards in our performance sections.

If you haven’t looked at the numbers yet, what you would find is that the RTX 2080 Ti is without a doubt the best performing card out there. In our synthetic tests and in-game benchmarks, it was around 20% faster than the GTX 1080 Ti in DX11. Getting into DX12 showed much bigger jumps in performance, at least in the synthetic tests with the gap being around 46/48%. With Nvidia bringing back asynchronous compute performance it really made a big difference. That is especially promising because DX12 is the future, like I’ve been saying in our previous AMD reviews where they have been focusing on DX12 as well. Overall in our gaming tests, the RTX 2080 Ti seems to be the sweet spot if you are looking to game at 4k. You can finally run 4k60Hz without any worries, not to mention 100/140Hz at 1440p being very doable as well.

The new Founders Edition design ended up looking amazing, I think this might be in my top two or three for cards. Part of the reason I like it so much is the all-metal construction, just like previous Founders Edition cards. It is a tank and is very heavy. But you know right away by holding the card that it is well constructed. The only downside to the design was them not giving the cards RGB right away. The clear logo has RGB lighting behind it from what Gamers Nexus has reported so maybe this will get fixed in the future. I do like the current clear logo better than the old green design, so if you do turn the lighting off it will blend in a little better. Speaking of, here is a look at the card lit up.

other 1

Cooling performance seems to be a big improvement over the past cards, but you do give up the blower design that was nice for SFF builds. They managed to keep that huge die with a 260-watt TDP cool in my testing. But there is room for improvement when it comes to noise from aftermarket cards. It will be interesting to see how they perform.

Now one of the biggest selling points for the new RTX lineup is the Ray Tracing performance as well as the addition of Tensor cores that allow AI and Deep Learning to be used in games and to help increase performance in the case of DLSS. Sadly though neither is really available yet. It is a chicken or the eggs situation right now so I am happy that Nvidia is pushing things along with the cards. Hopefully, the software will follow along. They do have a good-sized list of games that will be bringing support for ray tracing and a few are working on DLSS as well like Final Fantasy XV. But for the person buying today, you are spending a lot of extra money for performance that isn’t available yet.

The big issue with the RTX 2080 Ti is with its price. There are a lot of people upset because both of the new cards are priced up above the current lineup. This basically leaves room for the GTX 10 Series cards to stay right where they are price wise. It doesn’t help that there isn’t as much competition helping push those prices down. Even though the Vega 64 that I retested today is performing close to the 10 Series Nvidia cards, they are still harder to find on some of the models (though it is improving quickly). But the biggest kick in the pants of all is when you compare the performance increase with the price increase. Nvidia has never really been about value at the top end of their product line, and I don’t blame them for that. But at $1199, the RTX 208 Ti Founders Edition card costs as much as a good computer. I think you might even be able to build a GTX 1080 Ti-based system for about the same amount and currently in a lot of the games out today you see a 20% performance jump between the two cards.

Now anyone who isn’t worried about money isn’t going to have an issue with this card as an upgrade. When you ignore prices the RTX 2080 Ti is a monster. This is the fastest card out on the market. I also think you are going to see performance actually increase on this one for a long time. Historically asynchronous compute performance has helped with the longevity of cards performance wise and with it being key in DX12 and more DX12 games coming out it will help even more. Not to mention when games start taking advantage of ray tracing and DLSS. Those are two areas of the RTX 2080 Ti that aren’t even being used right now.

So should you buy one? Honestly, for most people, I think they will be waiting to see how the RTX 2070 performs or maybe even looking at GTX 1080 Ti’s on the market. But if you want that butter smooth 4k performance this is the only way to get it. Or if you want to get in on the Ray Tracing train early, again right now the 2080 and 2080 Ti are the only two ways to get that. You are spending more now, but with those benefits hopefully coming soon it might be more of an investment in the future.

fv5editorschoice

Live Pricing: HERE

Author Bio
garfi3ld
Author: garfi3ldWebsite: http://lanoc.org
Editor-in-chief
You might call him obsessed or just a hardcore geek. Wes's obsession with gaming hardware and gadgets isn't anything new, he could be found taking things apart even as a child. When not poking around in PC's he can be found playing League of Legends, Awesomenauts, or Civilization 5 or watching a wide variety of TV shows and Movies. A car guy at heart, the same things that draw him into tweaking cars apply when building good looking fast computers. If you are interested in writing for Wes here at LanOC you can reach out to him directly using our contact form.

Log in to comment

We have 951 guests and no members online

supportus