Nvidia GeForce RTX 2060 Review: Is Mainstream Ray Tracing For Real?

Controversy surrounding Nvidia’s Turing graphics architecture seems split into two camps. On one hand, critics are generally dissatisfied about the way GeForce RTX 20-series cards are priced compared to previous-gen boards. Or, they’re unhappy Nvidia committed so many resources to future-looking features that may not have a big visual impact or run smoothly, except on ultra-high-end models.

GeForce RTX 2060 sets forth to change minds on both fronts. At $350/£330, it lands below the GeForce GTX 1070’s launch price. And yet, the geometric mean of its average frame rates shows the GeForce RTX 2060 to be faster than GeForce GTX 1070 Ti! What’s more, recent updates to Battlefield V make Nvidia’s hybrid rasterization/real-time ray tracing approach playable on lower-end hardware. We averaged 68 frames per second (FPS) on the 2060 through our benchmark at 1920x1080 with graphics options, including DXR Reflection Quality, set to Ultra.

Meet TU106, Again: Impressive, Even After a Haircut

We were first introduced to the Turing-based TU106 processor when Nvidia launched its GeForce RTX 2070. The Founders Edition board was a bit faster than GeForce GTX 1080, but seemed overpriced at $600.

The version of TU106 used on RTX 2060 boasts 83 percent of the 2070’s CUDA cores, Tensor cores, and RT cores; 75 percent of its ROPs, memory bus, GDDR6, and L2 cache; and operates at 98 percent of its GPU Boost clock rate. Yet, at $350/£330, 2060 sells for roughly 60 perce of the RTX 2070 Founders Edition’s price. By the numbers, this card’s value story is bound to be a lot better.

Architecturally, TU106 is very similar to GeForce RTX 2080’s TU104 and GeForce RTX 2080 Ti’s TU102 processors. Nvidia simply builds the chip using fewer of the modular blocks than what go into its higher-end graphics cards. TU106 is composed of three Graphics Processing Clusters, each with six Texture Processing Clusters. Across the Turing family, TPCs include two Streaming Multiprocessors, adding up to 36 SMs in an untrimmed TU106. Nvidia disables six SMs for RTX 2060, so the card ends up with 1920 CUDA cores, 240 Tensor cores, 30 RT cores and 120 texture units. Those Tensor and RT cores are this generation’s signature features. So, if you missed our deep-dive into their purpose, be sure to check out Nvidia’s Turing Architecture Explored: Inside the GeForce RTX 2080.


GeForce RTX 2060 FE
GeForce RTX 2070 FE
GeForce GTX 1060 FE
GeForce GTX 1070 FE
Architecture (GPU)
Turing (TU106)Turing (TU106)
Pascal (GP106)Pascal (GP104)
CUDA Cores
1920
2304
1280
1920
Peak FP32 Compute
6.45 TLFOPS
7.9 TFLOPS
4.4 TFLOPS6.5 TFLOPS
Tensor Cores
240
288
N/AN/A
RT Cores
30
36
N/AN/A
Texture Units
120
144
80120
Base Clock Rate
1365 MHz
1410 MHz
1506 MHz1506 MHz
GPU Boost Rate
1680 MHz
1710 MHz
1708 MHz1683 MHz
Memory Capacity
6GB GDDR6
8GB GDDR6
6GB GDDR58GB GDDR5
Memory Bus
192-bit
256-bit
192-bit256-bit
Memory Bandwidth
336 GB/s
448 GB/s
192 GB/s256 GB/s
ROPs
48
64
48
64
L2 Cache
3MB
4MB
1.5MB2MB
TDP
160W
185W
120W150W
Transistor Count
10.8 billion
10.8 billion
4.4 billion7.2 billion
Die Size
445 mm² 445 mm²200 mm²314 mm²
SLI Support
No
No
NoYes (MIO)

Nvidia’s base clock rate for 2060 is 1,365 MHz (compared to 2070’s 1,410 MHz). Further, its GPU Boost rating is 1,680 MHz (versus 1,710 MHz). The Founders Edition card employs a 160W TDP, down 25W from the 2070, but up compared to 1060 (120W) and even 1070 (150W).

Uncut, TU106 maintains the same 256-bit memory bus as TU104. But 2060 loses two of the GPU’s 32-bit memory controllers, resulting in an aggregate 192-bit bus populated with 6GB of 14 Gb/s GDDR6 modules capable of moving up to 336 GB/s. Still, despite the pared-back configuration, 2060 serves up more memory bandwidth to its processor than last generation’s 1070 Ti.

In disabling those memory controllers, Nvidia also turns off a pair of TU106’s ROP partitions and two 512KB slices of its L2 cache. What starts as a GPU with 64 ROPs and 4MB of L2 is consequently trimmed to 48 ROPs and 3MB of cache for 2060. Like 2070, NVLink support is long gone at this price point.

Although TU106 is the least-complex Turing-based GPU thus far, its 445 mm² die contains no fewer than 10.8 billion transistors. That’s still enormous for what Nvidia might have once considered the middle of its portfolio. In comparison, GP106—“mid-range Pascal”—was a 200 mm² chip with 4.4 billion transistors inside.

Based on early benchmark results in Battlefield V with DXR enabled, it appeared as though real-time ray tracing wouldn’t be feasible on models below 2070. However, a combination of heavy optimization from EA DICE and relatively light cuts to TU106 mean 2060 maintains playable performance in the one game currently making use of Turing’s most touted advancement.

Meet GeForce RTX 2060 Founders Edition: The Little Engine That Could (Ray Trace)

From the outside, 2060 Founders Edition looks exactly like Nvidia’s own version of the 2070. It’s just as tall (11.5cm), wide (3.9cm) and long (23cm). Compact dimensions and a less massive cooler than the larger 2080 and 2080 Ti implementations also cut down on heft. 2060 Founders Edition weighs 968g, putting it within 2g of Nvidia’s 2070.

Up front, Nvidia utilizes the same pair of 8.5cm, 13-blade axial fans, which blow through a vertically oriented fin stack underneath. While we still miss the centrifugal fan and acrylic window of Nvidia’s previous reference cards, there’s no doubt that 2060 Founders Edition is built sturdily and with high-quality materials.

Flipped over, a plate covers the back of the PCB and wraps around, touching the front shroud at both ends to create clean, unbroken lines. The backplate makes direct contact with several board components with thermal pads in between to facilitate better heat transfer.

The top of this card is far less interesting than 2080 Founders Edition or 2080 Ti Founders Edition. Because 2060 does not support SLI over NVLink, there is no interface up there. Further, auxiliary power input moves to the back, presenting a cleaner aesthetic in windowed cases.

A single eight-pin power connector on the back should pose no fitment issues. Because 2060 Founders Edition is on the short side, a cable coming out of its back won’t bump up against low-hanging hard drive cages. And it sure looks better than a connector jutting forward.

Gamers buying the highest-end graphics cards are more likely to own monitors with modern display inputs. At least, that’s what we get from Nvidia’s decision to arm RTX 2080 and 2080 Ti with three DisplayPort connectors and one HDMI output, while 2060 Founders Edition sports two DisplayPort 1.4 interfaces, HDMI 2.0b and the rare DVI-DL port. It still supports VirtualLink over the card’s single USB Type-C connector, and, like the other RTX 20-series cards, you can use all four display outputs simultaneously for multi-monitor arrays.

With its fan shroud disconnected, RTX 2060’s heat sink stretches from one end of the card, past the 19cm-long PCB and to the stabilizing frame’s other side. The cable for both fans and lighting runs through that channel in the middle, where the circuit board is visible, to a connector underneath. Whereas RTX 2080 and 2080 Ti share a common thermal solution, 2060 Founders Edition employs a copper sink topped by pipes that dissipate heat through all of those fins.

That frame is held tight to the PCB by hex-head screws. Once you get them off, the frame can be carefully lifted up and away. Just be careful to not damage the thermal pads.

2060’s power requirements aren’t as exacting as 2080’s or 2080 Ti’s, so the smaller PCB is understandably less complicated. In fact, the 2060 Founders Edition is based on the same board as Nvidia’s 2070, minus a few surface-mounted components to account for two-fewer memory ICs and a lower thermal ceiling.

Given the 2060’s higher priced platform and mainstream performance level, we expect the Founders Edition card to be a stopgap until add-in board partners ramp up 2060 production of their own. If you remember back to 2016, Nvidia did something similar with its 1060 Founders Edition, which turned out to be a very limited product.

Like 2070 Founders Edition, the reference 2060 employs six power phases for TU106 and two for the GDDR6. While uPI’s uP9512 can technically control eight phases, it’s easily adapted to 2060’s less sophisticated power supply. A simpler uP1666Q two-phase buck controller is ample for the memory.

Gone are the ON Semiconductor FDMF3170 Smart Power Stage modules with integrated PowerTrench MOSFETs and driver ICs, found on  2080 and 2080 Ti. Even the cheaper NCP302155s utilized by 2070 (and capable of currents up to 55A) are gone, replaced by lower-end ON Semiconductor NCP302150s with high- and low-side MOSFETs, plus a driver, rated for 45A each. Current to the GPU is smoothed by the usual 220mH coils, while the memory similarly employs 470mH inductors.

Getting the power connector rotated 90 degrees and extended out the back of this card’s frame required an extension of sorts. Again, that’s identical to the 2070 Founders Edition we reviewed late last year. Rather than an eight-pin connector delivering ~11.1A through those three red +12V lines, though, we measured ~8.4A thanks to the 2060’s reduced power consumption.

How We Tested GeForce RTX 2060

2060 is a more mainstream graphics card than the other Turing-based boards we’ve reviewed. As such, our graphics workstation, based on an MSI Z170 Gaming M7 motherboard and Intel Core i7-7700K CPU at 4.2 GHz, is apropos. The processor is complemented by G.Skill’s F4-3000C15Q-16GRR memory kit. Crucial’s MX200 SSD is here, joined by a 1.4TB Intel DC P3700 loaded down with games.

As far as competition goes, the 2060 goes up against GeForce GTX 1070 and 1070 Ti. Of course, comparisons to GeForce GTX 1060 are inevitable as well. All of those cards are included in our line-up, as is GeForce RTX 2070. On the AMD side, we’re mostly interested in Radeon RX Vega 64 and Radeon RX Vega 56, although the more recent Radeon RX 590 seemed like an interesting comparison, too.

Our benchmark selection includes Ashes of the Singularity: Escalation, Battlefield V, Destiny 2, Far Cry 5, Forza Motorsport 7, Grand Theft Auto V, Metro: Last Light Redux, Shadow of the Tomb Raider, Tom Clancy’s The Division, Tom Clancy’s Ghost Recon Wildlands, The Witcher 3 and Wolfenstein II: The New Colossus.

The testing methodology we're using comes from PresentMon: Performance In DirectX, OpenGL, And Vulkan. In short, these games are evaluated using a combination of OCAT and our own in-house GUI for PresentMon, with logging via GPU-Z.

All of the numbers you see in today’s piece are fresh, using updated drivers. For Nvidia, we’re using build 417.54 for every card. AMD’s cards utilize Crimson Adrenalin 2019 Edition 18.12.3, which was the latest at test time.

MORE: Best Graphics Cards

MORE: Desktop GPU Performance Hierarchy Table

MORE: All Graphics Content

Create a new thread in the Reviews comments forum about this subject
78 comments
Comment from the forums
    Your comment
  • logainofhades
    I am not sure I would say it is faster, than a 1070ti. It seems that they trade blows throughout, at very similar FPS, for the most part. Price/performance, this is a winner, hands down, though, with the price being $50 cheaper than the cheapest 1070ti.
  • ammaross
    @Chris:
    "The Founders Edition card employs a 120W TDP, down 25W from the 2070, but up compared to 1060 (120W) and even 1070 (150W)."
    You typed "120W" when you meant "160W" as shown in your chart.
  • ingtar33
    this is exactly the same trick nvidia played with its other cards. all the cards this gen got pushed into a higher price bracket; it used to be we got 40% or so performance improvement in the SAME price bracket. now we're getting zero performance increase across the price brackets, however if we stick to the same product lineup we have to pay for a 50% increase in price.

    nice disingenuous framing of the problem.
  • cangelini
    Anonymous said:
    @Chris:
    "The Founders Edition card employs a 120W TDP, down 25W from the 2070, but up compared to 1060 (120W) and even 1070 (150W)."
    You typed "120W" when you meant "160W" as shown in your chart.

    Thank you, fixed!
  • teso.nagibator
    Comparing reference crappy vega cards to a dual (!) fan card? Riigght.

    Reference Vega cards can't sustain proper core clocks.

    Now, let's compare AIB Vega cards, undervolted and OCed to an OCed 2060.
  • NinjaNerd56
    What kind of test bed did you use?

    I have a GTX1060 6GB - EVGA SSC - on an I7 tower with an Optane board in front of a WD Black 10K RPM HD.

    Now, I tend to hypertune all of my boxes...and this one is par for the course.

    When I test Destiny 2, with Ultra settings, I get 200-240 fps at peak and around 130 at 96th percentile.

    Granted, 99.9% of the populace won’t tune a game tower like I do BUT considering none of the games I play use DXR, there’s ZERO value in my blowing $350.

    Yes, I could get EVEN MORE...to what end? At 120fps and above, I literally won’t appreciate any difference.
  • elroy.coltof
    In other words completely useless raytracing support. You'd be bat <mod edit> crazy to go 1080p with raytracing compared to 1440p without raytracing. All its direct competitors in the same price bracket can do good 1440p, no amount of raytracing achievable at 1080p is going to make up for the extra resolution.

    <Moderator Warning: Watch your language in these forums>
  • TCA_ChinChin
    Despite the confusing market shifts the Nvidia have done with the 20 series, I think that the rtx-2060 is still probably a good value relative to the rest of the RTX lineup. If the launch prices are right and availability is good, then even if it is more expensive than a gtx-1060, it's still alright since it compares decently fps/$ to gtx 1070/1070ti's. It wasn't the amazing value that the 1000 series had at their launch compare to the 900 series, but its at least reasonable. However, the gtx-1160 is probably gonna be a thing later, so idk about buying the rtx-2060 right now.
  • marcelo_vidal
    350US ITS A PREMIUM PRICE TO STILL PLAYING 1080P =) SORRY BUT I WILL PASS!
  • feelinfroggy777
    I'd like it better at $299
  • _Johnny5
    Anonymous said:
    I am not sure I would say it is faster, than a 1070ti. It seems that they trade blows throughout, at very similar FPS, for the most part. Price/performance, this is a winner, hands down, though, with the price being $50 cheaper than the cheapest 1070ti.

    Performance could improve with driver updates though. Still, this is the most expensive x60 card to come out *checks notes* ever. Seems kind of a wash when you could pick up a used 1080 for the same price. I thought graphics card prices were supposed to drop this year :D

    EDIT: FWIW the build quality on the FE 2060 looks top notch.
  • lojzemahnic
    Just buy it. The more you buy, the more you save, right TOM :)

    For no red/green fans price/performance of RTX 2060 still stinks.

    From Anandtech review: "The RTX 2060 (6GB) is simply no longer a ‘mainstream’ video card at $350... Against its direct predecessor, the GTX 1060 6GB, it’s faster by around 59%. In context, the GTX 1060 6GB was 80-85% faster than the GTX 960 (2GB) at launch, where presently that gap is more along the lines of 2X or more, with increased framebuffer the primary driver. But at $200, the GTX 960 was a true mainstream card, as was the GTX 1060 6GB at its $249 MSRP"
  • jgraham11
    Nice! In the final verdict " It largely outperforms them all and at a lower price point." Mean while, reality is, using Tom's own numbers: it beats the Vega 64 in only 3 titles and the rest, it gets destroyed by up to 30%. Not to mention the frame times, where the Vega64 provides a vastly superior experience...
  • salgado18
    Anonymous said:
    In other words completely useless raytracing support. You'd be crazy to go 1080p with raytracing compared to 1440p without raytracing. All its direct competitors in the same price bracket can do good 1440p, no amount of raytracing achievable at 1080p is going to make up for the extra resolution.


    Is it? Well, until now, you could only go with higher resolution. Now you can choose to go with raytracing. I'd go for raytracing, since my monitor is 1080p. I don't think there's such an easy answer, it all depends on the user.

    Also, remember people dismissed VR when it first appeared.
  • BulkZerker
    "Comparing reference crappy vega cards to a dual (!) fan card? Riigght.

    Reference Vega cards can't sustain proper core clocks.

    Now, let's compare AIB Vega cards, undervolted and OCed to an OCed 2060."

    They can though, when you manually set the money fan speed to 85% and scale to 100%
  • AgentLozen
    I'm sad to hear that the efficiency has dropped slightly. I was expecting the 2060 to be even with the 1070 worst case. What happened to the 12nm shrink? Do the ray tracing processors cause a dramatic drop in efficiency?

    On the other hand I'm glad to hear that we're finally seeing a value improvement over Pascal. Even if its only a few bucks, its better than what the other Turing cards offer. I suspect that the price may drop further still when AMD Navi comes around later this year.
  • InvalidError
    Anonymous said:
    350US ITS A PREMIUM PRICE TO STILL PLAYING 1080P =) SORRY BUT I WILL PASS!

    The joys of years of little to no meaningful competition. Hopefully, now that GPU crypto has mostly died, we'll see Navi drive performance at all price points up by a significant amount..
  • redgarl
    Value is not there... you can get a Vega 56 on sale with better value that this.
  • madbiker
    I've been running an overclocked (water cooled) 980ti for 4 or so years and I still cant justify an nvidia card. I paid $680 for my card, it gets about 18,000 3dmark in Firestrike (1080p). To me it's looking more and more like a Vega64 or AMD's next GPU are the only sensible upgrade.
  • nitrium
    Hopefully AIB versions will be at least $50 cheaper. Certainly that was case for the RTX 2070, where the FE is $599 whereas the AIBs are mostly around $499 - a full $100 cheaper. I'm on a GTX1050Ti (after my R9 390 died) and seriously want a higher end board; the RTX 2060 ticks all the boxes (except for 6GB VRAM - would have liked to see 8GB on it).