AMD vs Nvidia – what’s the best add-in GPU for you?

This post was originally published on this site

We both know the AMD vs Nvidia rivalry will never end.

For the longest time, we’ve only had to deal with this specific GPU tug-of-war, as the graphics scene is quite different now than what we experienced in the late 90s when GPUs were a new concept in PC gaming. Back then, multiple GPU brands saturated the market built by 3DFX, ATI (now owned by AMD), Diamond, Rendition, S3, Nvidia, and more. They each offered a unique take on rendering polygons, smoothing edges, and applying colored lighting. It was a fun (and annoying) time to be a PC gamer and a developer. A new frontier, if you will.

Right now, choosing the best graphics card for your gaming PC isn’t quite so convoluted, though each new generation produced by AMD and Nvidia contains multi-unit releases rather than just one card. Adding fuel to the fiery feud will be Intel’s entry into the GPU arena during 2020 with its Xe add-in graphics card spearheaded by former Radeon lead Raja Koduri.

So who wins in the current AMD vs Nvidia feud? We load up and aim to find out.

AMD vs Nvidia: This is AMD

AMD vs Nvidia

If you’re just tuning in, AMD and Nvidia have new generations begging for your hard-earned dollars. On Team Red, AMD just introduced the Radeon VII based on 7nm process technology. It’s a second-generation Vega chip (Vega 20) using the company’s fifth-generation GCN 1.5 architecture. It packs 3,840 streaming processors, 16GB of onboard memory (HBM2), and a 1TB/s memory bandwidth. It’s AMD’s answer to the RTX 2080, though slightly slower in some cases.

Moving down the line, AMD released the RX Vega 64 and RX Vega 56 cards in August 2017. Both are also based on the fifth-generation GCN 1.5 architecture and the first version of AMD’s Vega chip. The idea here was to release products comparable to Nvidia’s GTX 1080 and GTX 1070 at similar price points. For example, this PUBG benchmark shows the GTX 1080 and RX Vega 64 nearly neck-to-neck in performance at 1080p and 1440p.

AMD’s other more-recent GPU family is the RX 500 series covering the mainstream gaming market. It’s based on AMD’s fourth-generation GCN 1.4 architecture, with the RX 590 serving as the flagship using the 12nm Polaris 30 chip. Targeting the sub-$300 market enjoyed by Nvidia’s GTX 1060, this card benefits from the smaller process node to provide higher speeds than the RX 580. Along with these two cards, everything else listed in the RX 500 Series aims to bring affordable support for VR and Full HD graphics to the masses.

But a word of caution: If you’re shopping for an AMD card, take note of the listings without the “RX” prefix and/or the “X” tacked onto the numbers. Ignore these products, as this rebranding is for OEMs only and do not reflect hardware revisions. OEMs apparently require brand refreshes, thus the Radeon RX 580X is no different than the RX 580.

AMD vs Nvidia

We’d also like to note that, despite Nvidia’s saturation in the notebook market, AMD provides plenty of hardware too. The company delivers eight discrete GPUs ranging from the RX 580 to the 520. AMD also crams integrated graphics into its all-in-one chips (APUs) for desktops and laptops, and discrete graphics into Intel’s recent modules combining revised Kaby Lake CPU cores, Intel integrated graphics, HBM2 memory, and Vega M GPU cores.

Unlike Nvidia, AMD also has its foot in the desktop and laptop processor market, heavily competing with Intel. In comparison, Nvidia has an all-in-one processor as well, but it’s mainly used in automotive, set-top-box streaming devices, and the Nintendo Switch (Tegra X1 T210).

Despite Steam statistics showing otherwise, AMD still has a strong presence in the PC gaming market. In 2011 after one well-known publisher admitted in a roundabout way that PC gaming was dead and most of its revenue stemmed from consoles, AMD held a private meeting two years later to reveal its new gaming initiative: Unify the PC with consoles for better gaming.

AMD still has a strong presence in the PC gaming market.

That plan started with its Mantle API (now rolled into Vulkan) and developing APUs for the Xbox One and PlayStation 4 consoles. With this initiative, we wouldn’t see horrible ports across three platforms, but rather a unified, consistent experience for both gamers and developers. PC gaming is now at the forefront of high-definition gaming again partially due to AMD’s efforts.

But we can’t ignore AMD’s seemingly shallow GPU portfolio on the desktop front. PC gamers could argue that the company hasn’t produced a new GPU architecture since GCN 1.1 made its debut in 2011. Instead, AMD made revisions throughout the years to add support for new technologies.

Thankfully this could soon change. Navi is expected to be a completely new design built from scratch, just like AMD’s Zen design for CPUs. Navi will likely be announced during E3 2019, and part of AMD’s “Gonzalo” SoC for the next-generation Xbox Scarlett and PlayStation 5 consoles slated for 2020.

If AMD didn’t also focus on CPUs and console APUs, would the company have more time for GPU development? That’s hard to say, but note that AMD battles both Intel (CPU) and Nvidia (GPU) in the enterprise, so it’s resources are spread far and wide. Meanwhile, Nvidia won’t really compete with Intel until the Fe add-in graphics card arrives in 2020.

AMD vs Nvidia: This is Nvidia

AMD vs Nvidia

After sitting on its GTX 10 Series for two years, Nvidia finally launched its RTX 20 Series in August 2018 promising real-time ray tracing. In gaming, this technique tracks individual rays of light and how they interact with virtual objects. Instead of rendering a scene where a light source simply illuminates solids and casts shadows, ray tracing adds reflections and refractions to create a more realistic scene.

However, this process takes large amounts of computational power to render a single image. In the old days, waiting on a single ray-traced render required large amounts of time. This is why animation studios employed farms of computers to pre-render each frame in their animated movies. That said, ray tracing is far from new, but it’s new in gaming, which demands at least 60 images rendered each second in real time, not pre-rendered before you play.

To accomplish 60fps, Nvidia developed its “RT” core that calculates the intersection where a single ray of light collides with a large volume. If the ray doesn’t intersect, then the core moves on. If the ray does intersect, the core calculates the intersection again in greater detail. All of this work is done on the RT core to prevent bogging down the otherwise busy CUDA cores.

Backing the RT core is Nvidia’s Tensor core optimized for artificial intelligence. The idea is to use Tensor cores to enhance visuals and fill in the framerate holes, whether ray tracing is enabled or not. How? By using what Nvidia calls Deep Learning Super Sampling, or DLSS.

As the name suggests, Nvidia uses deep learning to train a neural network so it can intelligently render missing frames not generated by the graphics card. It also injects additional pixels into each rendered frame to increase the resolution. These improvements are provided through game profiles in GeForce drivers.

The current lead of the pack is Nvidia’s Titan RTX, costing a hefty $2,499. But you get what you pay for, such as 4,608 CUDA cores with a base speed of 1,350MHz and a boost speed of 1,770MHz. It also packs 72 RT cores, 576 Tensor cores, and a massive 24GB of on-board GDDR6 memory moving along at 14Gbps. It’s a performance beast that AMD currently doesn’t rival.

Other cards in the new RTX 20 Series family include the RTX 2080 Ti ($999), the RTX 2080 ($699), and the RTX 2070 ($499) launched in 2018 followed by the RTX 2060 ($349) introduced in January. These prices are suggested starting points, and Nvidia’s Founders Editions cost $100 to $200 more.

All five cards are based on Nvidia’s new Turing streaming multiprocessor architecture, 12nm process technology, and a new memory system design to support GDDR6 VRAM. Outside hardware support for real-time ray tracing and AI, Turing includes features pulled from Nvidia’s Volta design for data centers and enterprise along with improved shader execution efficiency and more.

AMD vs Nvidia

In a surprising-yet-not-surprising move, Nvidia revealed the GTX 1660 Ti graphics card in February, promising 1.5 more performance than the GTX 1060 (6GB) using the same 120-watt power draw, but for $30 more. It does not support SLI nor does it include RT cores or Tensor cores. Although we have yet to benchmark the card, it’s relatively on par with Nvidia’s older GTX 1070 in Ashes of the Singularity, Battlefield V, Far Cry 5, and a few others.

Other new cards supposedly in the works are the GTX 1160 and GTX 1650. An RTX 2050 was also briefly listed on a Dell G5 15 product page, but that was changed to the GTX 1660 Ti, indicating that Nvidia’s new RT- and AI-free GPU is eventually coming to laptops. The typo isn’t unexpected given OEMS, their PR firms, and webpage editors are never, ever on the same page.

Nvidia’s GTX 1660 Ti address the mainstream gaming market. Unlike AMD, Nvidia currently doesn’t set aside a specific family for budget-friendly add-in cards. For the older Pascal-based GTX 10 Series, Nvidia lumped all products into one family ranging from the Titan Xp to the GT 1030. You can still purchase these cards, as third-party OEMs have plenty of solutions up for grabs.

Nvidia currently doesn’t set aside a specific family for budget-friendly add-in cards.

On the mobile side, Nvidia introduced its RTX 20 Series for laptops in January. As with the 10 Series, you’ll see high-performance models and Max-Q variants with the power dialed down to prevent excessive heat buildup in thin form factors. The latest mobile family includes the RTX 2080 (vanilla and Max-Q), the RTX 2070 (vanilla and Max-Q), and the RTX 2060.

So what’s the difference between the desktop and mobile versions? All three are based on the same 12nm TU104 chip with 2,944 CUDA cores, 46 RT cores, and 368 Tensor cores. But the desktop add-in card consumes the most power of the three at 215 watts (225 watts for Founders Edition) followed by the RTX 2080 for mobile (150 watts) and the Max-Q version (80 watts).

That said, the base speed of the desktop variant is 1,515MHz while the mobile version ruins at 1,380MHz and the Max-Q version at 735MHz. According to Nvidia, the desktop variant is capable of 8 Giga Rays per second while the mobile version hits 7 Giga Rays per second and the Max-Q variant at 5 Giga Rays per second. Yay for a new class of benchmarking numbers!

Of course, given the RTX 20 Series for mobile is relatively new, you can find plenty of gaming notebooks with the older GTX 10 Series chips still saturating the market.

AMD vs Nvidia: Performance

AMD vs Nvidia

Real-time ray tracing in games didn’t start off on a positive note. When Nvidia initially revealed the RTX 20 Series, there were concerns that ray tracing made too much of a framerate hit, even on the high-dollar RTX 2080 Ti. With ray tracing enabled, the card couldn’t hit 60fps at 1080p in Shadow of the Tomb Raider, ranging between 33fps and 48fps instead. At the time, the developers said support was an early work in progress. Edios provided a polished update in a post-launch release.

By the time Nvidia’s RTX 2080 and RTX 2080 Ti arrived in September, gamers couldn’t experience ray tracing due to the wait for Microsoft’s Windows 10 October feature update (1809) containing the DirectX Raytracing (DXR) API. When it finally arrived, Microsoft pulled the update after reports of it deleting user data. The phased rollout through Windows Update didn’t resume until November, and only on devices believed to “have the best update experience.” Meanwhile, ray tracing support in games didn’t begin to roll out until November.

A recent revisit with Battlefield 5 showed the RTX 2080 Ti with ray tracing enabled (116fps) on par with AMD’s Radeon RX 570 (4GB) with no ray tracing (111fps), both tested at a Full HD resolution and Medium graphics settings. The RTX 2080 with ray tracing enabled (106fps) was slightly ahead of the GTX 1060 (3GB) without ray tracing (103fps) in the same test.

In the same batch of Battlefield 5 tests, the RTX 2080 Ti without ray tracing enabled hit an average 177fps running at a Full HD resolution and Ultra graphics settings. That level of performance is expected from a $1,199 card, but with ray tracing turned on, the framerate dropped down to 93fps (which is still good). The RTX 2080 saw a similar dip, hitting 155fps without ray tracing and 82fps with ray tracing.

AMD vs Nvidia

Meanwhile, you won’t find dedicated hardware acceleration for real-time ray tracing on AMD’s current graphics cards. Sure, you can run ray tracing on an AMD GPU, but it won’t support Microsoft’s DXR used in PC games. For AMD’s answer to real-time ray tracing, you’ll likely have to wait for the company’s Navi-based cards scheduled to arrive at the end of 2019. According to a recent interview, AMD won’t offer real-time ray tracing until it can be utilized across a wide range of low end to high-priced graphics cards.

Yet even without real-time ray tracing, AMD’s latest graphics cards hold up against Nvidia’s new RTX 20 series. In the same Battlefield 5 test using a Full HD resolution and Ultra settings, AMD’s older RX Vega 64 – with a $499 starting price – was on par with the newer RTX 2070 card with the same starting price. Meanwhile, AMD’s $399 RX Vega 56 outperformed Nvidia’s GTX 1080 graphics card by a two-frame average, which still sells for at least $549.

Even without real-time ray tracing, AMD’s latest graphics cards hold up against Nvidia’s new RTX 20 series.

But here’s the thing: AMD’s graphics cards typically consume more power. For example, the $399 RX Vega 56 consumes 210 watts while the $499 GTX 1080 consumes 180 watts ($549 at Nvidia). Or use newer cards for example: The $699 Radeon VII consumes 300 watts while the similarly-priced RTX 2080 uses 215 watts (reference design). Even if you’re not gaming, these cards consume power when idle.

AMD fans will argue that the company provides superior alternatives despite the power requirements. Sure, they may consume more voltage, but you can under-volt the cards to match the power required by Nvidia-based models and still get great performance. But if your monthly power bill isn’t a concern, under-volting isn’t necessary.

AMD vs Nvidia: Pricing

AMD vs Nvidia

AMD’s current strategy is to provide more performance per watt at a better price. As we’ve already stated, the new Radeon VII targets Nvidia’s RTX 2080. It’s a $699 card while the GTX 2080 has the same starting price.

AMD typically doesn’t sell graphics cards directly to customers, but instead provides a reference design for its hardware partners. That’s not the case with its Radeon VII. Right now you can even buy through AMD or a qualified retailer and receive a free game bundle including Tom Clancy’s The Division 2, Resident Evil 2, and Devil May Cry 5. Three solutions from AMD’s hardware partners range in price from $699 to $719 for now.

Nvidia sells graphics cards directly to customers under its Founders Edition brand at a higher price than the suggested retail price. The comparable RTX 2080 Founders Edition is $799 with an overclocked boost speed while add-in cards sold by hardware partners range from $699 to $899.

Moving down the line, AMD’s hardware partners sell variants of the Radeon RX Vega 64 reference design ($499) ranging from $409 to $806. The higher prices may stem from proprietary technologies, better fans, a better thermal design, added illumination, higher clocks, and so on. Higher prices could also be mere inflation due to limited product supply.

AMD vs Nvidia

Meanwhile, Nvidia sells the RTX 2070 directly to customers for $599, a $100 increase over its suggested retail price. Add-in cards from hardware partners like MSI, Gigabyte, Asus, EVGA, and Zotac range in price from $489 to $659. The older GTX 1080, which is in the same performance ballpark, originally had a suggested retail price for $499 but currently sells for $549 through Nvidia. You can still buy the GTX 1080, but they’re far from cheap ranging from $735 to more than $1,320.

For the RX Vega 56, AMD provides a $399 suggested retail price. Since it doesn’t directly sell this card to customers, third-party partners run with the design to add features, increase clock speeds, and jack up the suggested price. Currently you can get RX Vega 56 add-in cards ranging from $411 to $799.

Comparable cards provided by Nvidia would be the new RTX 2060 and the older GTX 1070 Ti. The latest edition to the RTX 20 Series has a suggested retail price of $349. Nvidia currently doesn’t sell a Founders Edition variant, but prices from third-party partners gravitate to the suggested price. The highest price for this model we’ve seen thus far is $399.

Meanwhile, the older GTX 1070 Ti has a $449 suggested retail price. Nvidia sells a Founders Edition model at that price while packing in a Fortnite game code. Cards from third-party partners aren’t quite so cheap, costing between $519 and $1,499. Used models float around the suggested retail pricing.

A complaint we’ve seen since the launch of Nvidia’s RTX 20 Series is that new cards under the $1,199 price tag don’t provide a significant, noticable performance boost over the previous GTX 10 Series generation. The argument is that the 1.32x 4K performance increase and shallow ray tracing support in games simply isn’t worth the large price hike. What you’re seemingly buying is support for real-time ray-tracing with a slight performance boost tacked on.

  Starting Price Founders Edition
Titan RTX $2,499
Titan Xp $1,200
RTX 2080 Ti $999 $1,199
GTX 1080 Ti $699 $699
RTX 2080 $699 $799
GTX 1080 $499 $699
RTX 2070 $499 $599
GTX 1070 $379 $449
RTX 2060 (6GB) $349 $349
GTX 1060 (6GB) $249 $299

Previous generational releases have always managed to provide SKUs that provide a significant performance increase, justifying the new cost. Currently, GTX 10 Series cards are thinning out on the market, thus you’ll see inflated prices especially after the troubles Nvidia experienced with the RTX 20 Series at the end of 2018.

AMD vs Nvidia: Who’s the winner?

Neither. Simply put: The GPU market is currently in a weird state.

AMD currently appears to be in defensive mode, releasing products that are nearly equal in performance with specific Nvidia cards already available at a similar price. The company’s release schedule in recent years has consisted of the RX 400 Series and RX 500 Series addressing the mainstream market, the two numbered Vega cards responding to Nvidia’s GTX 10 Series, and the Radeon VII targeting the RTX 2080. We’re still playing the wait-and-see game with AMD’s Navi architecture designed specifically for 7nm. Will it be AMD’s GPU equivalent to Zen?

AMD currently appears to be in defensive mode while Nvidia is clearly on offense.

Nvidia is clearly on offense, but we can’t help but wonder if it launched the RTX 20 Series prematurely. Outside of the initial launch issues, there are only four PC games on the market that actually support real-time ray tracing: Assetto Corsa, Battlefield 5, Metro Exodus, and Shadow of the Tomb Raider. Here are the other games slated to receive ray tracing support:

  • Atomic Heart
  • Control
  • Enlisted
  • Justice
  • JX3 Remake
  • Mechwarrior V: Mercenaries
  • ProjectDH

The list of PC games supporting Nvidia’s DLSS is quite longer, currently comprising of 28 titles including Anthem, Battlefield 5, Darksiders III, Final Fantasy XV, Playerunknown’s Battlegrounds, Serious Sam 4: Planet Badass, Shadow of the Tomb Raider, and many more. Remember, DLSS is separate from ray tracing, as it solely focuses on improving framerates and resolutions in supporting titles.

We believe Nvidia launched too early due to the GPU market’s previous state. Cryptocurrency miners were scooping up graphics cards, causing product availability voids and inflated pricing on the remaining hard-to-find units. To fill those voids, Nvidia and its partners produced additional cards, but the cryptocurrency mining craze abruptly ended, leaving Nvidia sitting on an add-in card stockpile.

Hacker Noon

To help clear out this backlog, Nvidia and its partners reduced the prices. The clear way to recoup that financial loss was to release the RTX 20 Series even though Windows 10 clearly wasn’t ready and there were only a few games supporting the hyped feature. Should you buy an RTX 20 Series card right now? If you’re moving from the GTX 900 Series or older, then yes.

Don’t take all of this as pro-AMD Nvidia bashing. It’s not. As we stated, the current state of the GPU market is simply weird, even more so given that Nvidia just launched a new graphics card devoid of its super-hyped ray tracing and DLSS features. But its affordable starting price may still not be enough to move PC gamers off the GTX 1060 and GTX 1050 Ti.

According to Steam’s Hardware & Software Survey for February 2019, the GTX 1060 dominates the GPU usage stats, commanding a 15.88 percent share. Following it is the GTX 1050 Ti, the GTX 1050, the GTX 1070, and the GTX 960.

As for Nvidia’s new RTX 2080 Ti, RTX 2080, and RTX 2070 add-in cards, they appear towards the bottom of the list, but still show some adoption.

Meanwhile, the Radeon RX 580 is currently AMD’s most popular product on the list with an unspecified Vega listing serving as the company’s sole non-Polaris product detected by Steam. Given it didn’t appear until January, this entry could be the Radeon VII.

  OCT NOV DEC JAN FEB
RTX 2080 Ti 0.15% 0.21%
RTX 2080 0.22% 0.31% 0.40%
RTX 2070 0.17% 0.33% 0.50%
RX 580 0.50% 0.63% 0.67% 0.82% 0.94%
RX 570 0.15% 0.23% 0.28%
RX 560 0.30% 0.25% 0.27% 0.28% 0.28%
RX 550 0.17% 0.18% 0.17% 0.19% 0.19%
RX Vega (?) 0.16% 0.19%

AMD vs Nvidia: Bottom line

Stats show that in February 75.02 percent of PC gamers accessing the Steam platform have Nvidia-based hardware installed while AMD commands a mere 14.68 percent. The numbers clearly show Nvidia wins in the GPU showdown, at least on Steam. But they also show that gamers are unwilling to upgrade from their GTX 900 Series and GTX 700 Series cards.

Gamers even hanging on to older Radeon R7 and Radeon R5 cards. Why is that? Perhaps PC gamers are just fine with Full HD graphics and don’t really want to take the 4K plunge, especially when you can get a great gaming desktop or laptop for the same cost as the Titan RTX or even the RTX 2080 Ti. Maybe 4K HDR is over-hyped and gamers aren’t ready to invest in OLED screens.

Still, 2.88 percent of the surveyed Steam members own the GTX 1080, which was $499 at launch. They also own the GTX 1080 Ti, which had a $699 launch price when it arrived in March 2017. Many PC gamers are willing to spend the money while others are just fine with average resolutions at high framerates.

PC gamers are definitely preferring Nvidia over AMD, but which do you prefer? Let us know down in the comments.

Be the first to comment

Leave a Reply