The World of Graphics Cards is Expanding

The end of 2022 has been an exciting year for gamers for quite a few reasons. First and foremost, the graphics card shortage has ended due to the crypto crash and Ethereum changing its Proof-of-Work mining protocol to Proof-of-Stake in what is called The Merge. This required validators to show they have a stake in Ethereum’s future by holding 32 ETH in their node to prove that they have Ethereum’s best interests in mind. This put an end to hardware mining using graphics cards, ultimately flooding the used markets with graphics cards that had been scalped or hoarded by miners. Along with this surge of used graphics cards, Nvidia has announced and began selling their 40-series cards, AMD had announced their Radeon RX 7000 series, and Intel has surprisingly stepped up with their ARC Graphics card to bring much needed competition into the market.

The Rise of the RTX 40 Series GPUs

Nvidia has been the first out of the gate with their new GeForce RTX series cards with a staggeringly high increase in performance, as well as price. The RTX 4090 was released on October 12^th and promised up to 2X the current performance of their current generation flagship, and up to 4X the performance with the addition of DLSS 3 and the new Ada Lovelace architecture. Nvidia also claims a 2X boost in performance for 3D rendering, AI tools, and video export speed for content creators and power users alike. These impressive numbers are contributed by the increase of CUDA cores, from 10496 on the 3090 to 16384 on the 4090, as well as the base clock speed going from 1395MHz (1695MHz with boost) on the 3080 to 2235MHz (2520MHz with boost) on the 4090. Naturally with a roughly 56% increase of CUDA cores comes an increase of power consumption. The 3080 was rated for 350W TDP while the 4090 got a 100W bump up to 450W TDP. Nvidia now recommends users to have an 850W PSU in their computer to power this card as well as the main system. Along with such a power demanding card comes a new 16-pin power connector to supply that power to it.

Those impressive numbers are a product of Nvidia’s new Ada Lovelace architecture. The 40 series GPU’s now come equipped with third-generation Ray Tracing cores to deliver up to 2X the performance compared to last generation GPUs to provide never before seen performance and photorealistic rendering. Also introduced is a new Shader Execution Reordering (SER) technology to optimize inefficient workloads to improve the shading performance and ultimately the speed of end-to-end traced image rendering performance. CUDA cores has also seen a boost in performance with the new architecture, now able to bring double speed processing for single precision floating point (FP32) over the last generation GPUs, effectively providing performance gains for graphics workflows and compute workflows like desktop simulation for computer-aided engineering programs. Along with the generational leap of the two previous components, the 40 series will now use PCIe Gen 4.0 to increase the bandwidth by 2X compared to PCIe Gen 3.0. This allows for faster GPU direct memory access transfers and data transfer speeds from CPU memory.  While this all sounds like a successful launch, it didn’t come without its woes.

The release of the new 40 series cards have been quite controversial. A major problem that consumers are facing with their new 4090 graphics cards are those new 12VHPWR power connectors burning up and melting the connector on the graphics card. This is thought to be due to the bends within the 35mm clearance that PCI-SIG had stated would cause thermal resistance, thus leading to damage. Other users have run tests to try and find the cause of the issue with some reporting that a loose connection between the card and the cable is likely the cause as temperatures would rise to 100 degrees Celsius. Fortunately, Nvidia and its AIBs have been working to get down to the root of the problem and fix it as soon as they can.

Burning cables aren’t the only backlash Nvidia was facing from its consumers. During its launch event back in September, Nvidia’s decision to launch two versions of their RTX 4080 graphics cards confused both gamers and PC enthusiasts alike. While it initially may seem that the differences in the two models was a downgrade of 4GB from the higher spec 16GB 4080 to the 12GB version, upon digging in further, it was revealed that the bottom tier 4080 card also had fewer CUDA cores. According to official Nvidia benchmarks, the bottom tier RTX 4080 performed up to 30% slower than the 16GB version. This caused confusion and evoked a community questioning as to why Nvidia hasn’t named their 12GB version the 4070 instead. Because of this, the response given from Nvidia was to “unlaunch” their 12GB 4080, stating that “The RTX 4080 12GB is a fantastic card, but is not named right. Having two GPUs with the designation 4080 is confusing, so we are pressing the ‘unlaunch’ button for the 4080 12GB. The RTX 4080 16GB is amazing and on track to delight gamers everywhere on November 16th”. This will make the cards easier to distinguish by consumers but those who hoped to get a sub $1,000 card will need to wait a little longer as the 4080 16GB card is still very expensive at $1,199.

Introducing the Radeon 7900 XT and XTX

Coming right for Nvidia’s throat is AMD with their new Radeon 7900 XT and their 7900 XTX GPUs to compete for the incredible yet expensive 40 series GPUs. While still unreleased until December 13^th, 2022, AMD has the 7900 XT priced at $899 while the stronger 7900 XTX will sit at $999 for their reference models. For $600 less than the 4090, the RX 7900XTX brings the same memory size, same memory bus, very similar clock speeds (2,300MHz, 2,500MHz boosted compared to 4090s 2,520MHz boosted clock speed), but brings the wattage down to just 355W compared to the 450W demanded by the 4090. It seemed that AMD was heavily targeting a power to performance gain for their RDNA 3 generation cards, sparing enthusiasts from having to upgrade their power supplies to meet the energy demands of their new card, unlike the 4090. The 7900 XTX is also rated up to 1.7X faster than their previous flagship GPU, the RX 6950 XT at 4K graphics. While AMD wasn’t too specific on the performance gains, they did release a few benchmark FPS numbers for select AAA games, most of which, however, are lower than the numbers put out by the 4090. Though unlike the RTX 40 series cards, the Radeon cards come with support for DisplayPort 2.1, allowing users to take advantage of more advanced displays, effectively future proofing the card.

Just like Nvidia’s DLSS technology, AMD has their own trick up their sleeve. New with the RDNA 3 generation cards is AMD Fidelity FX Super Resolution 2.1 (FSR) aimed to boost the framerates of select titles as well as deliver high-quality, high-resolution gaming without needing to upgrade to a new GPU. AMD is working to expand its FSR capabilities beyond the many games it has already supported, such as:

• A Chinese Ghost Story
• Arcadeggedon
• Back 4 Blood
• Baldur’s Gate 3
• Black Desert
• Chernobylite
• Deathloop™
• Dota 2
• Edge of Eternity
• Elite Dangerous: Odyssey
• F1 2021
• Far Cry® 6
• Gamedec
• JX3 Online
• Myst®
• Marvel’s Avengers
• Necromunda: Hired Gun
• NiShuiHan
• Resident Evil™ Village
• The Medium
• Vampire: The Masquerade – Bloodhunt
• World War Z®: Aftermath

FSR is a very useful tool to anyone using integrated to enthusiast grade graphics to boost native resolution in “performance” mode at 4K in these select games by an average of 2.4X of that without FSR enabled. This of course gives gamers a chance to enjoy 4K gaming at a much higher frame rate with ray tracing also enabled. This software is also open source, unlike Nvidia’s DLSS technology, allowing this technology to be expanded back to much older cards, as well as select Nvidia GPUs.

While we still must wait until December to truly test these new Radeon cards to see how they truly stack up to the competition, there is another competitor we can actively get our hands on to test and use from an unlikely company. I am talking about Intel’s ARC Alchemist.

A New Opponent

While Intel typically deals with integrated graphics with their processors, this year on October 12^th, they launched their own dedicated GPUs to provide fresh competition to the market. Unlike AMD, Intel isn’t trying to compete with the RTX 40 series cards. Instead, they are trying to appeal to the cheaper, 1080p gamers looking for an upgrade. There are 5 core cards on Intel’s ARC lineup, including the A770, A750, A580, A380, and A310. The A770 and A750 also have their own separate Limited-Edition GPUs available as well, kind of like Nvidia’s Founder Edition cards. Interestingly enough, the A770 Limited Edition sees a boost in GDDR6 memory to 16GB compared to the standard 8GB while the A750 Limited Edition sees no upgrade from its standard counterpart.

As mentioned previously, these cards aren’t competing with the top tier cards. The A770 only sports 8GB (16GB with the Limited-Edition card) of GDDR6 memory, 32 Xe cores, 32 ray tracing units, and a clock speed of 2100MHz. Xe cores are Intel’s core design, similar to how Nvidia has their Tensor cores on their RTX series cards. These cards were brought forward as a competitor of the last generations RTX 3060, even narrowly beating it by 1.12x in gaming performance, but at a lower cost. The ARC GPUs also have their own hardware acceleration features like DLSS and FSR 2.1 called Xe Super Sampling (XeSS). This works very similarly to the competition’s versions by rendering games at a lower resolution while using data from previous frames to make it resemble its native resolution. This allows you to have most, if not all the sharpness you would expect from a native resolution but at a higher frame rate. With all that said, the ARC GPUs are a very good consideration for mid-tier enthusiasts at the cost of $329 for the A770 and $289 for the A750.

Along with their gaming-oriented cards, Intel also released their Professional A-Series Graphics dubbed the ARC Pro A50 and ARC Pro A40. The A40 GPU supports ray tracing hardware, graphics acceleration, and machine learning capabilities in a compact factor, while the A50 GPU is fitted with more TFLOPs of performance and memory bandwidth but takes the form of a dual-slot card. Like the competitors, the ARC Pro series cards take advantage of the new PCIe Gen 4.0 to support 2x 8K displays at 60Hz, 1x 5K ultrawide at 240Hz, 2x 5K displays at 120Hz, and 4x 4K monitors at 60Hz. These cards also support full AV1 Hardware Acceleration as an industry first. This allows 4.4X faster hardware encoding compared to software encoding. While these cards are more geared towards content creators, the specifications aren’t all that bad either. The flagship ARC A50 comes supporting 8 ray tracing units, a base clock of 2000MHz, 6 GB of GDDR6 memory, and a Thermal Design Power of only 75W. The biggest difference between the A50 and A40 is that the A50 GPU has a peak performance of 4.8 TFLOPs at Single Precision rather than the 3.5 TFLOPs of the A40 GPU. While the pricing of these cards hasn’t been announced, taking a look at the pricing of their gaming line up does offer some insight into what they might end up costing.

Picking a new graphics card can be a very daunting task with all the competition from Nvidia, AMD, and now Intel. Though, as every PC enthusiast will tell you, picking a graphics card all depends on what you want to do with your machine. While some enthusiasts prefer to game at 4K at very high FPS, they would be better suited with a RTX 4090 compared to someone who needs something simple to run a basic CAD program or to use their system for photo editing. Though one thing is certain, increased competition in any market is always a good thing for the consumers. This often breeds more innovation, higher performance, and lower costs to try and gain the high ground of the market. This is the case with AMD’s new RDNA 3 cards coming for the RTX 40 series head on as well as Intel stepping in to shake the low-mid tier section. With this year’s launch events shaking the grounds for enthusiasts and pushing the boundaries of what we can do with this new technology, it only makes us more excited to see what will become available with the future generations of graphics cards.

This entry was posted on November 15th, 2022 and is filed under Uncategorized. Both comments and pings are currently closed.