PC gaming has underwent a lot of evolution over the last few years. On the graphics side, Nvidia was the first company to release raytracing-capable consumer graphics cards to the market. On top of that, the company introduced an innovative way to improve performance while maintaining high image quality via Deep Learning Super Sampling (DLSS). While the first iteration made images look like they were smeared with vaseline, version 2.0 made huge leaps and bounds.
On the CPU side, AMD has significantly improved its Ryzen chips each generation. The unification of the L3 cache in Zen 3 allowed the Ryzen 5000 series to make big improvements in gaming performance thanks to reduced cache latencies. On top of that, Smart Access Memory (SAM) provided a slight boost in gaming performance.
Over the next year, the PC gaming landscape will make even more progress. From Team Red, its DLSS competitor, FidelityFX Super Resolution (FSR), will allegedly launch next month and its Rembrandt APUs may bring impressive performance at the low-end. Meanwhile, Intel looks to move past the Rocket Lake debacle with its seemingly impressive Alder Lake chips and throw its hat in the GPU ring with DG2.
AMD's DLSS Competitor and Rembrandt APUs
Last November, AMD launched its RDNA2 graphics cards, notably the 6900XT, 6800XT, and 6800. Even though the cards "only" use GDDR6 memory in contrast to Ampere's faster GDDR6X VRAM, AMD made up the bandwidth gap with higher memory capacities and the Infinity Cache. Along with RDNA2's impressive clock speeds, Team Red has finally caught up to Nvidia in terms of rasterization performance.
However, Nvidia is still ahead in terms of features, notably on raytracing and DLSS. Thanks to the head start, Ampere has significantly better raytracing performance than RDNA2. In addition, DLSS allows Ampere to boost performance without sacrificing image quality to a significant degree. AMD cannot really do much on raytracing performance until RDNA3, but it will allegedly finally introduce its DLSS competitor in FSR next month.
Here are the rumored details from Coreteks:
If true, then FSR seems to tick the right boxes and looks to be on good track. Low overhead and ease of development are arguably the most important aspects for super sampling. However, the most interesting detail of the list is FSR's compatibility with Nvidia GPUs. It sounds very counterintuitive, but when you think about it in terms of market penetration, it makes sense.
Basically, making FSR cross-vendor will encourage developers to favor it over the Nvidia-exclusive DLSS by virtue of a larger userbase. RDNA2 is found in every PlayStation 5, Xbox Series X, and Series S. The PS5 has sold past 7.8 million units and while there are no clear numbers for the Xbox consoles, the total userbase should be around 13-14 million. However, while AMD has made improvements to their graphics cards, Nvidia remains the undisputed king with its large marketshare. By making FSR cross-vendor, it substantially enhances its market penetration and in the longterm, may become the de facto supersampling method. That said, it also largely depends on how well FSR performs and looks compared to DLSS.
On the budget end, AMD will finally implement RDNA2 into its APUs next year after sticking with the Vega architecture. Allegedly, Rembrandt will feature Zen 3+ cores based on TSMC's 6nm lithography and up to 12 RDNA2 compute units. According to Wccftech, compared to the best Zen 3/Vega Cezanne APU, the best Rembrandt APU will offer 50% more stream processors (768 vs. 512). In addition, Rembrandt will be on the AM5 platform, meaning it will take advantage of DDR5's improvements over DDR4.
Compared to the Xbox Series S, the best Rembrandt APU will feature significantly better CPU cores, but fewer CUs in the integrated GPU. While the RDNA2 compute units will likely be clocked much higher than the Series S's 1.565 GHz clockspeed, I don't think that is enough to make up for having 8 fewer CUs. Even then, I think Rembrandt will pack a respectable punch at the low-end.
Intel's Alder Lake CPUs and DG2 GPU
Over the past few years, Intel has struggled to move on from 14nm as its 10nm yields have not been up to snuff. Its Rocket Lake CPUs were not received well with notable YouTube channels like Gamers Nexus panning the 11700K as a "waste of sand" or the 11900K as "embarrassing". The chips sapped a lot of power and had pretty terrible cache latencies.
Intel seeks to launch their Alder Lake CPUs this November as their reverse UNO card and the specs look impressive. Unlike Rocket Lake that is based on 14nm, Alder Lake will be based on Intel's 10nm SuperFin lithography. According to Intel, this will introduce up to 20% more single thread performance (over Tiger Lake). The biggest innovation Alder Lake introduces is the big.LITTLE core layout. The 16 core Alder Lake CPU, for instance, will feature 8 Golden Cove cores and 8 more power efficient Gracemont cores.
Just like with AMD's Rembrandt, Alder Lake will support DDR5 for greater memory bandwidth and lower latency. Lastly, if everything goes as scheduled, Intel will beat AMD to the punch on PCIe Gen5. However, thus far, the increased bandwidth from PCIe Gen3 to Gen4 has not brought any tangible gaming performance boosts (yet).
What I'm most excited for is Intel's upcoming DG2 discrete GPU. Supposedly, DG2 will launch in the second half of this year on laptops and desktops. The discrete version will feature multiple SKUs going from as low as 86 execution units (EUs), 768 shading cores, and 4GB of GDDR6 to as high as 512 EUs, 4096 shading cores, and 16GB of GDDR6.
Specs of all DG2 SKUs from Wccftech.
What's encouraging is that for Intel's first foray into the discrete GPU market, the high-end DG2 card is rumored to not be a slouch. According to Moore's Law is Dead, the 512EU unit should perform around the 3070Ti. While it's not going to beat the 3080, achieving mid-high level performance with your best card is not bad. Plus, if FSR will be available on Nvidia GPUs, I would expect the same to happen with DG2, too, for a nice performance boost.
If the last bits of Moore's Law is Dead's info are correct, then Intel will also price their cards competitively against AMD and Nvidia's cards. With the chip shortages, the prices of both RDNA2 and Ampere cards greatly exceed their MSRPs. The caveat is that there isn't much incentive for Intel to price DG2 competitively, but I do hope that they follow through as more competition in the GPU market would be welcome.
Overall, I think AMD and Intel will push PC gaming forward substantially. Alder Lake looks to be a good comeback story with its impressive IPC improvements and enhanced 10nm node. Rembrandt may be a great budget option.
But most interestingly, FSR and DG2 may bring a whole lot more competition to the GPU market. If FSR performs and looks close to DLSS 2.0, and the feature is cross-vendor, then developers might prioritize FSR over Nvidia's proprietary solution. If DG2 offers decent performance and competitive prices, then this will incentivize AMD and Nvidia to adjust accordingly.