As 2020 attracts to an in depth, the matter of “which new $500+ GPU is best” has turn out to be moot for many potential PC players. Nvidia had a go at it beginning in September with the spectacular RTX 3000 sequence, providing on-paper worth that blew its RTX 2000 sequence straight into obsolescence. However up to now, its three RTX 3000 fashions have suffered from a mixture of low provide and savvy scalpers scooping up such scant stock.

This modifications the tenor of any dialog about AMD’s new RDNA 2 line of GPUs. In a extra steady supply-and-demand universe, I would watch out to warn patrons concerning the execs and cons of $500+ GPUs made by both producer this 12 months—every GPU has its personal clear victories. That is already excellent news for AMD’s two new playing cards happening sale this week, because it has been years for the reason that “Crimson Crew” has been this aggressive with Nvidia.

With no clear indication that AMD will deal with Radeon RX 6800 ($579) and Radeon RX 6800XT ($649) provides any higher than Nvidia and its RTX 3000 sequence, the decision is a bit wacky. The best choice for the remainder of 2020, actually, is no matter you possibly can really buy at an affordable retail value. These playing cards are duking it out sufficient to present both facet one thing in the best way of future-proofed gaming efficiency at high-but-fair costs. For those who’re extremely desirous to improve to this tier and see something in inventory this 12 months, outdoors of the worst resellers, shut your eyes and purchase.

Ought to either side be offered out for the remainder of the 12 months, after all, then selecting by the variations could alter the way you arrange on-sale notifications for GPUs within the subsequent few months. In that case, AMD could attraction you. Throughout the board, Crimson Crew’s playing cards have velocity to spare, with the $649 RX 6800XT specifically contending properly towards, if not outright surpassing, the $699 RTX 3080. And both new AMD card is perhaps your no-question champ if you happen to’re on the lookout for VRAM-intensive workloads and 1440p gaming—and if you’re unmoved by ray tracing.

I’m sorry, how a lot L3 cache?!

AMD Radeon RX 6800XT AMD Radeon RX 6800 Nvidia GeForce RTX 3080 FE Nvidia GeForce RTX 3070 FE Nvidia GeForce RTX 2080 Ti FE
Stream Processors 4,608 3,840 n/a n/a n/a
CUDA Cores n/a n/a 8,704 5,888 4,352
Texture Models 288 240 272 184 272
ROPs 128 96 96 96 88
Tensor Cores n/a n/a 272 (Third-gen) 184 (Third-gen) 544
Ray Accelerators 72 60 n/a n/a n/a
RT Cores n/a n/a 68 (2nd-gen) 46 (2nd-gen) 68
Core Clock 1,825MHz 1,700MHz 1,440MHz 1,500MHz 1,350MHz
Enhance Clock 2,250MHz 2,105MHz 1,710MHz 1,725MHz 1,635MHz
Reminiscence Bus Width 256-bit 256-bit 320-bit 256-bit 352-bit
Reminiscence Velocity 16GHz 16GHz 19GHz 14GHz 14GHz
Reminiscence Bandwidth 512GB/s 512GB/s 760GB/s 448GB/s 616GB/s
Reminiscence Measurement 16GB GDDR6 16GB GDDR6 10GB GDDR6X 8GB GDDR6 11GB GDDR6
TDP 300W 250W 320W 220W 260W
MSRP at launch $649 $579 $699 $499 $1,199

The breakdown between Nvidia’s RTX 3000 sequence and AMD’s RDNA 2 sequence sees both sides emphasize sure specs. VRAM is the obvious differentiator, with AMD getting a 16GB pool of GDDR6 VRAM into customers’ fingers at each tier of this 12 months’s line. Nvidia made its personal funky wager: much less of the same-specced VRAM (8GB) at its $499 RTX 3070, and a little bit extra (10GB) for its $699 RTX 3080 although that is bumped to a blistering GDDR6X configuration.

In any other case, the spec showdown between AMD and Nvidia differs on a foundation of compute items and clock speeds. AMD’s RX 6000-series playing cards exceed 2.1GHz, forward of the 1.7GHz vary on the RTX 3000 sequence, whereas their stream processor counts pale in comparison with Nvidia’s personal CUDA core counts. That is not an equal comparability, I admit, however when it comes to teraflop willpower, each of these values are used.

Just like the RTX 2000 sequence earlier than it, Nvidia’s RTX 3000 playing cards embody two sorts of proprietary cores: one pool devoted to ray tracing, and the opposite devoted to mathematical computation and juggling of Nvidia’s personal machine-learning fashions. AMD says that it has added its personal “ray accelerators” to the RX 6000 sequence’ boards (one per compute unit). By way of ray tracing-specific computation, AMD describes them as capable of calculate “as much as 4 ray/field intersections or one ray/triangle intersection each clock.” That is just like how Nvidia’s RT cores work, although their RTX 3000 sequence has upgraded the cores to moreover deal with the interpolation of triangle positions (used to effectively deal with curved mild trajectories) and to double the prior generations’ dealing with of triangle intersection charges.

AMD at present has no reply to Nvidia’s tensor cores—and their effectiveness at intelligently upscaling a shifting, lower-res picture. As an alternative, AMD reps hinted loudly to one thing referred to as FidelityFX Tremendous Decision coming to their newest GPUs at some undetermined level sooner or later. For now, nothing within the firm’s press supplies clarifies precisely how this method will work, the way it would possibly differ from its current FidelityFX Sharpening system, or what within the RDNA 2 structure can be leveraged by Tremendous Decision.

Weirdly, AMD had little particular to say about one of many extra unimaginable spec figures on its RX 6000-series GPUs unmatched by Nvidia: a whopping 128MB of L3 cache (surpassing the already staggering 64MB of L3 cache discovered on numerous AMD Ryzen CPUs). Precisely how present or future video games leverage that high-bandwidth chunk of cache, significantly with measures like tile-based or deferred rendering in a recreation’s juggle of GPU and CPU assets, stays unclear.


Please enter your comment!
Please enter your name here