Updated 17th November: Buy yourself an AMD Radeon RX 6900XT here.
Updated 13th November: We are a day closer and the search continues to find listings for the Radeon RX 6000 series GPUs.
Updated 12th November: With the Ryzen 5000 CPUs benchmarked, we cannot wait to get our hands on the RX 6000 GPUs to see what the full potential behind Smart Access Memory truly is.
Updated November 11th – Our team is constantly scouring the net for links, so make sure you check this page regularly as they can appear at any time during the day when we spot them. We are only a few days out from release now so excitement is beginning to hot up.
Following AMD’s big reveal of the 6000 series of GPUs on October 28th, one of the biggest subjects of interest is how the most powerful card of the range, the 6900 XT, popularly known as “Big Navi”, will compete with Nvidia’s most powerful offering, the RTX 3090. At present we can only go off the official data AMD provided on their launch and the earliest leaks from independent benchmarkers, but we will attempt to see how the red team’s challenger stacks up against the green team’s (largely AWOL) heavyweight champion.
If you are interested in looking at a more general comparison of the two ranges in their entirety, take a look at this article.
As we will discuss, not every subcomponent of the two cards can be directly compared (GPU cores, Ray tracing cores, Tensor cores and Memory categories included), however, the below table gives us a rough overview as to how the cards measure up on paper.
For those running larger resolution displays, 4K in particular, memory bottlenecking is proving to be an increasing issue in performance. No matter the clock speed and core power of a GPU, if the memory bandwidth isn’t sufficient this can frequently lead to low minimum FPS when benchmarked, even if the average FPS is high, leading to potential stuttering in gaming as the frame rate noticeably fluctuates.
The 6900XT falls below the green team’s flagship RTX 3090 in all three areas of VRAM, memory speed, and memory bus, however, AMD have developed a secret weapon. Based on technology developed for their Ryzen CPUs, AMD’s “Infinity Cache” apparently allows the GPU to operate at a bandwidth 2.17x greater than you would normally expect a 256-bit memory bus to be capable of. This means that, according to AMD, their card can operate at an effective memory bandwidth of 1,664GBps, nearly 1.8x the memory bandwidth of the RTX 3090. This means the 6900XT should outperform the Nvidia cards in this particular area using a smaller bus, saving on both power consumption and heat generated and contributing to the lower TDP and TBW of the AMD card.
Historically, software and drivers have been one area that AMD has lagged behind Nvidia, though with the 6000 series they seem to be narrowing that gap.
Both AMD and Nvidia cards are fully compliant with “DirectX 12 Ultimate”: this means they will support mesh shading, variable-rate shading, and sampler feedback. DLSS (Deep Learning Super Sampling) is an advantage that owners of the RTX 3090 will have over their rival, an exclusive proprietary technology that Nvidia is unlikely to give to their rival any time soon, though this will only be a factor of course in those games which use this technology. Still, the RTX 3090 currently has the edge in 8K gaming for this reason alone, and although AMD revealed they are working on their own version of A.I. upscaling tech called Super Resolution, we do not know when this will be available to owners of the 6900 XT, nor how effective it will prove.
Ray-tracing (also known as RTX – from where the Nvidia cards get their prefix) is another area in which Nvidia has the edge, though in this at least AMD have announced the 6900 XT will use their own version of the technology on release. Early tests, as reported by Videocardz, do suggest however that AMD’s implementation still lags behind that of Nvidia.
Besides the aforementioned Rage Mode and Smart Access Memory, owners of the 6900 XT will take advantage of the new “Radeon Boost” and “Radeon Anti-Lag” technologies to increase response time and reduce latency. In both their CPU and GPU launches AMD have really pushed the latency-reducing aspects of their technology, and it could be that this is an area they believe they can genuinely overtake Nvidia on. We look forward to seeing the third-party data on this in the near future.
AMD did not give us the traditional “Base Clock” speed measurement, opting instead to use their “Gaming Clock” measure which is supposed to be an indication of speeds the GPU will expect to reach during typical gameplay, rather than the extremely low rate that even an idle computer will rarely hit which “Base Clock” represents. We cannot do a direct comparison at this stage of both cards on a Base Clock or a Gaming Clock basis as Nvidia does not use the latter measurement, regardless, going off the Boost Clock figures alone we can assume that AMD’s card comes ahead in all these metrics.
Nevertheless, clock speed does not necessarily translate into performance – factors such as the number and power of the cores and the resulting IPC (Instructions Per Clock), combined with other metrics like the number and efficiency of ROPs (Render Output Units) in the rasterization process (turning the basic vector information from the CPU into fully fleshed out textures) all combine to produce real-world performance.
Unfortunately, we cannot directly compare the two measurements that both companies use for their GPU cores (Stream processors for AMD and CUDA cores for Nvidia) as they are each calculated differently. Similarly “Ray tracing cores” operate differently in both cards, and “Tensor cores” are a designation only really used by Nvidia.
Ultimately the only way to compare both cards is via benchmarking – both technical and in FPS terms on a variety of games and applications.
At this stage, the only benchmarks we have to go off are those supplied by AMD as part of their big reveal, and this only covered the performance of both cards at 4K resolution. This is however the display which both cards are engineered towards catering for, and the theatre in which Nvidia was expected to dominate. If you aren’t interested in operating a 4K display (or multiple 1440p screens) then we’d recommend considering the 6800 XT or the RTX 3080 (if you can get hold of either).
Tellingly, AMD only showed FPS performance in a variety of games with Rage Mode and Smart Access Memory turned on. We know that in Forza Horizon 4, the game in which they outperformed the RTX 3090 the most, both these features switched on contributed an additional 13% FPS uplift, meaning without either we could expect to see 150 FPS or thereabouts in this particular game. This is still above the RTX 3090 judging by the bar chart (assuming it is to scale) but this was the game in which AMD outperformed the most. For the other games, although the FPS increase will almost certainly vary from game to game, we could expect the RTX 3090 to pull ahead.
It may turn out that the biggest factor deciding whether people go with AMD or Nvidia is how easily they can buy either card. In theory, the earlier release of the 3090 RTX should have given it a head start in hoovering up the potential customers for cards in this price bracket, particularly as the target market of people willing to shell out for the premium of enthusiast cards is a lot smaller than that for the lower entries in both ranges. However, Nvidia’s 30 series release has been plagued by stock issues, as has been well documented, and it doesn’t look like this is likely to change before Spring 2021 according to Nvidia’s announcements on the subject, which may be the break AMD needs to convert people over from the green team to the red team.
Nevertheless, questions abound as to whether or not people looking to buy AMD’s 6000 series will face the same issues. It may be that AMD has higher stock built up for the initial launch, but given the unprecedented demand we are seeing this year for GPUs, our money is that stocks will soon run low.
Ultimately whether you decide to go with AMD’s flagship card or Nvidia’s depends on your circumstances. If you are looking to put together a new build and are considering buying a new 5000 series Ryzen anyway, then the performance boost that Smart Access Memory and Rage Mode will give you should be very tempting – offering comparable performance to the Nvidia RTX 3090 for $500 less. If you already have an Intel processor or an older AMD CPU and do not intend to upgrade then the decision is slightly less clear. Nvidia’s RTX 3090 would undoubtedly be the better performer, particularly if you are looking to play games where Ray Tracing and DLSS are supported, and if spending the extra $500 would be worth it for you to have to best of the best performance, then Nvidia’s flagship would seem to be the one for you. The caveats to all of the above are twofold: firstly, if you can wait for independent benchmarking to come out for the new AMD GPU (after December 8th for the 6090 XT) before making your choice then we highly recommend you do so, secondly whether or not you can actually buy either card from stockists in time for your needs may end up being the deciding factor between the two.