AMD Radeon 6800 XT vs 6900 XT ‘Big Navi’

Update: Check out our Where to Buy pages to get your hands on the latest GPUs from AMD:

• Where to buy 6900XT
• Where to buy 6800 Series
• Where to buy 6700 Series

So the big reveal for the AMD 6000 series graphics cards has come and gone and the general reaction from the tech community has been: we’re impressed, still slightly sceptical until we can get our hands on it, but quietly hopeful. We imagine the reaction of people over at Nvidia was somewhat less positive. “Rage Mode” and “Smart Access Memory” are two features in particular that have got everyone talking. Have a quick read on our rundown of these two features here.

Now we have some data we can compare the two heavy hitters of the 6000 range: the 6800 XT and the 6900 XT aka “Big Navi”. Specification wise there is not a great deal of difference between the two cards, but we will be examining the improvements AMD has made since the previous generation.

The main focus here will be the difference in FPS performance. There has been some talk of whether the FPS AMD showed represents an average FPS or a maximum FPS as it does not appear that this was explicitly stated in the presentation. For the sake of both the consumer and AMD we hope that the figures are indeed average FPS, otherwise they can expect a backlash once 3rd party benchmarks are released. For this reason we imagine these concerns are unfounded, however, as ever, take the official FPS figures with a pinch of salt. Our advice is to always wait for independent benchmarking before making a decision on which card is right for you. We will keep you updated with all the latest leaks as they happen.

Now, without further ado, let’s dive right in.

As earlier leaks suggested, the specification list for both the 6800 XT and the 6900 XT is remarkably similar, it is only in the key area of “compute units” that the major difference between the two cards lies.

As earlier leaks suggested, the specification list for both the 6800 XT and the 6900 XT is remarkably similar, it is only in the key area of “compute units” that the major difference between the two cards lies.

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 of a graphics card, 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. Both the 6800 XT and the 6900 XT come with an impressive 16GB GDDR6 VRAM, 16 Gbps memory speed, and 256-bit memory bus.

Although this would mean both AMD cards have a higher VRAM than Nvidia’s 3080 RTX at 10 GB GDDR6, they come in lower compared to its 19 Gbps memory speed and 320-bit memory bus. Additionally, both fall below the green team’s flagship 3090 RTX in all three of these areas, as it comes with 24 GB VRAM, 19.5 Gbps memory speed, and 384-bit memory bus.

Typically this would suggest that this means the 3090 RTX will still likely be the winner in terms of 4K performance compared to the most powerful AMD card however AMD claim to have a trick up their sleeve. Based on technology developed for their Ryzen CPUs, AMD’s “Infinity Cache” apparently allows the CPU 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 it can “apparently” outperform the Nvidia cards in this area using a smaller bus, and therefore saving on power consumption and heat generated. Regardless, both of the highest end AMD cards have the same memory specifications as each other.

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 of a graphics card, 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. Both the 6800 XT and the 6900 XT come with an impressive 16GB GDDR6 VRAM, 16 Gbps memory speed, and 256-bit memory bus.

Although this would mean both AMD cards have a higher VRAM than Nvidia’s 3080 RTX at 10 GB GDDR6, they come in lower compared to its 19 Gbps memory speed and 320-bit memory bus. Additionally, both fall below the green team’s flagship 3090 RTX in all three of these areas, as it comes with 24 GB VRAM, 19.5 Gbps memory speed, and 384-bit memory bus.

Typically this would suggest that this means the 3090 RTX will still likely be the winner in terms of 4K performance compared to the most powerful AMD card however AMD claim to have a trick up their sleeve. Based on technology developed for their Ryzen CPUs, AMD’s “Infinity Cache” apparently allows the CPU 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 it can “apparently” outperform the Nvidia cards in this area using a smaller bus, and therefore saving on power consumption and heat generated. Regardless, both of the highest end AMD cards have the same memory specifications as each other.

In a similar fashion to its Ryzen processors, AMD have managed to squeeze out great efficiencies from their latest generation of graphics cards. Part of this is due to the aforementioned use of the Infinity cache allowing for a smaller memory bus. In addition to this the improvements made to the manufacture of their compute units and other developments have led to a stated uplift of 54% performance-per-watt from the previous generation.

Now, with the claimed increase in performance between the 6800 XT and the 6900 XT (more on that later) you would expect a simultaneous increase in power consumption and with it, considering both GPUs use the same cooling solution, a higher output of heat. Indeed the lesser of the two cards, the 6800 XT comes in with a recommended PSU of 750W whereas the 6900 XT will require 850W. However, through some as yet unknown means of arcane science, both cards have the same “TDP” (Thermal Design Power). Whilst TDP is a questionable metric that cannot really be used when comparing AMD to Nvidia, between cards of the same series it generally is an accurate representation, so this identical figure between both cards is intriguing. Regardless, it is recommended to wait until independent testers have used both components themselves to see if this is truly the case, or if the 6900 XT will require an upgraded cooling solution compared to the 6800 XT.

In a similar fashion to its Ryzen processors, AMD have managed to squeeze out great efficiencies from their latest generation of graphics cards. Part of this is due to the aforementioned use of the Infinity cache allowing for a smaller memory bus. In addition to this the improvements made to the manufacture of their compute units and other developments have led to a stated uplift of 54% performance-per-watt from the previous generation.

Now, with the claimed increase in performance between the 6800 XT and the 6900 XT (more on that later) you would expect a simultaneous increase in power consumption and with it, considering both GPUs use the same cooling solution, a higher output of heat. Indeed the lesser of the two cards, the 6800 XT comes in with a recommended PSU of 750W whereas the 6900 XT will require 850W. However, through some as yet unknown means of arcane science, both cards have the same “TDP” (Thermal Design Power). Whilst TDP is a questionable metric that cannot really be used when comparing AMD to Nvidia, between cards of the same series it generally is an accurate representation, so this identical figure between both cards is intriguing. Regardless, it is recommended to wait until independent testers have used both components themselves to see if this is truly the case, or if the 6900 XT will require an upgraded cooling solution compared to the 6800 XT.

Despite leaks suggesting that the Big Navi card would come in at a higher gaming clock of 2,040MHz and a boost clock of 2,330MHz, both the 6900XT and the 6800XT actually have a gaming clock of 2,015 MHz and a boost clock of 2,250 MHz. AMD did not give us the traditional ‘Base’ clock speed measurement. In a similar way to how Boost speed can be a fairly technical measurement that rarely represents the frequency a GPU will reach when completing the majority of processes, ‘Base’ clock is an extremely low rate that even an idle computer will rarely hit. “Gaming” clock is AMD’s way of trying to give an indication of speeds the GPU will expect to reach during typical gameplay – obviously until third parties can test this it’s somewhat up in the air as to how this will translate in reality for most people.

Similarly how long either card will be able to maintain the boost clock is unknown until we see further testing data. Some cards of the previous AMD generation rarely achieved boost speeds and failed to be able to maintain them for any length of time to make a difference to performance. Hopefully the boost speeds of both new cards are not just marketing technicalities.

The number of Compute Units is the one area where the specifications of the two cards differ from each other, and from 72 cores in the 6800 RX to 80 cores in the 6900 RX it is a significant increase (over 11%). It is from here that the additional power of the more expensive card is derived.

As far as we are aware both cards will come with identical software. The aforementioned Rage Mode and Smart Access Memory will be features available for owners of both cards who also have the latest Ryzen 5000 series processors and an AM4 motherboard. Both GPUs will make use of the new “Radeon Boost” and “Radeon Anti-Lag” technologies to increase response time and reduce latency. Both cards will also be fully compliant with “DirectX 12 Ultimate”: this means compliant means they will support mesh shading, variable-rate shading, and sampler feedback. Although DLSS (Deep Learning Super Sampling) will remain an Nvidia exclusive technology, AMD revealed that they have developed their own ray-tracing technology, more details of which will emerge in the near future, and you can expect both cards to take advantage of this feature, though how it will compare to Nvidia’s own implementation remains to be seen.

At the end of the day, regardless of the technical similarities of both cards, what people are interested in is performance. At this stage we only have AMD’s own benchmarking, done on an FPS basis. Expect further technical benchmarking by third parties to emerge over the next month.

Unusually, AMD did not show 1080p performance for any of the cards revealed on October 28th, using 1440p as its minimum resolution. Whether this is an attempt to hide underperformance at this resolution versus Nvidia’s offerings, or simply a recognition that the next generation of GPUs will push the normal baseline gaming screen closer towards 1440p over the next year, only time (and independent 3rd party benchmarking) will tell.   

Below we can see the 1440p performance in FPS terms for the RX 6800 XT on a number of games, according to AMD. The $50 more expensive Nvidia RTX 3080 was used as a comparison. According to the end notes, both cards were tested using a Ryzen 5900 X with an AM4 motherboard.

Although we would like to give you a comparison of this performance alongside that of the 6900 XT playing the same games at 1440p, we were only shown FPS performance for the RX 6900 XT at 4K resolution; we will have to wait for the first independent benchmarks to come through before we can get an idea of this. 

Ultimately though this does reinforce the difference in intended purpose for the two GPUs – the 6900 XT, like its rival Nvidia RTX 3090, is designed with 4K use in mind – if you are a gamer playing on 1440p or 1080p, certainly using only a single screen, then at $350 more than the 6800 XT it would seem like an unnecessary extravagance. If you are thinking of operating two or more screens at 1440p then quite possibly the 6900 XT might prove to be worth it, however other factors come into play here: sampling and super sampling efficiencies determine how well things perform over multiple screens, not just the raw specifications, and this can change from game-to-game, application to application. 

We can safely assume that the 6900 XT will outperform the 6800 XT at 1440p and 1080p, given its higher number of Compute Units but by what margin is the $350 question. If previous releases by both AMD and Nvidia are any guide, the uplift between the Big Navi and its smaller brethren will be less significant the lower down the resolution scale you go – so you will likely get less of a difference in performance on 1440p than you will on 4k, and even less on 1080p than you will on 1440p.

Unusually, AMD did not show 1080p performance for any of the cards revealed on October 28th, using 1440p as its minimum resolution. Whether this is an attempt to hide underperformance at this resolution versus Nvidia’s offerings, or simply a recognition that the next generation of GPUs will push the normal baseline gaming screen closer towards 1440p over the next year, only time (and independent 3rd party benchmarking) will tell.   

Below we can see the 1440p performance in FPS terms for the RX 6800 XT on a number of games, according to AMD. The $50 more expensive Nvidia RTX 3080 was used as a comparison. According to the end notes, both cards were tested using a Ryzen 5900 X with an AM4 motherboard.

Although we would like to give you a comparison of this performance alongside that of the 6900 XT playing the same games at 1440p, we were only shown FPS performance for the RX 6900 XT at 4K resolution; we will have to wait for the first independent benchmarks to come through before we can get an idea of this. 

Ultimately though this does reinforce the difference in intended purpose for the two GPUs – the 6900 XT, like its rival Nvidia RTX 3090, is designed with 4K use in mind – if you are a gamer playing on 1440p or 1080p, certainly using only a single screen, then at $350 more than the 6800 XT it would seem like an unnecessary extravagance. If you are thinking of operating two or more screens at 1440p then quite possibly the 6900 XT might prove to be worth it, however other factors come into play here: sampling and super sampling efficiencies determine how well things perform over multiple screens, not just the raw specifications, and this can change from game-to-game, application to application. 

We can safely assume that the 6900 XT will outperform the 6800 XT at 1440p and 1080p, given its higher number of Compute Units but by what margin is the $350 question. If previous releases by both AMD and Nvidia are any guide, the uplift between the Big Navi and its smaller brethren will be less significant the lower down the resolution scale you go – so you will likely get less of a difference in performance on 1440p than you will on 4k, and even less on 1080p than you will on 1440p.

Whilst 4K performance is not a feature that is of practical interest to the vast majority of gamers in today’s market, it is becoming the new standard for manufacturers to show off the best of what your hardware can do at its limits. This was the area in which it was predicted AMD would struggle against Nvidia, whose Ampere architecture with its high shader count really excels on 4K resolution displays, though if AMD’s results are to be believed it does appear as if they are at the very least holding their own, and at best outperforming, their competitor.

Below are the performance figures that AMD has released for both cards. According to the end notes, both cards were tested whilst using a Ryzen 5900X CPU with an AM4 motherboard. Unfortunately AMD did not show the 6900 XT without Rage Mode or Smart Access Memory enabled (sneaky), which raises the question of whether this indicates its performance would be disappointing. Whatever the reason, we are not able to do a comparison of the performance of the two cards with these features disabled, based on the current information.

Whilst 4K performance is not a feature that is of practical interest to the vast majority of gamers in today’s market, it is becoming the new standard for manufacturers to show off the best of what your hardware can do at its limits. This was the area in which it was predicted AMD would struggle against Nvidia, whose Ampere architecture with its high shader count really excels on 4K resolution displays, though if AMD’s results are to be believed it does appear as if they are at the very least holding their own, and at best outperforming, their competitor.

Below are the performance figures that AMD has released for both cards. According to the end notes, both cards were tested whilst using a Ryzen 5900X CPU with an AM4 motherboard. Unfortunately AMD did not show the 6900 XT without Rage Mode or Smart Access Memory enabled (sneaky), which raises the question of whether this indicates its performance would be disappointing. Whatever the reason, we are not able to do a comparison of the performance of the two cards with these features disabled, based on the current information.

You will note that on the above images the FPS for the RX 6800 XT with Rage Mode and Smart Access Memory enabled is not specifically stated. However with a bit of numeric wizardry (basic mathematics) we can get an FPS comparison graph based on the percentages given, for the six games with comparable data. 

Unsurprisingly the RX 6900 XT performed better of the two, though whether or not these FPS gains are enough to justify the increase in cost is entirely up to you the buyer.

You will note that on the above images the FPS for the RX 6800 XT with Rage Mode and Smart Access Memory enabled is not specifically stated. However with a bit of numeric wizardry (basic mathematics) we can get an FPS comparison graph based on the percentages given, for the six games with comparable data. 

Unsurprisingly the RX 6900 XT performed better of the two, though whether or not these FPS gains are enough to justify the increase in cost is entirely up to you the buyer.

“Big Navi”, like its Nvidia counterpart, is the ultimate enthusiast card, for those who money is no object when it comes to buying a PC, and if (still a big if at this stage) this performance data is to be believed then it has certainly succeeded in its challenge to the 3090 RTX for $500 less. Rage Mode and in particular Smart Access Memory mean to fully realise this gain you have to pair the card with one of the latest Ryzen CPUs, however with the Ryzen 5900X currently priced at $549 that could be a total difference of $49 for what AMD are claiming is superior performance (vs Nvidia GPUs and Intel CPUs). 

When comparing the 6900 RX against 6800 RX though, it becomes slightly more of a personal choice. If your needs are purely for gaming at 1080p or 1440p then it would seem difficult to justify the expenditure of $350 for what will likely prove less of a gain in FPS performance compared to the 4K figures we have access to at the moment. Such money would be better spent elsewhere in full build, perhaps on a new SSD or going towards a new monitor.

For 4K gamers the question becomes: “is the additional FPS gain at this resolution worth my money?” For people thinking about operating multiple 1440p screens, perhaps even two 4K screens for workstation use and gaming, then we would strongly advise waiting for independent testing down the line, as this data just isn’t currently available.

As ever, the more time passes the clearer the picture will be, and we would always advise waiting for independent benchmarking. However if you have to make a purchase right now though, or are perhaps concerned with stock issues and want to snap these cards up before they run out: for those with very deep pockets that must have the best in tech then the AMD Radeon 6900 RX looks to be a stellar card, for everyone else looking to buy an enthusiast level card we’d recommend the AMD Radeon 6800 RX.