3 March 2021

RX 6700 XT announced... but is it really this good?


Today, AMD announced the RX 6700 XT - the card that a majority of people have been looking forward to. While the numbers provided by AMD look really good, I feel like we all need a good dose of skepticism until the proper reviews land. You'll see why in a moment...

Previously, on...

I was pretty lukewarm with my expectations for this card with calculated performance of around an RTX 2080 at 2.3 GHz. I expected it to be cheaper than $400 (MSRP) at that level of performance and without being pushed higher than around 2.3 GHz - due to the reasons stated in my prior articles on the subject. However, that's not the story being told by AMD's slides from its presentation...

Of course, I never delved into the discrepancy between the way I calculated for the Nvidia and AMD comparisons as I only ever focussed on the RTX 2080...

What's that, Pussycat..?

According to AMD, at 1440p, the RX 6700 XT will beat the RTX 2080 Super by about 10 fps (in the five selected titles that AMD chose to display) and will draw or beat the RTX 3070 (in the same five titles plus three more titles).

I think that's a pretty amazing achievement for a piece of hardware that has 66% the computing power of the RX 6800 - a card that ONLY JUST beats the RTX 3070 (it's around 8% more powerful according to TechPowerUp) and yet is only averaging around 200 MHz more in core clock frequency?

Waitaminute... that doesn't add up!

So, AMD are saying that the RX 6700 XT will do 87 fps at 1440p max settings* in Borderlands 3... but an RX 6800 will only manage ~95 fps on various setups. That it will manage 88 fps in Gears 5 when the 6800 will do 103 - 111 fps. In Valhalla it will get 82 fps when a 6800 will get 79 - 81 fps?!
*Not exactly sure what that means!
So what's going on?

These numbers are unbelievable! In a "they're true but not necessarily reflective of actual user experience" kinda way...

The answer is simple - AMD is pulling an Nvidia/Intel and obscuring their benchmarking settings in order to paint their hardware in the best possible light. This is especially shameful because AMD have been particularly good at this compared to both of those other companies when presenting their Ryzen, Threadripper and RX 5700/6800 lineups... but, of course, those products didn't need a leg-up!

It's clear that these comparisons to the RTX 2080 Super and the RTX 3070 are not on equal ground - AMD have clearly implemented SAM on the RX 6700 XT to get these numbers and that would be fine, except for the fact that they do not disclose it! The games that get a big boost from SAM are: Assassin's Creed Valhalla (up to 15 fps for the RX 6800), Hitman 2* (up to 22 fps), Borderlands 3 (up to 13 fps), Dirt 5 (up to 11 fps), Gears 5 (Unknown for 6800 but up to ~15 fps for 6800 XT). In contrast, WDL can gain around 5 fps with SAM... though I couldn't find any SAM benchmarks for Cyberpunk or Cold War to compare against. 
*Hitman 2 and 3's engine's share a lineage and it's likely that the performance gains observed for H2 will carry over to H3. 
Taking those into account and we're now looking at ~70 fps in AC:Valhalla, ~80 fps in Borderlands 3, 61 fps in Dirt 5, ~80 fps in Gears 5, ~110 fps in Hitman 3 and ~58 fps in WDL... and that's assuming that the only undisclosed performance feature is SAM.

Given that AMD decided to pit their RX 6700 XT with SAM enabled against Nvidia cards with no DLSS enabled I have to now also wonder which of these features might also be utilised in obtaining the benchmark numbers in their presentation...

For example, Gears 5, Dirt 5 and Hitman 3 all support VRS and this feature could add another ~5 fps to the performance of each game at 1440p, if Hitman 3 is anything to go by.

If I take both assumed improvements to performance from SAM and VRS away from these games, we return to performance of around an RTX 2080 or slightly above - which is very close to what I had predicted. 

Now, it *is* possible that these assumptions are incorrect and SAM and other FidelityFX features were not enabled - in which case this card is incredibly powerful (as powerful as the RX 6800 in many cases) and will be incredible value for money... but I'm very doubtful - the numbers just don't add up for my liking.

[UPDATE] Hardware Unboxed have confirmed that SAM was enabled in all of those benchmark numbers. (Around 7:55)

The truth will out...

I'm really looking forward to the actual reviews of this card and I think it'll be a very good performer but I am a bit concerned that AMD are trying to justify the price of this card (€479) based on highly specific results in their comparisons with competing cards. In the same way that I would not advise comparing DLSS-enabled games for Nvidia cards in an absolute comparison between Radeon and Geforce hardware, SAM is actually much more limited in its application and user ability than the RTX features are because all RTX cards can use them whereas only specific AMD CPUs and chipsets can currently use SAM.

Yes, AMD have now announced that they're bringing SAM compatibility to Ryzen 3000 series CPUs but they didn't announce which chipsets they will support (or when). X570 and B550 are certain to be a given in that those chipsets are already supported with respect to Ryzen 5000 CPUs but will X470/B450 chipsets also receive that feature? How many consumers can actually take advantage of those extra fps in their games?

Time will set this record straight.


In light of this, the RX 6700 XT is overpriced for its actual performance compared to the competition: with performance of around an RTX 2080, the $400 RTX 3060 Ti is better value. 

Of course,  you can make the argument that discussing MSRP is a moot point and you're right but that misses the underlying problem/point. The reason AMD have done this is because they want a piece of that extra money that AIBs and retailers have been getting, so they're charging the AIBs more for this product because they know the AIBs can and will absorb the price increase without the consumer knowing anything about it. 

At the same time, the move by AMD to mis-represent where in the performance hierarchy their products sit in order to justify higher MSRPs is a terrible move for consumers - all manufacturers can do this and both Intel and Nvidia have features that absolutely trounce and even nullify the SAM advantage that  AMD presented here. 

This makes reviewer independence even more important than ever because AMD is (as I explained previously) pushing the cost of the low end of PC hardware higher and higher and that is bad for consumers. 

In conclusion, the RX 6700 XT is poor value for the price, even beyond this temporary GPU price crisis... buyer beware!

No comments: