Did you look at the slides or presentation at all? Unless AMD is lying, they squarely beat the 3060 Ti in all but one game shown. I'm sure there are others that it'll lose to a 3060 Ti as well, because some games are just better optimized or run better on a particular architecture than another. Of the 8 games listed the 6700XT beats the 3070 in 5 games. I think it's clear why it's priced where it is.So we went from 2016 230USD midrange rx480 to 2019 400USD midrange 5700XT.And now we have 480USD 2021 6700XT "midrange".
6700XT will be by far worst card in lineup.It will not by faster than 3070.Probably it will be average at 3060TI level and without DLSS and RT performance.
Quite a few of these are obviously cherry picked, but the performance in Cyberpunk is a good sign. Navi22 is really fast for its size when you clock it high enough.
Yeah and they have there games that benefits from SAM more than 10% lolApparently, the benchmarks were with SAM enabled on the 6700 XT. Scummy of AMD not to point that out on the slides like they did with the other RDNA2 cards.
I was just saying that quite a few of the games were ones where AMD cards did well, the performance still looked fine when you looked at games like Cyberpunk and I did think it would be closer to the 3070.You might have a point if you showed that AMD lied about the results of any particular game, but you can probably get a 10% swing in either direction if your set of games includes enough outliers. A few other reviews concluded that the 6800XT was a few percent ahead of the 3080 on average, just because their set of test games leaned that direction.
AMD's results from their own slides show the cards about equal and several games within a margin of error most likely, but having even just one title that's AMD favored by a fair bit makes the average look a lot bigger than it really is.
The 6700XT is going to be much closer to the 3070 than to the 3060 Ti when your post suggested otherwise. The inclusion of the Cyperpunk results should tell you that even in games that run better on Nvidia hardware, AMD has a card that's closer to the 3070 than it is to the 3060 Ti.
¯\_(ツ)_/¯Apparently, the benchmarks were with SAM enabled on the 6700 XT. Scummy of AMD not to point that out on the slides like they did with the other RDNA2 cards.
The TPU test suite has a lot of DX11 games that have poor performance on AMD hardware at 1080p and 1440p due to driver overhead. Obviously this is still valid but it is not that reoevant for future games where the bulk will be DX12 or Vulcan.View attachment 40504
reality:
![]()
Its always PR best case scenario.You need calculate -10% from that and you get average neutral performance.So maybe 6700xt will be 5% faster than 3060TI.
We went from AMD/RTG desperately trying to survive to competing.So we went from 2016 230USD midrange rx480 to 2019 400USD midrange 5700XT.And now we have 480USD 2021 6700XT "midrange".
ReBAR is only enabled for RTX3060 + RTX 30XX laptops at the moment, so those comparisons in slides aren't valid for when actual consumers can get these cards, as both OEMs should have these features enabled by then.I'm also fairly sure that Nvidia has ReBAR (AMD just uses SAM for this feature) support as well now (a few of the 3060 reviews mentioned it at any rate) so unless they didn't use it for the Nvidia cards I don't think it's unfair.
Radeon VII was probably the biggest GPU turd of the past decade and anything will compare favorably against it.Radeon Vega 7 = 331mm2 @ TSMC 7nm , 16GB HBM2 , 300W TDP , MSRP $699
Radeon RX6700XT (NAVI 22) = 335mm2 @ TSMC 7nm , 12GB GDDR6, 230W , MSRP $479
At the same die size and same 7nm process , 26% faster with 70W less TDP and lower MSRP.
For reference ,
RTX3070 (GA104) = 392mm2 @ SS 8nm, 8GB GDDR6, 220W TDP , MSRP $499
Useless comparison in both cases as there are new functionalities which end up eating trasistor budget.Radeon VII was probably the biggest GPU turd of the past decade and anything will compare favorably against it.
Since it's basically the same performance as a 5700 XT you'd be better off comparing it against that. Of course a 33% larger die on the same process for 26% more performance and a 20% higher price doesn't look quite as good.
It's as useless (or useful) as any other comparison. I just think that a comparison against Radeon VII is a particularly bad one because it was an awful card, at least from a gaming perspective. That the 5700 XT basically equaled it easily shows as much. That's the card that the 6700 XT ultimately replaces in some respects, the main one being that it lines up against it based on name. As a consumer, I ultimately don't care about transistors anyway because I'm buying performance, not a bigger/smaller number of transistors.Useless comparison in both cases as there are new functionalities which end up eating trasistor budget.
Also only supports 8 games, not all of them like AMD's implementation does.ReBAR is only enabled for RTX3060 + RTX 30XX laptops at the moment, so those comparisons in slides aren't valid for when actual consumers can get these cards, as both OEMs should have these features enabled by then.
You are also buying 4 gbyte more VRAM, Ray tracing support, Variable rate shading support, Mesh shader support, better video processor, and so on and so on. Things that for yesterday's games mean nothing, and mean a lot for future games.It's as useless (or useful) as any other comparison. I just think that a comparison against Radeon VII is a particularly bad one because it was an awful card, at least from a gaming perspective. That the 5700 XT basically equaled it easily shows as much. That's the card that the 6700 XT ultimately replaces in some respects, the main one being that it lines up against it based on name. As a consumer, I ultimately don't care about transistors anyway because I'm buying performance, not a bigger/smaller number of transistors.
I think the more depressing fact is that the 5700 XT is going for $600 - $800 online right now. That really means that the 6700 XT reference card probably has a market price floor of $700 - $800 dollars.
I'm assuming they'll get there eventually with broader title support, and early testing (Guru3D SAM testing in 6900 XT review) showed that the gains were hit and miss and outside of one title, largely insubstantial, but I'm not overly worried about it. On some level it would be like complaining that a review shouldn't include DLSS results for the Nvidia card because AMD doesn't have their own equivalent capabilities yet. I don't think that argument stands up well either.Also only supports 8 games, not all of them like AMD's implementation does.
That's true whether you compare the 6700 XT to the 5700 XT or the Radeon VII though. Well, against the Radeon VII it's 4 GB less VRAM and it's a different type of memory, but I doubt it makes a practical difference. I'm not sure how much of a difference any features will ultimately make. Maybe they're really big and add a lot, or maybe they're like a lot of features that Vega was supposed to have which didn't really pan out. We can only really quantify their importance properly in hindsight.You are also buying 4 gbyte more VRAM, Ray tracing support, Variable rate shading support, Mesh shader support, better video processor, and so on ans so on. Things that for yesterday's gmaes man nothing, and mean a lot for future games.
Thread starter | Similar threads | Forum | Replies | Date |
---|---|---|---|---|
![]() |
Question RTX 4000/RX 7000 price speculation thread | Graphics Cards | 501 | |
R | Question vega vs RDNA2 iGPUs | Graphics Cards | 9 |
Similar threads |
---|
Question RTX 4000/RX 7000 price speculation thread |
Question vega vs RDNA2 iGPUs |