AMD Radeon "Where Gaming Begins Ep. 3 Liveblog:" Radeon RX 6700 XT Announcement

[btarunr]

AMD Radeon Technologies Group today is writing the third chapter in its return to competitiveness across the board, with the "Where Gaming Begins: Episode 3" online media event, which we are live-blogging here. We expect AMD to announce its much awaited Radeon RX 6700 XT graphics card that competes against NVIDIA's GeForce RTX 30-series performance-segment; as well as anticipate updates to the software feature-set of Radeon, and possible sneak-peaks to the future of the Radeon RX 6000 series and RDNA2 graphics architecture on the PC, including its mobile debut. Join us for more!



15:44 UTC: 15 minutes to go, all we're seeing is a slick render of what's possibly the RX 6700 XT reference-design board.

15:50 UTC: Scott Herkelman, GM and VP of AMD's Graphics Business, will take centerstage ten minutes from now.

16:00 UTC: Almost time:


16:01 UTC: Hello "Navi 22"


16:03 UTC: That box resembles a lot like leaks from months ago:


16:03 UTC: Build up to FidelityFX Supersampling?


16:05 UTC: Ryzen 3000 series supports R-BAR and Smart Access Memory:


16:06 UTC: The RX 6700 XT:


16:07 UTC: RX 6700 XT Specs:


16:07 UTC: 1440p gaming is the target for this card


16:08 UTC: The card is heavily noise-optimized, with 0 dBA idle:


16:09 UTC: 11-phase VRM, 8+6-pin power config:


16:09 UTC: DisplayPort 1.4 + HDMI 2.1

16:09 UTC: Beats RTX 2080 Super! Meaning AMD claims it beats RTX 3070!!


16:11 UTC: 12 GB matters:


16:11 UTC: RTX 3070 and RTX 3060 Ti bested: AMD


16:12 UTC: RE Village will be AMD optimized with raytracing and FidelityFX

16:13 UTC: Radeon Boost now supports VRS in DirectX 12 titles.

16:14 UTC: RX 6700 XT suitable for high-refresh rate gaming.


16:16 UTC: Mercedes AMG Petronas switches to an all-AMD machine:


16:16 UTC: Both reference and custom cards on day-one. Pricing and availability:


16:17 UTC: AMD teases Mobile RDNA2 family:


16:18 UTC: That's all, folks! It looks like the RX 6700 XT will turn up the heat against the RTX 3070, given its pricing/performance claims. We can't wait to get our hands on one!

View at TechPowerUp Main Site

 

[ZoneDymo]

I fail to see the point when nobody can get anything.

My friend ordered a rtx3060ti because it was finally time to upgrade from the hd7950....and Jan bece feb...Feb became march...March became may...where is April you might ask? Exactly!

 

[Raendor]

I fail to see the point when nobody can get anything.

My friend ordered a rtx3060ti because it was finally time to upgrade from the hd7950....and Jan bece feb...Feb became march...March became may...where is April you might ask? Exactly!

Ehm, between those two?

 

[phanbuey]

Maybe they can demo one of the 15 cards that they made total.

 

[ZoneDymo]

Ehm, between those two?

The point is that the company has so little faith in getting anything that they just decided to move the date by two months instead of 1

 

[Raendor]

The point is that the company has so little faith in getting anything that they just decided to move the date by two months instead of 1

Expectations management

 

[Xuper]

Isn't cheaper than 3070 ?

 

[dirtyferret]

Maybe they can demo one of the 15 cards that they made total.

today? they made one for this showing and maybe 15 more when it comes time to send them out for reviews

" much awaited Radeon RX 6700 XT graphics card that competes against NVIDIA's GeForce RTX 30-series performance-segment;"

Kind of like when comic book fans argue over which made up super heroes are more powerful?

 

[kayjay010101]

Beating the 3070 isn't that impressive considering it's only $20 less MSRP and doesn't have DLSS.
But good on ya, AMD. Not a bad card overall, all things considered. Now to see if they'll actually make more than a couple dozen.

 

[Zubasa]

Isn't cheaper than 3070 ?

It is slightly cheaper in theory, but of course currently everything goes for price.

 

[sepheronx]

They didn't say the price though this time. I think this is obvious after the fiasco with the 6800/xt launch and their claims of MSRP price.

 

[phanbuey]

DLSS has been the only way I can play cyberpunk - them bringing it to more games is a huge selling point as, especially now with cards being out of stock, it gives you a way to stretch the performance.

 

[Chomiq]

They didn't say the price though this time. I think this is obvious after the fiasco with the 6800/xt launch and their claims of MSRP price.

AMD is pricing the Radeon RX 6700 XT at USD $479 (MSRP)

 

[sepheronx]

yeah, I saw that later.

We all know that price is nonsense.

 

[Chrispy_]

Gamers aren't going to buy this because:

1) There won't be any in stock for the next 6 months.
2) It doesn't have DLSS
3) AMD's performance claims for RDNA2 were pretty cherry picked and included SAM and Rage Mode for the 6800/6900 launch. Real reviews showed performance ~10% slower.
4) The real-world competition for this has an MSRP of $399 and still includes desirable features that AMD haven't bothered addressing yet.

The miners will buy it though, and 95% of all 6700XT cards will go direct from distributors to miners, completely bypassing retail channels anyway....

 

[Anymal]

Where mining begins, AMD!

 

[evernessince]

DLSS has been the only way I can play cyberpunk - them bringing it to more games is a huge selling point as, especially now with cards being out of stock, it gives you a way to stretch the performance.

CP2077 has an option for AMD CAS with built-in resolution scaling.

I use it with my 1080 Ti and it works pretty good. It allows me to get good frames at 1440p.

I don't see DLSS becoming good until it just works in every game.

Where mining begins, AMD!

Either vendor really but Ampere is a more compute focused uArch than RDNA2 and is better for mining. You get 86.6 MH/s with the 3080 vs 59.2 MH/s on the 6800. The only AMD card that gets closer is the Vega VII, which is really not a surprise.

 

[phanbuey]

CP2077 has an option for AMD CAS with built-in resolution scaling.

I use it with my 1080 Ti and it works pretty good. It allows me to get good frames at 1440p.

I don't see DLSS becoming good until it just works in every game.

As a buyer your option is - buy card A that is about the same performance as card B and roughly the same price but is 40% faster in a few AAA games (with many more on the horizon), or buy card B. There's just no downside to that equation to go with card A.

 

[evernessince]

As a buyer your option is - buy card A that is about the same performance as card B and roughly the same price but is 40% faster in a few AAA games (with many more on the horizon), or buy card B. There's just no downside to that equation to go with card A.

That's an extremely misleading choice and yes there are downsides.

You aren't comparing two graphics cards with all the same options enabled, you are comparing them with one having DLSS enabled and the other without. That's not an apples to apples comparison and it's straight up misleading to anyone reading those numbers. If settings aren't the same, it's not a valid comparison, period. You'll find no professional reviews comparing two cards of mixed settings as part of the card's overall performance.

You are also ignoring the visual artifacts introduced by DLSS on top of that which again makes it apples to oranges. Graphics card A will give you more FPS at the cost of visual quality. Flickering, Ghosting, over-sharpening artifacts (DLSS does apply a light sharpen), loss of detail, ect. In addition there are certain restrictions you need to be aware of. Modding in particular posses a problem as the AI is only trained on the launch game. Any user added content or shaders has the potential to cause issues with DLSS.

CP2077 is the first game I've played that has supported DLSS and it supports AMD CAS + Dynamic / Static resolution anyways so really your point seems extremely moot. I'm getting performance equal to as if I had DLSS enabled anyways no thanks to Nvidia.

 

[Chrispy_]

As a buyer your option is - buy card A that is about the same performance as card B and roughly the same price but is 40% faster in a few AAA games (with many more on the horizon), or buy card B. There's just no downside to that equation to go with card A.

Meanwhile, the downside to card B is that you cannot run any CUDA app or RTX Voice, the video encoder quality is nasty, and game devs will develop and test on card A first and foremost because Team A is pumping way more money into developers in addition to making a bunch of free tools and libraries that leverage card A's features. Card B relies on Microsoft doing all the work and hoping that the Khronos Group and Open Source community pick up the slack.

I want good competition but AMD are still a long way behind Nvidia when you look at the complete package. Raster performance is a significant slice of the overall picture, but Nvidia are offering so much more than just good raster performance. I hate Nvidia, they are the dirtier, greedier company that does less for us consumers than AMD, but I can't ignore plain fact.

 

[evernessince]

Meanwhile, the downside to card B is that you cannot run any CUDA app or RTX Voice, the video encoder quality is nasty, and game devs will develop and test on card A first and foremost because Team A is pumping way more money into developers in addition to making a bunch of free tools and libraries that leverage card A's features. Card B relies on Microsoft doing all the work and hoping that the Khronos Group and Open Source community pick up the slack.

I want good competition but AMD are still a long way behind Nvidia when you look at the complete package. Raster performance is a significant slice of the overall picture, but Nvidia are offering so much more than just good raster performance. I hate Nvidia, they are the dirtier, greedier company that does less for us consumers than AMD, but I can't ignore plain fact.

RTX voice is CUDA so yeah, you stated the same thing twice. Remember when Nvidia lied about RTX voice being AI based? It's not, it uses CUDA and can be ran on older Nvidia cards.

It's really depends when you say "can't run any CUDA app" as well. Many apps can use more than one coding path so it isn't a big deal. You could maybe cherrypick specific examples but unless you are a professional with a specific workload this is a non-factor. It doesn't matter for the vast majority of people, you are talking a small subsection of the professional market. For gamers it's irrelevant.

"Free tools and Libraries"

I think you got confused in your own comparison. You can't even use all of Nvidia's features on it's own cards. AMD is the one that allows everyone to use their libraries and tools. The fact that I'm using CAS on an Nvidia card is proof enough of that. DLSS 1.0 ran on CUDA just fine, Nvidia just decided to not allow that anymore for 2.0. Whether they need turing cores for 2.0 is a question given that Nvidia have lied about DLSS 1.0 requiring turing cores in the past before. Of course, they could have just pulled a PhysX again and removed the CUDA codepath for DLSS from 2.0 simply to spite older Nvidia card owners. Dang, this just reminds me how much Nvidia lies. All the time apparently.

The only valid point in your entire comment is about the encoder. Nvidia's encoder is better. The whole "Nvidia's whole package is better" is a fallacy though. People used the same BS back in the day with PhysX to justify buying an Nvidia card and it was absolute trash. I remember enabling PhysX and HairFX on my 970 in Borderlands 2 and TW3 respectively and it always came with stuttering or performance issues. The only thing those techs were good for is convincing dopes to buy Nvidia all the while making them look good in benchmarks. Some things never change.

 

[phanbuey]

That's an extremely misleading choice and yes there are downsides.

You aren't comparing two graphics cards with all the same options enabled, you are comparing them with one having DLSS enabled and the other without. That's not an apples to apples comparison and it's straight up misleading to anyone reading those numbers. If settings aren't the same, it's not a valid comparison, period. You'll find no professional reviews comparing two cards of mixed settings as part of the card's overall performance.

You are also ignoring the visual artifacts introduced by DLSS on top of that which again makes it apples to oranges. Graphics card A will give you more FPS at the cost of visual quality. Flickering, Ghosting, over-sharpening artifacts (DLSS does apply a light sharpen), loss of detail, ect. In addition there are certain restrictions you need to be aware of. Modding in particular posses a problem as the AI is only trained on the launch game. Any user added content or shaders has the potential to cause issues with DLSS.

CP2077 is the first game I've played that has supported DLSS and it supports AMD CAS + Dynamic / Static resolution anyways so really your point seems extremely moot. I'm getting performance equal to as if I had DLSS enabled anyways no thanks to Nvidia.

You're not, I can enable both, and there is a massive difference... The visual artifacts argument against DLSS 2.0 is not something I find to be true while actually using it - if they are there then they are so hard to see that I actually don't notice them... meanwhile CAS looks like you just smeared vaseline on your screen - it's fine but it's not even close visually.

It's not about apples to oranges... it's that if you don't like it - turn it off, then you have the same performance as the other card... if you don't like DLSS don't use it... then it makes no difference what card you get. One side gives you options while the other simply doesn't have that feature. If you get a 3060ti and you want to use CAS (you won't) then go for it... nothing will stop you.

So when you're making a purchasing decision, other than you don't like one company personally, there is no reason to really limit yourself by getting a card with fewer features.

 

[mechtech]

Yep, another hard pass, $480US so about $624 cnd, then another 13% sales tax, so $705 cnd ball park.

That's ok, I will keep on waiting. Just prior to covid an RX 580 8GB was about $250 cnd, when we get something say double the performance at say $350 cnd, I will open my wallet, not before then.

 

[Chrispy_]

Yep, another hard pass, $480US so about $624 cnd, then another 13% sales tax, so $705 cnd ball park.

That's ok, I will keep on waiting. Just prior to covid an RX 580 8GB was about $250 cnd, when we get something say double the performance at say $350 cnd, I will open my wallet, not before then.

That may never come, unless you're talking about ex-mining cards dumped on ebay when the bubble bursts.

It's more than just COVID, Taiwanese drought, and Crypto mining though: Taiwan was exempt from the US-China trade war tariffs until December 31st last year, now they are not - and just like Brexit the price hike is bigger than just the tariff itself because it brings extra delays, inspections, administration - all of which adds cost to perform, and takes additional cost if you want to expedite that process.

Even when COVID's impact on manufacturing is 100% gone, and the droughts in Taiwan are over, and there is enough supply to go around because the miners aren't interested any longer - things will still be a bare minimum of 25% more expensive before tax and inflation are even considered, unless Joe Biden and Xi Jinping can come to an agreement.

It's not even as simple as the flat 25% price hike either, since some of the components that go onto those PCBs are from US-owned companies as well, and also subject to 25% tariffs. What if your GPU uses, say, Texas Instruments components that come from the US and get imported to Gigabyte? 25% tariff. Do you suck up that tariff or do you invest time and costs into finding a different solution using alternative Taiwanese parts instead? That's just one component in one direction at one stage of manufacture. Where do Texas instruments get their rare-earth metals from? Probably China, paying the 25% export tariff.

The "25%" tariff is just the raw value for a single jump across the divide. Everyone is making everything more expensive on both sides of the trade war and it's having a multiplicative effect. The more complex the end product the greater the likelihood that subcomponents have involved one or more additional US-China boundary hops :\

Thanks Donald, you 'helped' :\

 

[evernessince]

You're not, I can enable both, and there is a massive difference... The visual artifacts argument against DLSS 2.0 is not something I find to be true while actually using it - if they are there then they are so hard to see that I actually don't notice them... meanwhile CAS looks like you just smeared vaseline on your screen - it's fine but it's not even close visually.

It's not about apples to oranges... it's that if you don't like it - turn it off, then you have the same performance as the other card... if you don't like DLSS don't use it... then it makes no difference what card you get. One side gives you options while the other simply doesn't have that feature. If you get a 3060ti and you want to use CAS (you won't) then go for it... nothing will stop you.

So when you're making a purchasing decision, other than you don't like one company personally, there is no reason to really limit yourself by getting a card with fewer features.

I forget to mention earlier that DLSS 2.0's performance increase isn't anywhere near 40% on average either. In some tiles like Metro Exodus it's 4% and in other's 16% (Tomb raider).

I can get a slightly higher performance bump from CAS with the resolution scale set to 80% with minimal visual quality loss in CP2077. At 50% resolution scale it does look like vaseline but at 80% there's hardly any difference. DLSS is a good technology but it's in too few games and you can't make blanket statements about how much performance you'll get because it varies a lot based on the title and resolution and it does poorly at 1080p. There are a lot of caveats. CAS might not be as good compared to an ideal scenario for DLSS but it just works and it does so across a broad range of video cards.

"It's not about apples to oranges... it's that if you don't like it"

That's sidestepping the question. It IS apples to oranges and it is never acceptable to compare two video cards with different settings. That's benchmarking 101, you ensure the only variable that's different is the card you are testing. Failing to isolate means your data is worthless.