Where Gaming Begins - AMD RX6000 series Live Event

[Mark Kaine]

9 minutes ago, Aereldor said:

I hope you can flash a 6800 to a 6800xt like you can do with a 5700 right now. Otherwise it's not that much of a bargain - 18% faster than a 2080 ti means 18% faster than a 3070, but it's also 18% more expensive ($580 vs $500). 

And "available" (hopefully) 

[TrainFan475]

4 hours ago, BigDamn said:

She looks like Carol Stills from The Office

I knew she reminded me of someone...

[Lakobrija]

I just wonder, will their next apus be on rdna2

[tim0901]

7 hours ago, ScratchCat said:

The slide says 300W Total Board Power.

Is this with or without Rage Mode? It's basically overclocking, so it'll almost certainly result in a higher power consumption over stock.

 

Hopefully it's with Rage Mode, otherwise AMD is being a bit deceptive advertising performance that isn't available at the power consumption they're listing. But honestly I wouldn't be surprised if they were. AMD is no stranger to the deceptive marketing game.

[KeltonDSMer]

Does anyone know the HDMI spec of the new cards? I don't remember seeing that in the presentation, and with 4K 120 Hz TVs going way down in price this seems pretty important.

[AlTech]

10 minutes ago, KeltonDSMer said:

Does anyone know the HDMI spec of the new cards? I don't remember seeing that in the presentation, and with 4K 120 Hz TVs going way down in price this seems pretty important.

AMD's website says HDMI 2.1 VRR.

 

https://www.amd.com/en/technologies/rdna-2

[cj09beira]

13 minutes ago, KeltonDSMer said:

Does anyone know the HDMI spec of the new cards? I don't remember seeing that in the presentation, and with 4K 120 Hz TVs going way down in price this seems pretty important.

their site only says hdmi 4k: YES

¯\_(ツ)_/¯

[KeltonDSMer]

4 minutes ago, AluminiumTech said:

AMD's website says HDMI 2.1 VRR.

 

https://www.amd.com/en/technologies/rdna-2

Oh shoot of course I should have seen if they updated the site, thanks. That's a huge relief!

[Warin]

5 hours ago, derr12 said:

a virtual machine that simulates the old rage128 cards.  time to bust out your descent and descent2 floppies folks!

I cant believe that it took until page 9 for someone to make a comment like this. I totally love the shout out to old school ATI.

[AlTech]

1 hour ago, Lakobrija said:

I just wonder, will their next apus be on rdna2

Cezanne (Zen3 APUs) is still Vega 2.0 based according to leaks.

[GDRRiley]

8 minutes ago, AluminiumTech said:

Cezanne (Zen3 APUs) is still Vega 2.0 based according to leaks.

there is suppose to be 3-4 versions of APUS. some using RNDA2 some using VEGA with a mix of zen2 and zen3

[justpoet]

I like how competitive this seems, which gives me hope for future GPU prices.

 

I feel like AMD is going to milk the early adopter train here while NV doesn't have enough stock, netting them more income for future R&D (which they need to compete with NV long term), then when NV drops TI or Super cards, drop pricing to make the tiers change.

 

My assumption based on watching the presentation is that these are all the same GPU die, with a bunch of binning and fusing (much like they did with chiplets in CPUs), so BOM of each card level is probably similar.  That should also make driver work easier in general, so hopefully less issues there, and also prevent flashing something like a 6800 to a 6800xt.  It does, however, make one wonder if there will be another die for lower end cards, such as a 6700 and then 6500 being the binned down version, sometime next year, to perhaps target 1080/1440 vs the 4k main target of these cards.

 

It sounds like the AMD tech such as the cache and SAM are mainly only beneficial at high resolution/texture though.  So it may be quite interesting to see how this scales for fps when dropping settings/resolution vs a similar class NV card.  From the sounds of it, AMD slightly better 4k, similar 1440 (2k), and then probably worse 1080 (HD)…but of course we'll have to wait for real third party reviews and benchmarks to know.

 

RAGE Mode as described by GN isn't overclocking, but rather basically increasing power target/limit and only expecting a 1-2% increase, but on all titles (vs SAM needing title support, but able to do more).

 

RTRT (Real Time Ray Tracing) support is not RTX, but DX12Ultimate, so various other RT implementations won't automatically all work.

 

All in all, I'm happy with the announcement.  If they deliver on the expectations, things will be good.

[niofalpha]

Better than I expected. After the usual 5-20% less performance from the inhouse benchmarks will still be pretty good.

 

And is anyone else questioning the 6800XT to 6900 having the same TDP?

[BlackManINC]

2 minutes ago, niofalpha said:

Better than I expected. After the usual 5-20% less performance from the inhouse benchmarks will still be pretty good.

 

And is anyone else questioning the 6800XT to 6900 having the same TDP?

 You find something odd about that? I mean, the Ryzen 5 3600 and R7 3700X for instance have the same TDP so. 

[genexis_x]

I wanted to see real world testing/benchmarks when reviews are up.

[Ginger Penguin]

9 hours ago, mousesnob said:

I would think that the lack of cuda accel stuff would make that card pretty unattracrive.

That's pretty much my use case, unfortunately. I wonder if the reason they didn't announce any lower-end cards today is because of NVidia's hiding away of the 3060. I could be being a conspiracy theorist but Team Green could have been waiting to see what AMD was going to compete with in the traditionally held "budget" market so they could potentially undercut or match price for higher performance with the 3060 but now there's now 6700XT or lower, they might be forced to show their hand first so AMD can do the undercutting.

[VeganJoy]

dang i was hoping the 6900xt was gonna be a little more reasonably priced, just wanted it for the nice factor. so for the zen 3 launch where they teased rdna 2 specs, they only showed 6800xt numbers? also wasnt there some sorta rumor about a 6900xtx?

[Random_Person1234]

17 minutes ago, VeganJoy said:

dang i was hoping the 6900xt was gonna be a little more reasonably priced, just wanted it for the nice factor. so for the zen 3 launch where they teased rdna 2 specs, they only showed 6800xt numbers? also wasnt there some sorta rumor about a 6900xtx?

I would think the XTX models would be released after Nvidia releases their supers and Tis.

[RejZoR]

1 hour ago, VeganJoy said:

dang i was hoping the 6900xt was gonna be a little more reasonably priced, just wanted it for the nice factor. so for the zen 3 launch where they teased rdna 2 specs, they only showed 6800xt numbers? also wasnt there some sorta rumor about a 6900xtx?

Here we go again. 6900XT (for now at least as there are no reviews yet) trades blows with RTX 3090 at lower board power, a lot more memory and 500€ lower price tag. Yet people are saying things like "I was hoping it'll be more reasonably priced". It's as good as RTX 3090 and WAY cheaper yet people somehow expected it to be even cheaper. Just no. That's not how things work and never have and never will. Stop treating AMD like it's some sort of charity who needs to constantly undercut their own prices just because it's AMD. Sure, I'm a consumer and I'd love to have RX 6900XT for 500€. But I'm also a realist and I know that just can't and won't happen. If AMD has a competitive product, they should be allowed to price it competitively. If not, they'll have to lower the price to compete, because some don't care for 5fps less, but care for 100€ price difference.

 

Radeon cards of the past used brute force approach and it showed. They had all the power on paper, but never really showed that in real world applications. RX 6000 changes that as I can clearly see AMD put real proper effort in designing this thing. Infinity Cache to solve the bandwidth limitations allowing them to make cheaper memory interface (narrower physical bus is cheaper to manufacture) and also use a cheaper more readily available GDDR6 (non X) VRAM memory) which should help greatly in availability as they'll be easier to assemble with more readily available parts opposed to more exotic GDDR6X on RTX 3000 cards. RX 6000 cards seem properly fast as is and I suspect Infinity Cache may hold some "fine wine" performance as AMD tweaks and fiddles with it. Great thing about this is they already deliver and it's not like you're forced to wait for fine wine to kick in with time to even compete. They are already competing and we can see some further uplift through this which can and probably will be nice. Gives me the good old Radeon x1800 vibes with the Ring Bus that allowed ATi back then to pump out some extra performance as they fine tuned it better for games.

 

It's really interesting to observe AMD how massively it has matured in just very few recent years in all key areas. CPU's are highly competitive now and Radeon cards, despite not having the ultimate flagship, realistically, RX 5700 was a very competitive and well thought out design departing from old designs dramatically. And it showed with performance. RDNA2 just elevates that to a whole new level. AMD really put new income from CPU's to a good use and most certainly hired the right people. Or just gave enough resources for their existing engineers to really spread their wings. Because prior to this, they had underwhelming products for years despite all the hype. RDNA based Navi dropped basically from nowhere without much hype and it just blew everyone away.

[leadeater]

1 hour ago, RejZoR said:

6900XT (for now at least as there are no reviews yet) trades blows with RTX 3090 at lower board power, a lot more memory and 500€ lower price tag

3090 has more VRAM.....

[Drama Lama]

The cooler looks like any generic cooler from a board partner. I‘ll be honest Vega had much better looking coolers for the founders edition 

[Mark Kaine]

32 minutes ago, Drama Lama said:

The cooler looks like any generic cooler from a board partner. I‘ll be honest Vega had much better looking coolers for the founders edition 

In a way I like the design,, it's pretty minimalistic, but also a tad bit boring... 

 

 

[Loote]

1 hour ago, Drama Lama said:

The cooler looks like any generic cooler from a board partner. I‘ll be honest Vega had much better looking coolers for the founders edition 

I agree about the looks, but the temps and noise are much more important, I hope  it's going to be a decent performer.

[JediFragger]

7 minutes ago, Loote said:

I agree about the looks, but the temps and noise are much more important, I hope  it's going to be a decent performer.

Heard a figure of 37dB under gaming load somewhere (may have been the presentation itself). If you it seems comparable to FE/AiB models.

[porina]

9 hours ago, justpoet said:

It sounds like the AMD tech such as the cache and SAM are mainly only beneficial at high resolution/texture though.  So it may be quite interesting to see how this scales for fps when dropping settings/resolution vs a similar class NV card.  From the sounds of it, AMD slightly better 4k, similar 1440 (2k), and then probably worse 1080 (HD)…but of course we'll have to wait for real third party reviews and benchmarks to know.

I'm most interested to see how that cache works. Is it "dumb" and only stores more recently loaded data, or is it more selective? For example, if some data is frequently re-used, it makes more sense to keep that on hand than other data that once used can be dropped. I'd guess the opposite in a way, it might show most benefit at lower resolutions than higher. If data sets are big, proportionately less of it will be kept in the cache, so you become more memory bound.

 

SAM to me is a bit less interesting but still welcome. Less overhead access between CPU and GPU is not something I'd turn away, but seeing how PCIe bandwidth makes hardly any difference for current use cases, it feels more like a preparation for future use than something impactful in the near term.

 

7 hours ago, BlackManINC said:

 You find something odd about that? I mean, the Ryzen 5 3600 and R7 3700X for instance have the same TDP so. 

From the given information, the CU count is different, same stated clocks and power. That leaves the only variable being binning of the die. 5900XT must have higher "quality" die than 5800XT. If we assume they were of similar quality, then the clocks would be more different, lower on higher CU unit. It is also one thing to see paper specs, and another to see how they actually run. So running clocks in practice may still differ.