AMD Radeon RDNA2 Navi (RX 6500, 6600, 6700, 6800, 6900 XT)

Well if the rumors of the profit they made with the reference model at MSRP are true lets see how much stock would be.

 

Too bad NV didn't make the cut for the RT consoles then
Once they have ARM they might be in a better position to compete.
Though AMD have been making great progress in CPU and GPU the last few years too.

 

Huh? So AMD finally fights back at the highend and decides to not support OEMs and/or WCing?
This tweet seems to only be telling half the story.

AMD and their reference partner (still Sapphire?) will be continuously building/selling reference boards.
Whether or not other AIBs will offer actual reference cards or decide to build cheaper custom cards at reference specs is up to them.
Usually at least a few of the main/big AIBs will still be offering reference cards to consumers and OEMs.

These same stupid rumors seem to happen every launch and people seem to forget it has been debunked.
https://www.extremetech.com/gaming/296499-no-amd-hasnt-quit-making-reference-5700-and-5700-xt-gpus

Just like, only 10k cards available for launch, or AIB's complain about margins on reference cards.

AIB margins on reference cards are pretty much static and built into the MSRP. While I won't say the AIBs agree with it, it is normal.
Remember Nvidia AIB's crying about the margins on RTX30x0 reference cards or the RTX20x0 FE cards?

628mm2 vs 520mm2
384b/320b vs 256b
custom 8GiB GDDR6x vs 16Gib GDDR6
custom high density PCB vs standard PCB
custom designed cooler vs standard aircooler
MSRP close enough

Edit- All these weird/misleading tweets are coming from small-ish regions/markets.

 

I don't think nvidia is taking an unusually low cut here, rather that AMD is taking an unusually high one.

 

Impressive visuals. I assume this couldn’t be achieved with shadow maps because the main light source is modeled as an area light.

The RT numbers are interesting but what’s with the performance with RT off. Massive lead for RDNA 2 when games are tuned for IC?

 

I think that might not be the case ...

If they were using an area light source then the shadowed regions would have a penumbra or more commonly known as soft shadows and this effect should be made readily apparent with high intensity light sources that they used in their demonstration. I imagine it's more likely that they were ray tracing a spot/point light source but we can't totally rule out that the shape of the light source it might be some really small sphere with a near imperceptible penumbra ...

 

AMD Radeon RX 6800 series facing limited availability
November 27, 2020

AMD Radeon RX 6800 series facing limited availability - VideoCardz.com

 

„One thing was clear, it was certainly a lower number of cards at launch than the NVIDIA RTX 30 series.“

Interesting. So why the published numbers state contrary?

 

Wouldn’t the shadows fan out more if they were using a small volume light?

AMD and Nvidia haven’t published any numbers. Any other source is a waste of time.

 

@CarstenS Any update here? I'm realy intrested in the values of polgong throughouptut list and strip at 0% culling for 3090 and Navi 21. Thank you in advanced.

 

Well here's the basics for review and a conceptual illustration so that we'll be able to explore the extreme cases ...

The size of the umbra and penumbra are dependent on the surface area of the light source and the surface area of the object. If our object has a bigger surface area and completely covers our light source then only the umbra exists outside of the object. Consequently, if the light source covers our object then only the antumbra exists between the object and the light source since the object is entirely visible within the light source ...

In the case where the surface area of our occluding object is getting infinitesimally smaller then the umbra and penumbra are getting smaller while the antumbra is getting larger. For the opposite case where the surface area of our light source is getting infinitesimally smaller relative to the object then the antumbra and the penumbra will get smaller while the umbra will get larger as time goes on ...

Points lights in this scenario must fit in the last case that I described. For a game featuring area lights, the shadowing model doesn't seem to match it since a falloff from umbra to penumbra is seemingly absent ...

 

Definitely a point light with some fuzzing filtering at the edges. Which kind of makes raytracing half pointless for shadows to begin with. If you're going to go to the trouble of raytraced shadows at least make it shadow a largish area light.

 

https://videocardz.com/newz/amd-radeon-rx-6900-xt-has-a-gpu-clock-limit-of-3-0-ghz



According to Der8aur, 6800XT hits an internal artificial limit of 2800 MHz.
Patrick Schur, the guy leaking a lot of info on twitter and he has been right on the money consistently, said the 6900XT has a higher limit of 3000 MHz.
Getting 2.7-2.8 GHz on 6800XT seems trivial, or at least minor effort, with AIB cards. Lets see if the 6900XT can reach the 3.0 GHz limit.
Crazy that a GPU can clock higher than a bunch of CPUs out there.
6900XT@3.0GHz would be closing in on 31 TF. That is nuts. Even at 2.8 GHz it would produce 28.7 TF.

 

Damn that thing is flying. 31TF monster, AMD you got my attention Can't believe it, not far off from what some lower end CPUs clock.

 

Next year's 4th quarter is going to be amazing

 

I'm still not convinced these are "real" clocks per se - why AMD's own presentation mentions much lower clocks for both 6800 and 6800xt while in reality they are 200-300 mhz higher even with stock limits / heatsink. In poorly optimized titles (like FF XV demo with its "inside-out" tessellated cows), my Vega boosts upwards of 1700 while in more demanding games it goes down to 1650-1680, depending on the resolution etc. What I mean, if load is not high enough, the card boosts very high but it's not "real" frequency, as in it literally can't work at such high clocks in any real task.

Of course, this descrepancy between advertised and actual clocks can be just a mind trick to make people believe that card is overperforming their expectations (like nV's boost since Maxwell 2.0), but given it's the AMD we are talking about, it might just be a glitch or some sort of "peak" boost clock in a given second. Hopefully power tables will be available for use at some point, it seems the GPUs are power-limited most of the time, not just max-frequency limited.

 

Some folks seem to be reporting the “max” clock from the overclocking software and not the actual measured clock. While that’s helpful for comparison to other people’s max clock it may not tell you what clock the cards are actually running at.

 

There also is practical thing that max clock isn't necessarily the fastest for gaming. It seems that having highest possible clock speed can induce computing errors causing recalculation which slows things down. It pays off to find max clock and then test little bit lower clocks. In many cases that little bit lower clock produces better performance.

 

If the implementation in Navi22 is the same as navi21, it's unlikely that the problem on clocking high should cause errors as many things point to Navi22 cards having way higher clocks than navi21:

https://www.igorslab.de/en/amd-rade...oard-partner-biose-sample-beat-rates-and-tgp/

(Another hint was the Mac OS dumps).
TLDR: it seems that thermals or physical implementation could be the culprit here but the architecture by itself could reach even higher ceilings.