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

There doesn't need to be any mention of Nvidia for DavidGraham to post every single thing he finds that is pro-Nvidia. It's literally all he does here. It's boring.
It's pretty obvious as far as GPU's are concerned they have their strengths and weaknesses, and this might be indicative of his response. Once RE8 is reviewed we should get a better picture on how the different card tiers perform against recommended specs.
 
Why not? That Ryzen 2700 i have feels like a supercomputer to me :) Though i regret i have only 16GB, but with fast SSD out of core processing would more than compensate. I think it could be a nice workstation even.

hmmm... i don't think that's going to happen, but could PCs be cheaper if there would be less fragmented manufacturing? Meaning, one big company building 3 basic models at huge quantities for the masses?
I don't want that from economical perspective, but such company could ensure to get GPUs before those money diggers do.


That's not how economics works. More competition leads to lower prices and better products. look what is happening in the HDD market with those inflated prices, or in the CPU and GPU department with only 2 competitors...
 
There is no room for this kind of convoluted explanation when Capcom lists exactly what each AMD card is capable of, under the preset "Ray Tracing".

RX 6700XT: 4K/45fps same as RTX 2070
RX 6800: 4K/45fps
RX 6900XT: 4K/60fps same as RTX 3070

nBhuB2M.png

Unless you're claiming that RX 6800, while well nearly 40% faster than RX 6700 XT at 4K raytracing on average, is in this game magically just as fast as RX 6700 XT, there is. Clearly those numbers are nowhere near accurate.
 
Unless you're claiming that RX 6800, while well nearly 40% faster than RX 6700 XT at 4K raytracing on average, is in this game magically just as fast as RX 6700 XT, there is. Clearly those numbers are nowhere near accurate.

If the 6700XT does 45FPS at 4K Raytracing, then the RX 6800 is at worst very close to 60FPS, and the 6800XT is well above it (meaning the 6900XT isn't needed for 4K60).

Those performance estimates are wrong for Navi 22 or they're wrong for Navi 21, they can't be right for both. Either way, they're wrong.
 
Unless you're claiming that RX 6800, while well nearly 40% faster than RX 6700 XT at 4K raytracing on average, is in this game magically just as fast as RX 6700 XT, there is. Clearly those numbers are nowhere near accurate.
Same thing for NVIDIA as well, the 3070 is miles ahead of the 2070 in RT performance, yet it only delivers 15 more fps. Capcom is measuring baseline performance here, not the average fps, they clearly state that fps will drop in graphics heavy scenes with RT active, so they state the min GPU required to maintain fps at a specific target.

The whole point of the RE8 requirements discussion is to shed light on the HUGE disparity in hardware requirements between vendors, for 4K60 a 3070 is required from NVIDIA, while a 6900XT is required from AMD, the highest SKU available from them.

nor the fact that they think rasterization performance is more important and their RT performance is sufficient
That's funny, 2 years ago Turing RT performance was deemed not sufficient by AMD themselves and their supporters, now that AMD has a level of RT performance that is lower than even Turing, it suddenly becomes sufficient.
 
That's funny, 2 years ago Turing RT performance was deemed not sufficient by AMD themselves and their supporters, now that AMD has a level of RT performance that is lower than even Turing, it suddenly becomes sufficient.
One could argue AMD compensates with increased raster / general performance in comparison to Turing, so in total it's ok even RT perf is lower.
Personally i think RT at 4K and 60 fps is a myth (no matter how we look at it), and discussion is just to our own amusement. But i may be wrong. Hard to judge, knowing RT only from YouTube, grrr :mad:
 
That's not how economics works. More competition leads to lower prices and better products. look what is happening in the HDD market with those inflated prices, or in the CPU and GPU department with only 2 competitors...
Totally agree, but sometimes things are not logical. Mining is not, for example. At least i don't get it.
Also, there is only one Amazon, and it made things pretty cheap. Almost worthless, even :) And there is only MS and Sony competing on high end consoles, and they are cheaper than Dell, Asus, Saphire and all of them for the same stuff. That's what i meant.
But i don't want centralism. What i do want is: At least one chip fab per continent. Where i life we can do jodeling and belly splatting. But we can not make chips. That sucks.
 
One could argue AMD compensates with increased raster / general performance in comparison to Turing
That doesn't help in all cases though, Turing still comes ahead if multiple RT effects are used together (Black Ops, Control, The Medium), or when path tracing is used (MineCraft, Quake 2).
 
That doesn't help in all cases though, Turing still comes ahead if multiple RT effects are used together (Black Ops, Control, The Medium), or when path tracing is used (MineCraft, Quake 2).
This is the impression we get, but i guess it's not really the number of effects but simply the total number of rays. BlackOps for example does only shadows(?), so i would expect a smaller difference than we see.
Overall AMD performs better than i would have thought for what it is. You know i was optimistic about compute tracing, but not that optimistic.

On the other hand the RT experience surely is half assed. For Minecraft it's simply not fast enough, so i'd feel very disappointed. If NV releases super models with 16GB, AMD has not much of an argument left.
Thus they should release extensions early to expose flexibility on PC, even if currently that's just specs on paper for an unavailable product.
I assume such extensions could at first be used e.g. to remove smaller objects for rays which have traveled a long distance already, or similar tricks. Would help with perf. for little loss on IQ and is little work.
To motivate them, it surely isn't bad to point out NVs RT lead. If they don't expose still, that's a sign they plan to add fixed function traversal for RDNA3.
 
It is not like the 16GB of VRAM on the 6000 series is going to make up for the lacking RT performance...or lack of a DLSS'ish solution.
Besides you run out of GPU power long before you have fully utilized +10GB anyways.
I've seen some recent testing to show VRAM limits already affect current games, but can't remember where or what was the actual numbers.
However, all your statements relates HW to how current software is using it. It's difficult to say how this will change, and it's unclear how we should deal with this in discussions. Personally i tend to relate to future progress and 'what could be done', others ofc. relate to actual games.
While i don't think we want to increase demands on VRAM amount, there's a lot of applications which are possible beyond texture and geometry resolution. E.g. volumetric lighting (SDF, probe grids), realtime fluid simulation, general caching to avoid redundant work.
 
I'm very curious about they will fix RT with RDNA3. More units, or redesign the whole thing. A little OT but I'm curious about the Intel solution too.

Anyway, we can't buy anything for now...
Is there some indication that there's a need for redesigning the whole thing, rather than just make it beefier? We already know at least DXR is holding back it's current capabilities, not sure of Vulkan RT
 
Is there some indication that there's a need for redesigning the whole thing, rather than just make it beefier?
If you're okay with Navi 31 providing ~2080Ti RT performance in 2023 then no.

We already know at least DXR is holding back it's current capabilities, not sure of Vulkan RT
It's not holding back capabilities, it doesn't provide a low level h/w access. That's what APIs do generally. The extent to which this impact performance is highly unlikely to be above some single percent digits - or said capabilities would be included into DXR 1.1.
 
Back
Top