• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Navi 23 ‘NVIDIA Killer’ GPU Rumored to Support Hardware Ray Tracing, Coming Next Year

Status
Not open for further replies.
pete910 said:
AMD have supported VRR since their 2xx series cards, Not working on LG earier models it;s LGs issue. it works on the CX range(2020).
Click to expand...

CX support freesync. Tell me which AMD card has hdmi 2.1 and support hdmi forum vrr so I can test?

BTW, sitting on your behind waiting for someone else to do something so you can benefit is terrible way of doing business. Nvidia did the work and certified 2019 LG line, even with xbox you can use vrr.
 
Th0nt said:
Lol yeah, he is trying I guess, dont forget he's moving house! But yeah past three vids could have been in one short video.
Click to expand...

Yes full marks for trying but he doesn't half drag it out. If it's new genuine information, presented clearly, even rumour, I'm fully behind anyone making the effort to produce a video. I can't say I'd be motivated to do it.
 
Troezar said:
Yes full marks for trying but he doesn't half drag it out. If it's new genuine information, presented clearly, even rumour, I'm fully behind anyone making the effort to produce a video. I can't say I'd be motivated to do it.
Click to expand...

I think they drag it out as they need the video to be over a certain amount of time for royalties or something, but as its his 'news' update he pulls into it games etc. definitely needs condensing IMO.
 
Troezar said:
On that basis I'm thinking a smidgen below the 3080's performance assuming no secret sauce that boosts FPS further.
Click to expand...
AMD would win still if there big navi is just 1-5fps at most slower than the 3080 BUT widely and easily available..

Available product always wins against a product no one can easily get
 
The Old School Gamer said:
I'm interested to find out what the ray tracing performance will be like and if AMD has an answer to DLSS v2.
None of the rumours / leaks really touch on those subjects.
Click to expand...

No DLSS like support.. we haven't yet come across any rumours that big navi will have dedicated matrix operation units.. even the most simplest of AI models rely on matrix operations (again there's always that 1% chance of being proven wrong)

Regarding RT, AMD might have relatively optimised (read low IQ) but higher performing solution. Following are the differences from Ampere's implementation:
1. Tree traversals/triangle search controlled by shader. Nvidia has dedicated hardware. AMD RT will thus be more programmable.. you can skip triangle searches if they exceed a predefined depth
2. AMD will be implementing some kind of cache hierarchy optimisation scheme which would reduce cache misses even at lower level of cache duplication
3. Dont know how tree branches (bounding volumes) will be created on AMD hardware
4. RT uses monte carlo methods for ray origination, dont know how that would be implemented in AMD hardware
 
Circa 3080 raw performance and price at 1440p+ would do me (with more VRAM to lord it over 10gb peasants of course). Not fussed with RT or DLSS like functionality. :)
 
jonneymendoza said:
Its a given that RT will be in these cards. I mean its on the big NAVI console gpu's dont forget!!
Click to expand...

Exactly, its among all the detail if you check out the console spiel.

The question is not will it be in, its a question of how good is it. What would be a game changer is if its in lots of games as its part of the dev cycle. If it is available in lots of games out of the gate, with extra features like image sharpening etc. then it will be good for lots of gamers.

Ed - lol oversue of the word game doh.

chroniclard said:
Circa 3080 raw performance and price at 1440p+ would do me (with more VRAM to lord it over 10gb peasants of course). Not fussed with RT or DLSS like functionality. :)
Click to expand...

From what I am seeing it is going to be exactly what your wanting.
 
I cannot wait to see AMDs implementation of RT. From what we know of the xbox and ps5, the RT cores are hybrid, right? so when they're not being used for RT they're used for rasterisation and whatnot. Which I REALLY like the sound of. Unless I've got that wrong/mixed up. It means you don't necessarily have a chunk of the card doing sod all for most of its life, especially when so few titles support RT at the minute.
 
Phate said:
I cannot wait to see AMDs implementation of RT. From what we know of the xbox and ps5, the RT cores are hybrid, right? so when they're not being used for RT they're used for rasterisation and whatnot. Which I REALLY like the sound of. Unless I've got that wrong/mixed up. It means you don't necessarily have a chunk of the card doing sod all for most of its life, especially when so few titles support RT at the minute.
Click to expand...

the ray triangle intersection testing (RTIT) will still be dedicated hardware..

caches will be common to shader --its like the shader will make a function call to the RTIT unit and control tree traversals.. the shader cache will be overwritten with the RTITs output per triangle

then theres the infinity cache hierarchy rumour which suggests a highly efficient cache architecture (am blindly believing those rumours)

no clarity on how tree groupings will be generated yet

no clarity on how radiosity/illumination computations will be done yet.. these are multidimensional integrals that have to be numerically computed using monte carlo methods.. for normally distributed variables you'd generally need a cholesky decomposition which is again a matrix operation.. nvidia has those tensor cores to take care of such computations.. though i havent looked deep enough into nvidia's RT gems to conclude the structure of integrals and whether a normal distribution will be a tenable assumption.. but the normal distribution generally rears its head in some form in most cases

overall, i believe AMD's approach will be faster but some may call it cheating because AMD can explicitly program traversals to stop before finding a triangle..
 
Last edited:
Nablaoperator said:
the ray triangle intersection testing (RTIT) will still be dedicated hardware..
caches will be common to shader --its like the shader will make a function call to the RTIT unit and control tree traversals.. per tree branch the shader cache will be overwritten with the RTITs output.
no clarity on how tree groupings will be generated yet
no clarity on how radiosity/illumination computations will be done yet.. these are multidimensional integrals that have to be numerically computed using monte carlo methods.. for normally distributed variables you'd generally need a cholesky decomposition which is again a matrix operation.. nvidia has those tensor cores to take care of such computations.. though i havent looked into nvidia's RT gems to conclude the structure of integrals and whether a normal distribution will be a tenable assumption.. but the normal distribution generally rears its head in some form in most cases
Click to expand...

Tensor cores AFAIK aren't currently used as part of ray tracing on nVidia - while they can be used for a number of areas with a significant speed up actually doing so is a bit of a balancing act and can't just be used indiscriminately.
 
Nablaoperator said:
We have a much simpler way to estimate targeted performance:
50% improvement per W .. RDRs leaks are suggesting 60%
old navi tdp 225W
navi 21 rumoured tdp 300W
so navi 21 would be 2.1x (300/225*1.6) old navi (RX 5700 xt) fps throughput

since these are commercial targets i am assuming these translate to end performance and not how fast instructions are being processed, as many such instructions have to be thrown out and don't influence the final frame
Click to expand...

i don't agree with @humbug clock speed targets but the mood music from this AMD launch is really good now.

its going to hit all the performance/value targets most of us wanted at the beginning of the thread.

This is going to be a very solid launch from team red.
 
Jacky60 said:
The guys seriously unwell, seems he comes here to vent all his frustrations with everyday life on people in the graphics card forum. It is quite funny how he gets apoplectic about the most trivial things :D.
Click to expand...
He likes to come here and get his LeFury...ON!! :p
 
Status
Not open for further replies.
Back
Top Bottom