Metal_Spirit
Regular
15% or 25%? This is from AMD's
That’s RNDA 1. When I tunned in he was talking about gains from RDNA 2 over RDNA 1, and it seemed to me he mentioned 15% IPC gains.
Next Generation Hardware Speculation with a Technical Spin [post E3 2019, pre GDC 2020] [XBSX, PS5]
- Thread starter DavidGraham
- Start date
-
- Tags
- john sell
- Status
That’s RNDA 1. When I tunned in he was talking about gains from RDNA 2 over RDNA 1, and it seemed to me he mentioned 15% IPC gains.
That remember me, old time, when at the morning you're be awaked by a shining chrome sphere created by the Mighty Amiga!!!Chrome - the new lens flare/bloom...
Chrome - the new lens flare/bloom...
I just looked at that picture AMD used on the RT slide and oh my lord my eyes hurt. Mercury water, chrome balls, chrome walls - this is worse than lens flare.
Just for the sake of it I looked really hard at parked cars a few months ago. I realised it was the first time I ever noticed a reflected reflection. If RT hadn't been a topic I would have gone through life to my death never bothering to notice something my brain automatically discards. I can see the market concerns that drive these developments, but I can't for the life of me see the high power/cost path as a good evolutionary trajectory for gaming that I'm interested in subsidising. That's just a personal opinion, of course.
The point is such reflections are expensive, in contrast to AO or shadow rays. So the image indicates good performance.
Also, the picture shows lots of discontinuous depth at high frequencies. This hurts denoising and hints a somewhat larger number of rays per pixel. Really looks like denoising worst case. Also indicating good performance.
Makes me think it's not that different from RTX performance, at least not a bad surprise.
Though, we neither know resolution, framerate or how it looks in motion ofc.
Also, the picture shows lots of discontinuous depth at high frequencies. This hurts denoising and hints a somewhat larger number of rays per pixel. Really looks like denoising worst case. Also indicating good performance.
Makes me think it's not that different from RTX performance, at least not a bad surprise.
Though, we neither know resolution, framerate or how it looks in motion ofc.
If so, this could be about only the desktop GPU, but not necessarily consoles. Remembering, TMU patent lists CU interaction or optional fixed function processing of outer traversal loop. Probably having both is another option.
So there is still room for secret sauce or different implementations and possibilities, even if RDNA2 RT confirmed for both.
Metal_Spirit
Regular
I've been watching the presentation again, but skipping parts, since I do not have time... I seem to miss the affirmation of the 15% IPC gains... I heard a 50% (spoken by a Asian guy - David Wang?). Might I have missunderstood fifteen with fifty, when he was talking about perf/watt gains? If so, I'm sorry for misleading, but I always wrote "If i listened right" on my sentences...
Last edited:
There are many things the average person does not notice conciously about lighting fenomena, but that were them to be missing in a synthesized render it would feel uncanny, only for reasons one can't trully explain. That is even more true for highly dynamic scenes where artists can't band-aid odd looking places with hand placed artificial lights and other tricks to cover areas that are more pathological to modern real time renderers.
The promise of RT, although that utopia is still ways away, is that eventually engines will be able to do phisically correct path tracing of the entire scene in real time with enough precision that anything will look realistic no matter how you arange it, as long as it's made of proper PBR materials and lightsorces.
Last edited:
Miniature Kaiju
Newcomer
Chrome - the new lens flare/bloom...
The future is shiny and chrome.
DavidGraham
Veteran
FWIW, the demo AMD showed was prerecorded and not live.
Love_In_Rio
Veteran
RDNA performance at X clock = 100
RDNA 2 performance at X clock = 115
RDNA wattage = 100
RDNA 2 wattage = 77
RDNA 1 perf./watt = 100/100 = 1
RDNA 2 perf./watt = 115/77 = 1,5
That 15% sounds pretty equal to the gain Nvidia obtained from Pascal to Turing thanks to concurrent integer operations.
Last edited:
What do they mean with the m of 4096px/m ?
Pixels per minute...
/s
Pixel per meter?
Seems to be a typography unit ... https://www.unitconverters.net/typography/meter-to-pixel-x.htm
Great explanation about texel density
https://www.artstation.com/artwork/qbOqP
It means we will have huge quality texture everywhere. This is part due to the biggest RAM size and the better streaming from an SSD. It will probably be the same on XSX.
EDIT: Here it means the texel density will be perfect for 4k. You don't need to have bigger texture resolution it is perfect. Out of rendering at higher resolution no need for higher texture pack.
Last edited:
Pixel per meerkat?
thanks for this.Great explanation about texel density
https://www.artstation.com/artwork/qbOqP
It means we will have huge quality texture everywhere. This is part due to the biggest RAM size and the better streaming from an SSD. It will probably be the same on XSX.
EDIT: Here it means the texel density will be perfect for 4k. You don't need to have bigger texture resolution it is perfect. Out of rendering at higher resolution no need for higher texture pack.
we talk a lot about a bunch of stuff here that is technical, most of it around lighting and rendering etc, I would hope one day we get more texture artists in here to talk about their jobs etc. Seems interesting about how to get art right.
Shouldn't it mean that they are indeed using a narrow and fast design ?
We need Sebbi !!
- Status
Similar threads
- Locked
