Associate
- Joined
- 11 Sep 2020
- Posts
- 298
I too, think they will play it safe. It good actually have the same MSRP (but be available for that price unlike the 3080), and be £100/$150 cheaper...
-
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
- Thread starter Gregster
- Start date
- Status
Soldato
I have no insight on what AMD are bringing in this kind of detail - but I assume they won't be tied to the memory controller and cache the same way as before for better throughput.
EDIT: TBH this is kind of a given with the way architectures are going and I'm not sure I'd hold too much to that tweet. ROPs at the end of the day are ROPs there is a limit to how "special" they are.
Last edited:
Associate
- Joined
- 30 Aug 2009
- Posts
- 467
So what's this that I've missed about AMD's RT being inferior to nVIDIA's?
The reasoning is because Nvidia's Logo is green, while AMD's logo is red. Even if both cards were identical in terms of performance, drivers and software, Nvidia would still be leagues better in the eyes of many here, (D.P. etc) that are emotionally/financially invested in Nvidia.
Leak? have a link? i admit i havent been paying much attention to this thread today
Soldato
- Joined
- 16 Aug 2009
- Posts
- 7,740
Ah MSRP. Nvidia has nicely fudged the line on that one.
Green light rays are longer wavelength than red light so clearly superior lol.
Problem for me is - even if AMD knock it out the park with their approach and doubled 3090 RT performance - that approach is still second best to if the same effort had been put into implementing RT via a full range of fixed function hardware - while a general computer architecture in an overall sense is a much nicer thing to have some processing just needs something beyond that hence why things like Tensor cores exist.
(And I really like RT)
Associate
- Joined
- 30 Aug 2009
- Posts
- 467
Okay
Happy to dig the invoice but am sure it was £999 from ocuk. Ill confirm that a tad later
Associate
- Joined
- 30 Aug 2009
- Posts
- 467
The problem is that AMD's implementation is going to be heavily optimised for via consoles. Also, we've been here before with tessellation, where the boot was on the other foot where AMD had the dedicated hardware functionality.
Associate
- Joined
- 19 Jun 2017
- Posts
- 1,029
There's this leak about 128 ROPs.. if true, its going to be similar to 3090's 8K positioning.. this is the top alpha dog navi, not the one they showed in zen3 tease (assumption)
so we were speculating if they'd slap 32gb ram and parade it for $1,199
https://twitter.com/KittyYYuko/status/1316289459109851137
Not quite sure where you are going with that first bit - but it never turns out to be a thing - there are just too many differences between console and desktop architectures so console games coming to PC tend to either be optimised for PC with little respect to the console hardware or "ports" which run badly on any vendor regardless. I don't see that changing.
Tessellation is an interesting one - there was a bit of an awkward crossover there in that the amount of space that could be used for fixed function hardware didn't give much scope for it and ATI/AMD never really got behind pushing it anyway (they had it there but just sat on it - which is one of the reasons I don't have much time for ATI/AMD) and by the time things had moved to the point you feasibly could include enough fixed function hardware to do the job proper the general shader architecture and GPU performance in general had progressed to where it didn't need fixed function for broad usage (this won't be true for ray tracing any time soon as you'd need absolute monster shader architectures to do the job - 10 or even 20x or more more powerful than today).
Associate
- Joined
- 19 Jun 2017
- Posts
- 1,029
was brushing up some school math on parametric vector representations..
for ray box intersections you have got to do 3 series of vector operations and 2 sets of comparisons per face for determining hit
for ray triangle intersections you got to first determine the the ray's point of intersection with plane containing the triangle and somewhat complicated sign functions for determining if the point of intersection lies inside the triangle..
i believe this long sequence of operations will be fully accelerated in big navi with 1 instruction.
also, the hybrid arch makes sense cuz, the drop in RT perf can actually be load balanced (example, 40% fps hit for 40% RT allocation) which would make the architecture more scalable than Ampere..
overall i feel AMD is on the right path here
Soldato
- Joined
- 23 May 2006
- Posts
- 6,820
Sorry if it's already here and I missed it... Have there been any leaks WRT ray tracing on AMD bignavi ? I am fully committed to go AMD CPU in my next machine and would love to go full team red for the 1st time ever (last time I was close I had AMD CPU and ATI GPU)
I am fairly confident normal rasterisation should be ok for 4k 60:which is all I need.... But I really want ray tracing for this upgrade to really set it apart from my 1080ti
I am fairly confident normal rasterisation should be ok for 4k 60:which is all I need.... But I really want ray tracing for this upgrade to really set it apart from my 1080ti
ROPs can differ in their throughput - for example - outputting 8 or only 4 pixels per clock, 16 or 32 pixels per clock, etc.
It is more scalable in that regard but I don't see that as relevant any time soon - you'd need to scale up by orders of magnitude before that was an advantage.
Thing with ray tracing a lot of the meat of it is extremely simple calculations but lots of them - you want some kind of context switching or dedicated hardware so you can batch up vast amounts of them (or in some cases serially compute very fast) in a way that general purpose architectures are always second best for. A lot of vector operations are really simple a+b+c stuff until you hit plane intersections, etc. which can be really nasty - for some reason I always struggle a bit mentally with dot and cross product though they aren't really that difficult.
Years ago I found a way to cheat the first intersection test for each ray which meant even 10 year old GPUs went from like 10 seconds per frame to 300 FPS for that portion - but sadly never found a way to then implement the rest of ray tracing without all the normal performance hit.
https://www.youtube.com/watch?v=ydJCmXEHLrY
Last edited:
Soldato
rofl... and I believe that version of the card is being updated to a -1nm die platform in 2021... it's gonna be ace and the air off your case fans will power it as it'll not need PCI power supplies... and it'll have lots of RGB obviously rofl
Here's an oversimplified analysis of why I think AMD did not show their top card last week.
Since the 5700XT is about 30% slower than a 2080Ti we could assume that to match the 2080Ti we need to increase CU's by 30%.
So 40CU + 30% = 52CU.
The Xbox Seris X gpu is actually 52CU's and is supposed to be similar to a 2080Ti in performance which makes sense since it has exactly 30% more CU's than the 5700XT albeit at a lower clockspeed.
If we want to match the 3080 we need to add a further 30% more CU's.
52CU +30% = 68CU
This tells us that AMD has probably shown the mid level RNDA2 gpu with around 68CU's. The bigger 80CU chip(62CU +20%) will be shown on the 28th October imo and probably come close to the RTX 3090.
The 7nm process allows for an even bigger chip if AMD wanted since the 52CU Xbox chip is only 171mm² (smaller than the 5700XT's 251 mm²) and we know the GA102 in the RTX 3090 is a huge 628 mm². IF AMD made Big Navi 3x bigger than the XSX chip it would still be only 513mm² but pack in 156CU's
. We can only dream.....
Since the 5700XT is about 30% slower than a 2080Ti we could assume that to match the 2080Ti we need to increase CU's by 30%.
So 40CU + 30% = 52CU.
The Xbox Seris X gpu is actually 52CU's and is supposed to be similar to a 2080Ti in performance which makes sense since it has exactly 30% more CU's than the 5700XT albeit at a lower clockspeed.
If we want to match the 3080 we need to add a further 30% more CU's.
52CU +30% = 68CU
This tells us that AMD has probably shown the mid level RNDA2 gpu with around 68CU's. The bigger 80CU chip(62CU +20%) will be shown on the 28th October imo and probably come close to the RTX 3090.
The 7nm process allows for an even bigger chip if AMD wanted since the 52CU Xbox chip is only 171mm² (smaller than the 5700XT's 251 mm²) and we know the GA102 in the RTX 3090 is a huge 628 mm². IF AMD made Big Navi 3x bigger than the XSX chip it would still be only 513mm² but pack in 156CU's
. We can only dream.....- Status