• 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.
Caporegime
Joined
17 Mar 2012
Posts
47,656
Location
ARC-L1, Stanton System
Taking sales away from your competitor? Definitely a good think.

If AMD can do just half of what they are doing to Intel on the Desktop it would work for them, do you think AMD could? I don't, Nvidia are a very different kettle of Fish, Intel, Frankly are incompetent compared to AMD in CPU architecture design, this was almost inevitable, Nvidia are anything but incompetent.

Top12.jpg
 
Soldato
Joined
22 Nov 2009
Posts
13,252
Location
Under the hot sun.
Gents don't jump on the hype train. This card won't be cheap. The "rumoured" chip is too big for 7nm process. So are all the "rumoured" chips from Nvidia side some of like the 3080Ti 630mm2 at 7nm. There are impossible to be made, as neither Samsung nor TSMC can manufacture such chips and anything over 500mm2 will have sub 20% if not lower yields.

If AMD pulls a 350mm2 chip yeah, we should be looking for a £700 5900XT. MCM is the only way now.
 
Soldato
Joined
30 Jan 2007
Posts
15,436
Location
PA, USA (Orig UK)
@humbug : no.. I don't think AMD can do the same to Nvidia right now. Graphics are a much harder work load. If they could do a successful chiplet design (unlikely).. Then maybe....or something else smart. They already tried wider memory and it didn't really help.
 
Last edited:
Caporegime
Joined
17 Mar 2012
Posts
47,656
Location
ARC-L1, Stanton System
@humbug : no.. I don't think AMD can do the same to Nvidia right now. Graphics are a much harder work load. If they could do a successful chiplet design... Then maybe....or something else smart. They already tried wider memory and it didn't really help.

AMD's GCN shaders are nothing like as efficient as Nvidia's, there is massive bottleneck somewhere in the front end.

RDNA is a massive ground up rework, its a much much better architecture, better than Pascal but only about 95% that of Turing, i say "only" that's not bad and a massive step up from GCN, RDNA2 will improve that again, its shaders may even surpass Turing.

The problem is Nvidia are willing to make massive GPU's because they know they can sell them by the millions at very high prices, AMD can't do that, which is one reason the 5700XT is so small and clocked to the limit which utterly destroys that half decent the power efficiency built into them.

And Nvidia, they are still on 12nm, watch what happens when they move to 7nm.

AMD are doing a very good job with RDNA, it is an example of AMD's great capabilities in GPU design, but they don't have the market behind them to make Nvidia sweat, which is the least they need to do.

PS: HBM was not created for Gaming, its a workstation thing, in that its good, even Nvidia use it.
 
Man of Honour
Joined
13 Oct 2006
Posts
91,163
Gents don't jump on the hype train. This card won't be cheap. The "rumoured" chip is too big for 7nm process. So are all the "rumoured" chips from Nvidia side some of like the 3080Ti 630mm2 at 7nm. There are impossible to be made, as neither Samsung nor TSMC can manufacture such chips and anything over 500mm2 will have sub 20% if not lower yields.

If AMD pulls a 350mm2 chip yeah, we should be looking for a £700 5900XT. MCM is the only way now.

Samsung has test chips (relatively simple but not super simple as they are a benchmark for improvements) at close to the reticule limit (slightly over 800mm2) on 7nm + EUV with yields supposedly now in excess of 50% - sure not suitable for large GPU production at this time but give it another year or so it will be perfectly possible (risk production might even be possible sooner). TSMC last I heard is struggling a bit with bigger chip yields but mostly seems to be down to some contamination issues rather than technical ones.
 
Soldato
Joined
14 Aug 2009
Posts
2,784
People do by Radeon cards but at the high end nothing like the numbers that's worth it, AMD would have to be selling a lot more cards at a low price than they do at high prices if still lower than Nvidia, if they are to undercut Nvidia by $500 at the 2080TI end they have to sell at least as many as Nvidia if not more for it to make sense, in reality they would be doing well to get more than 30% share. they would still get 20% share at 90% of Nvidia's price.

Is it worth selling more at a lower price than it is less at a higher price.

AMD needs to have strong products, generation after generation to actually get some solid ground in the brand perception. That means solid drivers from the start, solid cards at good prices. That means putting good coolers on them - not the mistake they've made with R290(x), and if they're not up to speed with nVIDIA - like it was with Fury (X) and Vega, just sell at a competitive level - even if that means making almost no money out of some products.

Also get their tech in the software department up and going - AI (in games), physics and all those other technologies they've made free, but with no incentive for the developers to put in games.

Until they'll do this, I don't see them getting market share fast.
 
Caporegime
Joined
17 Mar 2012
Posts
47,656
Location
ARC-L1, Stanton System
AMD needs to have strong products, generation after generation to actually get some solid ground in the brand perception. That means solid drivers from the start, solid cards at good prices. That means putting good coolers on them - not the mistake they've made with R290(x), and if they're not up to speed with nVIDIA (like it was with Fury X and Vega), just sell at a competitive level - even if that means making almost no money out of some products.

Also get their tech in the software department up and going - AI (in games), physics and all those other technologies they've made free, but with no incentive for the developers to put in their games.

Until they'll do this, I don't see them getting market share fast.

I agree with all of that.... however.

I get why AMD put blower coolers on their reference cards, OEM's like them and right now AMD need to do everything they can to keep OEM's happy, Nvidia are in a potion where they can just say "you're not getting blowers anymore they are crap" at least Nvidia's blowers were ok, when you have enough R&D to spent you can come up with something half decent despite it being fundamentally flawed.

Drivers... again its down to R&D, but Nvidia's drivers are far from perfect, i've had issues with image corruption even on the desktop when using global sharpening, odd CPU spiking when using G-Sync, and for some strange reason my GPU will not load above 90% in Star Citizen, if i overclock the GPU i get high FPS relative to the overclock, if i run lower res my FPS jump up dramatically, but always the GPU is 90% or less. it started doing this a few drivers back...

 
Caporegime
Joined
18 Oct 2002
Posts
29,866
I find the 5700XT to be an excellent 1440p card, and even a 350nm version will be kicking ass and on the tail of the 2080Ti :eek:

Just slot it in at the 2080S pricing (retaining slightly higher-clocked 8GB) and job done :cool:
 
Soldato
Joined
10 Oct 2012
Posts
4,421
Location
Denmark
If AMD can do just half of what they are doing to Intel on the Desktop it would work for them, do you think AMD could? I don't, Nvidia are a very different kettle of Fish, Intel, Frankly are incompetent compared to AMD in CPU architecture design, this was almost inevitable, Nvidia are anything but incompetent.

Top12.jpg

My thoughts as well. Intel is taking a beating mainly due to the rather unlucky timing of all the security issues. If they hadn't had happened Ryzen wouldn't have gained the same mindshare, at least that is my theory. AMD will need to either be vastly superior on performance alone at the very top without pulling any tricks with dual GPUs, water cooling or melting PCIe Power plugs :p (price doesn't matter here i think)m while offering solid performance/dollar cards at the mid and low end or have another Intel disaster happen but for nvidia. I don't believe they will gain the mindshare they need until either of those 2 happened. I'm not betting on either though.
 
Soldato
Joined
22 Nov 2009
Posts
13,252
Location
Under the hot sun.
My thoughts as well. Intel is taking a beating mainly due to the rather unlucky timing of all the security issues. If they hadn't had happened Ryzen wouldn't have gained the same mindshare, at least that is my theory. AMD will need to either be vastly superior on performance alone at the very top without pulling any tricks with dual GPUs, water cooling or melting PCIe Power plugs :p (price doesn't matter here i think)m while offering solid performance/dollar cards at the mid and low end or have another Intel disaster happen but for nvidia. I don't believe they will gain the mindshare they need until either of those 2 happened. I'm not betting on either though.

Actually Intel takes a beating because 10nm failed, as it was scheduled for 2015!!!!!!!
 
Soldato
Joined
22 Nov 2009
Posts
13,252
Location
Under the hot sun.
Intel's architecture is beaten, AMD's architecture is infinitely more scalable.

Sure, monolithic chips at 10nm were going to be expensive also. Very expensive, hence AMD has MCM design and laughs.
Same thing would happen on GPU side. The first on MCM GPUs going to win also. If one of either AMD or Nvidia stumbles on the transition, going to be left behind for long time.

Personally here I believe AMD has better knowhow how to do this and I would be surprised if the upcoming Navi cards aren't MCM but monolithic.
 
Soldato
Joined
20 Apr 2004
Posts
4,365
Location
Oxford
AMD's GCN shaders are nothing like as efficient as Nvidia's, there is massive bottleneck somewhere in the front end.

RDNA is a massive ground up rework, its a much much better architecture, better than Pascal but only about 95% that of Turing, i say "only" that's not bad and a massive step up from GCN, RDNA2 will improve that again, its shaders may even surpass Turing.

massive ground up change or not it still has some of the high level GCN issues.A; Bandwidth staved B: Power-hungry vs the competition . It took a "massive ground up rework" and a massive node advantage to reach the PPW of Pascal.

I suspect they wher hand strung with a lot of GCN baggage to keep compatibility from the console semi custom customers.

PS no show of the mythical unicorn nvidia killer big navi gpu during there main press conference I see as well
 
Last edited:
Permabanned
Joined
27 Sep 2019
Posts
2,570
RTX2080Ti is 2 year old card also, soon to be replaced...............

2000 series launched September 20, 2018

It was on the market October 2018..... It's replacement is 3 months away according to rumours.
And if we keep the optimistic (muahaha) rhetoric of the rumours, 50% faster and 50% more power efficient (muahaha)

Hope you getting my sarcastic tone.

Seems about 15-16 months to me.
 
Status
Not open for further replies.
Back
Top Bottom