• 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.
The 5700XT has 40 CUs and is only 15-25% behind a 2080 Super. Why would something with 60% more CU's only match it?

It shouldn't just "match" it. Because of:

1 Refined process able to bring the clocks up;
2 Microarchitecture improvements from RDNA 1 to RDNA 2;
3 Variable Rate Shading;
4 New Ray-tracing capable hardware which is currently unknown but nothing stops it to be far superior than nvidia's RT cores.

Variable Rate Shading is particularly interesting. This technology on its own should bring very significant speed boosts.

Anyone to forecast the impact of VRS?


And, the difference between the Radeon RX 5700 XT and RTX 2080S is around 10-15%.

Untitled.png

https://www.techpowerup.com/review/sapphire-radeon-rx-5700-xt-nitro-special-edition/27.htmlimage sharing sites
 
Last edited:
The 5700XT has 40 CUs and is only 15-25% behind a 2080 Super. Why would something with 60% more CU's only match it?
1: it's more than 15-25% behind a 2080 Super
2: CU performance doesn't scale linearly
3: it's AMD, they're going to drop the ball somewhere and not get 60% more CUs worth of performance out of 60% more CUs

Of course, if you don't want me being conservative then I'll also go on record and say

5900 XT: 80 CUs, 16GB HBM2, TBP 350W, utterly crushes RTX Titan for £1,500.
 
1: it's more than 15-25% behind a 2080 Super
2: CU performance doesn't scale linearly
3: it's AMD, they're going to drop the ball somewhere and not get 60% more CUs worth of performance out of 60% more CUs

Of course, if you don't want me being conservative then I'll also go on record and say

5900 XT: 80 CUs, 16GB HBM2, TBP 350W, utterly crushes RTX Titan for £1,500.

It isn't more than 25% behind in the vast majority of games, CU count doesn't scale linearly sure on GCN but we don't know how well it'll scale on RDNA, especially RDNA1 to RDNA2.

Nice of you to assume they're going to drop the ball without knowing anything about RDNA2 or the new card coming.
 
the problem is in the grand scheme of things is big navi is a generation behind though, its not going up against 2080's it will be matched against 3080's.

just have to wait and see when either card ships but i have a feeling amd is going to have to be very hard on its profit margins to shift large numbers of big navi especially if the 3000 series have shipped or will be shipping within a month of it.
 
It isn't more than 25% behind in the vast majority of games
Fair enough, I thought it was. 64 CUs would quite easily be up there with the Ti then.
CU count doesn't scale linearly sure on GCN but we don't know how well it'll scale on RDNA, especially RDNA1 to RDNA2.
Nothing scales linearly, it's safe to say RDNA won't either.
Nice of you to assume they're going to drop the ball without knowing anything about RDNA2 or the new card coming.
Oh come on, AMD have done nothing but drop the ball to varying degrees with their GPUs for quite some time. Radeon is still skint, it'll take a long time for Su, Wang and the Zen engineer recently moved over (I've forgotten her name) to turn them around. Until then I'd like to temper my expectations.
 
And if Turing is anything to go by, Nvidia will be doubling down on their RT performance at the expense of traditional raster. Which means AMD's moving target is nowhere near as moving as it once was.

even if they dont move the architecture along its supposed to be on 7nm, turing was on 12nm so you will have a jump due to die shrinkage so even if they literally just crank out turing 2.0 with just more rt cores it should be a faster product clock for clock with the die shrinkage.

as for the moving target sure not as much as it was but il keep saying it, big navi IS a generation behind and yes when it drops if it can beat overclocked 2080's good for amd, its about time but some people have to be realistic about where the product will sit, its going to be upper mid table maybe in a couple of more years amd will be going clock for clock with nvidia in the same year but for now its more of the same so im expecting around 3070 performance, but i doubt it will trouble the 3080. again its a nice jump up from amd in performance but just not there yet where we'd all like it to be.
 
even if they dont move the architecture along its supposed to be on 7nm, turing was on 12nm so you will have a jump due to die shrinkage so even if they literally just crank out turing 2.0 with just more rt cores it should be a faster product clock for clock with the die shrinkage.

as for the moving target sure not as much as it was but il keep saying it, big navi IS a generation behind and yes when it drops if it can beat overclocked 2080's good for amd, its about time but some people have to be realistic about where the product will sit, its going to be upper mid table maybe in a couple of more years amd will be going clock for clock with nvidia in the same year but for now its more of the same so im expecting around 3070 performance, but i doubt it will trouble the 3080. again its a nice jump up from amd in performance but just not there yet where we'd all like it to be.

Big Navi is supposedly RDNA 2 so how do you know it's a generation behind. We know literally nothing about it bar it can do RT.
 
Fair enough, I thought it was. 64 CUs would quite easily be up there with the Ti then.

Nothing scales linearly, it's safe to say RDNA won't either.

Oh come on, AMD have done nothing but drop the ball to varying degrees with their GPUs for quite some time. Radeon is still skint, it'll take a long time for Su, Wang and the Zen engineer recently moved over (I've forgotten her name) to turn them around. Until then I'd like to temper my expectations.

I don't think it'll scale completely linearly but I do feel it should scale a fair bit better than GCN.

AMD have had some good hits that people have called disappointments purely because they are overhyping them. The RX480 was a good card for its price, the 470 was great for the money and the 5700/XT and 5600XT are great little cards.

The RX500 series were very mediocre and Radeon 7 was too so I would say they have hits and misses.
 
Oh no, does not bode well for @LoadsaMoney then. 90% of his quotes are from that fake new site :p

:( :D

AMD hardly have a share, if any now, as when they bring a card out, it only matches a card from Nvidia, that was released 1/2 years prior, so no one will buy it, as no one downgrades, whats the point in a card, that is 1/2 years older, and slower, than the cards we're are all now.

They needed to bring out matching performance cards, at the same time as Nvidia, but they couldn't.
 
:( :D

AMD hardly have a share, if any now, as when they bring a card out, it only matches a card from Nvidia, that was released 1/2 years prior, so no one will buy it, as no one downgrades, whats the point in a card, that is 1/2 years older, and slower, than the cards we're are all now.

They needed to bring out matching performance cards, at the same time as Nvidia, but they couldn't.
Funny thing is, when they release these cards that they finally match after all that time they price it the same. Lol.
 
Anyone who quotes WCCFTECH on here should have a 12 month ban from here, without question. That's like quoting that rag of a paper The Sun for journalistic factual evidence.

#1: the source is WCCFTech. Not a reliable source.

I do think Wccftech quotes and links needs to be moderated. They quite clearly generate rumours and sell it as news.

WCCFTech has never been the source. The source is Chiphell, Twitter and numerous other sites who have been only joined by WCCFTech later.

You are too harsh, WCCFTech is not worse than any other rumour/leaks publications site.
 
Status
Not open for further replies.
Back
Top Bottom