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

[Calin Banc]

Don't pull that one By any measure the 20 series was higher priced, by a lot. Don't you think if they could have charged the same they would have? This time Nvidia have some competition. Just be glad of that and improved pricing. We're all winners then.

AMD are far from perfect, the issue is Nvidia can do no wrong even when they get it very, very wrong. Balanced praise or criticism is much better.

What do you mean higher priced? Compared to previous generations (before Turing), then yes.

https://www.techpowerup.com/review/nvidia-geforce-rtx-2080-founders-edition/
https://www.techpowerup.com/review/nvidia-geforce-rtx-3080-founders-edition/34.html

https://www.techpowerup.com/review/msi-geforce-rtx-2080-ti-gaming-x-trio/
https://www.techpowerup.com/review/msi-geforce-rtx-3090-gaming-x-trio/32.html

Ampere to Turing is about the same for xx80 part and more expensive for top end (at least for now). Of course, going by MSRP.

Bot Turing and Ampere were waaaay to high priced, but don't expect AMD by default to come on lower.

 

[jonneymendoza]

This sounds like an nvidia disciple. He's basically said "awww well its ok isnt it?". Completely blinded I would say.

Anyway, decoder info.

When does big navi get revealed?

 

[Mr Oblong]

When does big navi get revealed?

28th October.

 

[gpuerrilla]

When does big navi get revealed?

28th October, two weeks

 

[jonneymendoza]

Thanks

 

[LePhuronn]

By any measure the 20 series was higher priced...Just be glad of that and improved pricing. We're all winners then.

But there is no improved pricing, that's the point. "Any measure"? The 3080 in the real world costs the same as the 2080 and the 3090 is £200 more than the 2080 Ti. The 3070 sure as hell won't be cheaper than the 2070 Super.

Yet another example of how much Nvidia screwed the pooch on this one, and yet it's all perfectly fine because "3080 is better than a 2080 Ti for half the price". Idiots will never open their eyes.

 

[Mumee]

Navi 21 has 128 ROPs

 

[gpuerrilla]

Rops mean though? The nerds will just say it doesnt matter. "Look at Vega" etc.

 

[pugheaven]

Navi 21 has 128 ROPs

I reckon it['s got 256... any advances?

 

[Rroff]

128 ROPs so what?

 

[LePhuronn]

I reckon it['s got 256... any advances?

ONE HUNDRED MILLION!!!?!?

 

[Vince]

128 ROPs so what?

So take my money... obviously!!

 

[Rroff]

So take my money... obviously!!

I'm confused by the reactions positive or negative or acting like this is new information.

 

[Vince]

I'm confused by the reactions positive or negative or acting like this is new information.

Doesnt matter at all does it... at the end of the day you could make a gpu with a billion rops but only 2 shaders and 2 tmus.. it would be junk. Its all about balancing that pipeline.

 

[Phate]

2 weeks todaaay

squeeeee

 

[g67575]

128 ROPs would make a significant difference to performance, but is there any truth in the claim?

There's no 'burden of proof' or an 'obligation to prove one's assertion' when it comes to random claims made on the Internet. But, why not apply the same standard anyway? Instead of assuming what someone says is true, maybe because it's what we want to believe?

 

[humbug]

It's strange.

I bet the people currently hyping up Navi to beating the 3080 will quickly change their tune and praise it even if it's ~5 behind a 3080 after the fact (Which price depending, is still a very good GPU, not sure I'd find it an exciting event though).
But not long to find out I guess.

I think 5% slower than the 3080 is ok, AMD have to tempt buyers with price for that tho, so i would expect it to be £100 to £150 cheaper.

 

[Vince]

128 ROPs would make a significant difference to performance, but is there any truth in the claim?

There's no 'burden of proof' or an 'obligation to prove one's assertion' when it comes to random claims made on the Internet. But, why not apply the same standard anyway? Instead of assuming what someone says is true, maybe because it's what we want to believe?

Its only going to make any difference if the pipeline is staved by a lower rop count. I dont think we can say with any certainty that is the case. As before its a balancing act to keep the pipeline as full as possible to get the most possible througput, if that means 128 rops are required then that is what we get. Like memory bus bandwidth though there might be some magic you can do in there to maximise utilisation of any given part in the pipeline.

I dont think any given spec actually matters, say for example we got a card with 128bit bus, 2000 odd shaders 64 rops etc etc, pretty low numbers, and that card ends up being faster than anything from the conpetition would anybody still not buy it because the numbers look low? Even if performance is there?

 

[Nablaoperator]

So take my money... obviously!!

If thats true.. then we are looking at another pseudo 8K card priced at approx. $1200..
though i like what i see..
need to be generous with the wallet (post reviews ofc)

 

[humbug]

128 ROPs would make a significant difference to performance, but is there any truth in the claim?

There's no 'burden of proof' or an 'obligation to prove one's assertion' when it comes to random claims made on the Internet. But, why not apply the same standard anyway? Instead of assuming what someone says is true, maybe because it's what we want to believe?

A lack of ROP's can bottleneck a GPU, if a GPU has 128 ROP's it may be because its bottlenecked with 96, its not necessarily an indication of performance, if its true it indicates that AMD know their GPU is bottlenecked with anything less than 128 ROP's.

I'm stating the obvious, the only thing i can say is 128 ROP's is a lot, a hell of a lot.