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

[BigBANGtheory]

I think some of the big navi expectations are just bonkers so many people are over hyped now.

 

[humbug]

I think some of the big navi expectations are just bonkers so many people are over hyped now.

Define over hyped, is it unreasonable to expect AMD to compete with the 3080?

 

[gpuerrilla]

This kind of thing has happened before every AMD release. Remember Polaris? AMD came out and said it was a mainstream card replacing the R9 380. But some info leaked out that showed the RX 480 was performing around 1080 levels. Humbug and a few others were out with their napkin math and worked out it was faster than a 980Ti. Gibbo had to come in and calm the place down.

Or how about Vega, Same thing, info leaked, napkin maths, Vega 64 was going to be faster than the 1080Ti. Even after the Vega FE edition was launched and tested with Game drivers, Some people here still believed it was going to be faster than the 1080Ti. Remember people saying, no way was the Gaming Vega card going to be just above GTX 1080 levels of performance, no way!!

So, No thanks, I am not getting on that train. AMD's statement was up to 50% performance per watt. And until a review comes out that shows otherwise, that's the estimate I am basing my expectations on.

I would rather be wrong and AMD have higher performance than my expectations.

I am not putting all my eggs in one basket either. I just cannot see a new tech with huge effort to uplift from previous gen, with consoles using a cut-down version semi-gimped being on par with a 2080 as evidence enough, being a terrible performer. I would sense a near to 3080 card likely called the 6800XT being the target, and I am waiting to see otherwise that they have something concrete that it is a flop.

You do realise there are smear campaigns all the time aswell as the hypers?

 

[five8five]

Well so far the rumoured performance looks good.

Which rumoured performance though because it varies wildly.

 

[Nablaoperator]

Which rumoured performance though because it varies wildly.

Right now we only have clarity on 2.1x RX 5700 XT for the top part (assumed at 300w TDP) basis power targets.. we have used the more optimistic 60% IPC per watt uplift leaked by Paul
But there are some zealots (in ChipHell) suggesting that biggest navi will be a 340w part

 

[LePhuronn]

See above, it's up to 50% improvement in performance per watt. Both AMD and Nvidia have used these kind of statements before and it's usually in a very specific cases that it turns out to be true. Look at how Nvidia's 1.9x increase in power efficiency has played out.

Hence me saying "let's see what AMD manage to achieve in the real world"

 

[Martini1991]

Define over hyped, is it unreasonable to expect AMD to compete with the 3080?

No, and I'd be happy with a 7970/680 situation again.
That said, there was a lot of pro AMD users whom seem disappointed with the 3080 performance and said it was overhyped etc. By the same token a GPU that only competes with the 3080 must therefore be a disappointment from those people.

From a showcase POV, I think AMD competing with the 3080 (Price aside) a month or two later isn't exactly stellar, performance parity at a later date doesn't get me excited. I'll certainly buy the AMD over the Nvidia, but it'd still make AMD second place in mind.

 

[Dave2150]

Define over hyped, is it unreasonable to expect AMD to compete with the 3080?

Historically, yes.

Radeon VII didn't compete with 2080
Vega64 didn't compete with 1080 (slightly slower with massively more power draw)
RX480 (polaris) competed well against the 1060. Polaris had no high end offering
FuryX didn't compete against the 980ti. Driver issues, massive power draw.
290X competed against the 780, had much more power draw for similar performance

 

[TNA]

What would you expect from me if AMD are competitive?

That would be fine too to be honest. But would rather see them take the performance crown

 

[Dave2150]

That would be fine too to be honest. But would rather see them take the performance crown

Taking the performance crown, with no driver issues, would be absolutely amazing for all of us.

 

[Martini1991]

Just had a thought about Nvidia moving the 3070 date
It well be that the new AMD card is between a 3080 and 3070. It'll look worse for AMD if they're behind the only new card, rather than in between two new cards...

 

[TNA]

Taking the performance crown, with no driver issues, would be absolutely amazing for all of us.

Indeed. Won’t hold my breath, but it is possible

 

[Nasher]

Historically, yes.

Radeon VII didn't compete with 2080
Vega64 didn't compete with 1080 (slightly slower with massively more power draw)
RX480 (polaris) competed well against the 1060. Polaris had no high end offering
FuryX didn't compete against the 980ti. Driver issues, massive power draw.
290X competed against the 780, had much more power draw for similar performance

Vega64 overtook the 1080 in many things. Also most of the benchmarks were using the flawed reference or Asus cards with badly installed coolers, so the results were dud.

 

[thenewoc]

Yep, depends if comparing launch day drivers or the later fine wine.

 

[gpuerrilla]

Im partial to a fine wine. Typically as when I get round to ordering a new GPU its months after the actual release.

This time is different though, I am keen to get on the fresh gen although nvidia made that almost impossible due to **** stock and the scalpers.

 

[melmac]

You do realise there are smear campaigns all the time aswell as the hypers?

Smear campaigns? Oh, come on give me a break. Anyone doing a smear campaign on AMD for the last few years would be out of a job. AMD are have been doing very well to smear their own products.

 

[K1LaW8tt5]

Historically, yes.

Radeon VII didn't compete with 2080
Vega64 didn't compete with 1080 (slightly slower with massively more power draw)
RX480 (polaris) competed well against the 1060. Polaris had no high end offering
FuryX didn't compete against the 980ti. Driver issues, massive power draw.
290X competed against the 780, had much more power draw for similar performance

power draw doesn't matter anymore though

 

[pugheaven]

If Big Navi came out running 500W and bleww the compeition into the weeds, I couldnt' give a monkey's, I'll be buying it. I've never even looked at wattage ever when buying a GPU or CPU... my requirement is, is it a weapon and is the price v's performance there, if so, I buy it. Christ when I ran my two 295x2's, that was pulling over 900W - 1000W... overclocked and BF4 ran like a weapon so don't care lol

 

[Firegod]

I still don't get the problem with Vega that people have. Here is a review of the card I have for example - https://www.tomshardware.com/uk/reviews/asus-rx-vega64-strix-oc,5520.html and it beats the 1080 in the benchmarks shown, and is sometimes within touching distance of the 1080ti but mostly it is 2nd place all the way at 1440p. I think it's a good 1440p card. And as documented, in balanced power mode (which is what I use) tomshardware record a power draw of 275w. Why people keep mentioning the bad power consumption of Vega yet the RTX 3080 averages 335w (some sites show higher figures again) - confuses me. It's like oh, we'll just ignore that...

 

[Davedree]

Vega64 overtook the 1080 in many things. Also most of the benchmarks were using the flawed reference or Asus cards with badly installed coolers, so the results were dud.

Gp104 was a small Mobile and Desktop die. It was cheaper to produce, no expensive Hbm process, and the inflated pricepoints in mobile and desktop made big profits.
Vega was a failure and I'm not going to repeat myself from the posts that I made at the time of speculation and when I proved myself correct after release.
The other side which gets conveniently forgotten is the unstable crashes, people daisy chaining the 12v rail causing instability issues, having to upgrade their inadequate psu.

Yes benchmarks were flawed, early drivers were poor and broken, the reference card needed tweaking, Asus Strix was a poorly bodged recycled heatsink.
However no this didn't make the results a dud, this was a repeat of a repeat of the failed marketing via Amd where the product wasn't ready or it's potential wasn't shown at the most important time (Review DAY).
This is why Amd have a few objectives to meet this time round
Don't rush the launch, ensure the drivers are solid, design a decent range of coolers or build a batch of review cards with a cooler similar to Aibs, then tune the bios to ensure the reviewer will get great out of the box performance.