RDNA4

[RobertR1]

My take is that AMD should focus on a specific price point rather than going toe to toe with nvidia across the range. Similar to what Intel did with Battlemage and focused on $250.

AMD winning outright in the 350-500 lineup would do wonders for their market share and mine share. It would also take the stress of competing at a technical level with Nvidia. Both consumers and tech media is way more responsive to bang for buck price points.

Then hopefully with udna they can start clawing back on the tech side. If they carry on with rdna4 as they have, it’ll be rough.

 

[IQandHDR]

They tried that with the HD 4000, HD 5000 and HD6000 series, never really manifested (and thus they gave up that approach):


But I understand they have to try something, because the trend has been clear since NVIDIA launched the GeForce 8 series.
They kinda keep status quo until they fumble and lose market share, then rinse and repeat.
It seems the most recent blunder/oversight were RT.

The numbers tell a story.
Geforce 8 series hurt them. (Excellent performance from team Green)
Geforce 900 hurt them even more (Excellent performance and power consumprtion from team Green)
GeForce 20 series hurt them again. (RT and Upscaling from team Green)
GeForce 40 series hurt them bad. (Even more RT and Upscaling from team Green)

Despite people saying that the GeForce 40 series is "bad value", customers disagree.
AMD has a 2 sided problem.
1. Their hardeware trails in RT performance, even Intel is surpassing them in RT in the same performance segment.
2. Their software is chasing NVIDIA/Intel, offering less features/image quality).

They cannot solve this by soley going "mid range", their software stack is "killing" them as much as their hardware is.

 

[trinibwoy]

4080/Super performance sounds high end now but keep in mind last gens high end does get moved down the stack. Nvidia's 5070 or 5070ti at the very least will likely be where the 4080/s sits now. So AMD's biggest die for RNDA4 would only need to be competiting against Nvidia's 2nd biggest die cut down, or even it's 3rd biggest die (and GB205 itself seems like it will sit lower relative to previous 04 dies)

Yeah GB205 seems less ambitious than AD104 so maybe AMD will have an easier task going up against Nvidia’s #2 die.

They need to do a lot more than compete though. They need to absolutely demolish Nvidia at whatever price point they’re targeting. Similar performance for $100 less isn’t going to cut it.

 

[DegustatoR]

https://videocardz.com/newz/amd-reportedly-discontinues-radeon-rx-7900-gre

Could probably mean something for the possible launch price of N48 card(s).

 

[Cappuccino]

Despite people saying that the GeForce 40 series is "bad value", customers disagree.

It is poor value, compared to previous Nvidia generations, hence why 30 series generally outsold it.

 

[IQandHDR]

It is poor value, compared to previous Nvidia generations, hence why 30 series generally outsold it.

And still it gained more marketshare against AMD than the 30 series (posting never graph):

 

[Cappuccino]

And still it gained more marketsahre agsinst AMD than the 30 series (posting again:

This is literally the last image you posted, you didnt need to post it twice lol.

Measures of value arent predicated on how it does vs the competitor. Both releases were incredibly poor value (particularly at first). 40 series was pretty poor value, 7000 series was even worse value.

 

[IQandHDR]

This is literally the last image you posted, you didnt need to post it twice lol.

Measures of value arent predicated on how it does vs the competitor. Both releases were incredibly poor value (particularly at first). 40 series was pretty poor value, 7000 series was even worse value.

Just updated the graph with 2 more quarters, source:

Die Grafikchip- und Grafikkarten-Marktanteile im dritten Quartal 2024 | 3DCenter.org

Die Marktforscher von Jon Peddie Research zeigen die neuen Zahlen zum Markt der Grafikchips für Desktop-Grafikkarten (Add-in Boards) für das dritte Quartal 2024. In selbigem ging der weltweite Grafikkarten-Absatz der Grafikchip-Entwickler AMD, nVidia

www.3dcenter.org

 

[Albuquerque]

Measures of value arent predicated on how it does vs the competitor. Both releases were incredibly poor value (particularly at first). 40 series was pretty poor value, 7000 series was even worse value.

You'll have to explain the measurable criteria you're using to define "value." Like it or not, one person's value is another person's subjective reasoning. I value the color purple, so neither product line has a lot of "value" to me. I'm simultaneously right, and obviously tainting the argument to favor my specific opinion. Value comes in a lot of dimensions, all of which require definition to facilitate reasonable conversation.

Nevertheless, regardless of your definition of "value", the data strongly suggests many people feel it had value. The 4000-series product offerings were purchased at a higher rate than the 3000-series predecessors, and certainly more than the competitor's offerings.

 

[BRiT]

Would you kindly remain on topic. Market Share seems to be entirely off topic for this Architecture and Product discussion.

If you want to discuss that, go find another thread for it likely in the Graphics and Semiconductor Industry section or some place which isn't here. This thread should not devolve into yet another tired Nvidia vs Amd cycle, so lets cut that off before it continues.

 

[Albuquerque]

I for one want to see updated RDNA4 cores in a mobile X3D processor part. Imagine the gaming laptop goodness with a lower core-count X3D part (say, single CCD six cores / twelve threads) and perhaps a 24CU RDNA4 chiplet. It wouldn't be a world beater, but it might be able to TDP out around 35W during gaming scenarios and compete well with like the NVIDIA 4060 mobile offerings.

 

[DegustatoR]

Nevertheless, regardless of your definition of "value", the data strongly suggests many people feel it had value. The 4000-series product offerings were purchased at a higher rate than the 3000-series predecessors, and certainly more than the competitor's offerings.

I have no idea where the "worse value" argument comes up from. It's not worse even in MSRPs and then you'd have to remember that 30 series has been selling during crypto boom at outrageously high prices. For me personally my old 3080 had cost just a tad less than my current 4090 (!) So yeah.

 

[Scott_Arm]

I’d like a compelling software story with rdna4. Mantle 2.0 or something leading. I’d like a story about improving reliability in software and enabling new performance etc. AMD has great engineers but their launches always feel like they’re just playing catch up with better pricing. I think @RobertR1 is right, and a focus on that $400-500 range would be the best bet, but it has to be more than that. Rdna4 hopefully has some interesting story to tell about ray tracing, or work graphs. Some new vlkan or dx12 features that’ll be exciting for amd.