AMD Radeon RX 8000 "RDNA 4" GPU Spotted on Geekbench

[gffermari]

• TSMC 4N process (5 nm custom designed for Nvidia)[1] – not to be confused with N4

 

[ARF]

TPU database is wrong
https://en.wikipedia.org/wiki/Ada_Lovelace_(microarchitecture) . Wikipedia correctly lists it as 4N so a custom 4nm variant.

Let's see:

TPU:

Wiki:

Both share the same information.

Still N4P would offer advantages even over this because 4N is years old at this point.

Wrong. The "P" is a very minor process modification and won't result in any advantages.

• TSMC 4N process (5 nm custom designed for Nvidia)[1] – not to be confused with N4

Virtually one and the same thing. A "plus" is replaced by a new digit. Meaningless.

N5 density:

N4 density:

5 nm process - Wikipedia

en.wikipedia.org

 

[ymdhis]

Problem with RX 7000 is the power consumption in media playback scenarios.

I can just disable hardware video playback and have a ryzen use 2% cpu to decode video at the cost of maybe 5 watts, plus it also improves seeking latency considerably.

 

[LabRat 891]

16GB 256-bit for the larger core? that's it?
I've got that with a golden rabbit...

Unless RT performance is *greatly* improved, I 100% expect the 7900XT and XTX to continue to be sold and/or get a re-brand (ala Radeon VII sold alongside 5700XT, or R9 290->390x)

 

[Makaveli]

This is rough napkin math. If RDNA4 is 15% faster than RDNA3 per compute unit, at 56 compute units RDNA4 would be close to 4070 super in speed. The 4070ti performance may be out of reach, unless RDNA4 is a lot more significant of an improvement.

If they can get it to 3Ghz+ clock speeds should be decent.

 

[DemonicRyzen666]

16GB 256-bit for the larger core? that's it?
I've got that with a golden rabbit...

Unless RT performance is *greatly* improved, I 100% expect the 7900XT and XTX to continue to be sold and/or get a re-brand (ala Radeon VII sold alongside 5700XT, or R9 290->390x)

from what rumors say it only double the Ray accelarator per shader unit, so don't expect too much maybe 33% increase RT?

I doubt it would eve get that it will be straved for instructiond anyways.

 

[oxrufiioxo]

from what rumors say it only double the Ray accelarator per shader unit, so don't expect too much maybe 33% increase RT?

I doubt it would eve get that it will be straved for instructiond anyways.

The latest rumors which should be taken with a huge grain of salt as amd ones tend to be straight out of dreamland is that it will be 4080 raster with 4070ti super RT performance for 500 bucks.

I'll believe it when I see it.

Although 500 is probably the most made up part they will price is 10-15% lower than it's closest performance rival from the 50 series so if the 5070 is 699 expect 599-649..... Assuming that's the card it's most similar to perfomance wise.

Once we see the PS5 pro which likely shares some of the DNA with RDNA4 we will see hopefully it has the dedicated hardware like the PS5 pro for upscaling and that isn't a Sony only thing.

 

[TheinsanegamerN]

I'mma place my bet now: performance will be within 5% of the 7800xt in both raster and RT. There will be no 7900 series replacement. Nvidia marketshare will increase.

 

[Tomorrow]

Let's see:

TPU:

View attachment 360753

Wiki:

View attachment 360754

Both share the same information.



Wrong. The "P" is a very minor process modification and won't result in any advantages.



Virtually one and the same thing. A "plus" is replaced by a new digit. Meaningless.

N5 density:
View attachment 360755
N4 density:
View attachment 360756

5 nm process - Wikipedia

en.wikipedia.org

I posted a link where TPU says 5nm for 4090. You keep posting the same 4N FinFET image but not the link where this originates from.
As others have said it's really still 5nm and i would not call two years of process advancements as "minor modification".
Zen 5 already proved the efficiency advantage of N4P. Let's see if Nvidia uses the extra budget for efficiency or performance.

Blackwell already *IS* N4P. There's nothing to argue here. Gaming variants of this architecture will also be N4P. Anyone hoping otherwise is deluding themselves. Rubin will release next year with N3 and HBM4 and the next gaming architecture is likely 2026 on N3.

I think we will be lucky if Nvidia finally decides to include DisplayPort 2.1 UHBR20 on gaming cards tho i dont expect any VRAM increases.
It will likely be the same meh generation as 20 series was.

 

[Minus Infinity]

16GB 256-bit for the larger core? that's it?
I've got that with a golden rabbit...

Unless RT performance is *greatly* improved, I 100% expect the 7900XT and XTX to continue to be sold and/or get a re-brand (ala Radeon VII sold alongside 5700XT, or R9 290->390x)

Ok grain of salt, but MLiD is claiming his sources have said 8800XT will run at 2.9GHz+, be roughly equal to 7900XTX in raster and about 50% faster in RT and 16GB on 256bit bus.

 

[Neo_Morpheus]

Will be waiting to see how much power consumption is improved over predecessor.

Ahh, the time when power consumption matters (when it can be used against AMD) but doesnt matter otherwise, as told many times in here. See all the excuses of intel cpu owners and 3090 and 4090 gpus on Ngreedia side on the subject.

Amyways…wondering if they plan to follow through with the lower price as it has been rumored before, like the rumored 500 bucks unit with better performance than the current 7900xtx.

I've been thinking about what might be going on with all the next gen discrete GPU releases. We have been on a GPU release schedule of about two years for a while now. Here are the dates of the last releases:

A770 - Oct 2022
4090 - Oct 2022
7900XTX - Dec 2022

So there is still time for the next gen to come out this year. Rumors swirling around the tech media might be due to filling that 24/7 news cycle. The sites are probably getting bored and jumping at every bit of information. However, if the rumors are true and the next gens are delayed, all three families most likely are delayed for very different reasons:

Blackwell - Nvidia has all the resources in the world to release whatever they want whenever they want regardless of AI/Data center prioritization rumors. Therefore, the delay is most likely external. I think Nvidia needs TSMC 3 nm in order to produce Blackwell. I don't think 4 nm will yield the number of units and power consumption levels that will allow them to go above the 4090. Blackwell for the client might be delayed until 2025H2 since 3NP and 3NX are not ready until later this year or the middle of next depending on which node version is needed. With rumors suggesting that neither Intel or AMD will reach or exceed the 4090, there is just no rush for Nvidia to push out high end Blackwell.

Battlemage - Intel still has resources but they are dwindling fast. It is still possible that they may discontinue their entire GPU division for both data center and client. Rumors indicate that the GPU is still coming and Intel has a variety of choices when it comes to process node. But their financial situation is dire and there has been very little information regarding their next gen data center GPU which would get higher priority than the client.

RDNA4 - AMD has the least resources of all three companies so they probably do need to prioritize product releases based on the most important markets. As some have pointed out, AMD is focusing on the data center so any delays to RDNA4 would most likely be caused by data center GPU product schedules. The process node is not a problem as AMD might be leaving the top end GPU market for a while and they have yet to utilize 4 nm.

Sometimes we gamers ignore the fact that we are not the biggest money makers of these companies. I dont like it but understand why a company with still limited budget will dedicate more of that money to enterprise components, like zen 5.

Problem with RX 7000 is the power consumption in media playback scenarios.

Lord, this again. It’s tiresome. I’m pretty sure that you dont care about how much power your intel/ngreedia combo is using when gaming.

Adding the speculation of higher RT performance, this needs to be priced at $399 and a possible 192bit cheaper model with, for example, 44-48 CUs at $299 to really make some noise.
If AMD comes out with a 56 CUs model at $499 or more, it will mean that they are not ready yet to offer anything competitive to Nvidia and only wish to maintain a position in the discrete market to keep the Radeon brand alive.

I love the idea of prices coming down, but given that everything has simply become more expensive, we shouldn’t expect the return to those prices of old. Hell, TSMC alone has raised the price per wafer a couple of times just in the last couple of years.

Again, dont get me wrong, i dont like the msrp of a 7900xtx anymore than the already overpriced 4090, but doubt we will see a negative adjustment in a meaningful way (lower prices)

 

[john_]

I love the idea of prices coming down, but given that everything has simply become more expensive, we shouldn’t expect the return to those prices of old. Hell, TSMC alone has raised the price per wafer a couple of times just in the last couple of years.

Again, dont get me wrong, i dont like the msrp of a 7900xtx anymore than the already overpriced 4090, but doubt we will see a negative adjustment in a meaningful way (lower prices)

Those prices aren't an example of prices "coming down" but more of an example of "getting more from a new generation at current prices". What was the normal for over a decade and until a couple years ago. Are we already educated to expect higher performance from a new generation product, to also come at a higher price? Hope not.

$399 for the top model wouldn't be an example of prices going down considering RX 7700 sells even lower than that.
$299 for a 40+ CU model with 12GBs of VRAM wouldn't either, considering that RX 67x0 XT sells at that price range.

Anyway, as I said those prices I mentioned would be an indication that AMD wants to increase it's market share and it is ready to do a limited price war with Nvidia at the sub $500 market, while also maintaining a healthy advantage over the future Intel cards. If AMD doesn't have the wafer supply, or the will to try to increase it's market share, we can expect prices to be again relative to prices of Nvidia competing models.

 

[ARF]

I posted a link where TPU says 5nm for 4090. You keep posting the same 4N FinFET image but not the link where this originates from.

What chip does "4090" include ?

NVIDIA AD102 GPU Specs

18432 Cores, 576 TMUs, 192 ROPs

www.techpowerup.com

 

[john_]

TSMC 5nm

 

[Neo_Morpheus]

Those prices aren't an example of prices "coming down" but more of an example of "getting more from a new generation at current prices". What was the normal for over a decade and until a couple years ago. Are we already educated to expect higher performance from a new generation product, to also come at a higher price? Hope not.

$399 for the top model wouldn't be an example of prices going down considering RX 7700 sells even lower than that.
$299 for a 40+ CU model with 12GBs of VRAM wouldn't either, considering that RX 67x0 XT sells at that price range.

Anyway, as I said those prices I mentioned would be an indication that AMD wants to increase it's market share and it is ready to do a limited price war with Nvidia at the sub $500 market, while also maintaining a healthy advantage over the future Intel cards. If AMD doesn't have the wafer supply, or the will to try to increase it's market share, we can expect prices to be again relative to prices of Nvidia competing models.

To a point, you could say that AMD is already doing a mini-tiny price war since their offerings are indeed cheaper.

Again, dont get me wrong, I would love lower prices, i just dont know how realistic that could be with today’s money devaluation. Maybe a small reduction, but seriously doubt we will get even near to what it used to be.

And i do wish to be wrong on that, for the sake of our pockets.

 

[Tomorrow]

What chip does "4090" include ?


View attachment 360826

NVIDIA AD102 GPU Specs

18432 Cores, 576 TMUs, 192 ROPs

www.techpowerup.com

From your own link: Process Size 5 nm

You seem to be under the impression that Nvidia already uses pure 4nm and thus there's now way they could possibly improve anything by moving to N4P?

I would like to remind you that both Kepler and Maxwell were produced on the same 28nm node with roughly 1,5 years apart.
Despite the same node Nvidia managed to:

Improve performance by 28% (780 Ti to 980 Ti).
Double the L2.
Quadruple the L1.
Increase transistor count by a billion (chip size increased by only 40mm² or 7,1%).
Double the VRAM within the same 250W power.
Increase both base and boost clocks.
Lower the price by $50.

There's a lot that can be done even on the same node. Even more on a two year newer node.
Would i as tech enthusiast be happy if they used 3nm for gaming cards? Yes of course i would. But im also a realist and gamers are not their biggest customers. Same with AMD and Intel.

 

[ARF]

Would i as tech enthusiast be happy if they used 3nm for gaming cards?

Especially that 3nm is not TSMC's latest & greatest. That is the new 2nm node.

TSMC begins trial 2nm production for Apple, say reports

Chip foundry TSMC is beginning production of chips made using its 2nm manufacturing process this week, according to reports from Taiwan.

www.eenewseurope.com

[News] TSMC Reportedly Plans to Commence Trial Production for Apple’s 2nm Chips Next Week | TrendForce News

After Samsung announced a major breakthrough in 2nm, securing the first batch of orders from Japanese AI company Preferred Networks, its rival TSMC is...

www.trendforce.com

https://www.patentlyapple.com/2024/07/tsmc-is-set-to-begin-trial-production-of-2nm-chips-next-week-ahead-of-mass-production-in-2025-for-devices-like-the-iphone-17.html

All delays by AMD, nvidia or intel, and keeping the products on the older 7nm(6) and 5nm(4) will lead to declining sales, and users finding other areas where to spend money.

 

[Tomorrow]

Especially that 3nm is not TSMC's latest & greatest. That is the new 2nm node.

TSMC begins trial 2nm production for Apple, say reports

Chip foundry TSMC is beginning production of chips made using its 2nm manufacturing process this week, according to reports from Taiwan.

www.eenewseurope.com

[News] TSMC Reportedly Plans to Commence Trial Production for Apple’s 2nm Chips Next Week | TrendForce News

After Samsung announced a major breakthrough in 2nm, securing the first batch of orders from Japanese AI company Preferred Networks, its rival TSMC is...

www.trendforce.com

https://www.patentlyapple.com/2024/07/tsmc-is-set-to-begin-trial-production-of-2nm-chips-next-week-ahead-of-mass-production-in-2025-for-devices-like-the-iphone-17.html

All delays by AMD, nvidia or intel, and keeping the products on the older 7nm(6) and 5nm(4) will lead to declining sales, and users finding other areas where to spend money.

Most people dont buy these products based on process nodes. Most people dont even know what a process node is.
People primarily buy on price and performance. If the product is the same or lower price and higher performance with same or better efficiency then it sells regardless if the process node is same or not. Declining sales have been because there has been mostly bad pricing for most products and the economic situation is also not good which makes people decide twice before pulling the trigger.

2nm is years away. When words like trial and apple are thrown around then you know it will be years before big several hundred mm² chips could be produced on it with acceptable yields.

I would much more prefer at least flagship gaming cards move to HBM. This is not something that's just starting or super expensive (contrary to what others say). Even using older mass produced HBM3 (current is HBM3e and HBM4 next year) would offer advantages in terms of area, power and performance.

 

[john_]

From your own link: Process Size 5 nm

Most people dont buy these products based on process nodes. Most people dont even know what a process node is.

Haven't you realized by now that it is pointless to expect a conversation with that account?
You are just losing your time, while he is probably just having fun.