Monday, January 29th 2024
Top AMD RDNA4 Part Could Offer RX 7900 XTX Performance at Half its Price and Lower Power
We've known since way back in August 2023, that AMD is rumored to be retreating from the enthusiast graphics segment with its next-generation RDNA 4 graphics architecture, which means that we likely won't see successors to the RX 7900 series squaring off against the upper end of NVIDIA's fastest GeForce RTX "Blackwell" series. What we'll get instead is a product stack closely resembling that of the RX 5000 series RDNA, with its top part providing a highly competitive price-performance mix around the $400-mark. A more recent report by Moore's Law is Dead sheds more light on this part.
Apparently, the top Radeon RX SKU based on the next-gen RDNA4 graphics architecture will offer performance comparable to that of the current RX 7900 XTX, but at less than half its price (around the $400 mark). It is also expected to achieve this performance target using a smaller, simpler silicon, with significantly lower board cost, leading up to its price. What's more, there could be energy efficiency gains made from the switch to a newer 4 nm-class foundry node and the RDNA4 architecture itself; which could achieve its performance target using fewer numbers of compute units than the RX 7900 XTX with its 96.
When it came out, the RX 5700 XT offered an interesting performance proposition, beating the RTX 2070, and forcing NVIDIA to refresh its product stack with the RTX 20-series SUPER, and the resulting RTX 2070 SUPER. Things could go down slightly differently with RDNA4. Back in 2019, ray tracing was a novelty, and AMD could surprise NVIDIA in the performance segment even without it. There is no such advantage now, ray tracing is relevant; and so AMD could count on timing its launch before the Q4-2024 debut of the RTX 50-series "Blackwell."
Sources:
Moore's Law is Dead (YouTube), Tweaktown
Apparently, the top Radeon RX SKU based on the next-gen RDNA4 graphics architecture will offer performance comparable to that of the current RX 7900 XTX, but at less than half its price (around the $400 mark). It is also expected to achieve this performance target using a smaller, simpler silicon, with significantly lower board cost, leading up to its price. What's more, there could be energy efficiency gains made from the switch to a newer 4 nm-class foundry node and the RDNA4 architecture itself; which could achieve its performance target using fewer numbers of compute units than the RX 7900 XTX with its 96.
292 Comments on Top AMD RDNA4 Part Could Offer RX 7900 XTX Performance at Half its Price and Lower Power
If only a 5% price hike in the mid-range graphics card sector was my biggest problem these days...
Looking at product names alone is foolish. One should look at specs, performance, power consumption and price. Then, one can see that the 7800 XT is the obvious upgrade over the 6700 XT, not the 6800 XT.
Edit: If one compares the 7800 XT to the 6800 XT (at launch), then it is obviously cheaper. If one compares it to the 6700 XT, then it is obviously faster. Take your pick. :D
I'd assume that if AMD added matrix cores to RDNA they'd use the design that's already in CDNA2 / 3; the presence of those is indicated by FeatureMAIInsts in the gfx9 MI FeatureSet in AMDGPU.td.
gfx12 doesn't have this feature or derive from the single featureset that does and there'd be no real point in implementing it differently when pretty much everything else relevant is shared between the ISAs.
So that will most likely go RDNA 3 to 5 just like i'm doing Zen 3 to Zen 5
GFX11 has AI acceleration.
but the modes it supports is both more and less limited than CDNA1. better bfloat16, CNDA1 supports it at halfspeed, RDNA3 at equal vs fp16... but then int8 receives no doubling in performance on RDNA3, but it does on CDNA.
There are also 2 WMMA instructions... WMMA and rocmWMMA, the first will use accelerators if they exist, the 2nd requires them.
RDNA3 supports both instructions.
gpuopen.com/learn/wmma_on_rdna3/#:~:text=AMD%20GPUs%20based%20on%20the,clocks%20of%20optimal%20work%20scheduling.
github.com/ROCm/rocWMMA
6900 XT +47% -> 7900 XTX
6800 XT +36% -> 7900 XT
6700 XT +48% -> 7800 XТ
6700 +42% -> 7700 XT
6600 XT +10% -> 7600
Speaking of RX 6600 XT and RX 7600, they are essentially same thing with quite small feature set updates.
Navi 33 is a direct rebrand of Navi 23 which is a direct rebrand of Navi 10.
Navi 10, 23 and 33 is basically same thing.
2019 Navi 10 251 sq. mm, 10.3B transistors, TSMC N7
2021 Navi 23 237 sq. mm, 11B transistors, TSMC N7
2023 Navi 33 204 sq. mm, 13.3B transistors, TSMC N7+ (aka N6)8900 won't be a worthy upgrade over your 6900 XT. Maybe the power consumption, but you can underclock and undervolt your card without losing performance now, don't wait for the Navi 48-based small 8900 XT.
The guy said:You replied:So, it is not right to claim that 7600 is a successor for RX 6600 XT, when in fact it is nothing more than a little improved overclocked version on a "plus" TSMC process.Maybe the NeXT gen afterwards with RDNA 1.5. That will probably be a Radeon with a new branding scheme if we follow the HD 7970 -> R9 290X evolution.
AMD are also doing amazing with their custom soc's for xbox and playstation as well as car GPU's, so the gpu division has been a major success for them in the past 5 years.
Performance has to increase generation over generation or you're advocating for stalling the performance tiers and then just introducing more and more price tiers above to get that benefit. Isn't that why people call Nvidia "nGreedia" after all?
Previously the normal generational uplifts were likely 40, 50 and in some cases 60%.
So yeah, given that the 6500 XT is a rebrand of the previous 5500 XT which offers the RX 480 performance from year 2016.
This is 8 years gone by...
The RX 6400 somewhat "justified" its existence despite being even worse because it remained relatively affordable even during the worst of the crisis, and it was a low profile, single slot, slot powered GPU, the first in a very long time (since the GT 1030).