Exynos 2400: Xclipse 920 GPU lags behind Snapdragon 8 Gen 3's Adreno 750 due to RDNA 2 limitations

[Redaktion]

The Exynos 2400's RDNA 2-based Xclipse 920 GPU has its clocks readjusted from 1.7 GHz to 1.2 GHz due to limitations with the architecture. In its current state, it is 20-30% slower than Qualcomm's Adreno 750.

https://www.notebookcheck.net/Exynos-2400-Xclipse-920-GPU-lags-behind-Snapdragon-8-Gen-3-s-Adreno-750-due-to-RDNA-2-limitations.744845.0.html

[DantePierttyr]

sorry for the sincerity, but this is basically spreading fake news.

This is not a limitation of RDNA2, it is a limitation of Samsung's manufacturing process efficiency curve. Just compare SD 8 Gen 1 vs SD 8 Gen 1 plus or even Vega 8(12nm) vs Vega 7(7nm) to see what effect the best manufacturing process has on maximum clockrate and overall efficiency.

If the SD 8 Gen 3 was in an inferior process it would also suffer from clockrate limitation.

[SamsungForLife]

sorry for the sincerity, but this is basically spreading fake news.

This is not a limitation of RDNA2, it is a limitation of Samsung's manufacturing process efficiency curve. Just compare SD 8 Gen 1 vs SD 8 Gen 1 plus or even Vega 8(12nm) vs Vega 7(7nm) to see what effect the best manufacturing process has on maximum clockrate and overall efficiency.

If the SD 8 Gen 3 was in an inferior process it would also suffer from clockrate limitation.

Plus variant usually have higher clock rate.  That's why they are called as plus.

nvidia 2000 and 3000 series are made by Samsung. So I don't think it's their manufacturing issues.
(Don't compare 3000 series and 4000 series because the node difference is significant between them .)

samsung has improved their manufacturing process recently.  Just compare newer Exynos 1380 and old 1280.  You can see more efficiency of 1380 even though both are same node process.

Also newer snapdragon 4gen2 is made by Samsung 4nm . It's efficiency is also good .

[Suiying]

It could be RDNA2 limitations indeed.
Don't forget that this thing is designed for a Laptop in the first place, not a mobile phone.
Plus 680M (12CU RDNA2) on 6800U consume like 15W while clocking at 1.1GHz, which is fine for a Laptop. But 15W in the phone? That's a big no-no.

[Tommo918]

Just gonna stick with the S23 Ultra, if they're gonna pull the Exynos crap on us again, it's just simply not good enough end of. Snapdragon is the way to go fedup with being fobbed off with subpar chipset.

[Dave.XT]

sorry for the sincerity, but this is basically spreading fake news.

This is not a limitation of RDNA2, it is a limitation of Samsung's manufacturing process efficiency curve. Just compare SD 8 Gen 1 vs SD 8 Gen 1 plus or even Vega 8(12nm) vs Vega 7(7nm) to see what effect the best manufacturing process has on maximum clockrate and overall efficiency.

If the SD 8 Gen 3 was in an inferior process it would also suffer from clockrate limitation.

Plus variant usually have higher clock rate.  That's why they are called as plus.

nvidia 2000 and 3000 series are made by Samsung. So I don't think it's their manufacturing issues.
(Don't compare 3000 series and 4000 series because the node difference is significant between them .)

samsung has improved their manufacturing process recently.  Just compare newer Exynos 1380 and old 1280.  You can see more efficiency of 1380 even though both are same node process.

Also newer snapdragon 4gen2 is made by Samsung 4nm . It's efficiency is also good .

Nvidia Turing was made by TSMC.
Not Samsung.
However, look at 8g1 vs 8g1+ literally the same design on different fabs. Normally a + is a fan revision but with Samsung, it was so poor battery life was 20% more than it should and 20% less performance than it should have