Discussion Samsung Exynos 2400 SoC

FlameTail · Jan 25, 2024

The Samsung Exynos 2400 SoC is out, and it powers the Galaxy S24 and Galaxy S24+ smartphones in select regions worldwide.

CPU
1x Cortex-X4 @ 3.2 GHz
2x Cortex-A720 @ 2.90 GHz
3x Cortex-A720 @ 2.6 GHz
4x Cortex-A520 @ 2.0 GHz
GPU
Xclipse 940
12 CU RDNA3

Process
Samsung SF4P (4LPP+)

Exynos 2400 | Mobile Processor | Samsung Semiconductor Global

Unleash mobile excellence with the Samsung Exynos 2400. Featuring Xclipse 940 GPU for immersive gaming and on-device AI for efficiency, the Exynos 2400 elevates your mobile experience.

semiconductor.samsung.com

Golden Reviewer has tested the CPU and GPU performance/efficiency.

https://x.com/Golden_Reviewer/status/1750213147582193908?s=20

https://x.com/Golden_Reviewer/status/1750551301279334710?s=20

Panino Manino · Jan 25, 2024

For what reason CPUs on Exynos consume so much more energy?

gdansk · Jan 25, 2024

All the latest phone SoCs (8g3, D9300, E2400) are good enough for a low power laptop these days in my estimation.
I'm sick of fan noise on Windows laptops. I hate it. I've tried 7840U, 1360P, 1250U laptops. All screamers. I'm still stuck with my M1 Pro MacBook.
Maybe Qualcomm will save me. But I bet Exynos can't be used in laptops because of licensing terms of the RDNA IP.

FlameTail · Jan 25, 2024

gdansk said:
All the latest phone SoCs (8g3, D9300, E2400) are good enough for a low power laptop these days in my estimation.
I'm sick of fan noise on Windows laptops. I hate it. I've tried 7840U, 1360P, 1250U laptops. All screamers. I'm still stuck with my M1 Pro MacBook.
Maybe Qualcomm will save me. But I bet Exynos can't be used in laptops because of licensing terms of the RDNA IP.

Also most Windows laptops have terrible fan optimisation.

FlameTail · Jan 25, 2024

gdansk said:
All the latest phone SoCs (8g3, D9300, E2400) are good enough for a low power laptop these days in my estimation.
I'm sick of fan noise on Windows laptops. I hate it. I've tried 7840U, 1360P, 1250U laptops. All screamers. I'm still stuck with my M1 Pro MacBook.
Maybe Qualcomm will save me. But I bet Exynos can't be used in laptops because of licensing terms of the RDNA IP.

Yes, Exynos cannot be used because of RDNA IP and not Mediatek because there are no Mali drivers for Windows.

Qualcomm is the only option for now and the near future.

FlameTail · Jan 26, 2024

What do you guys think about the Xclipse RDNA3 12 CU GPU?

It's basically a Radeon 780M in a mobile SoC.

Tup3x · Jan 26, 2024

Probably lack of native OpenGL ES driver will hurt the performance and cause compatibility issues. Efficiency and performance might vary greatly from game to game.

hemedans · Jan 26, 2024

Tup3x said:
Probably lack of native OpenGL ES driver will hurt the performance and cause compatibility issues. Efficiency and performance might vary greatly from game to game.

Is it confirmed E2400 still use Angle? There was reports last Year AMD was working directly to create software for this soc.

hemedans · Jan 26, 2024

Panino Manino said:
For what reason CPUs on Exynos consume so much more energy?

Take it with grain of salt, 8 gen 3 and D9300 use same cores and manufacturing process but those data show huge efficiency difference, more than 40% which doesn't make sense at all.

FlameTail · Jan 26, 2024

hemedans said:
Take it with grain of salt, 8 gen 3 and D9300 use same cores and manufacturing process but those data show huge efficiency difference, more than 40% which doesn't make sense at all.

Waiting for Geekerwan review.

The ultimate smartphone SoC Reviewer.

Shivansps · Jan 26, 2024

FlameTail said:
What do you guys think about the Xclipse RDNA3 12 CU GPU?

It's basically a Radeon 780M in a mobile SoC.

The only thing this tells me is that if a phone can have 12CU RDNA3 then AMD APUs are extremly underpowered.

Shivansps · Jan 26, 2024

FlameTail said:
Yes, Exynos cannot be used because of RDNA IP and not Mediatek because there are no Mali drivers for Windows.

Qualcomm is the only option for now and the near future.

There are rumors of Mediatek working on a Windows Mali driver, but to be honest this is ARM responsibility, they should be the ones doing it.

There are also efforts to port panthor/panfrost to Windows. But i dont expect anything out of that.

FlameTail · Jan 26, 2024

Shivansps said:
The only thing this tells me is that if a phone can have 12CU RDNA3 then AMD APUs are extremly underpowered.

Pretty much yeah. Even the base M chip is miles ahead.

SarahKerrigan · Jan 26, 2024

FlameTail said:
Yes, Exynos cannot be used because of RDNA IP and not Mediatek because there are no Mali drivers for Windows.

Qualcomm is the only option for now and the near future.

What about Nvidia, with or without the Mediatek alliance?

FlameTail · Jan 26, 2024

SarahKerrigan said:
What about Nvidia, with or without the Mediatek alliance?

Sure they can, and probably will. Just not yet- as the Microsoft-Qualcomm WoA exclusive agreement is not yet expired.

In fact, the biggest challenger to Qualcomm in Windows On ARM would be Nvidia, I reckon.

SarahKerrigan · Jan 26, 2024

FlameTail said:
Sure they can, and probably will. Just not yet- as the Microsoft-Qualcomm WoA exclusive agreement is not yet expired.

In fact, the biggest challenger to Qualcomm in Windows On ARM would be Nvidia, I reckon.

Obviously, but since you were talking about Mali drivers as a blocker, it seemed you weren't really including the exclusivity agreement as a factor.

Shivansps · Jan 26, 2024

SarahKerrigan said:
Obviously, but since you were talking about Mali drivers as a blocker, it seemed you weren't really including the exclusivity agreement as a factor.

Im not even sure if that agreement even exist. The other day i read somewhere about the ARM CEO complaining about that agreement, BUT i dont think thats the problem, ARM is the problem here, how they expect anyone to use ARM SoCs for Windows when there is a not a pc-compatible UEFI for ARM SoCs. If you see Windows running on a ARM soc other than Qualcomm it is because someone took the time to port EDK2 to that particular SoC on their free time and it they make it work as best as they can.

They have a GPU IP, but they are not providing drivers outside android, they have reached an agreement, recently, with collabora to make a driver for Linux for their new Mali GPUs, if they are not making drivers for Windows, how they expect for this to even work?

If you want to make a ARM SoC for Windows, right now you need to code your own UEFI, code your own video drivers or use the AMD GPU IP. The cost is just too prohibitive. THIS is the reason, not an agreement that we are not even sure it exist.

This on top of having to code even more drivers, wifi, ethernet, as arm64 drivers for windows are rare.

Nothingness · Jan 26, 2024

Shivansps said:
Im not even sure if that agreement even exist. The other day i read somewhere about the ARM CEO complaining about that agreement, BUT i dont think thats the problem, ARM is the problem here, how they expect anyone to use ARM SoCs on Windows when there is a not a pc-compatible UEFI for ARM SoC. If you see Windows running on ARM socs other than Qualcomm it is because someone took the time to port EDK2 to a particular SoC on their free time and it they make it work as well as they can.

It's not Arm responsibility to implement UEFI for each SoC. And you know what? Arm even is a promoter of UEFI along with AMD, Intel and MS: https://uefi.org/members

In fact it's likely that if you don't see Windows on SoC not from Qualcomm from an official source, it's because they can't, and guess why they can't? Most likely because they are not allowed to do it.

SarahKerrigan · Jan 26, 2024

NXP continues to be the weird outlier in all of this - not only do they ship a BSP for Win10 on their hardware, they have accelerated drivers for the Vivante GPU IP it uses.

I can only assume there's some kind of carve-out for non-consumer devices in the exclusivity agreement.

Shivansps · Jan 26, 2024

Nothingness said:
It's not Arm responsibility to implement UEFI for each SoC. And you know what? Arm even is a promoter of UEFI along with AMD, Intel and MS: https://uefi.org/members

In fact it's likely that if you don't see Windows on SoC not from Qualcomm from an official source, it's because they can't, and guess why they can't? Most likely because they are not allowed to do it.

Im not saying it is their responsabiity to do it for each soc, but there has to be a better option than "go and fork EDK2" and try to make it work for Windows with each OEM making their own effort on their own on. I doubt AMD and Intel do that.

There is no proof that this agreement exist, no party admited it, it is just a rumor, it always was. The reality here is that there cant be Windows/ARM systems whiout doing some hard, expensive work that no one is interested in doing.

SarahKerrigan · Jan 26, 2024

Shivansps said:
Im not saying it is their responsabiity to do it for each soc, but there has to be a better option than "go and fork EDK2" and try to make it work for Windows with each OEM making their own effort on their own on. I doubt AMD and Intel do that.

There is no proof that this agreement exist, no party admited it, it is just a rumor, it always was. The reality here is that there cant be Windows/ARM systems whiout doing some hard, expensive work that no one is interested in doing.

Intel and AMD most assuredly have their own internal EFI reference sources. There's no magic "x86 EFI" that works for everything with no vendor support.

I really don't know what you expect ARM to do here.

Nothingness · Jan 26, 2024

Shivansps said:
Im not saying it is their responsabiity to do it for each soc, but there has to be a better option than "go and fork EDK2" and try to make it work for Windows with each OEM making their own effort on their own on. I doubt AMD and Intel do that.

There is no proof that this agreement exist, no party admited it, it is just a rumor, it always was. The reality here is that there cant be Windows/ARM systems whiout doing some hard, expensive work that no one is interested in doing.

So at the same time you say some hackers could make Windows work on some SoC on their free time, and at the same time you need "hard, expensive work" a company could not justify?

Shivansps · Jan 26, 2024

Nothingness said:
So at the same time you say some hackers could make Windows work on some SoC on their free time, and at the same time you need "hard, expensive work" a company could not justify?

None of those options are even remotely close to be fully functional. No one can get EDK2 to fully work with all devices on the socs that are currently avalible.

And there is no way to get around the lack of drivers, specially gpu drivers. There are some efforts, yes.

Windows does not block anything on other socs than Qualcomm and you can even activate your copy if you want, there is no evidence to sustain that this agreement exists.

Im also going to add that ARM cores started to be fast enoght to run Windows on A76, so this is still a, more or less, recent achivement.

adroc_thurston · Jan 26, 2024

FlameTail said:
It's basically a Radeon 780M in a mobile SoC

Now compare the clock rate.
(PHX one is kinda bust so only clocks >2.5x higher instead of 3.3x higher. Sadge).

moinmoin · Jan 26, 2024

Shivansps said:
The only thing this tells me is that if a phone can have 12CU RDNA3 then AMD APUs are extremly underpowered.

You wouldn't want a fire brick in your hand. To avoid that you need to keep clocks lower. To keep performance up you then need to add CUs. The result is what you are observing, a phone with many CUs to keep clocks down, and an APU where fewer CUs (which saves silicon area and as such cost) can fly due to much higher clocks being usable.

Discussion Samsung Exynos 2400 SoC

Platinum Member

Senior member

Platinum Member

Platinum Member

Platinum Member

Platinum Member

Senior member

Member

Member

Platinum Member

Diamond Member

Diamond Member

Platinum Member

Senior member

Platinum Member

Senior member

Diamond Member

Platinum Member

Senior member

Diamond Member

Senior member

Platinum Member

Diamond Member

Platinum Member

Diamond Member