Samsung’s New Exynos 2200 Processor with AMD RDNA 2 GPU Reportedly a Huge Disappointment

Tsing

The FPS Review
Staff member
Joined
May 6, 2019
Messages
12,595
Points
113
samsung-exynos-2200-render-1024x576.jpg
Image: Samsung



Samsung used the phrase “game changer” to describe its latest Exynos processor with AMD RDNA 2-based Xclipse GPU in an announcement last month, but there is now growing indication that the chip may have been majorly overhyped. Numbers such as those shared by tech YouTuber TechAlter have suggested that while the Exynos 2200 features significant improvement in the neural processing department, the new mobile processor can only claim 5 percent greater CPU performance and 17 percent greater GPU performance versus its predecessor, the Exynos 2100. The numbers for the Exynos 2200’s Xclipse GPU are certainly surprising, as many had assumed that AMD’s RDNA 2 graphics technology would have delivered a much greater graphics boost alongside features such as hardware accelerated ray tracing (RT) and variable rate shading (VRS).



The Exynos 2200 numbers we got yesterday...

Continue reading...


 
I don't see the disappointment... What kind of weird framing.... The coverage on these confuse the crap out of me. What are people expecting, 2x gains, I mean wth. Is intel behind this?
 
Yea I'm kind of on board with Uvilla. Perhhaps it was over hyped or people just simply didn't understand the technology.

Phone's are running at **** near native 4k these days, and you expect a mobile GPU to deliver full screen native resolution gaming performance and Hardware ray tracing...

Have you NOT seen what is happening in the PC market? Are you REALLY that dumb of a reviewer mr phone guy that you can't piece 2+2 = not gonna happen?
 
Gotta admit I was expecting more than a 17% bump from RDNA. I think that's the part that stings. And no power/thermal metrics released yet.

Not that 17% is insignificant, but you'd think jumping to a big name brand would net you some fantastic gains. The CPU gain is ... anemic, but ok. Most things aren't CPU bound. And not many people get too excited about a neural engine. The GPU was the real draw here, and it's... just ok.
 
Have you NOT seen what is happening in the PC market?
The bar here is definitely Apple. I know I was hoping that maybe, with AMD's muscle, we could have another ARM chip that would catch up to Apple and enable some better low power devices - not just phones, but devices like the new Macbooks - laptops with all-day-and-then-some batteries, settop boxes like the Shield, etc.

This maybe catches up to the A12 on the CPU side, which is a 2018 release, but still impressive, but Apple absolutely crushes it on graphics. Apple made a ~huge~ jump from the A12 to A13 - which brought it near the the low end of discrete cards, and it kept moving forward with the A14/M1, which is today's current generation, and due to be replaced this year. Even Intel is competing against Apple - they had to make a huge deal about how Rocket Lake will beat an M1 (they just don't mention it uses much more power to do so, and only with select benchmarks).

Even with AMD, Samsung isn't even in the same ballpark. It's like Samsung is just aiming for just a better phone SoC, while Apple is aiming for the whole enchilada and trying to compete with full x86 machines with discrete graphics. Apple isn't anywhere near challenging a 3090 anytime soon, but they are making huge gains in that direction, and it's not like anyone is running a 3090 on a all-day battery either.
 
M1 Max and 3090 are the benchmarks, respectively, and realistically they're both old hat today.

I think that the best way to look at the Exynos 2200 release is in the perspective that the product has successfully launched. Since this is an integration of an entirely new IP, it's a win for AMD and Samsung for sure.
 
M1 Max and 3090 are the benchmarks
Depends on the intended use case really. Even though they share a lot of the same architecture and design, there's a reason Apple kept a distinct naming convention between the A14 and M1.

My understanding is Exynos is intended for phones and other similar devices, not necessarily a laptop. Although it would be nice if they had looked at that - think of what a Steam Deck running something like that could do. Neither of these examples is going in a tablet or phone.

I think it's entirely appropriate to compare this to the A14, and even against just that the Exynos appears woefully behind. That said, you did mention it's a win because they actually shipped something, and with that, I agree entirely. I had hoped to see RDNA scale a bit better on the low end than we are seeing here, but that doesn't mean they didn't learn something, and can crank out better results in the next revision, or that RDNA3 (or some other variant) will be able to scale a bit better on the lower end.
 
Last edited:
What on earth is an NPU?

That said, I too was kind of expecting more than a 17% improvement over the existing Samsung GPU designs given how much experience AMD has in GPU manufacture.

That said, we are talking a COMPLETELY different power envelope than AMD is used to working under in even their low end power saving devices. It should maybe not be a surprise that the scaling downwards in power wasn't there.

Sometimes you need a dedicated architectures for low and high power designs.
 
That said, I too was kind of expecting more than a 17% improvement over the existing Samsung GPU designs given how much experience AMD has in GPU manufacture.

That said, we are talking a COMPLETELY different power envelope than AMD is used to working under in even their low end power saving devices. It should maybe not be a surprise that the scaling downwards in power wasn't there.

They swapped architectures for the Exynos to an architecture that's new to both AMD and Samsung in terms of implementation, drivers, operating system etc., and I'd give them credit even if there was no performance improvement.

Like, sure, faster is better, but just getting it all working on the first go is a win!
 
Become a Patron!
Back
Top