Samsung Exynos 5250 - production starting in Q2 2012

  • Thread starter Deleted member 13524
  • Start date
Ailuros said:
Nice read. Does anyone have a clue what this is about?

http://www.theinquirer.net/inquirer/news/2249399/samsung-develops-a-programmable-mobile-gpu
Click to expand...

Configurable parallel processor not unlike what an FPGA can do: I posted about it 2 months ago:
Furthermore during my kernel hackings, I wondered what a dedicated memory space "srp" stood for in the 4412. I researched a bit and it raised a few eyebrows: Samsung Reconfigurable Processor

http://web.yonsei.ac.kr/wjlee/document/HPG2011.samsung.wjlee.paper.pdf
http://www.highperformancegraphics.org/media/Posters/HPG2012_Posters_W-J.Lee.pdf

The 4210 and 4412 (And 5250 presumably) ship with an SRP based audio "unit".
Click to expand...
 
Nebuchadnezzar said:
The kernel is the only one aware of the total entities/threads running on the system and is the one who schedules them on the different cores. Your hypothetical hardware thread bouncing between different cores on load would wreak havok on caching for one, secondly it would make the kernel scheduler obsolete, and because it obsoletes the scheduler, how complex would the hardware logic need to be? It needs to be at least RT capable and have priorities on threads, and who knows what else. That's also why I mentioned preemption because it's screwing with it. The whole thing is on several levels beyond what's currently possible not only from a hardware standpoint but also software.
Click to expand...
I wasn't arguing for a full blown replacement of the kernel scheduler. What I had in mind, and I should have been more clear about this, is that hw exposes, say 4 a15 cores. Kernel schedules threads with all the priority, affinity knowledge. The hw just transparently switches to a lower power core when it sees the current load is too low.
 
rpg.314 said:
I wasn't arguing for a full blown replacement of the kernel scheduler. What I had in mind, and I should have been more clear about this, is that hw exposes, say 4 a15 cores. Kernel schedules threads with all the priority, affinity knowledge. The hw just transparently switches to a lower power core when it sees the current load is too low.
Click to expand...
I understand what you mean, but where would the advantage lie in this? You lose heterogeneous multiprocessing, and you can't have the advantages of extremely low granularity as in Haswell-like DVFS because that would still wreck havok on the caches. If it weren't for that, then I would see it possible.

I'm getting a bit afraid of the Snapdragon 600 rumours on the SIV. That would be a huge let-down. Not that the Krait 300 is bad, but it's just a relatively boring evolutionary step.
 
Nebuchadnezzar said:
Configurable parallel processor not unlike what an FPGA can do: I posted about it 2 months ago:
Click to expand...

Interesting. And I thought that Samsung had abandoned their so far experiments for their own GPU. This and a few other details that aren't publicly known yet could direct into a quite different landscape in the SFF mobile in the less foreseeable future...:cool:
 
Isn't the Creative X-Fi also reconfigurable to some degree? We had to choose "games", "music" or "audio editing" in the drivers.
 
Nebuchadnezzar said:
I'm getting a bit afraid of the Snapdragon 600 rumours on the SIV. That would be a huge let-down. Not that the Krait 300 is bad, but it's just a relatively boring evolutionary step.
Click to expand...

Now that ZTE announced a smartphone with Tegra 4, im getting a bit more confident that Samsung can do it
 
Jubei said:
Now that ZTE announced a smartphone with Tegra 4, im getting a bit more confident that Samsung can do it
Click to expand...
We'll see. I actually started to think about constraints in different markets such as US CTIA specifications clashing with design decisions. Does anybody know more about this? In any case, there's too many rumours and leaked benchmarks involving the S600, things might tip in favour of Qualcomm?

Also found a nice Q&A session with ARM & Linaro in the form of a Google hangout: https://www.youtube.com/watch?v=CSaoRPmDOWM

It's quite old by now but very informative and bullshit-free.
 
Relatively reliable even though their writers are sometimes retarded. They're probably right on this one. All those leaked benchmarks pointed towards it too. It's a pity.
 
Always wondered about the possibility of full hd amoleds in 5inch sizes...untill recently it wasnt possible to do it...I dont know whether thats beem solved or not..maybe just a manufacturing difficulty.

I think ill skip on a galaxy s4 this time around..ill keep my powder dry for either note 3, or asus padphone 3.
 
Nebuchadnezzar said:
Relatively reliable even though their writers are sometimes retarded. They're probably right on this one. All those leaked benchmarks pointed towards it too. It's a pity.
Click to expand...

I had quite a hard time myself imagining how a 4*A15+4*A7 config would work out for a smartphone design in terms of power consumption. I didn't want to risk any bold projections in order to not make a fool out of myself, but my gut feeling was telling me up to now that a quad Krait with it's asynchronous frequency scheme could be a more power sensible idea.

If true I doubt they'll miss the performance projections for the octacore and if definitely not by all that much, always under a perf/mW perspective.
 
Ailuros said:
I had quite a hard time myself imagining how a 4*A15+4*A7 config would work out for a smartphone design in terms of power consumption. I didn't want to risk any bold projections in order to not make a fool out of myself, but my gut feeling was telling me up to now that a quad Krait with it's asynchronous frequency scheme could be a more power sensible idea.

If true I doubt they'll miss the performance projections for the octacore and if definitely not by all that much, always under a perf/mW perspective.
Click to expand...
It doesn't make sense, they could still miss their clock target by a lot and still remain in front of the Snapdragons in performance. The Tegra 4 benchmarks are quite stunning, compare them to the S600 and there's a almost 30% performance gap that you could eliminate for power savings in peak TDP, and still remain way ahead on average consumption due to the bulk work being done on the A7.

The asynchronous frequency planes advantage of Krait is also diminished against a big.LITTLE design since they also at least have differing frequency planes for the clusters.

And there's the part of the fab: let's disregard that insane 100M figure that popped up a few weeks ago, even if they match the S3's 40M sales that's a shit load of money going Qualcomm's way, what does a high end SoC go for? $30? That alone is over a billion getting outsourced instead of staying in-company. Does TSMC even have the capacity to supply basically what would be a quasi-monopoly of Qualcomm in the mobile space?

That last video I posted with the ARM representative, he pointed out that consumers would see the chip this spring.

Was ARM that jaded? Or do we need to wait for the next process shrink for A15 chips to be viable in the mobile space?
 
Last edited by a moderator:
Nebuchadnezzar said:
It doesn't make sense, they could still miss their clock target by a lot and still remain in front of the Snapdragons in performance. The Tegra 4 benchmarks are quite stunning, compare them to the S600 and there's a almost 30% performance gap that you could eliminate for power savings in peak TDP, and still remain way ahead on average consumption due to the bulk work being done on the A7.
Click to expand...

I won't until I see a full Tegra4 in a smartphone design with comparable power consumption/battery file than a S600.

The asynchronous frequency planes advantage of Krait is also diminished against a big.LITTLE design since they also at least have differing frequency planes for the clusters.
Click to expand...

It's the rumors that are indicating that it had a power problem. Bad implementation from Samsung's side then? On the other hand it's not that we've seen A15 smartphone configs flooding smartphones yet either.

And there's the part of the fab: let's disregard that insane 100M figure that popped up a few weeks ago, even if they match the S3's 40M sales that's a shit load of money going Qualcomm's way, what does a high end SoC go for? $30? That alone is over a billion getting outsourced instead of staying in-company. Does TSMC even have the capacity to supply basically what would be a quasi-monopoly of Qualcomm in the mobile space?
Click to expand...

I thought Qualcomm planned to dual source it's manufacturing?

That last video I posted with the ARM representative, he pointed out that consumers would see the chip this spring.

Was ARM that jaded? Or do we need to wait for the next process shrink for A15 chips to be viable in the mobile space?
Click to expand...

I'm not jumping to any conclusions yet. I'm merely wondering what the heck is going on exactly.
 
I'm more disappointed about no AMOLED. S600 will be a bit underwhelming in 2 years time, but it's still a great kit. No AMOLED is no deal though... and with a complete lack of alternatives I guess I'll be holding onto my S2 a few extra months in the hope that the Note 3 delivers.
 
Everything is pointing towards the S600 but im still hesitant in believing it. Last year Samsung went to great lengths hiding S3 from leaking, they made a fake shell to hide the design even. Of course this could just be wishful thinking on my part
 
http://www.anandtech.com/show/6787/...ture-deep-dive-plus-tegra-4i-phoenix-hands-on

Well I for one am not convinced about tegra 4 being in many smsrtphones and performing anywhere near the stunning bemchmarks we saw for the tablet.

Anand tech showed a nvidia slide where tegra 4 on 28nm was able to consume 40% less power than tegra 3 on 40nm with equal performance....a bit misleading when tegra 4 had to clock down to something like 800mhz to achieve it compared to tegra 3 1600mhz!!

So whilst outright performance in a tablet is everything we had hoped it would be...and more...power consumption remains questioned. ..especially in smartphones.

Also I was suprised to to learn that tegra 4 doesn't independently clock each core..it only power gates them off...have I got that right?...
 
You must log in or register to reply here.

Similar threads

F
Exynos 5450
2 3
Replies
43
Views
14K
Lazy8s
L
D
Creative ZMS-20 and ZMS-40
2 3 4
Replies
75
Views
24K
darkblu
D
M
Samsung Orion SoC - dual-core A9 + "5 times the 3D graphics performance"
16 17 18
Replies
358
Views
113K
Nebuchadnezzar
N
A
Holy Wall of Text: My impressions post Playstation 4 (PS4) unveil
2 3 4
Replies
70
Views
20K
(((interference)))
I
Back
Top