Samsung Exynos 5250 - production starting in Q2 2012

  • Thread starter Deleted member 13524
  • Start date
D

Deleted member 13524

Guest
Source: Samsung

What is said in the press release:

- Dual Cortex A15 @ 2GHz
- 32nm low-power High-K Metal Gate
- "More than 4x the 3D performance of existing Cortex A9 solutions" (not sure if this means 4x Mali400 MP4 in Exynos 4210 or something else)
- "Panel Self Refresh feature" for reduced power consumption (says something about keeping still images within the SoC, I wonder if it's low-power eDRAM for framebuffer to avoid system RAM consumption).
- SoC's memory bandwidth of 12.8GB/s (dual 32-bit first-gen LPDDR3 @ 800MHz / 1600MT/s?)
- Mass production starting in 2Q 2012, sampling to partners now.



CPU and memory bandwidth wise, this will already compete with AMD's lower power Fusion.

So what could be the GPU in this? Series 5XT or Series 6? Too early for Series 6? Mali T604?
 
Whatever will be integrated into that one in terms of a GPU, I've personally become pretty used to funky Nx times performance claims from different sides nowadays. Everyone does it, so it's hardly a crime ;)
 
Dual A15 should be a pretty big jump.

Overall it might very well be that the SoC is =/>2x times faster than its predecessor. However typical creative marketing would tell you >6x times (2x times CPU + >4x times GPU).
 
Well, I guess the Galaxy S 2 can be considered ~3x faster than the Galaxy S.
During the Exynos 4210 (then "Orion") announcement, they claimed a 5x performance bump over previous architectures..

Maybe we can guesstimate the Exynos 5250 will have a 3D performance advantage of 2.5x over the current Exynos 4210, which should also be the same performance bump between the dual Cortex A9 @ 1.2GHz and the dual Cortex A15 @ 2GHz.
 
Well, I guess the Galaxy S 2 can be considered ~3x faster than the Galaxy S.
During the Exynos 4210 (then "Orion") announcement, they claimed a 5x performance bump over previous architectures..

No wonder if you have 4 fragment cores and twice the TMUs at a higher frequency; in any case that 40% difference between marketing and reality isn't exactly small.

Maybe we can guesstimate the Exynos 5250 will have a 3D performance advantage of 2.5x over the current Exynos 4210, which should also be the same performance bump between the dual Cortex A9 @ 1.2GHz and the dual Cortex A15 @ 2GHz.
Depends which GPU it really is, amount of cores, frequencies and what not, but still under 32nm. If it ships in real products before the end of 2012 it'll compete with SoCs like OMAP5@28nm (if TI is on track obviously). I'd expect OMAP5 in terms of GPU power to be roughly 2x times iPad2 at least.

Before that Samsung is also going to release a 32nm Exynos with 50% higher frequencies (CPU@1.5GHz, Mali400MP4@400MHz) which will be roughly by that amount faster than the original 1GHz Exynos. OMAP4470 should be IMHO roughly on iPhone4S level and the former about 20% ahead in terms of GPU performance.
 
Don't for get the Exynos 4412 with quad A9s. I think that will be the SOC with the T604 described in ARMs roadmap.
 
Don't for get the Exynos 4412 with quad A9s. I think that will be the SOC with the T604 described in ARMs roadmap.

Well the list of possible SoCs from Samsung could even be longer than any forementioned so far. Who guarantees that Samsung won't be using besides it's own SoCs also Qualcomm and NVIDIA SoCs as in the recent past?

Besides the 4412 appeared so far only in a kernel patch from Samsung and hasn't been officially announced yet. One would expect that a quad A9 SoC would be announced before a dual A15 SoC. Of course is anything possible, but this does sound quite weird.
 
The biggest improvement for me is the memory bandwidth. It's good to see embedded SoC's leving behind the days of virtually non-existent bandwidth. If anything, interposers/TSVs/die-stacking should come here first since there is no heat dissipation to worry about.
 
Samsung's Soc roadmap seeing agressive/strange to me.

4210...existing dual core Mali400MP4 @267Mhz

4212...apparently its a die shrink to 32nm, with "50% graphics performance increase" which would mate with he Mali being clocked @400Mhz
..........Out Q1 2012 ?

4412...Seen in some soft dev, maybe a quad core A9

5250...dual core A15, volume production Q2 2012. Some sites are reporting this as being Mali-T604, but Samsung hasn't said.

I'm trying to figure out where the SGX 5XT licence fits in, that was recently announced. In terms of performance, if its for handset/tablet, one would imagine it must appear within 12 months, as samsung has access to T-604 which is at least competitive with 5XT, if not more. I assumed at the time (and still do) that this licence was to get a chip out in time for win8 devices Q3/Q4 2012, on the basis that ARM may not have stable win DX drivers for T604.

Of the 3 chips mentioned above, the one we know least about is the 4412, so is this the target for 5XT ? If not and there is a fourth samsung soc in 2012, it would seem very agressive. Does samung really have the market volume for 4 very different families of soc in 2012 ?, remember we are talking different processor cores, different number of cores, and 3 different graphics cores. That must take significant finances/resources to develop, that requires big volume to recoup, and even then they may well still be limited in foundry capability based on Apple volume (even allowing for Apple to move some capacity next year).


Its not at all clear, and is in stark contrast to Apple's stance, where its unlikely they'll design more than 1 significant Soc (ignoring slight variations) next year, and ship incredible volume of it, and others than have already been in production for a year or two.
 
As I indicated above that infamous 4412 might have been just an internal "experiment" that they probaly dumped. It doesn't make sense to have within a comparable timeframe a 2*A9@1.5GHz, a 4*A9 and a 2*A15 SoC.

The 4212 could be a direct shrink to 32nm serving mainstream with 5250 being meant for high end.

As for the Series5XT license: it was only recently announced and while there can be N time frame between real licensing and public announcements I wouldn't suggest that it was time enough for integration into 5250 if it was originally planned for late 2011 sampling as it is now and hasn't been delayed. A delay scenario is highly unlikely considering the A15 CPU cores.

By the way: if Apple is manufacturing its next SoC under Samsung 32nm, it should be obvious that Samsung knows what Apple was/is exactly planning. In that regard I wouldn't be one bit surprised if Samsung changed its roadmap somewhere in the past and replaced the 4412 with the 5250.
 
IF this chart is in anway accurate then I think it reveals a lot about Samsungs plans:

Screen%20Shot%202011-11-09%20at%202.06.20%20PM.png

Source: Anandtech

Notice the Quad-A9 with T604 due sometime in 2012? I think that is the 4412. Also notice the lack of any A15 chip containing a Mali GPU until the latter part of 2013. The exynos 5250 is scheduled for production in Q2 2012, so if that isn't using Mali what could it be using? That's where the PowerVR 5XT license comes in.

Why would Samsung switch from Mali to PowerVR for this paticular SOC? If you read the original press release, it clearly states that the 5250 is aimed at 'high end tablets'. IMO, this SOC is aimed at Windows 8 tablets. IMG have experience building Windows drivers for their GPUs (see Intel GMA), whereas there are no current Windows implementations of Mali AFAIK. They may well be looking for something stable and mature for their initial Windows 8 SOCs.

I think it may be too power hungry for use in phones. The lack of any mention of big.LITTLE in the press release corroborates with this.

In summary:

4212 (Die shrunk 4210 with higher clocks): Replaces 'Hummingbird' in mainstream smart phones.

4412 (Quad A9): High end smartphones such as Galaxy SIII and tablets.

5250 (Dual A15, PowerVR 5XT) High end tablets. Windows 8, Android.
 
IF this chart is in anway accurate then I think it reveals a lot about Samsungs plans:


Source: Anandtech

Roadmaps can change. Especially if Apple is still manufacturing at Samsung's foundries. Would you suggest that Samsung doesn't have an insight in such a case what Apple might be cooking? Behold a dual A15 is by a magnitude more efficient than a quad A9.

Notice the Quad-A9 with T604 due sometime in 2012? I think that is the 4412. Also notice the lack of any A15 chip containing a Mali GPU until the latter part of 2013. The exynos 5250 is scheduled for production in Q2 2012, so if that isn't using Mali what could it be using? That's where the PowerVR 5XT license comes in.
Rumors have it that the 5250 was being planned with a T604, but since Samsung isn't mentioning any specifics about the graphics block it might not be definite at the time of the SoC announcement.

Why would Samsung switch from Mali to PowerVR for this paticular SOC? If you read the original press release, it clearly states that the 5250 is aimed at 'high end tablets'. IMO, this SOC is aimed at Windows 8 tablets. IMG have experience building Windows drivers for their GPUs (see Intel GMA), whereas there are no current Windows implementations of Mali AFAIK. They may well be looking for something stable and mature for their initial Windows 8 SOCs.
It's unusual that you'll see me biting IMG's butt (in a virtual sense) but it must be that "experience" that delayed Intel's Cedar Trail platform twice.

I think it may be too power hungry for use in phones. The lack of any mention of big.LITTLE in the press release corroborates with this.

Just because they don't mention anything about it, doesn't guarantee it's absence either. Besides there's no need to use the same frequencies both for tablet and smart-phone platforms.

In summary:

4212 (Die shrunk 4210 with higher clocks): Replaces 'Hummingbird' in mainstream smart phones.

4412 (Quad A9): High end smartphones such as Galaxy SIII and tablets.

5250 (Dual A15, PowerVR 5XT) High end tablets. Windows 8, Android.
Possible. However if Apple is still manufacturing its next SoC at Samsung 32nm, who's to guarantee that they won't again run into supply constraints with their own SoCs. In fact considering the rumor mongering about Apple's future plans they might even need higher quantities then before.

I'm still expecting an answer though as to why Samsung hasn't announced the 4412 yet, while for 4212 and 5250 they already did.
 

So let me understand that one: Samsung has officially announced that Exynos 5250 will go into mass production in Q2 2012, yet that site claims that a product based on it will appear already in February?

February doen't belong to Q2 either way one tries to twist it, and even worse start of mass production for a SoC doesn't under any circumstance equal a product hard launch. Ok they might announce it in February, but if it really goes into mass production in Q2 expecting final products even in Q3 could be optimistic.
 
Just because they don't mention anything about it, doesn't guarantee it's absence either. Besides there's no need to use the same frequencies both for tablet and smart-phone platforms.

I agree, but the fact that they talk about 'high end tablets' gives me the hunch that this is a relatively power hungry SOC. Also the A7 core was announced very recently so it's probably be too soon to see that in shipping silicon. That doesn't exclude the possibility of them using some other type of low power core like in OMAP 5 though.

Apple can be conservative with their clocks whereas Samsung can't to the same extent due their competitive environment. We are seeing something akin to the old 'gigahertz race' amongst android SOC vendors.

I'm still expecting an answer though as to why Samsung hasn't announced the 4412 yet, while for 4212 and 5250 they already did.

I'm not sure. You could be right and we may never see the 4412. Or they could be waiting till MWC to do a big reveal alongside the GSIII?

The quad core marketing factor shouldn't be discounted either. We all know Dual A15 > Quad A9 for most use cases, but to the layman 'quad core' carries a more immediately understandable message.
 
So let me understand that one: Samsung has officially announced that Exynos 5250 will go into mass production in Q2 2012, yet that site claims that a product based on it will appear already in February?

February doen't belong to Q2 either way one tries to twist it, and even worse start of mass production for a SoC doesn't under any circumstance equal a product hard launch. Ok they might announce it in February, but if it really goes into mass production in Q2 expecting final products even in Q3 could be optimistic.

I dont think the headline matches the text, where they only say that the tablet will be "unveiled" at MWC.

FWIW, last year Samsung said that their dual-core Exynos chip would go into mass production in the 1st half of 2011 and it showed up in the Galaxy S2 which was released in May.
 
I agree, but the fact that they talk about 'high end tablets' gives me the hunch that this is a relatively power hungry SOC. Also the A7 core was announced very recently so it's probably be too soon to see that in shipping silicon. That doesn't exclude the possibility of them using some other type of low power core like in OMAP 5 though.

Apple can be conservative with their clocks whereas Samsung can't to the same extent due their competitive environment. We are seeing something akin to the old 'gigahertz race' amongst android SOC vendors.

Exclude us humble "citizens" here (me and you included). How often have rumor mongering sites (let's get a gazillion more hits) been wrong in the past? Hell I wouldn't even be surprised if Samsung announces a Tegra3 (or any other SoC) tablet in February after all and all the so far stuff turns out to be just bullshit.

I'm not sure. You could be right and we may never see the 4412. Or they could be waiting till MWC to do a big reveal alongside the GSIII?

The quad core marketing factor shouldn't be discounted either. We all know Dual A15 > Quad A9 for most use cases, but to the layman 'quad core' carries a more immediately understandable message.
If Apple still manufactures at Samsung you can take it for granted that they most likely will again be supply constraint with their own SoCs and they'll use a multitude of SoCs from other manufacturers for different markets, exactly as they do right now (OMAP3, Tegra2, OMAP4, Qualcomm S3 etc.).

As for the quad vs. dual core thing: consumers have started reading reviews these days. Good luck getting higher performance numbers out of a quad A9 in mainstream mobile benchmarks.
 
I dont think the headline matches the text, where they only say that the tablet will be "unveiled" at MWC.

FWIW, last year Samsung said that their dual-core Exynos chip would go into mass production in the 1st half of 2011 and it showed up in the Galaxy S2 which was released in May.

Fudzilla has also copied some of that stuff in a newsblurb and they also mention a "launch". Either some of the wannabe authors need to find out what the difference between a product introduction and launch is or they need to get a proper calender. Either way I'm gonna bold and claim that 9 out of 10 are just parotting bullshit.
 
Back
Top