Samsung Exynos 5250 - production starting in Q2 2012

Engadget:

http://www.engadget.com/2012/08/10/samsung-exynos-5-dual-white-paper-specs/

here is the whitepaper:
http://www.samsung.com/global/busin...mate_WQXGA_Solution_with_Exynos_5_Dual_WP.pdf

attention, this file loads as a zip-file despite the *.pdf in the link. Therefore download the file, rename the file from *.zip to *.pdf then you can open it.

 

Maybe a few digits of the codenumber would change, but I'd still expect to name it Exynos 4. There's a way easier explanation for the sudden increase in fillrate synthetic benchmarks. 32nm Mali400MP4@440 (4 TMUs) should have only a slight difference in fillrate performance compared to a 4TMU T604@500MHz. What else could there be in a Samsung SoC that suddenly could give better fillrate performance in such a test?

 

Another Slim Chance of Hope that iPhone will get Cortex A15. If not then it just sucks that others will have a MUCH better CPU only 1 - 2 months after its introduction.

 

Was/is it all that different with the iPhone4S vs. competing smartphones?

 

Guys this new Samsung chip is going to rock the world...I'm not sure I like the idea of back from 4 cores room just 2...even though I know overall performance will be better...I would have hoped to have a couple of cortex A7s in there as well?

Regardless, the Mali t604 really does look amazing.. unless there has been a few miss types...it has all of the next gen apis including halti, open cl, and dx 11!?....I'll believe it when I see it.

Here is a brilliant demo of what the beastie can do...
http://www.engadget.com/2012/08/07/arm-mali-t604-hands-on/

If only the galaxy note 2 would have such a chip with 2 gb lpddr3..I might buy one
I wonder which will be the better chip...a Qualcomm snapdragon s4 pro quad core...or the exynos 5250?...

 

A15 CPU IP is a whole new generation compared to A9; it might be a good idea to not judge CPU efficiency just by sterile CPU core amounts since a dual core A15 is going to bitchslap any quad A9.

All next generation GPUs are going to be amazing; which will be better time will tell.

 

Guess they couldn't 2ghz out of the A15

 

Qualcomm couldn't get higher with its Krait either and I assume it's more a yield/28nm problem for now than anything else.

 

According the Charlie d..it's not yield...at least with krait...so who knows? I'm thinking more along the lines of power consumption personally.

 

IIRC, 4S was one of the first dual core cortex A9 phones on the market. The Ipad 2 certainly was one of the first on the market. Tegra 2 was available but no devices had US availability.

 

Most upcoming GPU seem to be focused around the OpenGL ES 3.0/DirectX 9_3 feature set, but there are some with DX11 support. I'm guessing the idea is to have easier code portability between DX11 based PC games and upcoming consoles. Some work will already be required to move from x86 or perhaps PowerPC on upcoming consoles to ARM for mobile. But does the Mali-T604 have the performance to match low-end desktop DX11 GPUs? Ivy Bridge HD4000 would probably be the low-end reference point. If these DX11 mobile SoC are far slower than low-end PC DX11 GPUs then DX11 support might turn out to be more of a marketing feature than a common use case because PC DX11 games aren't easy to bring over requiring work to simplify the graphics and mobile games are focused on OpenGL ES3.0/DX9_3.

Of course, DX11 will only be available in Windows RT and Windows 8 anyways. Since OpenGL ES 3.0 was just released I'm presuming OpenGL ES 4.0 is still a while away. I wonder if Khonos will formalize some of this > ES 3.0 functionality as an optional Extension Pack.

 

Well I would bet more on power consumption that anything else wrt to clock speed.
If they were to leave passive cooling behind as MS with its X86 they would most likely clock the chips higher.
I found 1.7GHz pretty good already I mean one year ago or so Apple shipped the most successful mobile device in history with CPu clocked below 1GHz.

 

Krait's pipeline is significantly shorter than A15's though. I'm frankly very surprised they're only running at 1.7GHz.

Perhaps it's a power consumption issue. A15's kinda power hungry.

 

Metafor...Samsung did state that they would be utilising cortex a7s end of this year?...I would have thought that chip would have been a prime candidate?...seems unbalanced to me.

Also what happened to the other statement Samsung made about having their own lte modem this year and not having to rely on Qualcomm? ..unless they just meant utilising Intel hsdpa ip..

 

They started this SoC likely around the time ARM first announced the A7. I think you severely underestimate how long it takes to integrate something like that into an SoC; especially considering you're supporting a hybrid coherency model.

I believe the Korean version of the Galaxy S3 has both the Exynos and their own LTE chip. The problem is that qcom's is the only one that's truly multi-mode at the moment. And if you're going to use a qcom LTE chip, you might as well use the integrated version; which due supply constraints, probably has way more supply at 28nm than the discrete 28nm modem chip.

 

Given that this SoC is most likely for their upcoming 11 inch tablet i dont think they need Cortex A7 yet, the tablet should be big enough to mitigate most heat issues and they can use a large battery for the increased power consumption.

 

Yes I forgot about the Korean version... surely an integrated lte modem is what they are aiming for in the long run..perhaps waiting till 28nm...

I read somewhere that Qualcomm has most of the baseband ip...hopefully that will change in future...I know nvidia has grey coming next year..st errickson was supposed to have lte integrated..

I'm not only surprised by the low clockspeed..but also by the 1mb cache especially with this cache coherency,perhaps there is no need for more.
Having read the white paper I'm very impressed with the innovative power saving techniques Samsung has developed...the static display tech is clever.

Someone wrote on anandtechs comment section that Mali t604 runs at 533mhz...whether that's true or not?...but I would doubt it's even twice the power of galaxy s3...I could be underestimating though.

 

Regarding the whole LTE on chip topic, can you guys explain this diagram to me? On the bottom of the second page of this product information sheet, (http://www.samsung.com/global/business/semiconductor/minisite/Exynos/data/Exynos 5 DUAL.pdf) on Samsung's Exynos 5 website there is a diagram with an arrow going to an "LTE Modem". I am unclear how that works. Does it simply mean that the Exynos 5 works well with a separate LTE modem, or that it can be integrated on the SOC? There seems to be some confusion over this from commenters on the popular blogs, too. Thanks guys.

 

Well, it should be a good deal faster than the MP400 in GS3, even if 5x is stretching it. At 500 MHz the t604 is 68 Gflops vs ~16 Gflops for the 440 MHz MP400. I would hope that it is more than 2x as fast on average.