Low-cost emerging market SoC/phone discussion

  • Thread starter Deleted member 13524
  • Start date
For me, not being able to be competitive for volume mobile mediatek socs seats in an 18 month period, was a major misstep by the company.

Hum.. the fact that IMG didn't get a design win for Mediatek's low/mid-end SoCs doesn't mean they weren't competitive. It just means they didn't win..

Besides, I think I'm yet to see a Mali Midgard implementation that wasn't a bit of a letdown, performance-wise. Kirin 920 and most Exynos 5 models have a hard time matching their Qualcomm, apple and nVidia counterparts in 3D performance.
 
Hum.. the fact that IMG didn't get a design win for Mediatek's low/mid-end SoCs doesn't mean they weren't competitive. It just means they didn't win..

Not winning the various seats that were available would suggest that their offering overall was not as good (i.e. as competitive) as Mali. Overall comprising price/performance/area/support etc. To assume they lost out on a sequence of soc graphics seats when their offering overall was better, would suggest Mediatek made a number of entirely illogical choices, which would be a nice thought for IMG, but hardly likely.

Besides, I think I'm yet to see a Mali Midgard implementation that wasn't a bit of a letdown, performance-wise. Kirin 920 and most Exynos 5 models have a hard time matching their Qualcomm, apple and nVidia counterparts in 3D performance.

"A bit of a letdown" for an enthusiast like yourself likely equates to "good enough" for the markets they are addressing. I'd suggest that very much applies to the low end to which I am referring.
 
Mediatek chose an option between several offerings. This doesn't mean that IMG's offering was inherently worse or non-competitive, it just means Mediatek found Mali to be the best option for those specific SoCs.
PowerVR enjoys a brand recognition that Mali does not. I could see IMG charging more money for a GPU that provides about the same performance than Mali's offering. Mediatek might have just chosen to pay for said brand recognition in their higher-end SoCs and pass that bit for their lower-end chips.

I'm not really an enthusiast in the smartphone/tablet scene. I don't really play games in my smartphone so my 2 year-old HTC One is likely to last me at least another year or so.
 
I could see IMG charging more money for a GPU that provides about the same performance than Mali's offering. Mediatek might have just chosen to pay for said brand recognition in their higher-end SoCs and pass that bit for their lower-end chips.

If your musing is correct, it amounts to IMG not being able/prepared to compete (on price) at the lower end. Sounds like not being competitive to me.

My understanding is the mali400 had a good size/performance ratio, and that Powervr series 5 was unable to get similar performance for the same die size. in low end chips, die size is all important.
 
tangey, you're practically saying that being competitive is the same as winning. Being competitive by definition means having a chance to win and to lose.
If you're automatically destined to win, then it's not a competition.
 
You either get a design win or you don't. In IMG's case, for several high volume, low cost, Mediatek mobile Socs, they did not get them. Whatever their "chance" was of winning we don't know. We do know that they didn't win them. One assumes, if Mediatek took rational decisions, that the determination was that the Mali option was better than the PowerVr option. For those chips that would be primarily based on cost, where die-size is a crucial factor.

I have seen enough die-shots of Mali400 implementations for it to strongly suggest to me that Mali400 is very good use of die-area at that performance level. Evidentially, I'd say there is a good case to suggest that series5 was not competitive at the low end. The argument for implementing T-series seems less clear to me.

In the broader picture I would suggest that IMG was, understandably, more focused on R&D-ing high performance solutions, given that their major customer, Apple, who went on to provide 40% & 45% of their entire technology income in FYs 2012 & 2013 respectively, was keen for performance and perf/watt and seemingly not too concerned about die area.

So the R&D priority was the right one. What am I saying ? I'm saying they should have R&D-ed low cost solutions with equal vigor, using the money there, rather than throwing money away in their CE division that just keeps on / and will continue to, lose many millions a year. in the two years mentioned above, the CE division lost £11.8M.....and another 8.6M in the most current year.
 
Last edited:
How would you know what would be best to invest on? You'd need to get a clear look at their mid and long-term plans to see if it was a bad call or not.
And you keep mentioning Mali400 but what was the last smartphone SoC using that? Mediatek's low and mid-end SoCs for smartphones all use Midgard GPUs, and the architecture is so different that I don't think you can use Mali 400's die-size as reference.
 
Well he has a point for the low end stuff. Besides there's not only Mali400, but quite a few devices that ended up with Mali450s.

MT6572: Mali400MP1@500MHz
MT6582: Mali400MP2@500MHz
MT6592: Mali450MP4@700MHz

....and there are quite a few application and tablet processors with Mali4xx.
http://en.wikipedia.org/wiki/MediaTek#cite_note-GPU_GFLOPS-80

In fact the THL4000 I've ordered for my wife recently has a 6582 (have to pick it up from the post office tomorrow heh...)
 
I assumed tangey was talking about the newer SoCs (MT67xx) since the development of those earlier ones predate the MIPS acquisition.
 
I assumed tangey was talking about the newer SoCs (MT67xx) since the development of those earlier ones predate the MIPS acquisition.

Up thread I specifically said I felt the issue pre-dated MIPS.

"I don't think it is a MIPS issue. The problem occured before the MIPS acquisition"
 
Last edited:
How would you know what would be best to invest on? You'd need to get a clear look at their mid and long-term plans to see if it was a bad call or not.

That is the crucial element in IP design. You need to forecast what IP, end prodcuts will be using 4-5 years in advance, so you can R&D the correct IP. And that's not my assessment. That's the assessment of the IMG CEO.

http://synopsys1.http.internapcdn.net/synopsys1/svsnug2013-imagination-keynote/main.htm#
See from 5 mins onwards
 
If the following article is correct, the first Soc from the Sofia line is a rockchip Soc with Atom CPU and Mali450.

I'm guessing this is the first appearance of x86 Mali drivers
http://www.unlockpwd.com/rockchip-at-ces-2015-rk3368-sofia-3g-r-tv-stick-for-9-and-chrome-os/
The slide the article refers to would seem to confirm it.

Given Both rockchip and Intel have PowerVr relationships, and both series5 and rogue have previously been mated to Atom/x86, must be a real strong reason to go with Mali450 and design x86 drivers from scratch. Once again I suggest it is related to die size.

The article also mentions the RK3368, although it states the GPU is unknown, IMG have confirmed that they are the GPU supplier in the following PR.
http://www.imgtec.com/news/detail.asp?ID=951

It is possible IMG is also supplying the video decode.
 
As tangey mentioned, the GPU in the Kirin 920 is a Mali T628MP4, not a G6200 (that's Mediatek's high-ends).
IIRC, the G6200 at 600MHz will be quite a bit more powerful than the T628 in Kirin 920. It seems that particular implementation was terribly done at least for the 3D part.

I'm not complaining about the 1.7GHz limit even one bit. My wife's Honor 6 has a spectacular performance and a very long battery life. Anandtech's review really doesn't make justice to the device (also, it seems they had a unit with broken WiFi antennas).

The Wifi problem wasn't just limited to my unit but many reported it on the Chinese version. The Mate 7 also has sub-par reception.

But I stick to my conclusion on the Kirin, it's just not an efficient chip in my book. The 5430 is in the MX4Pro is leagues ahead for example. I'll be getting an MX4 soon so we'll get some good PowerVR numbers out of that I hope.
 
I've fooled around a bit with the Meizu MX4 (nice device despite a few rough edges on the sw side) and it's still too expensive for my taste. Give me the sw maturity and stability I usually get from devices in that price range and I'll gladly pay as much.
 
I don't find the software to be bad at all, just the Chinese store / theme engine seemed not finished for the western market. Otherwise it is an extremely solid phone and beats out the Huawei offerings by quite a bit. Hope to have the review up within a week.

Yes they're kinda losing their edge by pricing it higher in western markets. But I still think it's a good deal.
 
I don't find the software to be bad at all, just the Chinese store / theme engine seemed not finished for the western market. Otherwise it is an extremely solid phone and beats out the Huawei offerings by quite a bit. Hope to have the review up within a week.

Yes they're kinda losing their edge by pricing it higher in western markets. But I still think it's a good deal.

Compared to the Zopo Z999 carrying the same SoC the "good deal" is highly relative; the MX4 might have a small edge for the sw against the fore mentioned Zopo but nothing immediately noticable to even remotely justify the price difference. Plus I have the impression from the quick looks I had that the Zopo has a better display.
 
Back
Top