Qualcomm Krait & MSM8960 @ AnandTech

Hasn't Qualcomm always looked at as many different sources as possible? The key problem being that often you can only get a certain process node at just one foundry.

 

It's digitimes after all so a grain of salt is always advisable to start with. The headline in question uses "shift" to describe what QCOM supposedly intends to, so it could imply something else.

 

Qualcomm has been rumored for years to be switching to other foundries, but has there ever been a confirmed case of a Qualcomm product not made at TSMC?

 

http://chipworks.force.com/catalog/ProductDetails?sku=QUA-MDM9215M&viewState=DetailView

The MDM9215 is dual-sourced from TSMC and Samsung 28 nm. It's only a modem though.

 

Just read the article with googling it. It says Qualcomm is pretty pissed because Apple got the capacities für 20nm to produce so early. As QC being TSMCs largest customer i can understand they're pretty unhappy about that. But TSMC denied the news. As for Digitimes, mostly you're right, but for TSMC they don't have so bad contacts.

Lower-End SoCs are also coming from Gfs and UMC i'm pretty sure, but as for higher end it's all TSMC at the moment.

 

Thanks for the tip Samwise.

 

Ok first entry was more than just dissappointing:

http://gfxbench.com/device.jsp?benchmark=gfx30&os=Android&api=gl&D=LG G3 (F460)

Second entry from Samsung this time a whole damn lot better:

http://gfxbench.com/device.jsp?benchmark=gfx30&os=Android&api=gl&D=Samsung SM-G901

Is Samsung playing tricks with frequencies again (since those are higher than even the MDP tablet results) or do they simply use a performance optimized newer driver?

 

"performance optimized" (*cough*)

 

In the Samsung Galaxy S5 chassis, presumably this type of performance can only be maintained for a very short period of time.

 

Not enough data yet unfortunately; too early for me to jump to conclusions and as they say better safe than sorry.... *snif snif* but something is smelling funny

 

The Galaxy benchmark is valid (and not "cheating"). Seems those G3's scores are funky.

 

When Anand gets on a MDP tablet with a S805 17.7 fps I am raising an eyebrow when I see 19.5 fps in a smartphone not without reason. I left a blank spot open in case the Galaxy uses faster drivers. With Samsung's recent track record don't be surprised that many end up suspicious

 

We got 18.8 fps in our benchmark, article will be up soon. Samsung stopped all benchmark shenanigans since the S5.

 

Who's "we"?

 

Not to be rude, but how could you possibly know?
I'm honestly interested if you do have non obvious means, I have a longstanding interest in benchmarking since the formation of SPEC back in the day.

 

AnandTech.

Because I've uncovered the cheating in the first place and know Samsung's mechanisms. Samsung is still doing user-space DVFS (thermal limiting) in the S5 and even tries to hide it by obfuscating the application names in the source code (decompiled myself) so that people don't find it, however this is limited to games which is a very valid use-case. All benchmarking apps were removed from this mechanism after the Note 3.

 

Thanks. Is that score sustained over multiple, continuous runs?

 

No. There's not a single device out there that is able to do this as far as I know.

 

Is the GPU higher clocked than in the MDP tablet?

 

Then Kishonti needs to change testing methods. The medium results aren't much of a help.