Intel Q1 2015 earnings, still losing ~1B per Quarter from Mobile

[Guest1]

I don't think an upclocked Airmont would be able to hold up; Cherry Trail is likely to be outperformed by the Cortex A72/A53 big.LITTLE solutions that the likes of Qualcomm and MediaTek will be shipping later this year.

Assuming they can be delivered in time. Please see the link about yield issues above.

If SoFIA MID is a 2016 part, and if Intel wants it to be competitive with its ARM contemporaries, it really needs to pack Goldmont CPU cores. Given that the original SoFIA wasn't on the roadmap ~15 months ago, I'd assume SoFIA MID is a new addition to the family that was defined fairly recently. Intel's product planners should know what they're up against, and if they decide to put in a ~2.7GHz Airmont, SoFIA MID will not hold up well against TSMC 16FF+ A72s.

SoFIA MID is to compete with the A53 AP's not the A72's that's what Morganfield was designed for so why are you comparing it to a high performance AP? Again if TSM was tasked with A9 at the last minute it seems capacity at the bleeding ARM edge has all been allocated to Samsung and Apple. The rest of the ARM vendors won't have anything at 14nm / 16nm until 2016 at which time Intel says it will be close to bringing 10 nm online.

 

[Arachnotronic]

Assuming they can be delivered in time. Please see the link about yield issues above.

28nm A72s coming this year, courtesy of Qualcomm and MediaTek. Geekbench results for the MediaTek part are quite promising (as always usual caveats around Geekbench apply).

SoFIA MID is to compete with the A53 AP's not the A72's that's what Morganfield was designed for so why are you comparing it to a high performance AP? Again if TSM was tasked with A9 at the last minute it seems capacity at the bleeding ARM edge has all been allocated to Samsung and Apple. The rest of the ARM vendors won't have anything at 14nm / 16nm until 2016 at which time Intel says it will be close to bringing 10 nm online.

Nah...SoFIA LTE 2 will be going up against the A53-class chips. SoFIA MID gets to play ball with these A72/A53 big.LITTLEs + whatever ARM markets its Artemis core (successor to A17) as.

 

[RU482]

Unknown, but they are selling 15 W Pentium and Celeron Broadwell U parts.

Unlikely. the whole point of celeron is to offer up the i3/5/7 parts that weren't 100% functional. I guess a pentium is a higher featured celeron, which has value too depending on how good yields are.

making bay trail class celeron and pentium parts was just a bad/confusing idea

 

[Guest1]

28nm A72s coming this year, courtesy of Qualcomm and MediaTek. Geekbench results for the MediaTek part are quite promising (as always usual caveats around Geekbench apply).

Ok if they are on 28nm. It's interesting that they would back port it to 28 nm when it was designed for 16nm.

Nah...SoFIA LTE 2 will be going up against the A53-class chips. SoFIA MID gets to play ball with these A72/A53 big.LITTLEs + whatever ARM markets its Artemis core (successor to A17) as.

Wow that will make Morganfield with Goldmont very interesting. Intel 14nm vs ARM on 14 nm / 16 nm FF+. I wonder how efficient the SoFIA baseband on 14nm will be. In the age of good enough computing it will be interesting to watch this will play out with the OEM's.

 

[jpiniero]

Unlikely. the whole point of celeron is to offer up the i3/5/7 parts that weren't 100% functional. I guess a pentium is a higher featured celeron, which has value too depending on how good yields are.

making bay trail class celeron and pentium parts was just a bad/confusing idea

Bay Trail is something like 40% cheaper for Intel to make, not to mention they are giving it away to finish off AMD.

IIRC, Core Celeron and Pentium are different dies than the i3.

 

[Arachnotronic]

Bay Trail is something like 40% cheaper for Intel to make, not to mention they are giving it away to finish off AMD.

Do you have evidence that Intel is "giving away" Bay Trail-M/D and Braswell?

 

[jpiniero]

Do you have evidence that Intel is "giving away" Bay Trail-M/D and Braswell?

How else can you explain how quickly it came to dominate consumer retail sales? If they were making any kind of money on it Intel would have used that to cover up the contra-revenue.

 

[Arachnotronic]

How else can you explain how quickly it came to dominate consumer retail sales? If they were making any kind of money on it Intel would have used that to cover up the contra-revenue.

It's a good low-cost PC product.

 

[DrMrLordX]

http://appleinsider.com/articles/15...tsmc-for-30-of-a9-chip-orders-for-next-iphone

Thanks for the link. Hmm! Samsung prioritizes 14nm fab capacity for their own products, putting Apple in an uncomfortable position, so they turn to GF who is struggling to get 14nm yield up right now (which is really not that surprising, GF wasn't supposed to be fabbing 14nm parts until 2016!), and then they turn to TSMC. I guess TSMC gets a reprieve after losing so much of Nvidia's business.

There's SoFIA LTE2 and SoFIA MID both going to be 14 nm not 28nm X3. That would be noncompetitive. I am curious about what SoFIA mid is going to be packing. Morganfield will be on Goldmont I would expect SoFIA mid to be between Goldmont and Airmont. So maybe a down clocked Goldmont or an upclocked Airmont? Cherrytrail is already launched with Surface 3. Morganfield is supposed to be after that if I remember the roadmap properly. SoFIA is just a phone AP not for tablets of course that wouldn't stop an OEM from squeezing it in one.

Now we're talking a moving target, but you and Arachnotronic have already hashed that out well enough. I'm looking at immediate successors to the Bay Trail chips going into cheapie tablets.

It's a good low-cost PC product.

Ummmm really? What's the fastest Bay Trail-D out there, the J2900? It's better than the anemic J1900, but not by all that much. I guess it depends on your definition of "good".

 

[dealcorn]

Ummmm really? What's the fastest Bay Trail-D out there, the J2900? It's better than the anemic J1900, but not by all that much. I guess it depends on your definition of "good".

I am curious what use/thought pattern you have that caused you to characterize the quad core J1900 performance as "anemic" when the more obvious choice would have been the cheaper, but faster, dual core, J1800? I have been running a J1800 almost 24/7 since shortly after release. With 4 GB DRAM, a SSD, a second NIC and Debian Jessie it handles multiple home server roles such as firewall/router, DNS and DHCP server and Rsync transfers. It also handles GUI desktop roles for daily use such as browsing, media consumption and AisleRiot Solitaire. I characterize the performance in my use pattern as quite reasonable and surprisingly livable considering the price. Is your selection of the word "Anemic" code speak for a suggestion that it is a poor choice for Crysis? The overwhelming market acceptance of Baytrail M/D at full margin with no contra revenue suggests you are out of touch with mainstream demand.

 

[coercitiv]

it handles multiple home server roles such as firewall/router, DNS and DHCP server and Rsync transfers. It also handles GUI desktop roles for daily use such as browsing, media consumption and AisleRiot Solitaire

and how would you characterize the J1800 performance in Solitaire ?

 

[IntelUser2000]

You gotta be kidding. Did you notice what those 2 nodes did for mobile and server? We now have 18 cores servers and 4.5W mobile Core chips.

Server guys always executed well, and 4.5W Core M is basically Atom Pro @ $300.

 

[DrMrLordX]

I am curious what use/thought pattern you have that caused you to characterize the quad core J1900 performance as "anemic" when the more obvious choice would have been the cheaper, but faster, dual core, J1800? I have been running a J1800 almost 24/7 since shortly after release. With 4 GB DRAM, a SSD, a second NIC and Debian Jessie it handles multiple home server roles such as firewall/router, DNS and DHCP server and Rsync transfers. It also handles GUI desktop roles for daily use such as browsing, media consumption and AisleRiot Solitaire. I characterize the performance in my use pattern as quite reasonable and surprisingly livable considering the price. Is your selection of the word "Anemic" code speak for a suggestion that it is a poor choice for Crysis? The overwhelming market acceptance of Baytrail M/D at full margin with no contra revenue suggests you are out of touch with mainstream demand.

Oh for goodness' sake.

There have been entire threads in this forum complaining about Bay Trail-D and Kabini as being too weak compared to low-end Pentium/Celeron chips based on Core architectures, or even older architectures like Conroe et al. They are mobile/tablet CPUs being sold in desktops where we should be seeing chips like Haswell Pentiums/Celerons or the x4 840 instead. J1800, J1900, J2900, makes no difference. It's still Bay Trail-D. And those tasks you listed could be handled by a CPU from about a decade ago.

If you got the chip knowing full well what it was and what it could do, and found that it was a better option than some used refurb box from TigerDirect or your closet (or wherever), then hey good for you. Most consumers that wind up with these things get them as a part of an AIO machine from Best Buy or Staples. And many of those consumers suffer accordingly. OEMs are shlepping this crap on desktop consumers because they're inexpensive to buy and inexpensive to cool. Then they offer similar units with the Pentiums or Celerons for more money, when those chips don't necessarily cost anything more as a part of the system BoM (though they might cost a few dollars more to cool).