The Intel Atom Thread

Page 156 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
S

Shivansps

Diamond Member
Sep 11, 2013
3,851
1,518
136
The eMMC issue reported by Anadtech and Notebook check is very strange, it does not happen to my BT tablet for sure, i mean, yeah its slower than a SSD but it should not feel slower in any case, its actually faster than most common harddisks out there.
 
S

Sweepr

Diamond Member
May 12, 2006
5,148
1,142
131
Some bits about Cherry Trail from the review:

IntelAcceleration_575px.png


AirmontModuleSize_575px.png


Cherry Trail itself is a fairly small chip. While Intel hasn’t released an official die size for it (no more than they did Bay Trail), from various IDF presentations they have released official numbers for the complete package, and decent photos as well. As a result we can take a decent stab at die sizes, and from Intel’s photos we’d estimate that the die size is around 83mm2. Unfortunately we don’t have anything quite comparable to Bay Trail, though we believe it to be a good bit smaller than Intel’s previous generation tablet SoC. At 83mm2 this would put the die size at quite a bit smaller than Apple’s A8X tablet SoC, and actually is just a hair smaller than the A8 phone SoC’s 89mm2.

Diving a bit deeper, Intel has also released some size data for individual Silvermont/Airmont CPU modules. The dual core modules, which contain the 2 CPU cores along with L2 cache and appropriate glue, have shrunk significantly from Silvermont to Airmont. Overall the Airmont CPU module is 64% smaller than the Silvermont module. And to be clear that does not mean Airmont is 64% of the size of Silvermont, that means that Airmont has been reduced by 64%; relative to Silvermont it is only 36% of the die size. Intel has achieved a better scaling factor than 22nm to 14nm alone, indicating that they have almost certainly enacted further optimizations to bring down the die size as an architectural level.
Click to expand...
 
frozentundra123456

frozentundra123456

Lifer
Aug 11, 2008
10,451
642
126
That first slide is a joke. CT may be closer to broadwell, but they conveniently forgot that BW is around a year late, so CT is in fact late too.
 
R

Roland00Address

Platinum Member
Dec 17, 2008
2,196
260
126
Intel did actually reveal the die size of Baytrail they just did not highlight it to the press. It has been available for free on intel's website for years in the engineering documents that anyone can download for free.

The die size of the 3700 series of baytrail parts was 9.723mm by 10.477mm which is about 101.87 mm2
 
A

Abwx

Lifer
Apr 2, 2011
10,939
3,440
136
bullzz said:
you mean it is similar to ipad? not bad i say

http://www.notebookcheck.net/Apple-iPad-Air-2-A1567-128-GB-LTE-Tablet-Review.129396.0.html
Click to expand...

Dont know comparitively with an iPad, OS and benches are too different, comparatively to a Baytrail the CPU has about the same IPC, progress is obviously the GPU.

Real power comsumptions number cant be deduced from this review since the adaptator is only 13W, i guess that the chip has an efficency that is about the one of Baytrail if we are to interpret the stress test numbers and the voltages used, notably higher than BT, but surely compensated, at least partly, by smaller parasistic capacitances of the 14 nm process.

Overall it s not the CPU that motivated MS to use this chip but rather its GPU since BT s one is simply unsoldable as such...
 
J

jpiniero

Lifer
Oct 1, 2010
14,584
5,206
136
Oh I see what the problem is with the Surface 3 - the idle power consumption is terrible. I wonder why.
 
cbn

cbn

Lifer
Mar 27, 2009
12,968
221
106
Arachnotronic said:
No clue, sadly. I would expect more EUs + better architecture with Gen. 9 to deliver a nice boost, though.
Click to expand...

Hopefully the Goldmont cores are a good deal beefier. (re: the current 14nm SoC seems too light on CPU relative to iGPU).

With regard to Braswell, I would hope we see a 10 watt SKU to help increase the CPU to iGPU performance balance. But maybe Intel plans on using chips like Pentium 3825U (2 core/ 4 thread) at 10 watt cTDP instead?
 
S

Shivansps

Diamond Member
Sep 11, 2013
3,851
1,518
136
cbn said:
Compared to my Athlon 5350 at 1080p low, the results shown in the video at 1280 x 720 low are about 10 FPS lower (on average).

My guess is CPU bottleneck.
Click to expand...

Im 100% sure that CPU turbo does not work while the IGP is in use, BT work like this, i dont see how CT could be any different.
 
T

twjr

Senior member
Jul 5, 2006
627
207
116
Arachnotronic said:
One more big leap in graphics, and a solid jump in CPU, should make Broxton-based 2-in-1s quite decent Windows gaming machines.
Click to expand...

I'd just like to be able to play Civ 5 in touch screen mode with a few more civs. My baytrail tablet just isn't quite there.
 
cbn

cbn

Lifer
Mar 27, 2009
12,968
221
106
Shivansps said:
Im 100% sure that CPU turbo does not work while the IGP is in use, BT work like this, i dont see how CT could be any different.
Click to expand...

So 1.6 Ghz quad core with 16 EUs for gaming loads.

And the iGPU clock runs full speed?
 
Last edited:
S

Shivansps

Diamond Member
Sep 11, 2013
3,851
1,518
136
cbn said:
So 1.6 Ghz quad core with 16 EUs for gaming loads.

And the iGPU clock runs full speed?
Click to expand...

For like 10mins, BT-T cant mantain 100% cpu and igp AND IGP turbo, but it depends on cooling. Not sure how realistic is that either, you are kinda never at 100% load on everything.

BT-T where able to maintain full 100% CPU turbo or 100% IGP turbo for a very long time (+1h easily).

Anyway, that was with BT, it needs to be re-checked with CT, but its probably the same. So yeah, CPU turbo is unlikely to work while gaming, it is disabled to enable IGP turbo on tablet SKUs.

And thats the main difference with Braswell parts, just like BT-M, cpu and igp turbo can be used as long there is good cooling.
 
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom