Qualcomm Krait & MSM8960 @ AnandTech

[Wishmaster]

Anand compared Krait directly with Tegra3 running Transformer Prime
http://www.anandtech.com/show/5563/qualcomms-snapdragon-s4-krait-vs-nvidias-tegra-3
Those results show that Krait can easily beat Tegra 3 in CPU performance, and match if not surpass it in GPU department(wonder how it looks like at higher resolutions like 1080p?).
Hopefully this time ODM's won't cripple those Krait based devices just like they did last year with msm8660/8260.

 

[metafor]

Performance varied depending on device for 8x60 devices. Sensation XL, Amaze and some of the smaller brand phones scored far better than, say, the rezound. I suspect it has to do with throttling the CPU harder to conserve power to make up for the LTE modem.

 

[Mariner]

Here's TI throwing their hat into the ring in as well:

http://www.anandtech.com/show/5571/...hz-dual-core-a15-vs-quad-core-a9-13-ghz-video

We're expecting Krait to be comparable to the A15, I believe? This indicates that both ought to do well against Tegra3.

NV have done well to get the T3 to market first once again, however. OMAP 5 looks to be some time away from the market so I suppose it will be up to Qualcomm to get Krait in devices as soon as possible to take the performance crown. The other A15 implementations are still some way off as well, I believe?

 

[Helmore]

We're expecting Krait to be comparable to the A15, I believe?

I believe most people here expect Cortex-A15 to be faster than Krait clock for clock, but to consume a bit more power as well. We won't know for sure unless we have benchmark numbers of course.

 

[Deleted member 13524]

Humm, so the Cortex A15 is dependent on heterogeneous computing (with a Cortex A7 made on LP?) in order to achieve smartphone power-levels, but Krait may not need such a thing?

 

[dagamer]

Humm, so the Cortex A15 is dependent on heterogeneous computing (with a Cortex A7 made on LP?) in order to achieve smartphone power-levels, but Krait may not need such a thing?

I wouldn't extrapolate that far. TI has more recently started to stick in Cortex M3 cores to help with some tasks, with at the OMAP4470 being the most newest example. Samsung however does not have any "little" cores as far as I know.

I would hope that a dual-core Cortex A15 beats the Tegra 3, since those chips won't be in shipping products en masse for at least another year.

 

[Deleted member 13524]

I wouldn't extrapolate that far. TI has more recently started to stick in Cortex M3 cores to help with some tasks, with at the OMAP4470 being the most newest example. Samsung however does not have any "little" cores as far as I know.

I would hope that a dual-core Cortex A15 beats the Tegra 3, since those chips won't be in shipping products en masse for at least another year.

In OMAP4, the Cortex M3 are "hardwired" to the imaging and video DSPs, they're not accessible to the O.S. in any way and we can't call it "heterogeneous computing". The same happens with OMAP5 having two Cortex M4.

 

[Exophase]

Humm, so the Cortex A15 is dependent on heterogeneous computing (with a Cortex A7 made on LP?) in order to achieve smartphone power-levels, but Krait may not need such a thing?

According to who? At least three A15s have been announced, none of them have big.LITTLE as far as we know, and I'm sure all of them are intended for phones.

We also don't have any indication that the Cortex-A15s won't be LP, I'm pretty sure many Cortex-A8s and A9s in mobile SoCs were, not counting nVidia's.

 

[Wishmaster]

Don't know if anyone's interested but still I thought that those informations should reach out to more people:
http://www.qualcomm.com/media/blog/...dview-snapdragon-qualcomm-introduces-brite-an - similar to the highly praised technology used in tegra3
http://www.qualcomm.com/media/blog/...andles-51-surround-sound-capture-and-playback - not revolutionary, but still good to know that it is there ;-)
http://www.qualcomm.com/media/blog/2012/02/21/s4-s-adreno-gpu-more-just-gaming - it confirms that S4 uses HDMI 1.4 and that it is WiFi Display compatible.
http://www.qualcomm.com/media/blog/2012/02/17/snapdragon-s4-camera-revealed - image stabilization in action, gesture controls à la kinect

 

[metafor]

Localized dimming and color/intensity modulation has existed for a while. I believe Intel was the first to show it off in laptops way way way back a decade ago.

 

[Wishmaster]

Localized dimming and color/intensity modulation has existed for a while. I believe Intel was the first to show it off in laptops way way way back a decade ago.

I never said it is revolutionary, it just shows that qualcomm has saw the power of marketing(similar to what nvidia is doing with their tegra family).

 

[Laurent06]

Out topic, but as it was discussed in this thread, here is an article discussing Linux support for ARM big.LITTLE.

 

[metafor]

Interesting. How does the Hypervisor determine when to swap clusters? Is there some form of real-time profiling and information from the guest OS's scheduler?

 

[Exophase]

Interesting. How does the Hypervisor determine when to swap clusters? Is there some form of real-time profiling and information from the guest OS's scheduler?

According to that article it doesn't, it's told to by the OS.

 

[metafor]

According to that article it doesn't, it's told to by the OS.

I was talking about ARM's prototype hypervisor. Although the solution they put in there by mapping to dynamic frequency is fairly clever, yes.

 

[Exophase]

I was talking about ARM's prototype hypervisor. Although the solution they put in there by mapping to dynamic frequency is fairly clever, yes.

I was too.

"With a single call, the hypervisor can atomically suspend execution of the whole system, migrate the CPU states from one cluster to the other, and resume system execution on the other cluster without the underlying operating system being aware of the change; just as if nothing has happened."

The code still has to be instructed to perform the switch, it doesn't happen automatically.

 

[wco81]

So the HTC One has Tegra 3 in the international versions and a Krait-based SOC in the US one, with integrated LTE modem.

People are already making comparisons to the quad-core T3 versus the dual-core in the US model, wondering if it's the same situation as Samsung phones which have better SOCs in the international versions than the US version with LTE.

Is this the first high-profile Krait design?

 

[Wishmaster]

So the HTC One has Tegra 3 in the international versions and a Krait-based SOC in the US one, with integrated LTE modem.

People are already making comparisons to the quad-core T3 versus the dual-core in the US model, wondering if it's the same situation as Samsung phones which have better SOCs in the international versions than the US version with LTE.

Is this the first high-profile Krait design?

Depends. Does asus padfone and all LTE version of Transformer Pads count? And don't forget about HTC One S with msm8260a(basically 8960 withouth LTE).
And don't forget about panasonic eluga and many other phones running on snapdragon s4, much more than on the tegra3.

 

[metafor]

Interestingly enough, the benchmarks of the HTC One S are worse off compared to a 1.5GHz A33 Tegra. Albeit Anand didn't have time to benchmark Basemark (which Adreno performs great in) nor Vellamo and Linpack, which Krait excels in.

But I do find it disturbing that there is quite a difference between the MDP and OEM software. Hopefully other manufacturers will improve on this.

 

[Helmore]

Interestingly enough, the benchmarks of the HTC One S are worse off compared to a 1.5GHz A33 Tegra. Albeit Anand didn't have time to benchmark Basemark (which Adreno performs great in) nor Vellamo and Linpack, which Krait excels in.

But I do find it disturbing that there is quite a difference between the MDP and OEM software. Hopefully other manufacturers will improve on this.

Anand stated that he thinks software will improve in future builds, which will have an impact on those benchmarks they ran. I'd say we wait until we have retail handsets to see how they compare performance-wise.