NVIDIA Tegra Architecture

There are plenty of wide aspect ratio tablets available, there will be plenty more. This one just fits someone else's requirements.

I know, but this one has most of the stuff I wanted: new and long-lasting 64bit SoC with great performance, new OS, good quality stereo speakers, high-DPI IPS screen, right size...

Almost everything but that damn aspect ratio...

I guess I'll just hope for a new TF7xx will come out with stereo speakers and a Denver TK1. My TF700 is showing its age.
 
Google is able to provide CPU performance, GPU performance, screen quality, audio quality, OS quality, and aesthetics that are as good or better than the next gen iPad Air 2 for $100-$120 less money, so that is a pretty good accomplishment.

I love how you can make this confident statement of fact before the next gen iPad Air 2 is even announced, and none of the devices have been tested.
 
Excellent Integer and Memory performance against A8. Some of the floating point performance is significantly lower, however. Any idea why that might be?
 
I love how you can make this confident statement of fact before the next gen iPad Air 2 is even announced, and none of the devices have been tested.

The performance of the Denver CPU core is a pretty well known entity now (shown at Hot Chips), as is the performance of mobile Kepler GPU (TK1 32-bit), enhanced Cyclone CPU (A8), and GX6650 GPU (50% faster than GX6450 in A8).
 
Excellent Integer and Memory performance against A8. Some of the floating point performance is significantly lower, however. Any idea why that might be?

As mentioned above by Nebu, this Nexus 9 Geekbench 3 data appears to be using AArch32. So measuring using AArch64 should give a huge boost to some of the floating point numbers.

When Anandtech reviewed the iPhone 5s, moving from AArch32 test to AArch64 test in Geekbench 3, they measured an improvement of +25% in BlackScholes, +16% in Blur Filter, +195% in DGEMM, +119% in SFFT, +26% in N-Body, +51% in RayTrace ( http://www.anandtech.com/show/7335/the-iphone-5s-review/4 )

Here is a Geekbench 3 comparison between HTC Volantis Nexus 9 and iPhone 5s, both using AArch32: http://browser.primatelabs.com/geekbench3/compare/1014854?baseline=1002621

The single-core performance in this comparison is 80-90% higher than first gen Cyclone! Now, exactly how much improvement there will be from AArch32 to AArch64 with the Denver core is unknown at this time (the Cyclone core shows an improvement of ~ 35% overall). It is also unclear exactly what CPU clock operating frequency is being used for the Denver cores benchmarked here.

At Hot Chips, Denver was said to have a Geekbench 3 single-core score that was ~ 1.65x higher than R3 Cortex A15 in Tegra K1, and ~ 1.375x higher than Cyclone in A7 iPhone 5s. So I don't know exactly what to make of these two new AArch32 data points. The only way for these numbers to make sense is if NVIDIA showed AArch32 Geekbench 3 data for Denver at Hot Chips while comparing it to AArch64 Geekbench 3 data for Cyclone.
 
Last edited by a moderator:
As mentioned above by Nebu, this Nexus 9 Geekbench 3 data appears to be using AArch32. So measuring using AArch64 should give a huge boost to some of the floating point numbers.

When Anandtech reviewed the iPhone 5s, moving from AArch32 test to AArch64 test in Geekbench 3, they measured an improvement of +25% in BlackScholes, +16% in Blur Filter, +195% in DGEMM, +119% in SFFT, +26% in N-Body, +51% in RayTrace ( http://www.anandtech.com/show/7335/the-iphone-5s-review/4 )

Here is a Geekbench 3 comparison between HTC Volantis Nexus 9 and iPhone 5s, both using AArch32: http://browser.primatelabs.com/geekbench3/compare/1014854?baseline=1002621

The single-core performance in this comparison is 80-90% higher than first gen Cyclone! Now, exactly how much improvement there will be from AArch32 to AArch64 with the Denver core is unknown at this time (the Cyclone core shows an improvement of ~ 35% overall). It is also unclear exactly what CPU clock operating frequency is being used for the Denver cores benchmarked here.

At Hot Chips, Denver was said to have a Geekbench 3 single-core score that was ~ 1.65x higher than R3 Cortex A15 in Tegra K1, and ~ 1.375x higher than Cyclone in A7 iPhone 5s. So I don't know exactly what to make of these two new AArch32 data points. The only way for these numbers to make sense is if NVIDIA showed AArch32 Geekbench 3 data for Denver at Hot Chips while comparing it to AArch64 Geekbench 3 data for Cyclone.
Anand really messed up the descriptions in that article, I think there's many cases of confusion between ARMv7 and AArch32, for example the integer tables are a clear error as you can take advantage of crypto instructions in AArch32 - which is clearly not represented in that table.

The end result between AArch32 vs AArch64 should be much less than what's depicted in the Cyclone comparison.
 
Anand really messed up the descriptions in that article, I think there's many cases of confusion between ARMv7 and AArch32, for example the integer tables are a clear error as you can take advantage of crypto instructions in AArch32 - which is clearly not represented in that table.

The end result between AArch32 vs AArch64 should be much less than what's depicted in the Cyclone comparison.

That makes some sense, because the HTC Volantis Nexus 9 crypto integer scores in AArch32 are already much higher than iPhone 6 in AArch64.

It appears that the primary area of benefit in AArch64 will be in improving some of the floating point scores.
 
I find it odd that this claims 2.5GHz while Google says Nexus 9 is running @2.3GHz.

Why is FP better? Has the ARMv8 instruction set improvements in that area?
AArch64 supports IEEE FP numbers in the SIMD engine. So this means some form of FP vectorization is now possible.
 
I find it odd that this claims 2.5GHz while Google says Nexus 9 is running @2.3GHz.
The maximum frequency populated in the CPUFreq table is not the same as the maximum CPUFreq policy frequency - this is just something you'll have to consider and filter when looking at clock rates reported by some apps.
 
Back
Top