Quick Video AT&T One X vs Tegra 3 One X Benchmarks and Look

[superscientific]

Here is a video where I got to compare the AT&T One X at a local AT&T store with my Tegra 3 International One X and got to run a few benchmarks simultaneously. I also got to play with the AT&T version and filmed a super quick UI speed comparison. I am impressed with the S4 variant. I think once LTE becomes live in my area then I will purchase one. Enjoy the video!
[YT]http://www.youtube.com/watch?v=Y6xb8t3FoFI[/YT]

 

[fernandez21]

Cool video, thanks. I wonder how battery will compare between them?

 

[superscientific]

Yeah I am wondering too. I am thinking about buying this version to compare the two and keeping the one I like best but I would have to buy it out right. I just don't think there would be enough difference to make me spend that much money. I have really good battery life on my international version and 32gb of storage. LTE would be the big winner here but I don't have LTE yet in my area.

 

[bplewis24]

Great video, but I only take exception to one thing you said in the video around the 2:10 mark. You mentioned that Tegra 3 has more cores and some would assume that it should perform better in the Nenamark test. However, Tegra 3 only has more CPU cores, not GPU cores (that I'm aware of). So the number of cores in the SoC wouldn't be relevant to a GPU frames per second benchmark.

Brandon

 

[RonD]

I get an error when trying to view the video. Has the video been pulled? What were the results.

 

[neoenigma]

I get an error when trying to view the video. Has the video been pulled? What were the results.

Quadrant Standard
Tegra 3: 4988
Qual S4: 4794

Linpack Multithread
Tegra 3: 132, 133, 132
Qual S4: 171, 214, 206

Nenamark 2
Tegra 3: 48 FPS
Qual S4: 58 FPS

 

[MattMJB0188]

I had originally wanted the Tegra but after seeing that the S4 is better for real world performance, I'll take that any day now. Really glad to see that the device is more smooth with the S4.

Was the total opposite regarding the Galaxy S II line last fall.

 

[Berzerker]

Great video, but I only take exception to one thing you said in the video around the 2:10 mark. You mentioned that Tegra 3 has more cores and some would assume that it should perform better in the Nenamark test. However, Tegra 3 only has more CPU cores, not GPU cores (that I'm aware of). So the number of cores in the SoC wouldn't be relevant to a GPU frames per second benchmark.

Brandon

I'm fairly certain the Adreno in the S4 doesn't have a 32-core GPU lol.

 

[superscientific]

The Tegra 3 has a 12 core GPU

 

[I Monarch]

I own a quad-core One X, and work at at&t. I've spent plenty of time playing with the S4 version, and I can't see any differences between my phone's performance and the Tegra 3 in my phone. The one thing I DO notice, however, is some ICS features have been stripped out of the at&t version. I love being able to uninstall apps from the app drawer, and you can't do that with the S4 One X. Little things like that make me glad I went unlocked.

 

[superscientific]

Plus we can play tegra optimized games! I always favor unlocked international variantsof phones but LTE is enticing

 

[UCLA 15]

The Tegra 3 GPU is nothing special. Back when I was thinking about getting a Transformer Prime, the general consensus about its weak spot (after weeks of technical discussions on XDA) in terms of performance was the GPU. The iPad 2's GPU is better (2, not the new iPad, which of course beats it by an even larger degree). Thing is, the S4's GPU is also nothing special (Adreno 225). The S4/Tegra 3 CPUs are significant improvements over 2011 CPUs, but unfortunately they are still using GPUs with 2011 performance. These processors really need the next gen GPUs to shine. Adreno 3xx, the Mali that will be coming with Samsung's dual-core A15 exynos, and the next gen PowerVR will be worth talking about.

As for now, it seems like the S4 is actually the better chip for real world performance, which is what really matters. Battery life will also be interesting to see, for me I think the S4 will win that too.

 

[Kevin OQuinn]

The Tegra 3 GPU is nothing special. Back when I was thinking about getting a Transformer Prime, the general consensus about its weak spot (after weeks of technical discussions on XDA) in terms of performance was the GPU. The iPad 2's GPU is better (2, not the new iPad, which of course beats it by an even larger degree). Thing is, the S4's GPU is also nothing special (Adreno 225). The S4/Tegra 3 CPUs are significant improvements over 2011 CPUs, but unfortunately they are still using GPUs with 2011 performance. These processors really need the next gen GPUs to shine. Adreno 3xx, the Mali that will be coming with Samsung's dual-core A15 exynos, and the next gen PowerVR will be worth talking about.

As for now, it seems like the S4 is actually the better chip for real world performance, which is what really matters. Battery life will also be interesting to see, for me I think the S4 will win that too.

I'm a technical person by nature, so I feel the need to point this out. The Adreno 225 is computationally as powerful as the PowerVR SGX543MP2. The problem is that Qualcomm felt the need to have other bottlenecks hinder the performance (it's very memory bandwidth restricted). This problem will be solved with the Adreno 3xx because it will be a completely new architecture that won't be as constrained by other bottlenecks in the system.

 

[I Monarch]

I use my device for Twitter, Reddit, Google searches, music, and photos. Either version of the One X fits my needs, and I just so happen to have caught the Daily Steals sale for the international version.

 

[SlimJ87D]

I've dived a little deeper and looked at the benchmarks and I've noticed that the Tegra 3 actually beats the S4 in the CPU department, but the S4 gets bumped up by "memory" in the benchmarks.

Can anyone explain to me why the "memory" benchmark part for the Tegra 3 is not so impressive as the S4?

This was similar to the Galaxy S2 SoC S3 vs Exynos where the S3 Qualcom chip had a major boost in "Memory" but the Exynos had a higher CPU benchmark score.

 

[UCLA 15]

I'm a technical person by nature, so I feel the need to point this out. The Adreno 225 is computationally as powerful as the PowerVR SGX543MP2. The problem is that Qualcomm felt the need to have other bottlenecks hinder the performance (it's very memory bandwidth restricted). This problem will be solved with the Adreno 3xx because it will be a completely new architecture that won't be as constrained by other bottlenecks in the system.

Hmm, I remember that the weak spot with Tegra 3's GPU was also a bottleneck. The big complaint on the TF Prime forums on XDA was the 32 bit wide single channel interface, when the other big players were using dual channel. It doesn't get more disappointing than when you have a lot of potential but bottlenecks like that.

 

[SlimJ87D]

Here's an example. The Tegra 3 scores higher in CPU and the S4 in memory.

S4 HTC ONE XL

S4 HTC ONE S (left) Tegra 3 HTC ONE X (right)

So if we did a partition change on the Tegra 3 to increase it's memory performance lets say with an Ext 4, are you telling me that the Tegra 3 will outperform the S4?

Can someone explain to me more clearly and technically as possible why the memory is what makes the S4 beat the Tegra 3 and why it matters?

EDIT: Sorry, I am a fool. I read the chart wrong, I thought the red area was I/O and the green area was memory, I got it mixed up.

 

[SlimJ87D]

Here are more test done by Slash gear HTC One X Review (AT&T) - SlashGear:

 

[SlimJ87D]

Here are more test done by Slash gear HTC One X Review (AT&T) - SlashGear:

"The Tegra 3 model has significantly stronger CPU performance, unsurprisingly, almost double what the S4 can manage; however, the Snapdragon delivers roughly twice the memory I/O. That?s in no small part down to its dual memory channels, and means data read/write is faster."

"With the display set to be permanently on, along with GPS, and with mixed use of WiFi and cellular data, playing videos and browsing the internet, both versions of the One X managed to run for over eight hours with 26-percent power left. Bear in mind this is an extreme case: with more typical ? though still heavy ? use, we went for nearly a full day on a single charge."