First NVIDIA Tegra 4 benchmarks are in, lags behind the iPad's A6X SoC
posted by Nick T. / Jan 15, 2013, 8:26 AM
Apparently, someone with access to a Tegra 4 evaluation board decided to take it for a spin and see what NVIDIA's new mobile chip was capable of. And what better way of doing that than running a few benchmarks aiming for a new high score? Well, whoever tested the Tegra 4 didn't quite get there, despite posting some pretty promising results on the GLBenchmark database.
The GLBenchmark results you see here were achieved using Dalmore, which is one of the two Tegra 4 evaluation boards that select developers get to play with. Android 4.2.1 was installed on it, and the display used during testing had a resolution of 1920 by 1128 pixels, as reported by the app. The Tegra 4 chip ran at a maximum clock speed of over 1.8GHz.
Still, keep in mind that the Tegra 4 SoC was not tested on a smartphone or a tablet bound to hit the market. When fitted inside an actual product, its performance could increase after the necessary optimizations are applied. Feel free to browse the results yourself, but keep in mind that they have been taken down from the GLBenchmark database. Images, however, are still available in Google's cache. The links are below!
Posts: 1482; Member since: Sep 30, 2009
Benchmarks are NOT law. If benchmarks did matter more companies would use more methods to manipulate them like they used to.
posted on Jan 15, 2013, 8:29 AM 16
Posts: 1488; Member since: Dec 29, 2011
i think the problem is Android 4.2.1 , there is something wrong with the code , even the Tegra 3 -nexus 7 has shuttering issues ,they should try to run it on jellybean and rescore
posted on Jan 15, 2013, 10:18 AM 3
Posts: 2236; Member since: Feb 14, 2011
I haven't really read up on the Tegra 4 chipset, but did they fix the memory bandwith issue that plagued the Tegra 3? Perhaps that has to do with some of the lagged performance. Or it could be simple optimization problems.
posted on Jan 15, 2013, 7:15 PM 0
It's optimization. And perhaps the drivers are not ready yet. Once it's all done. I expect the performance to be atleast more that of Nexus 10. BTW I must say ARM did an incredible job with the Mali-T604. Only 4 shader cores and it beats the hell out of Adreno 320 and A6!
posted on Jan 16, 2013, 4:05 AM 0
when benchmarks are concerned, Android will always be a=in disadvantaged position against iOS. iOS is a very light OS and the CPU does not have to work hard at all.
posted on Jan 15, 2013, 11:33 AM 1
Posts: 403; Member since: Oct 05, 2012
You are right But I am willing to get those disadvantages over Apple's too simple and boring iOS which doesn't allow me to do muti-tasks and have freedom It's like doing one job at 95% vs doing many jobs at 90%
posted on Jan 15, 2013, 6:12 PM 1
Posts: 3; Member since: Jan 18, 2013
Tell that to people who have gotten iOS5 on iPhone 3GS or iOS 6 on iPhone 4. There's not much difference at this point between the performance. Maybe 5% at most, but definitely not 30% or more, which is what people here seem to be implying.
posted on Jan 18, 2013, 4:37 PM 0
Posts: 1143; Member since: Oct 05, 2011
i still think 72 GPU cores is a bit of a Roid rage.
posted on Jan 15, 2013, 8:37 AM 3
Posts: 685; Member since: Nov 23, 2010
What do you expect from a company that specializes in GPUs? They can't do anything else other than pump up graphics. It's a bit sad because until new consoles come out, improving graphics will be pointless and that's just a maybe.
posted on Jan 15, 2013, 8:52 AM 0
Posts: 174; Member since: Nov 21, 2011
That word "cores" is misunderstood at a level over 9000. It's not actual GPU cores. It's CUDA cores like the desktop GPUs. A single GTX680 may have well over 1000 CUDA cores, however a GTX690 is a dual GPU.
posted on Jan 15, 2013, 8:56 AM 1
Posts: 148; Member since: Jan 25, 2011
They are 72 GPU CUDA cores. The CUDA cores don't have much processing power per unit. They are basically designed to executing many concurrent threads at the same time. But the execution is slower than the traditional processors. In the end the throughputs are equivalent.
posted on Jan 15, 2013, 9:47 AM 0
Posts: 1337; Member since: Oct 05, 2011
Let's wait for the score on the final and shipping product rather then an evaluation board. PA please note that readers are not initially aware of the fact this is a pre-production board and the title of the artical is at best very Mis-leading.
posted on Jan 15, 2013, 8:39 AM 2
Posts: 3722; Member since: Aug 16, 2011
Maybe if they used "pre-production" instead of "first" in the title, but I have a hard time believing anyone will actually be misled by this title (I'd say it is at WORST misleading, not at best.)
posted on Jan 15, 2013, 10:01 AM 0
Posts: 484; Member since: Jan 15, 2012
Tegra 4 never aimed to be as powerful as the A6 is: that thing is a graphics monster designed for very optimized products, while the Tegra 4 is a chip that will be used universally on mid- to high-end devices of all kinds.
posted on Jan 15, 2013, 8:47 AM 0
The reason why a6 chips fast is not because the chip is powerful, is because the utilization of the ios...
posted on Jan 15, 2013, 8:54 AM 1
Posts: 121; Member since: Jan 05, 2013
No a6 chip is fats because of the gpu ipad 4 uses sgx554mp4 which is never beaten by other gpu for now if you want beat ipad 4 just use powervr 6 series (rogue) on the geekbench score im sure the tegra 4 will beat a6x soc
posted on Jan 15, 2013, 4:45 PM 0
Posts: 3; Member since: Jan 18, 2013
God so much misinformation in these comments. ALL OF YOU ARE WRONG. The A6X is TWICE if not THREE TIMES larger than the Tegra 4 GPU. The whole SoC is 120mm2 (A5X used to be 160mm2), compared to Tegra 3 and Tegra 4's 80mmw die size. But that's for the whole chip - the GPU, and all the other components. So that means all that extra 40mm2 is completely dedicated to EXTRA GPU cores, which means it could be twice as large, if not more, than Tegra 4's GPU. That still doesn't make me happy about the fact that Tegra 4 doesn't support OpenGL ES 3.0 and OpenCL (neither does A6X), but I just hate seeing so much false information, that then everyone takes for granted.
posted on Jan 18, 2013, 4:35 PM 0
Posts: 2156; Member since: Oct 17, 2011
"The leaked GLBenchmark results are apparently from a much older silicon revision running no where near final GPU clocks."
posted on Jan 15, 2013, 8:53 AM 4
Posts: 271; Member since: Oct 19, 2012
Exactly, I wanted to post this too. NVIDIA has worked on their GeForce ULP, we have too hope that power efficiency is good too. What really bugs me is that it is not unified architecture, it will be like mali 400, which is also non-unified.
posted on Jan 15, 2013, 10:42 AM 1
Posts: 418; Member since: May 04, 2012
This is why I really like PowerVR Graphics, I just don't like the fact that the more powerful ones are on iDevices, and not on Androids. But we'll see if the Tegra 4 can beat the SGX when it's already being sold on official devices.
posted on Jan 15, 2013, 8:54 AM 0
Send a warning to post author
Send a warning to Selected user.
The user has 0 warnings currently.
Next warning will result in ban!
Ban user and delete all posts
Message to PhoneArena moderator (optional):