AnTuTu and Intel

[R0H1T]

Never forget, Intel's SDP > AMD's ACP.

Yeah sure if it makes you sleep well at night :thumbsup:

 

[Zucker2k]

Linus Torvalds' take on Antutu and Neon:
www.realworldtech.com/forum/?threadid=133996&curpostid=134835

 

[Nothingness]

Linus Torvalds' take on Antutu and Neon:
www.realworldtech.com/forum/?threadid=133996&curpostid=134835

It's a take on NEON. If you want his take on AnTuTu, you should read this: http://www.realworldtech.com/forum/?threadid=133996&curpostid=134950

Ha yes, it's not the link Francois Piednoel put on his twitter account as it paints a less rosy picture for Intel.

 

[Exophase]

Linus Torvalds' take on Antutu and Neon:
www.realworldtech.com/forum/?threadid=133996&curpostid=134835

Linus is dead wrong here, many sane people use NEON. I think he's pretty out of touch on this one. It's pretty outrageous to say it's insane to go for huge performance benefits because some people have or had tablets with Tegra 2, especially with all the apps that are only available on iOS. The arguments about compatibility with Cortex-M series devices is just silly, no one cares if your software can easily be ported to a microcontroller.

I'm releasing Android software soon that is utterly dependent on NEON for performance, to the extent that Tegra 2 devices will be blacklisted outright because it's not worth their money to run a version using the vastly slower non-NEON paths. They shouldn't be wasting their money on it (and I shouldn't have to deal with angry people who want their money back).

nVidia blew it with Tegra 2 but fortunately there weren't others dumb enough to repeat this mistake in the mobile space, and nVidia fixed it with Tegra 3 onwards. At this point I doubt an awful lot of the market is using Tegra 2.

Stuff like SPEar on the other hand is a great justification for why ARM made NEON optional with Cortex-A9 - it's an embedded product meant for running fixed software and they evaluated that NEON wasn't worth the investment for this space.

 

[ChronoReverse]

I'm releasing Android software soon that is utterly dependent on NEON for performance, to the extent that Tegra 2 devices will be blacklisted outright because it's not worth their money to run a version using the vastly slower non-NEON paths. They shouldn't be wasting their money on it (and I shouldn't have to deal with angry people who want their money back).

A good decision. You'll be finding out soon that Android users are often very vocal complainers in the least useful way. If their device can't run it for a good reason, they won't hesitate to post a flaming review.

I'm assuming your piece of software will be DraStic

nVidia blew it with Tegra 2 but fortunately there weren't others dumb enough to repeat this mistake in the mobile space, and nVidia fixed it with Tegra 3 onwards. At this point I doubt an awful lot of the market is using Tegra 2.

A lot of Tegra2 devices got left behind with the jump to ICS so by now their users are pretty used to incompatibility

 

[Exophase]

A good decision. You'll be finding out soon that Android users are often very vocal complainers in the least useful way. If their device can't run it for a good reason, they won't hesitate to post a flaming review.

I'm assuming your piece of software will be DraStic

Yeah it's DraStic. My contribution to PCSX-ReARMed is also highly dependent on NEON.. but that's not the only part of that codebase that utilizes it. Mupen64plus also uses NEON. Lots of people are asking NEON questions on StackOverflow, I think it's a given that it's used in a variety of games and libraries. Avoiding NEON just because of Tegra 2 seems pretty daft to me. Calling anyone who uses NEON insane is way out there.

Too bad Google refused to make an armv7a-neon target for their fat binaries, maybe to avoid pissing nVidia off.. instead they insist you check for it at run time and have fallbacks. It's funny because there are a ton of other flags you can use to block platforms with, like not supporting certain texture compression formats. Good thing they let you block individual products.

 

[krumme]

No harm done repeating it. Damn good, solid work Exophase - i would have liked to understand more than 10% of it, but i know that if it had been wrong the sky would have fallen on you by now.

 

[pelov]

Stuff like SPEar on the other hand is a great justification for why ARM made NEON optional with Cortex-A9 - it's an embedded product meant for running fixed software and they evaluated that NEON wasn't worth the investment for this space.

The majority of ARM cores go into embedded products. The bare bones SoCs/core designs are still very much influenced by that. They design vanilla cores purposely because the chips will eventually end up in everything from those sweet Japanese toilets to iPads.

Does anyone else find it comical that he seems to have completely forgotten about 3DNow!, SSEx>SSExx.x, x87, and the slew of other now defunct x86 extensions/ISAs?

 

[lamedude]

AFAIK 3DNow! and Cyrix's Extended MMX are the only defunct extensions on x86. They were barely relevant and Intel never supported them so requiring them was never an option. Skyrim has proved you can require SSE2 without AthlonXP guy coming out of the woodwork to complain that you broke his software.

 

[Idontcare]

Yeah it's DraStic. My contribution to PCSX-ReARMed is also highly dependent on NEON.. but that's not the only part of that codebase that utilizes it.

Kudos and congrats Exophase :thumbsup:

 

[galego]

Qualcomm official reaction to Intel-Antutu issue:

Q. Intel put out a press release last month about an Atom Z2580 outperforming several ARM-based competitors while using less power. As an incumbent in this space, what was Qualcomm’s reaction based on the metrics and methodology used?

A. Recent articles confirm what we have been saying for some time. First, analysts have consistently reported that Qualcomm Snapdragon processors lead in overall mobile performance when you look across a broad range of benchmarks and outperform the competition, including Intel in virtually all of them. Second, and more important a lesson from this week is that looking at performance through a single benchmark can lead to inaccurate conclusions. With the corrections AnTuTu made this week there were swings in some of their measurements (one as much as 50%) and the new results are now showing that ARM processors lead Intel processors. Finally, it's important to look beyond CPU–specific benchmarks, where the CPU is only about 15% of the SoC and it is the other 85% of the chip that is important to deliver great gaming, camera, web and multimedia experiences.

http://www.tomshardware.co.uk/qualcomm-ama-toms-hardware,review-32736.html

 

[Arachnotronic]

Qualcomm official reaction to Intel-Antutu issue:



http://www.tomshardware.co.uk/qualcomm-ama-toms-hardware,review-32736.html

[redcated]

Personal attacks will not be tolerated
-ViRGE

 

[krumme]

Qualcomm official reaction to Intel-Antutu issue:



http://www.tomshardware.co.uk/qualcomm-ama-toms-hardware,review-32736.html

I dont see anything wrong with this response from qualcomm. But obviously its hard for some to accept Intel is both cheating and using all their ressources at the 15% in a market where nobody wants to buy their products. What a waste of talent.

 

[beginner99]

I dont see anything wrong with this response from qualcomm. But obviously its hard for some to accept Intel is both cheating and using all their ressources at the 15% in a market where nobody wants to buy their products. What a waste of talent.

Besides the fact that I often have trouble understand what you want to say, I just got myself the RAZR I. I'm pretty content so far. I must admit however that my purchase was driven by the form factor (4.3 in a small chassis compared to other phones), design and good pricing rather than "tech". But reviews said it was one of the snappiest phones they ever tested and snappier than other low-midrange products wit usually run low clocked "no-name" ARM CPUs.

 

[krumme]

Besides the fact that I often have trouble understand what you want to say, I just got myself the RAZR I. I'm pretty content so far. I must admit however that my purchase was driven by the form factor (4.3 in a small chassis compared to other phones), design and good pricing rather than "tech". But reviews said it was one of the snappiest phones they ever tested and snappier than other low-midrange products wit usually run low clocked "no-name" ARM CPUs.

Just shows the SoC is only 10% - at most - of the experience. The cpu part is 15% of the 10%. I doubt i would notice if the s600 in my phone secretly was swapped with a quad a7.

 

[Nothingness]

Just shows the SoC is only 10% - at most - of the experience. The cpu part is 15% of the 10%. I doubt i would notice if the s600 in my phone secretly was swapped with a quad a7.

Indeed. As an example latest Galaxy Tab 3 with Intel inside is getting bad reviews all around the web. If razr i was snappy thanks to its processor, then the Tab 3 should be good too.

Example: http://www.laptopmag.com/reviews/tablets/samsung-galaxy-tab-3-10-1.aspx

 

[StrangerGuy]

Indeed. As an example latest Galaxy Tab 3 with Intel inside is getting bad reviews all around the web. If razr i was snappy thanks to its processor, then the Tab 3 should be good too.

Example: http://www.laptopmag.com/reviews/tablets/samsung-galaxy-tab-3-10-1.aspx

Conspiracy: What is there to stop Samsung from putting Intel in a bad light by deliberately sabotaging the performance of Atom products? It's not one failed model out of a several dozen Galaxy products is gonna hurt in the long run.

"You see, we actually had our own Atom tablet but it sucked hard, so no more Atom again!"

In the same vein of the complier tricks and strongarming tactics Intel used in the past. The circle is complete.

 

[Nothingness]

Conspiracy: What is there to stop Samsung from putting Intel in a bad light by deliberately sabotaging the performance of Atom products? It's not one failed model out of a several dozen Galaxy products is gonna hurt in the long run.

"You see, we actually had our own Atom tablet but it sucked hard, so no more Atom again!"

In the same vein of the complier tricks and strongarming tactics Intel used in the past. The circle is complete.

I have another possible explanation: Razr I used an Intel designed platform both HW and SW and was hence highly tuned.

 

[StrangerGuy]

I have another possible explanation: Razr I used an Intel designed platform both HW and SW and was hence highly tuned.

Which raises the question how is Intel going to ensure the performance when they don't even have control of the phone? They might as well release their own branded phones right? Whoops can't do that, Lenovo & co won't be happy.

Uphill battle indeed.

 

[monstercameron]

I have another possible explanation: Razr I used an Intel designed platform both HW and SW and was hence highly tuned.

but Stanger's reasoning is more entertaining...

 

[Nothingness]

Which raises the question how is Intel going to ensure the performance when they don't even have control of the phone? They might as well release their own branded phones right? Whoops can't do that, Lenovo & co won't be happy.

Uphill battle indeed.

Yes, that's one of the issues they have. Note that if Merrifield is as great as Intel claims it is, this will be less of an issue.

@monstercameron: definitely

 

[jhu]

Just shows the SoC is only 10% - at most - of the experience. The cpu part is 15% of the 10%. I doubt i would notice if the s600 in my phone secretly was swapped with a quad a7.

Not quite. The A7 is really slow. About on par with an A8. What you might not notice is dual core A9. The performance difference between my Barnes & Noble Nook (Cortex A8) and Nook Tablet (dual-core Cortex A9) was night and day.

 

[monstercameron]

Not quite. The A7 is really slow. About on par with an A8. What you might not notice is dual core A9. The performance difference between my Barnes & Noble Nook (Cortex A8) and Nook Tablet (dual-core Cortex A9) was night and day.

and a tablet with an allwinner a31? could you really tell the different in normal use cases?

 

[StrangerGuy]

Not quite. The A7 is really slow. About on par with an A8. What you might not notice is dual core A9. The performance difference between my Barnes & Noble Nook (Cortex A8) and Nook Tablet (dual-core Cortex A9) was night and day.

Funny because my Mediatek quad A7 phone feels really fast. If ~20% slower than a Galaxy S3 is slow that is so off the mark it's ridiculous.

 

[jhu]

and a tablet with an allwinner a31? could you really tell the different in normal use cases?

My dad has one. It's really slow.