It seems to me that Primate Labs has selected a strange combination of tests and combine them using an arithmetic mean method that does not give reliable results. I wish you would talk more about some of the individual scores rather than the composite.
Which of the 4 Geekbench sections do you think has most relevance to a smart phone and also if the same is true for tablets.
Some of the tests seem reasonable but other choices seem strange. How are the 4 sections weighted when they are combined? 50% of the integer result is based on compress/decompress. Does a smartphone really do a lot of encrypt/decrypting? Lua instead of Java?
How do the memory read and write tests differ from the stream copy?
Geekbench combines 4 benchmark sections some way to generate a composite number.
1. Integer: encrypt/decrypt, text compress/decompress, image compress/decompress, Lua script to compute prime numbers
2. Floating point: Mandelbrot, DOT product, LU decomposition, Lucas-Lehmer prime number test , Sharpen and Blur Image
3. memory performance: sequential memory read, write, system malloc()/free(), memset(), memcpy()
4. stream performance: computes copy(a[I]=b[I]), computes scale a(i) = q * b(i), computes add a(i) = b(i) + c(i), and computes triad a(i) = b(i) + q * c(i).
"How much faster is Clovertrail plux vs Clovertrail ?? Current benchmarks has NVidia's Tegra 4 at 3x faster than clovertrail.."
[However, this year's flagship smartphones from HTC, Samsung, and Sony aren't using the Tegra 4 and we all know Apple won't be using anything but its own custom chips for the next iPhone. That's practically the entire handset market. Even for tablets, we have yet to see any big Tegra 4 design wins, though they may arrive later this year. The shift toward smaller, more affordable slates, may account for the seemingly lackluster interest in a high-end and more expensive chip like the Tegra 4.]
[Tegra 4 had zero non-reference design wins announced at launch.]
From Charlie at SemiAccurate who said that Nvidia has a pattern of big promises followed by substantially lessor performance...
"Come on Wallis, put your balls on the line, what will Clovertrail+ score?:)"
[You do a half-azzzed comparison to older hardware, get called on it, and now you want to talk smack. Pretty funny. Ask me a reasonable question and I'll answer it. I don't blame you for being upset over the end of the ARM era. It has to be a massive disappointment. Deal with it. You really thought ARM was going to continue to sit on top with the "just good enough" attitude, failing fabrication, insufficient R&D, and the limited funding that the fabless concept insures? ]
You think you're going to get mileage off a chart with the Atom Z2480? LOL
The Z2580 supports four times as much storage, double the cores/threads, keeps a 2GHz boost clock, doubles the maximum RAM, improves the camera imaging sensor and capabilities, and adds support for Android 4.2 (Jelly Bean).
My bad...I had taken the "monk" on his word and considered him as an unbiased geek. Looks like he is just a basher posing as a geek...!
My post headline was true. The A15 has been beaten (for now) and the Intel SoC has the highest clock (but not the highest performance).
As for unbiased, perhaps not, but I am not totally in the ARM camp, I just don't agree with Wallis and others regarding Intel's potential in the App processor space. Much of what is written about the ARM/Intel 'battle' is missing some of the story. As an example, all this 'hype' about the App processor in the Galaxy S4 missed the big win for Intel - it's got the baseband for the S4 international version.