Replies to post #130399 on Intel Corp (INTC)

[simplytom]

@ibc

Yes Geekbench is stupid...

at the Moment I'm looking at the 3dmark ice storm extreme physics score.

It seems to be the most realistic number crunching Benchmark.
If this score is high the tablet is working great witch a browser.

[Dmcq]

Also, Intel is not immune to benchmark-eting. Check out the latest slide on the 64-bit 'advantage' where they show crafty and eon performing much better on 64-bit vs. 32-bit. This is pretty extreme cherry picking as they don't show the benchmark that slow down due to pointer expansion.

What would have allowed a bit of reasonable comparison is if they had shown their figures compares to a Z3770 or some specific clovertrail+ design. Then people could use their own relative figures for a general comparison. The manufacturers will make their own assessments rather than using Intel's figures. Anyone expecting Intel to act disinterested or unbiased..., well as Grove warned only the paranoid survive!

[This Causes an Error]

I think Intel needs to stop hyping up these chips that don't win designs with useless, made-up benchmarks and start winning some designs.

[biosbob]

Also, Intel is not immune to benchmark-eting. Check out the latest slide on the 64-bit 'advantage' where they show crafty and eon performing much better on 64-bit vs. 32-bit. This is pretty extreme cherry picking as they don't show the benchmark that slow down due to pointer expansion.

Well it is nothing remotely similar to the BAPCO fiasco. At least the flattering 64-bit benchmark score slide contains a _big_ box stating "Z3480(Merrifield) in 64-Bit mode can deliver increased performance on certain application usages when 64-bit Android applications are available".

[biosbob]

I have to correct you when you say ARM flushes to zero. It does not.

Though I am not an ARM floating point expert, I am pretty sure it is possible make am ARM processor flush to zero on floating point underflow. I also believe that an app can control the setting.

For VFP9 and VFP10, ARM has "full IEEE754 compliance with ARM support software, or near IEEE754 compliance with hardware only". I suspect the support software needed for full IEEE754 compliance causes a performance impact when underflow occurs or denormals are loaded from memory.