[TechReport]ARM unveils Cortex-A72 CPU, Mali-T880 graphics, and more

[Exophase]

Intel always puts its testing methodology on the slide, so not sure why you're complaining.

Saying what benchmarks they ran in a footnote at the bottom hardly means they're above any possible criticism. For example, when talking about SPEC things like compiler and flags are very relevant (not to mention a list of sub scores) - SPEC2k6 results would mean they have to be on the SPEC site but conveniently SPEC2k is obsoleted and therefore unlisted.

Or you know, not actually saying what they're testing against - not only not specifying the hardware used and important things like clock speeds but not even naming the competitor, instead just referring to them with things like "Comp1"...

And then there was the the report that they got a firm to do by proxy, that claimed that Saltwell had a ridiculously huge perf/W advantage over some ARM competitors. This was of course because they used a version of AnTuTu compiled with version of ICC that badly gamed the CPU part of the benchmark. That is a pretty deceitful marketing attack.

But I guess it's not enough that I said ARM and nVidia are worse than them, Intel must be above all criticism...

 

[Arachnotronic]

So...looks like the A72 is a solid core from ARM. IPC nicely up, power down, what's not to like?

 

[NEDM64]

So...looks like the A72 is a solid core from ARM. IPC nicely up, power down, what's not to like?

We, geeks would like. Marketing on the other hand...

 

[imported_ats]

Saying what benchmarks they ran in a footnote at the bottom hardly means they're above any possible criticism. For example, when talking about SPEC things like compiler and flags are very relevant (not to mention a list of sub scores) - SPEC2k6 results would mean they have to be on the SPEC site but conveniently SPEC2k is obsoleted and therefore unlisted.

Its pretty standard practice for Intel to include even more information on the backup slides of the slide deck. Intel pretty much always details everything about their performance comparisons to a fairly reasonable level of detail.

And then there was the the report that they got a firm to do by proxy, that claimed that Saltwell had a ridiculously huge perf/W advantage over some ARM competitors. This was of course because they used a version of AnTuTu compiled with version of ICC that badly gamed the CPU part of the benchmark. That is a pretty deceitful marketing attack.

Um, the issue with AnTuTu was that the benchmark was actually broken and wasn't actually doing anything. It was literally a small loop with no output. The compiler reasonably just did dead code elimination and poof.

 

[Nothingness]

Um, the issue with AnTuTu was that the benchmark was actually broken and wasn't actually doing anything. It was literally a small loop with no output. The compiler reasonably just did dead code elimination and poof.

That's not at all what was done. Search this forum, Exophase and I posted several messages that IMHO proved icc was specifically tuned for AnTuTu and that the AnTuTu author properly optimized Intel binary while not doing any effort for the ARM binary.

 

[ancientarcher]

That's not at all what was done. Search this forum, Exophase and I posted several messages that IMHO proved icc was specifically tuned for AnTuTu and that the AnTuTu author properly optimized Intel binary while not doing any effort for the ARM binary.

I remember that well. All the tech forums were awash in the 'fact' of Intel's superiority and how it will soon drown ARM in mobile. such a hoax!

The logical question to ask is, why did the AnTuTu author do that. Did he get paid for that? Why would he do that otherwise if there was no monetary or other benefit? And of course, who would pay the bribe (monetary or otherwise) but Intel itself. Not that I am saying it hasn't done that before (wink wink AMD) but alas this tried and tested 'strategy' didn't succeed this time around. such a shame!

 

[soccerballtux]

That's not at all what was done. Search this forum, Exophase and I posted several messages that IMHO proved icc was specifically tuned for AnTuTu and that the AnTuTu author properly optimized Intel binary while not doing any effort for the ARM binary.

they did this with the icc and AMD and Via processors way back too--
http://arstechnica.com/gadgets/2008/07/atom-nano-review/6/

 

[Idontcare]

I remember that well. All the tech forums were awash in the 'fact' of Intel's superiority and how it will soon drown ARM in mobile. such a hoax!

The logical question to ask is, why did the AnTuTu author do that. Did he get paid for that? Why would he do that otherwise if there was no monetary or other benefit? And of course, who would pay the bribe (monetary or otherwise) but Intel itself. Not that I am saying it hasn't done that before (wink wink AMD) but alas this tried and tested 'strategy' didn't succeed this time around. such a shame!

Bapco?

 

[Exophase]

Its pretty standard practice for Intel to include even more information on the backup slides of the slide deck. Intel pretty much always details everything about their performance comparisons to a fairly reasonable level of detail.

So you think they named the competitors "Comp1" but named them in some backup slide? Why?

Um, the issue with AnTuTu was that the benchmark was actually broken and wasn't actually doing anything. It was literally a small loop with no output. The compiler reasonably just did dead code elimination and poof.

Completely wrong, read my analysis on the subject:

http://forums.anandtech.com/showthread.php?t=2330027

 

[kpkp]

First announced product using A72 appears to be the Amazon TV box with a Mediatek MT8173, shipping October 5.

 

[geoxile]

First announced product using A72 appears to be the Amazon TV box with a Mediatek MT8173, shipping October 5.

I'm surprised how cheap it is considering the A72 is the next big core from ARM.

 

[jhu]

First announced product using A72 appears to be the Amazon TV box with a Mediatek MT8173, shipping October 5.

Awesome! Now someone just needs to get root on this thing when it arrives and we can do some CPU benchmarks.

 

[Systems analyst]

ARM INVESTORS CONFERENCE
http://phx.corporate-ir.net/Externa...WxkSUQ9LTF8VHlwZT0z&t=1&cb=635778262245579577

ARM report that 'ARTEMIS', the follow on to MAIA - the A72, will be the new top-end mobile core. It will be on 10nm. p.21.
They stated that there will be an annual cadence to the top-end mobile cores. Q&A session.
They also stated that they are working on 7nm 'physical' IP; they provide optimised versions of their cores, and other IP, for a particular production process, TSMCC, Samsung and Global Foundries. p.48.
They named a new set of Mali cores, under development. p.21
They will reveal something about their server-specific cores, currently under development, later this year. Q&A session.
They, in conjunction with their partners, have reset their 2020 target for the server market to 25% of units. p.52.
A variety of servers are pictured on p.45.
Their new target for the network market is 45% by 2020. p.41.
Intel is a partner. p.22.

It would be interesting to see an AnandTech review of the servers on p.45.