I can understand why GB included AES, SHA1, SHA2 as they (along with compression) algo's are a good proxy for general integer performance. The problem with the AES test is it uses the special AES instructions (when available) so it doesn't become an integer test any more. My understanding is that the SHA1 and SHA2 don't use special instructions, but I'm a little suspicious of the high scores of the A7 as V8 does have instructions that support the SHA family (compiler optimization?). There are still lots of questions.
One of the reasons why ARM "won" against other RISC families (such as MIPS and SuperH) is that it's easier to code at the assembly level which is important when you are deeply embedding.
For those interested A7 is now on GB3. Usual benchmark caveats apply, as well as a few more. The Baytrail tablet is 32bit (you would expect it to be faster in 64bit) and it's the reference design so real products could be faster (or slower). There are still questions surrounding the A7's clock, and the listing for the A7 say 32bit, which I think is wrong.
A7 V Baytrail (Z3770)
A7 V MacBook Pro (Intel Core i7-3615QM)
The A7 is Apples design to the ARM64 ISA, which is a new ISA, not an extension to ARM32. Is it to the full ISA, yes, it looks like it is.
Ashraf, type the following into Google
geekbench3 compiler options
And it will give you more detail on the compiler options used across various OS's for GB3. It'll explain, in part, why ARM scores have 'jumped' relative to x86.
As for Apple, there are still lots of unknowns. If the performance is down to a good optimizing compiler (as you suggest in the article), then that team need a massive pat on the back. It normally takes a long time for a compiler to produce decent code for a new ISA.
This thread started with the statement from you:
"We are now getting a strong dose of reality as to why many thought 64-bit was a waste of time on a smartphone."
The fact is many people were wrong - as Anand pointed out in the article ("64 bit" not worth it). ARM64 is not just about addressability.
"But for ARM smartphones the benefits are few and will take years to realize". You'll be able to buy an ARM 64-bit phone next week and you'll benefit from the new ISA as soon as you open a web page, especially if it's SSLed.
No Wallis, I wasn't being churlish, childish or petulant. This is turning into a religious debate, devoid of clear facts and reasoning. The facts speak for themselves. If you choose not to "believe", then that's OK, I'm fine with it, honest:)
As Anand said:
"The immediate upsides to moving to 64-bit today are increased performance across the board as well as some huge potential performance gains in certain FP and cryptographic workloads."
There are very, very few downsides.
No wallis, it's facts. QCOM has got 30+ billion in cash. Google it.
The last question was:
"If Qualcomm's cash cow can be impacted, how much can they sustain their SoC development efforts and compete with Intel?"
Which is a very valid question (considering what Intel did to AMD), but with 30billion in cash, QCOM's revenue could drop to zero and still fund years of SoC development.
(additional) memory is a liability for any low power device -
Very true, but thats the *point* you don't need any extra memory to benefit from a recompile!
On the 64 bit I'd describe it another way. It's not that 64 bit for a smartphone is a waste of time, it's that for Apple's 32 bit IOS-based ecosystem 64 bit is so far ahead of the need that users won't see the benefit for years. And there is a cost to add the additional RAM to support 64 bit and power efficiency concerns.
Did you read the anandtech article? There are other benefits to Apples 64bit SoC than just memory accessibility. The entire Apple software stack is 64 bit - the kernel, OS, and default Apps. For 3rd party apps it can be as simple as a right click for the developer.
So Qualcomm seems to be the main competition for Intel in the mobile space.
If Qualcomm's cash cow can be impacted, how much can they sustain their SoC development efforts and compete with Intel?
A while:) QCom has over 30 billion in cash to spend - more than Intel.
If x86 starts taking share, perception could change and the bubble would pop before any financial impact. As a side, even if x86 managed double digit market share in both smartphone and tablets in the next year or two it would have little impact on royalty revenues (due to the small amount ARM actual earn per SoC and the volumes numbers it ships).
There are lots of things 'supporting the bubble', some of them based on facts such as the correlation between licence revenues (at record highs) to downstream royalty revenues. The bulk of ARM's future (royalty) growth is coming outside of mobility.
People misunderstand the reasons for the price jump due to Apples 64bit SoC. Yes ARM will generate more in royalties and it speeds up QCOM and #$%$ their roadmaps, but it lifts a lot of the 'fear' surrounding ARM64 bit. There were real concerns that the ISA wouldn't get traction, would be hard to implement, wouldn't have the performance, would be hard to write for etc. Apple answered all those questions.
Then you have the real bubble concepts, internet of things, wearable computing and servers blah blah blah.
This rumor hasn't come from ARM.
ARM have answered this question many times over the years, basically there are lots legals in place to stop a 'hostile' partner taking over the governorship of the IP/ISA to the detriment of others. This includes existing partners (such as the big three, Apple, QCOM and Samsung [remember, they compete directly but work within the ARM ecosystem]), or a new outsider such as Intel.
Once you are working on a specific ISA version (such as Apple) you are really free to do what you want.
Intel inside: US chip giant said to be in talks to take over ARM
"ARM’s shares surged on Wednesday following murmurings the electronics company will be taken over by Intel."
Did RT fail because of price and/or performance? Or did it fail for other reasons, ie, because it was a dead duck without legacy?
The BOM question is important. Why can't you have $100 clovertrail tablets?
Look at the BOM (Bill of Materials) for a cheap tablet and then you'll see how.