In defense of Jon, I don't think he is breaking any NDA here. He is not giving us any more details on Intel's Bay Trail -- other than saying that he thinks it is a game changer. And of course, he were to share any inside knowledge of Bay Trails (which he couldn't), he could have defended his position better in this forum -- in my opinion.
I don't want to get you in trouble but if you were generously provided with a look at Bay Trail covered by a non-disclosure agreement ....how is it you can write and publish an article about it, which is surely disclosure?
Gondalf, from your reply the only conclusion one can make is that you must be confusing Geekbench with Dhrystone. Geekbench does not only get 50% of its score from memory tests (including Stream), but it even includes a primality test that you seem to regard so highly... Please do try to understand it before dismissing it.
I agree SPEC is good for evaluating server CPUs, and when 64-bit ARM servers appear I'd imagine we'll see some ARM scores (expect the ICC vs GCC compiler issue to stir up more discussions in the future as AMD, Transmeta and VIA have/had the same issues). But it's just not useful for mobile SoCs as they do not run the same floating point heavy workloads.
It's true that older ARM SoCs suffered from memory bandwidth and prefetching isssues, but given the much improved memory scores (the GT-I9500 scores 2.5 times that of the S3 on Stream), it looks like A15 SoCs have now solved those issues, and are pretty much level with Bay Trail.
Who says I don't trust QC? I discuss and base my opinion on hard facts, not anonymous claims without evidence. If you do have links that prove Anand's numbers are incorrect, then I'd be interested to see it. However talking about TDP numbers between different SoCs is going to be fruitless due to the completely different methods and definitions used.
So sorry wilco but your useless Geekbench is more like MANY other syntetic tests around the web, it is popular because is cross plataform, not because is good.
The main concern in this so so (many so) bench is that it doen't stress the L2 cache and even less the memory controller with its own memory subsystem.
The main advantage of SPEC is that it is able to stress a lot "at the cpu level" the device (soc level if you prefer), showing the average performance of core, L1, L2, L3, memory controller, main memory ALL together.
SPEC is done for this and its widely used to judge a server cpu (not the core).
You are almost wrong saying that a poor syntetic bench is better in a phone or in a tablet. It's not an accident that the old Atom is able to perform ,in real world software, much much better as indicated by a poor, small synthetic test. In a real world workload is the cpu(soc) that counts, the core is only a component of the equation. Its not a secret that the main defect of ARM socs is the memory subsystem, and ARM is trying to address this .....but needs time...many years to match Intel or Amd expertise.
About power consumption (that is disclosed now thanks to some official Qualcomm slides), call me when a review will show an ARM SoC running a real stress test ala Prime95, not a very light bench done so so.
I do not comment on your rather unfair reasoning on TDP, capable of triggering a stupid flame war.
Its funny to see ARM followers not even trust in Qualcomm. I pretty believe that this ARMmania is like a religion.....no comment.
I have read a lot of discussion & comparison on this forum of Intel & ARM chips, lkg power etc...
In my experience Low power design & architectures are not the same ones one can use in High performance applications and vice versa. Same goes for Process technology. A process optimized for Low power will be most efficient at slower Freq. Freq/Power curve for this tech would not look the same at 4GHZ and vice versa.
So when somebody quotes the xyz chip has best Perf/watt and much better than some other chip, that is not entirely true. This xyz chip may be best in a certain freq range, but will not be best option in High Perf needs and vice versa.
( I have worked in IDM,OEM, ASICs and foundry business and seen multiple gens of Process tech and Design ).
We can compare the desktop with laptop, while laptop are lot more capable than tablet still they are not able to fully replace desktop, however comparing laptop in early days to now, they have gained lot more power and becoming close enough to desktop where they replace its for large number of use cases.
Tablet is in early stage with limited capabilities. It certainly have replaced desktop/laptop for sales people who mostly use devices to check email/document preprations and for school kid who use it for social media or may be for non-technical people who use it for staying connected with people. However its still far away from use cases like editing photos e.t.c.
It will be interesting to see how it evolves and how many usecases it can beat laptop/desktop. may be they come with enough powerfull CPU with wireless monitor/keyboard hookup where you feel like using desktop and having the system around tablet. Syrface is so far closest to this among all tablet, however "windows" itself is loosing edge (in my opinion).
Current Tablets are mostly good for entertainment and may be reading occasional mail.
I still cannot open 2-3 apps at same time, so I cannot listen to music and also surf the web or write a email or a Doc.
I can't imagine using a tablet for work. I need a real keyboard and mouse and 2 screens and lot of DRAM.
Even when I am crunching data on a remote server or open a Design layout from secure server, I need my big screen to look at Design on 1 screen and keep an eye on a code execution on other, all the while listening my music.
"Like, one example, would I feel confident moving around my investments online, on a smartphone or tablet? Uuuh, no.
Or another one. We recently did some serious shopping for rugs and rug pads, including research on what kind of pads work on what kind of floors. Would I have selected rugs based on what a smartphone screen showed? Uuuh, no again."
I would use my tablet in both of the above cases. As Junko said, its just a personal mental distinction created by people.
I'm told Intel planned to talk about and annouce Bay Trail at Hot Chips this week. A presentation was all approved and ready to go, but got pulled at the last minute. Apparently they are saving the news for their IDF event in two weeks.
Bert, I don't believe that PCs will disappear either.
You make a good case that how some of us make a mental distinction what tablets are for and what PCs are for.
You talk about the ergonomic shortcomings of smartphones and tablets for the stuff you would normally do on PCs; for me, I found the erognomic shortcomings in PCs. WhenI want to do a quick search on Google, or catch a glimpse of video clips on Youtube, do I go to my PCs to do that? No, I prefer the ergonomic convenience of tablets -- just lying down on my sofa.
Replay available now: A handful of emerging network technologies are competing to be the preferred wide-area connection for the Internet of Things. All claim lower costs and power use than cellular but none have wide deployment yet. Listen in as proponents of leading contenders make their case to be the metro or national IoT network of the future. Rick Merritt, EE Times Silicon Valley Bureau Chief, moderators this discussion. Join in and ask his guests questions.