Which kind of work? Tape out means the design is already implemented for a process. The functional design is done already before the tape out, e.g. when ARM presented their Cortex A5X cores last year.Tapeout means the initial design is done. It means you have a functioning simulation that you are now going to attempt creating functional silicon from. Your design team in most cases still has at least a year of work ahead.
The design team has imo only work to do, if the first chips that come back from the fab dont work. Do you mean that debugging phase? Then ok.
@IDC:
I guess he meant *after* tape out, didnt he? Then the next task would be to manufacture the mask, no design team is needed then, isnt it?
I don't think anybody has ignored it. It's well known that tapeout to production silicon is two years - assuming everything goes well.
Sure, but the original design team wont work on the mask, will they? Are there overlappings possible? Just thought that producing a mask is a totally different task.It still applies, just change the name of the team to "mask", then "production", then "validation", etc.
Every step of the way in the process, the exact same set of questions apply. Time versus cost versus scope.
Sure, but the original design team wont work on the mask, will they? Are there overlappings possible? Just thought that producing a mask is a totally different task.
Bottom line, I think the days of huge increases in CPU and GPU power on phones are pretty much already over, and you can start getting used to much more modest increases that are driven by process improvements first and design upgrades second, while keeping similar peak power consumption. In other words, much like how things have been for desktop processors for the last several years. People talk about ARM quickly catching up, thinking they have a lot of easy wins to cover because their single threaded performance is way below. That isn't the point, it's about where they their power wall and since these devices are much smaller form factor they hit it at a much different place. ARM is pretty much already "caught up" in that regard (in that they compete decently with Intel who also caught up with power consumption with Atom)
I think both you and Exophase are exaggerating a bit. The truth lies between both posts. Mobile chips will continue to improve but not with drastic improvements as they have been. And while Intel certainly entered the market too late and was too complacent with Atom, they are not out yet. There is nothing inherently bad about x86 when it comes to low-power processing and Intel is still ahead in process. Intel is only vulnerable until it hits the 14nm node later next year where it will be uncontested for a while.I agree with you on all except this.
I think you underestimate the competence inside Intel - or AMD - for that matter, when comparing to ARM. And the complexity involved in designing and getting to production, a huge modern cpu.
I think its just now that Qualcomm, ARM starts to get the sort of competence inhouse, that is even comparable to Intel. The performance gains we have seen have been a consequence of not so much easy architectual wins as of opportunity to expand and build competences. Therefore i think we will continue to see huge power efficiency mm2 gains on the ARM side compared to x86.
You just cant dump more gpu mm2 into the phones from now on, but ARM and Qualcomm will dump more competence into the mix. And Intel dont have the same opportunity because they have had this effective and huge cpu and process machine for years.
I think the situation now just shows how bad, bad it was for Intel - and AMD - to come to late to the mobile market. TSMC, Qualcomm and Samsung is no small players. They dwarf Intel the same way Intel dwarf AMD.
And the new test here by anand, just show Intel misses the most important part for mm2, namely the gpu part. IB might look strong compared to 320, but when jaguar hits, it will show Intel really needs something new on the arch. side for the gpu, else the rest is waisted.
But there is no indication that they'll be selling 14nm mobile parts anytime soon ! What you are also neglecting is that the cortex A15 & Krait parts will be refined over a period of time & unlike the slow progress in x86 world there is a whole lot of wiggle room for ARM chipmakers with FinFET, SOI, bulk not to mention Mali/PowerVR graphics thereby undercutting Intel yet again, unless of course they're willing to forego those insane margins !I think both you and Exophase are exaggerating a bit. The truth lies between both posts. Mobile chips will continue to improve but not with drastic improvements as they have been. And while Intel certainly entered the market too late and was too complacent with Atom, they are not out yet. There is nothing inherently bad about x86 when it comes to low-power processing and Intel is still ahead in process. Intel is only vulnerable until it hits the 14nm node later next year where it will be uncontested for a while.
But there is no indication that they'll be selling 14nm mobile parts anytime soon ! What you are also neglecting is that the cortex A15 & Krait parts will be refined over a period of time & unlike the slow progress in x86 world there is a whole lot of wiggle room for ARM chipmakers with FinFET, SOI, bulk not to mention Mali/PowerVR graphics thereby undercutting Intel yet again, unless of course they're willing to forego those insane margins !
But there is no indication that they'll be selling 14nm mobile parts anytime soon ! What you are also neglecting is that the cortex A15 & Krait parts will be refined over a period of time & unlike the slow progress in x86 world there is a whole lot of wiggle room for ARM chipmakers with FinFET, SOI, bulk not to mention Mali/PowerVR graphics thereby undercutting Intel yet again, unless of course they're willing to forego those insane margins !
As someone said earlier this is a niche market so its not wrong to assume that with multiple ARM vendors coming in & certain other costs going down it'd be naive to think that ARM servers would remain prohibitively expensive as they are now !
I'd like to see when/where is the clover trail comparable to Exynos Octa or Snapdragon 600 ?
In the best case that will be the year of TSMCs 16nm Finfet process, too, if I count the months correctly:
http://www.eetimes.com/electronics-news/4398727/TSMC-taps-ARM-V8-in-road-to-16-nm-FinFET
Nov. 2013 + 4 Quarters = Nov. 2014.
I don't think that calxeda truly represents the juggernaut ARM has become, ignoring raw performance metrics(cause there is no working ARM Cortex-A57 chip yet) I can't see how a monolith like Samsung or even lesser rivals like Qualcomm/Nvidia won't make cheaper ARM servers if & when they decide to enter the market. Yes it is a long shot atm but I don't see any logic in "ARM can't produce cheap servers" even though their eventual performance is what'll really matter in the end !Yes, you are right, two choice for them much cheaper or die. But it's unbelievable prejudgment to believe they will definitely become cheaper, and have no chance to die. It's niche market, and will be crawled by many ARM vendors and Intel (ATOM). How much market share do you think each of those ARM vendors has? This is the market need lots of R&D and operational cost, and how can they even it up with the small volume? By share the chip into mobile phone or share the server camping with mobile phone?
And after all, how can they got the market share, if there is already Intel solutions with compatible energy consumption, better performance, better performance/watts, easier to use, better support of software and community, and cheaper/much cheaper price?
Interesting, I'll wait for a bit of dissemination on these raw numbers because some of'em aren't directly comparable to one another. Something like a cross platform PCMark would be nice for starters but kudos to Intel, anyways I hate Android these days cause anything less than a quad core & 2GB RAM onboard makes the phone super laggy especially with those 1080p screensThis is the link of earlier comparison of K900 and Galaxy S4.
http://www.google.com/url?sa=t&rct=...=NlEKjFX1tkGJngsmIfuZyg&bvm=bv.44770516,d.bGE
I don't think that calxeda truly represents the juggernaut ARM has become, ignoring raw performance metrics(cause there is no working ARM Cortex-A57 chip yet) I can't see how a monolith like Samsung or even lesser rivals like Qualcomm/Nvidia won't make cheaper ARM servers if & when they decide to enter the market. Yes it is a long shot atm but I don't see any logic in "ARM can't produce cheap servers" even though their eventual performance is what'll really matter in the end !
Yes, you are right, two choice for them much cheaper or die. But it's unbelievable prejudgment to believe they will definitely become cheaper, and have no chance to die. It's niche market, and will be crawled by many ARM vendors and Intel (ATOM). How much market share do you think each of those ARM vendors has? This is the market need lots of R&D and operational cost, and how can they even it up with the small volume? By share the chip into mobile phone or share the server camping with mobile phone?
This is true, but this is where the IP business model becomes efficient: the costs are mostly shared between a lot of different players at every level, you only have to invest where you think it is useful to differentiate, whereas Intel has to do everything in house, and cannot share any cost with the rest of the industry.
Both the big players (Samsung, Apple, QC) and the small ones (Calexda/Applied Micro or the Chinese vendors) can survive provided they do not target the same performance/cost/schedule.
Obviously any company can fail and die, this has nothing to do with being ARM or not.
Interesting, I'll wait for a bit of dissemination on these raw numbers because some of'em aren't directly comparable to one another. Something like a cross platform PCMark would be nice for starters but kudos to Intel, anyways I hate Android these days cause anything less than a quad core & 2GB RAM onboard makes the phone super laggy especially with those 1080p screens
Yes .. but ... ARM is a much bigger fish than MIPS. Yes ARM will have soon 64bit CPUs for server purposes, that is similar to MIPS, but there are still lots of differences.You can compare ARM with MIPS. The MIPS really have anything ARM should have and could have. (64bit, low power, RSIC, high performance, software support, vendor support and the IP business mode) What hunting MIPS will also hunt ARM in the server market.
Yes .. but ... ARM is a much bigger fish than MIPS. Yes ARM will have soon 64bit CPUs for server purposes, that is similar to MIPS, but there are still lots of differences.
Just think about who has licenced the Cortex A57 design and will provide chips:
AMD, Broadcom, Calxeda, HiSilicon, Samsung and STMicroelectronics.
Additionally there are nvidia, AppliedMicro and 2 anonymous companies (probably Qualcomm and Apple) who will design their own ARM64 cores.
Together these companies have a "critical mass" and there will be several, different kinds of SoCs aiming at different market niches. The customer can choose among several solutions and get the best one.
Even Intel will learn to fear that momentum. Over a longer time period, Intel could even be pushed out to the big-tin server / HPC market and compete with IBM/Oracle only. But lets see how their Server-Atom SoC will perform. Intel's always good for a surprise, too, and they have a manufacturing advantage.
Servers, while profitable for Intel, isn't what pays the bills. It's Joe Blow on his laptop who uses his $200 CPU to write emails and use MS Office.