Update:

> in terms of 64-bit ARMv8 (AArch64) cores, there is announced so far:
> - Cortex-A53 and Cortex-A57 from ARM Ltd.
> - Denver from nVidia
> - X-Gene from Applied Micro
> - Thunder from Cavium

- Cyclone from Apple

http://en.wikipedia.org/wiki/Apple_A7
http://www.anandtech.com/show/7335/the-iphone-5s-review/3
http://www.anandtech.com/show/7335/the-iphone-5s-review/4


Edit: added more links

[ Edited by Andreas_Wolf 18.09.2013 - 12:21 ]

This might be of more interest for desktop users:

AMD Hierofalcon

Up to eight 64-bit ARM Cortex-A57 cores
Up to 2GHz.
Dual 64-bit memory channels DDR3/DDR4
10-gigabit Ethernet
PCI Express 3

The current high end ARM chips are all designed for phones or at most tablets. This is more like a desktop part.

> This might be of more interest for desktop users: AMD Hierofalcon
> [...] This is more like a desktop part.

""Hierofalcon" is the first 64-bit ARM-based platform from AMD targeting embedded data center applications, communications infrastructure and industrial solutions."
http://www.amd.com/us/press-releases/Pages/amd-details-embedded-2013sept9.aspx

So it is a server oriented processor.

> So it is a server oriented processor.

Oriented at the tasks required between the servers, to be precise. I think it's targeted at the same fields as for instance the IBM PowerEN.

Update:

>> They could co-develop PPC and ARM SoCs so all they have to do is drop in whatever
>> cores the market requires.

> Yes, that's what the announced QorIQ Layerscape product line will be all about.

First chips of ARM-based QorIQ LS1 series announced:

http://media.freescale.com/phoenix.zhtml?c=196520&p=irol-newsArticle&ID=1864869

Regarding future QorIQ LS chips based on Power Architecture:

"Freescale isn’t abandoning the PowerPC architecture it has championed in the past, Bustami said, but it has decided to create a parallel ARM-based product line because of the direction the market heading."
http://gigaom.com/2013/10/16/freescale-puts-arm-cores-to-work-running-the-internet-of-things/

Quote:

Or rather - That's the markets targeted by AMD (who has other CPU's in their portfolio that they want to sell to the desktop market).

It doesn't seem to have an embedded GPU (as expected of course). "Weak GPU's" (everything is relative) of the current ARM chips is an argument that some people frequently use to discard ARM for desktop applications.

It does have third generation PCI-e though, so whether it's suited in desktop contexts or not may come down to the total PCI-e bandwidth/number of lanes, etc. Who knows, but if you'd be able to use 16 lanes for a state-of-the-art graphics card (or a budget one for that matter), and a few other lanes for some additional peripheral controllers/slots, then I don't think it's out of consideration for desktop applications. 8x 64-bit Cortex-A57 cores with dual-channel DDR 2/3 etc will perform quite well, and not every desktop application/customer needs or wants a Core-i7 level gaming system anyway (in fact, most probably don't).

Looking forward to see more information about this one. I think it can become quite interesting!

> Or rather - That's the markets targeted [...]. It doesn't seem to have an embedded
> GPU (as expected of course). [...] It does have third generation PCI-e though, so
> whether it's suited in desktop contexts or not may come down to the total PCI-e
> bandwidth/number of lanes, etc. Who knows, but if you'd be able to use 16 lanes
> for a state-of-the-art graphics card (or a budget one for that matter), and a few other
> lanes for some additional peripheral controllers/slots, then I don't think it's out of
> consideration for desktop applications. 8x 64-bit [...] cores with dual-channel DDR
> 2/3 etc will perform quite well, and not every desktop application/customer needs
> or wants a Core-i7 level gaming system anyway (in fact, most probably don't).

This sounds suspiciously similar to the reasoning of those strange folks who are in favour of using current and future Power Architecture communications SoCs in desktop-oriented systems ;-)

Update:

> in terms of 64-bit ARMv8 (AArch64) cores, there is announced so far:
> - Cortex-A53 and Cortex-A57 from ARM Ltd.
> - Denver from nVidia
> - X-Gene from Applied Micro
> - Thunder from Cavium
> - Cyclone from Apple

- Vulcan from Broadcom
- K12 from AMD

http://www.broadcom.com/press/release.php?id=s797235
http://www.linleygroup.com/newsletters/newsletter_detail.php?num=5065
http://www.amd.com/en-us/press-releases/Pages/ambidextrous-computing-2014may05.aspx


Edit: added K12

[ Edited by Andreas_Wolf 08.05.2014 - 14:00 ]

These chips all look very interesting but unless they get used on a standard form factor motherboard I can't see them as being very useful for desktop computing.

Some of those are designed for networking infrastructure type gear so we may never see them in a desktop form factor.

OTOH they do need development machines so there will be boards of some description out there.


AMD might be more interesting because they have 64bit ARM based parts on the way and I fully expect you'll find at least some of these on desktop type boards.

Then again my "desktop" has been a laptop for over 10 years. You can already get ARM laptops and more are beginning to appear.

I quite fancy the Chromebook Pixel as a Linux Laptop, much nicer than the Dual Core Penryn Macbook Pro I've currently got.

I don't think well ever see Laptops with a PPC CPU ever again though.

> Some of those are designed for networking infrastructure type gear so we may
> never see them in a desktop form factor. [...] AMD might be more interesting
> because they have 64bit ARM based parts on the way and I fully expect you'll
> find at least some of these on desktop type boards.

In case these AMD parts you refer to happen to include Hierofalcon, I'd say this actually is "designed for networking infrastructure type gear". At least that's what AMD says:

https://morph.zone/modules/newbb_plus/viewtopic.php?forum=3&topic_id=7675&start=474

> I quite fancy the Chromebook Pixel as a Linux Laptop

This one hasn't an ARM CPU, opposed to the Samsung Chromebook and the HP Chromebook 11, right?

Quote:

Yes you are right, I hadn't looked into it in detail and just assumed it was ARM-based like the others.

Still looks lovely though. :)

> just assumed it was ARM-based like the others.

Wikipedia has a nice table on the Chromebook models:

http://en.wikipedia.org/wiki/Chromebook#Commercial_machines

Apparently, only a minority of them is ARM-based.

Quote:

In case these AMD parts you refer to happen to include Hierofalcon, I'd say this actually is "designed for networking infrastructure type gear". At least that's what AMD says:

Funnily enough they said it was primarily for servers just today but they mentioned other stuff as well.

4 - 8 Cores, 2GHz and PCIe3.0. A very high spec for an ARM.
Things are about to get mighty interesting.



[ Edited by minator 24.10.2013 - 21:10 ]

You probably weren't expecting this:

IBM Licenses ARM for Custom Networking, Comms Chips

XMOS also threw in a Cortex M3 to one of their designs.

> You probably weren't expecting this:
> IBM Licenses ARM for Custom Networking, Comms Chips

Indeed. Let's see what the implications will be for IBM's PPC4xx and PPC A2 business(es). IBM's move is probably not very encouraging for existing and potential PPC4xx core licensees.
Any idea what IBM's customers will need the Mali GPU for in their routers, switches and base stations?

Edit: An attempt to explain:

""This deal makes a lot of sense, and saves IBM the trouble of maintaining its own CPU cores," said Linley Gwennap, senior analyst of the Linley Group. "IBM has been offering PowerPC cores (e.g., 405, 440) through their ASIC business, but it has not updated these cores for years, so they are getting stale [meanwhile] ARM has become quite popular," he said in an email exchange. "IBM also does ASICs for set-top boxes which is probably why they picked up a Mali license," he added."
http://www.eetimes.com/document.asp?doc_id=1319884

This would answer my Mali question. And indeed, the PPC4xx core series has last been updated in 2009. And the most recent chip they sell with PPC A2 cores (PowerEN) has been introduced in early 2010.

[ Edited by Andreas_Wolf 25.10.2013 - 11:28 ]

Addendum:

> Regarding future QorIQ LS chips based on Power Architecture:
> "Freescale isn’t abandoning the PowerPC architecture it has championed in the past"
> http://gigaom.com/2013/10/16/freescale-puts-arm-cores-to-work-running-the-internet-of-things/

http://www.freescale.com/files/training/doc/dwf/DWF13_APF_NET_T0795.pdf is a nice presentation from April 2013 I just stumbled upon now. Interesting things in it (among others):

Page 8:

"We absolutely stay committed to our Power Architecture and our new e6500 core delivers an unmatched performance/power ratio for the years to come. We continue to invest in this area. [...] We will pick the right core for the right segment/application in alignment with our customers and networking expertise. [...] ARM high end cores have been part of FSL portfolio for years and require minimal incremental investment to leverage for Networking
Freescale will use standard ARM offering where possible
Freescale has a fully negotiated architectural license ready to exercise if need"

Page 10:

Core roadmap until 2015 shows Cortex-A53 ("Apollo") and Cortex-A57 ("Atlas") as well as the e6501 core (adding virtualization interrupts) for 2013/2014, all in the current 28nm process. For 2015, two new cores in 20nm process are shown. I believe that at least the higher performance one ("NGC" with multithreading, higher single-thread performance, directory snoop filters, low-latency clustered accelerators) is supposed to be a new Power Architecture core succeeding the e6500/e6501 core. The lower performance one ("Helios") may be ARM (which would mean a not yet announced core by ARM Ltd.) or Power Architecture.

Page 22:

QorIQ LS2045P:
- 4 x e5500 at 1.5 GHz
- 2 x DDR4
- 6 x PCIe 3.0
- 2 x SATA3
- 2 x USB 3.0
- 8 x 10GbE
- 16 SerDes lanes (8 for networking, 8 for PCIe)
- LCD controller (DIU)

Page 24:

QorIQ LS3400P:
- 20 x e6500 at 1.5 GHz
- 4 x DDR3
- 6 x PCIe 3.0
- 2 x SATA3
- 2 x USB 3.0
- 2 x 100/40GbE
- 24 SerDes lanes (12 for networking, 12 for PCIe)

Page 26:

Mention of T3-Series.

Page 33:

Mention of QorIQ T4400, which should have 20 e6500 cores.


Edit:
http://2014ftf.ccidnet.com/pdf/0016.pdf (page 14/15) from May 2014 has the information that QorIQ LS3 series will have "ARM and Power Architecture cores".

[ Edited by Andreas_Wolf 29.10.2014 - 11:31 ]

How much longer will we have to wait for Qualcomm's Snapdragon APQ8084 chipset?

> How much longer will we have to wait for Qualcomm's Snapdragon APQ8084 chipset?

Wikipedia says Q1 2014 for samples:
http://en.wikipedia.org/wiki/Snapdragon_%28system_on_chip%29#Snapdragon_805

Qualcomm announced last month that the Snapdragon 805 (which includes the 8084) was "sampling now and expected to be available in commercial devices by the first half of 2014":
http://www.qualcomm.com/media/releases/2013/11/20/qualcomm-technologies-announces-next-generation-qualcomm-snapdragon-805

The Linley Group reported three weeks ago that the chip was "already sampling to lead customers" and that they expected it "to appear in phones and tablets in 3Q14":
http://www.linleygroup.com/newsletters/newsletter_detail.php?num=5086

So about the same time as the ARM Cortex A53 based products appear.
Interesting, so the 32 bit products will lead in performance.

Update:

> Figure for Swift substantiated and figures for Cortex-A53 and Cortex-A57 added

New list of DMIPS per MHz and core figures (added Cortex-A12 and specified different Krait types):

ARM Cortex-A7: 1.9
ARM Cortex-A8: 2.0
Qualcomm Scorpion: 2.1
ARM Cortex-A53: 2.3
Marvell Sheeva PJ4: 2.4
ARM Cortex-A9: 2.5
Marvell Sheeva PJ4B: 2.6
ARM Cortex-A12: 3.0
Qualcomm Krait 200: 3.1
Qualcomm Krait: 3.3
Qualcomm Krait 300/400: 3.4
Qualcomm Krait 450: 3.5
Broadcom Brahma15: 3.5 (unsure whether or not this core is just ARM Cortex-A15)
Apple Swift: 3.5
ARM Cortex-A15: 3.5
ARM Cortex-A57: 4.1...4.8 (depending on implementation)

Does anybody have DMIPS figures for the Apple Cyclone or the Applied Micro X-Gene?