It’s a hardware day today, and since AnandTech is the most authoritative source on stuff like this, we’ve got more from them. Arm announced its next big micro-architecture – which will find its way to flagship smartphones soon.
Overall the Cortex-A77 announcement today isn’t quite as big of a change as what we saw last year with the A76, nor is it as big a change as today’s new announcement of Arm’s new Valhall GPU architecture and G77 GPU IP.
However what Arm managed to achieve with the A77 is a continued execution of their roadmap, which is extremely important in the competitive landscape. The A76 delivered on all of Arm’s promises and ended up being an extremely performant core, all while remaining astonishingly efficient as well as having a clear density lead over the competition. In this regard, Arm’s major clients are still heavily focusing on having the best PPA in their products, and Arm delivers in this regard.
The one big surprise about the A77 is that its floating point performance boost of 30-35% is quite a lot higher than I had expected of the core, and in the mobile space, web-browsing is the killer-app that happens to be floating point heavy, so I’m looking forward how future SoCs with the A77 will be able to perform.
As linked above, the company also announced its next-generation mobile GPU architecture.
The snapdragon 855 is already up to 30 faster than all intel U cpu’s. competing agains the i5 even though thermals are way down and uses a LOT less watts.
Icant wait to see what qualcom, samsung and sony (and to some extent apple) can do with this baby.
What is Sony going to do with Arm architecture? They don’t make any chips right?
https://www.sony-semicon.co.jp/products_en/spresense/
I didn’t think so, but I looked it up, and well, apparently they do.
Buuuut I don’t see any indication that the ARM cores they use are the higher-end, higher-performance cores such as these. They build chips using Cortex-M 32-bit microcontrollers, running at about 150-200Mhz.
I’m not sure where you’re getting that from. Most performance comparisons say snapdragon 855 is about equal to Core i5-8250U; losing some benchmarks by a little and winning some benchmarks by a little.
Of course Intel designs their cores for high performance desktop/server and then tries to make them work for something they were never designed for (ultra low power) by “tweaking” (e.g. downclocking, reducing cache size and disabling cores to force it into a certain TDP; without changing anything in the core itself); so being equal to Intel’s ultra low power CPUs isn’t really any kind of victory (it’s as meaningless as giving a snapdragon 855 a huge water-cooled heatsink and over-clocking it as much as possible and finding out that it loses severely against 105W Intel chips in a desktop/server space that snapdragon cores weren’t designed for).
I think it’s of note that ARM parts are now competitive IPC-wise with Intel. Mobile SoCs actually provide more functionality than the Intel parts (AI co-processors, modem, etc), and in some cases they have more performant GPUs. And they do so with lower power and smaller dies, thus at a significantly lower cost.
It always happens like this; Minis started to replace Mainframes, then high end Workstations/Servers started to replace Minis, then PCs started to replace the previous high end, and now it is time for the mobile parts to replace PC CPUs.
It’s a matter of economics, money is now on the mobile space. Intel wont be able to compete against a bunch of fabless ARM vendors eventually. They no longer have the node nor the architecture leadership.
It’s interesting how the same story keeps repeating.
As a matter of fact my main desktop has been an ARM single-board computer for the past two years. While I was skeptical at first, the software has improved to such an extent that I can do pretty much everything on it that I can do on my powerful x86 laptop and that includes C/C++/assembly development and watching videos on YouTube.
And this is just a low-end device with Cortex A53 cores and Mali 400 graphics. Newer boards such as Odroid N2 and Khadas Vim3 will be an absolute pleasure to work with, if I have so little to complain about the performance of this system. The only thing that is really important is to use a modern microSD card, because the speed of the card has an immense impact on the overall performance of the device.
Can you tell us a little bit more about your setup ? Chose of device, etc.
Like what Linux distribution (or other ?) you are using and what Mali 400 driver did you end up with ?
I’ve seen SBCs which take regular SSDs, I think those might even be more suitable as a desktop.
Lennie,
I concur with psychicist’s opinion that ARM computers really can be effective as desktop replacements. My most recent purchases are an odroid xu4 and banana pi m3. As long as you have an appropriate distro installed, the user experience is practically indistinguishable. Even if you are programming, it’s all the same tools and environment. These SBCs are cheap enough that you might as well just get one to tinker with
However the update situation is similar to android (ie terrible), and you’re often locked to a specific ARM kernel provided by the manufacturer as the mainline kernel isn’t supported. You’ll install a prebuilt image (of ubuntu or raspbian) from the vendor and you’ll be dependent upon them for updates, in my cases I was running older kernels right out of the box. This is much less of an issue with x86 vendors. It’s usually not the vendor’s fault, they’re just using the latest kernel from the manufacturers. I’m extremely eager to get away from this problem, I’ll be very happy if someone’s finally cracked it.
I was drawn by the notion of running octo-cores at gigahertz speeds on passive cooling, but if you place constant heavy loads on these ARM SBCs with passive cooling, they will overheat/throttle. I have oversized passive heatsinks on mine and though they absorb a lot of heat, the dissipation is still insufficient for steady state loads. I’ve ended up needing fans on my “passive” heat sinks. It’s not a big problem for normal desktop use, just don’t expect to run all cores at 100% without cooling. When I do throw heavy loads on my banana pi, the desktop becomes unstable when it throttles. That said, my use cases are more demanding than ordinary desktop users who are much less likely to overload these with normal desktop use cases.
Alfman,
Yeah, I had an ARM Chromebook in the past I was able to install a regular Linux distribution on that also mostly worked.
My main issue as you mentioned is kernels and GPU drivers.
The recent announcement by Samsung of them using AMD GPU IP. People are saying: they will probably use that to replace the ARM Mali GPU IP on their Exynos SoC. The current AMD GPUs actually have good open source drivers which could mean that the new Exynos SoC with AMD GPU could in theory also get good open source drivers (!). That would mean Samsung ARM devices could have open source GPU drivers. That could possibly change the game. Do you think this might happen ?
Lennie,
Honestly, your guess is as good as mine, but here’s hoping. If that could be solved, finally we might have ARM SBCs that would be a delight to work on.
I had a Debian 8 image from Orange Pi that I immediately upgraded to Debian 9 running for a year, which did have various shortcomings.
Last year I found they had uploaded another image with Ubuntu 16.04 so I upgraded it to 18.04 and respectively 18.10 and recently 19.04. The kernel is old and a bit problematic and I am only using fbdev for graphics at the moment.
But to my surprise audio works using ALSA and JACK after disabling Pulseaudio and Firefox also started working. So gradually as software is getting fixed and improved with newer distributions the device also becomes more and more useful.
A few days ago I compiled QEMU 4.0.0 and installed Debian amd64 in it. I need a bit more space to install Ubuntu 19.04 but that also works using a Tianocore EFI.
I couldn’t get the Linux 5.2-rc3 kernel I built to boot but that must be an error on my side. The Lima driver for the Mali 400 graphics is also upstream, which should result in better graphics performance.
The Armbian image I tried a few days ago still has problems so I am not using it at the moment. All in all the device has become more useful over time.
I believe you are talking about the 8cx and not the 855 compared to a 2017 Kaby Lake-R, and at least according to this https://www.pcworld.com/article/3397783/qualcomm-snapdragon-8cx-benchmarks-vs-core-i5.html
At times, students like it when teachers make an effort to prepare a graph or a certain image on their own to help them learn they are always available to help you out with assignments by offering brilliant cheap dissertation writing service. So, make sure you invest in their services without thinking twice.
Can we get a button for reporting spam comments?
Look, this way :
https://www.youtube.com/watch?v=JEwtWVTRW4c
I love spam.
I’ve gotten a kick out of the Tocino Spam lately. Sooo gooood.
How about adding login captcha to all accounts younger than 30 days?
dsmogor,
I hope you’re not suggesting google recaptcha, we already had it enabled on the wordpress site and it was a nuisance. To be fair, I don’t know if there was a time limit. It could be tolerable for new accounts/creation, but when it was originally enabled for the new wordpress site it was prompting me frequently. I pushed strongly to have google’s re-captcha disabled for us because it seriously degrades the website usability for users like me in particular. For better or worse google’s captcha is based on off-site activity and it makes the captcha especially difficult for users who don’t use google products, regularly clear cookies, block google tracking scripts, and don’t login to google accounts from their IP. It’s bad for privacy (I’m not even a fan of wordpress auto-embedding these youtube clips, since that’s another tracker for google). Beyond the annoyance and privacy factors, there are numerous occasions when I am denial of serviced by google re-captcha and it refuses to let me through regardless of how many times I answer captcha questions on websites that I’m a legitimate user of.
It’s tough to create an unobtrusive captcha that works long term, and IMHO google’s failed even though many people still use them by default. Ironically AI is to the point where computers are even better than humans at solving these turing tests. So how do we fix that? Well, I know osnews isn’t looking to do any development whosoever, but theoretically a custom captcha solution could do the trick not because it’s difficult to crack, but because it’s simply not worth the spammer’s time to retool for small sites. In other words, one doesn’t necessarily have to come up with a captcha that can’t be solved by computers, just exploiting the lack of developer resources to break the captcha may be enough…at least until the AI advances to the point where it’s able to replace us developers, haha.
A very good article. Thanks to the author. It is very important to support and help students. If you need help writing an essay, come to my site homeworkfor.me .
https://www.youtube.com/watch?v=JEwtWVTRW4c
This article will help to architecture students. Architecture students can go on pay for research paper for receive research paper help for architecture assignments.
https://www.youtube.com/watch?v=JEwtWVTRW4c
Given the recent Huawei treatment and the fact that majority of top 10 phone brands are from China, Arm designs over all just got A LOT less relevant. We should be now listening for the news from the https://en.wikipedia.org/wiki/Loongson front because that’s what these producers will be using in the future.
It’s not easy to port an OS (Andriod) to a different CPU, especially when you start looking at a high-performance (and therefore instruction set dependent) Dalvik JIT that’s needed for almost all third party apps.
Huawei can continue using their existing “already licenced” ARM designs, so they won’t really have much reason to change CPU for about 3 years. If the trade wars last longer than that; ARM is a UK company and not a US company, and they like money, so it’s likely they’ll find a way to licence new ARM designs to Huawei. From what I’ve heard ARM actually has 3 CPU design teams where one is in the US and the other 2 aren’t, so maybe it means Huawei only get access to 2 out of every 3 new ARM designs. Of course smartphones don’t really need faster CPUs anyway, so maybe nobody cares in the first place.
Their main problem for smartphones (if trade wars continue) is that they won’t by able to load privacy violating spyware from the world’s biggest spammer on their products (no Google Play). Quite frankly, in the long term, I think that’d be a huge benefit (lots of people would prefer a good alternative, and Huawei already have alternatives for most of it that they use for the local Chinese market).
Brendan,
Yeah. Obviously there’s short term pain for huawei, but I think this could ironically help elevate a huawei alternative in the long run. More mobile platform competition could actually be good for consumers, but in terms of the whitehouse’s trade war strategy, this is emblematic of short term thinking. That we’re on the fringe of a new trade war with mexico is reckless, our manufacturing/farming markets are highly integrated. Both countries need each other and by the time the fighting is done we’re just going to be left with casualties and little to show for it.