Arm Neoverse N1 & E1 Platforms Announced For Cloud To Edge Computing

[phoronix]

Phoronix: Arm Neoverse N1 & E1 Platforms Announced For Cloud To Edge Computing

Arm announced today their Neoverse N1 7nm platform catering towards cloud workload performance as well as the Neoverse E1 platform for high-efficiency infrastructure...

http://www.phoronix.com/scan.php?page=news_item&px=Arm-Neoverse-N1-E1

[numacross]

AnandTech has an interesting analysis of the announcement. All in all it looks very promising and more competitors on the server CPU market are always welcome

[wizard69]

AnandTech has an interesting analysis of the announcement. All in all it looks very promising and more competitors on the server CPU market are always welcome

Why limit this archetecture to servers though? Plop a GPU onboard and use a on implementation that can use a higher cLock rate and you have an excellent multi core workstation.

[BillBroadley]

Why limit this archetecture to servers though? Plop a GPU onboard and use a on implementation that can use a higher cLock rate and you have an excellent multi core workstation.

An excellent multicore workstation... that can run very few OSs and has few native apps. Want to run Ubuntu-18.04 or similar, sure. Google earth, google chrome, skype, netflix, slack, etc... nope.

Without being able to run OSX, or Native x86 windows 10 the volume is going to be small, which means the price is going to be huge. You'd be much better off with a top of the line Intel or AMD Epyc.

[coder]

Why limit this archetecture to servers though?

Because servers operators are often concerned with throughput-oriented performance, very sensitive to energy efficiency, and most willing to recompile their code to reap those benefits. Many of the same characteristics as embedded, where ARM is already a mainstay.

Plop a GPU onboard and use a on implementation that can use a higher cLock rate and you have an excellent multi core workstation.

You mean so they can grab a small slice of the small Linux workstation market? Yeah, it's hard to see why they'd pass by that money tree and go straight to the massive server market, first.

Most professional workstation apps are not supported on ARM, limiting users to what they can get via their distro or what they're willing to build themselves. So, out of hand, you're already losing out on the most lucrative segment of the workstation market. As ARM gains a foothold in servers and laptops, that will start to change, and you should start seeing some of these server chips trickling into the workstation & desktop market.

[wizard69]

An excellent multicore workstation... that can run very few OSs and has few native apps. Want to run Ubuntu-18.04 or similar, sure. Google earth, google chrome, skype, netflix, slack, etc... nope.

Workstation kinda implies professional use where some of those apps don’t matter. Beyond that most Linux distrust are ported to ARM already. In a nut shell I’m thinking a Linux only machine here but hey if Microsoft can get crappy Windows 10to run on it why not.

Without being able to run OSX, or Native x86 windows 10 the volume is going to be small, which means the price is going to be huge. You'd be much better off with a top of the line Intel or AMD Epyc.

As long as people are looking for alternatives to Windows 10 or high priced Macs we have potential. Beyond that there are a number of reasonably priced ARM boards out there already. I see little reason why a $1200 workstation couldn’t be made with this archetecture. Of course you would need to have somebody implement the architecture in a workstation suitable SoC.

[wizard69]

Because servers operators are often concerned with throughput-oriented performance, very sensitive to energy efficiency, and most willing to recompile their code to reap those benefits. Many of the same characteristics as embedded, where ARM is already a mainstay.

lit the right chip is brought to market it could easily serve both the workstation and server markets. Frankly ARM has Thad a lot of success in the server world so you would think any partners they might have would do what is possible to drive volume. I really don’t see ARM having a lot of advantage in the server space until the put the same effort into infra structure and support the way Intel or AMD does. I just don’t have the warmfuddy feeling that ARM grasps what the server user community really wants and frankly need. That is where custom spins are simply unacceptable as you end up with too much invested with small untried suppliers.

You mean so they can grab a small slice of the small Linux workstation market? Yeah, it's hard to see why they'd pass by that money tree and go straight to the massive server market, first.

The ARM position in the server world is just too small to even consider at the moment. Workstations, even Linux only could easily outstrip today’s server sales.

Most professional workstation apps are not supported on ARM, limiting users to what they can get via their distro or what they're willing to build themselves.

Yes a true chicken or egg problem. You will never solve that problem until actual ARM workstations arrive. On the flip side today’s distrust are fairly complete with their ARM ports. Beyond that it isn’t uncommon for workstation users to write their own software. You must realize that the intention here is not the gaming market.

So, out of hand, you're already losing out on the most lucrative segment of the workstation market. As ARM gains a foothold in servers and laptops, that will start to change, and you should start seeing some of these server chips trickling into the workstation & desktop market.

Herein lies the problem ARMs model makes it very difficult for a smAAll outfit to go out and buy their chips. This architecture will likely end up largely in custom SoC design for or buy companies that will keep the chip internal. If Google or Facebook uses this design, do you think the SoC will becomes widely available? I doubt it.

[andrei_me]

Neoverse is cool and all, but I think Ares was a great name already

[coder]

lit the right chip is brought to market it could easily serve both the workstation and server markets.

Meh, the ARM cores are still optimized for size (so they can pack lots per die) and energy efficiency, whereas x86 cores are still much better on single-thread performance.

In case you haven't noticed, server chips tend to run into higher core counts and lower clock speeds. For most workstation apps, single-thread performance is still relevant. That means a lot less overlap between the markets than you seem to suggest.

Frankly ARM has Thad a lot of success in the server world

The ARM position in the server world is just too small to even consider at the moment.

You're contradicting yourself.

Workstations, even Linux only could easily outstrip today’s server sales.

That's an interesting opinion. Got any facts?

Did you know that Amazon even has their own ARM-based SoCs? To use in their own ARM-based servers?
https://www.phoronix.com/scan.php?pa...formance&num=1

Huawei also makes some:
https://www.phoronix.com/scan.php?pa...ei-Kunpeng-920

Not to mention efforts from players like Qualcomm and Cavium. I don't know how much traction those have gotten, but just how big do you think the Linux workstation market is? I could believe that Amazon already has more ARM-based servers than the entire Linux workstation market.

Yes a true chicken or egg problem. You will never solve that problem until actual ARM workstations arrive.

Well, laptops could start the ball rolling. Qualcomm looks to have a pretty decent Coretex-A76 based SoC for this segment.

You must realize that the intention here is not the gaming market.

Who said anything about gaming? People dropping fat wads of cash on true workstations are mainly doing it for CAD/CAM, visualization, etc. Those ISVs are not looking to support any more than the minimum number of hardware configurations. Most probably won't touch ARM until it already has a significant market presence and there's real demand among users.

Herein lies the problem ARMs model makes it very difficult for a smAAll outfit to go out and buy their chips. This architecture will likely end up largely in custom SoC design for or buy companies that will keep the chip internal. If Google or Facebook uses this design, do you think the SoC will becomes widely available? I doubt it.

ARM's model is its strength. They don't really care where their cores end up, as long as revenues increase. And they sell their cores to anybody - so Google or Qualcomm using these cores is not going to preclude somebody else coming along and doing something different with them.

Honestly, ARM has been one of the break-out success stories, in the past decade, with few signs of the momentum slowing. A lot of that is specifically down to their business model.

[Skwaloo]

In Top500, an ARM server is 204. Not bad !

https://www.top500.org/search/?q=arm