Isn't an issue here that Arm is RISC? As far as I understand, that takes way more complicated software code to do what you can do with less complicated software code on x86 and those instruction sets.

Sure, x86 might be outdated and prime for a rethink, but RISC the way to accomplish that? Or is it better with something more similar to x86? There is probably a reason there are not so many Arm based servers.

From your link, the ARM desktops are fairly expensive. Are ARM desktops generally more expensive at the moment?

All modern CPUs contain similar hardware, independent of the instruction set. All modern software runs on any CPU instruction set. The RISC versus CISC debate became irrelevant in the 1990s.

High-performance ARM CPUs are a recent development. We will see if they gain any market share compared to x86. I liken the high-performance CPU business to the Olympics - there is a reason why not many people win.
Ed

Yep. They don't have the advantages of bulk buying at the moment. The thing from solid-run.com is the cheapest @$750, but then it's probably beatable with x86 equivalents. It's going to be close to $1500 by the time you put flesh on the bare bones. And it's not a finished product, it's a usable design project.

....but the Raspberry Pi400 or 4B are a lot cheaper, & very usable as desktop computers; I'm using one right now!

These are all low power machines. There isn't a desktop ARM chip yet that can compare with high end AMD64 such as Core i5/7/9 or AMD Ryzen.

I remember hearing this talk in the early 90s when Apple switched to the PowerPC. In 2005 Steve Jobs announced Apple was switching to Intel i.e. back to CISC. Nearly 30 years later that "outdated architecture" dominates both desktops and servers with cheaper AMD64 servers such as Xeon or Opteron replacing RISC offerings from Sun Sparc and HP PA-RISC.

ARM may be the future but Intel is not going away soon. ARM dominates mobile devices due to the lower power requirements but those concerns mean little to most desktop users. Also unlike servers desktop users don't need 16-core CPUs as desktop software isn't that multi-threaded yet.


The following quote needs some major explanation.

I was unaware that Audi and VW made computer CPUs. I do remember a major auto emissions scandal where VW cars were intentionally programmed to cheat US auto emissions tests. There are photos of millions of unsaleable VWs in US deserts.

1 members found this post helpful.

Back in 2014 HP released its ARM powered moonshot servers, a 4.3U rackmount enclosure housing up to 45 Proliant sever cartridges. See the HPE Moonshot 1500 Quickspec. Current server cartridges are Xeon powered but back in 2014 you could get ARM powered ones; M400 and M800 now EOL

I can't see anything as a current offering but remember seeing a system back in 2015 when I retired, looked like a box of mini blade servers. Neat!

2014 Computer Weekly article.

Play Bonny!

That's approximately what speculative execution is about, thus the comparison. Speculative execution also introduces several distinct classes of unpatchable security flaws, with many exploits in each category, that Wintel has chosen to brush under the rug using their control of what's left of the trade press. On top of that an article from the 1980s surfaced where they warned not to pursue speculative execution for more or less those reasons.

1 members found this post helpful.

So speculative execution is about criminal fraud in order to bypass US auto emission standards ?!

You are somehow equating a potential security flaw in Intel chips with actual intentional programming by VW to cheat US emission tests. Wow. Was Intel intentionally programming their chips to bypass US auto emission standards ?!

This is not a comparison. There is nothing there.

No there is no issue. Your understanding is incorrect. You seem to be conflating RISC with VLIW. RISC code is about 30% larger due to the need for separate load and store instructions to memory however it is not more complicated.

RISC has been around in servers for over 20 years with Sun Sparc, IBM POWER, and HP PA-RISC which used proprietary UNIX. However they are far more expensive than AMD64 servers and have fallen out of favor.

Both are about cheating on measured performance.

1 members found this post helpful.

IMHO, RISC only made sense with 8 bit computers. 8bit lasted longer than they should have for a silly variety of reasons.

In something like the Z80, you had 1, 2, 3 & perhaps 4 byte instructions; You were probably running CP/M and if you needed more than 64k, you had to use memory pages, and either switch over for handling one program or duplicate the OS in a "Multiverse" type situation. You then had to keep careful watch since you had zero memory management in the Z80

In this situation, RISC was a great idea. It meant neat speedy execution. Coming from the Z80, it was attractive. But with 32bit databuses & a possible 4 Gig of separate instructions available, it's irrelevant.It's useful for academics, who keep it alive, but like other stuff academics have, it's a fossil.

it's a highly political and highly charged question. i won't answer it or i'll be ejected off linuxquestions again!

- ARM is not UK it is china

- ARM is not open: most of it is closed

... undocumented cpu instructions big tech has you don't

... the SOUTH BRIDGE of ARM is totally closed to you: how to access video low level, even what kind of bus - all "under a blanket of details they hope you don't ask before you buy" they say open, but it is in fact the most closed platform ever - more closed than apple iMac running on intel even. it's asian, probably sony vr (virtual reality) video

- ARM is on small platforms with cell call ability and Intel is not (as far as i know) (ie, pads) - currently that is not something that is avoidable or arguable. it just is fact.

- ARM is susceptible to "you lack the newest instructions" issues. exmaple, first thing ARM ?1.2 does is add compat instructions to support "older ARM 1.2". result? you get UBUNTU (old versions not supported and "discouraged"), so you get new ubuntu: it needs the the 1.2 instruction to support 1.1, but your running 1.1 and don't have the new instruction? gotcha. I'm not saying AMD and Intel never had an issue that way.

- getting an ARM that doesn't have integrated (and fully working) 3D video and cell call ability is probably a rip-off. if your "going that way", then don't go 1/2 way. ARM won't be better than AMD or Intel if your getting "an un-armed arm"

- ubuntu google is deep state into "pressing you to use ARM". so be aware if you don't you might "get punished"

Just preserving this.
I hope somebody comes along and debunks most, if not all of the above.