MDD G4: $70 a good price?

[Cory5412]

when there was some talk about what it would take to make a viable not-Intel platform for the data center.

Itanium is only just spinning down, SPARC and POWER have been being sold continuously, and are still being developed (even if not as much as before) and sold by their respective vendors. Neither is even that expensive compared to similar grade server hardware. I.e. if you're maxing out a Dell R740 or R7415 or an HP DL380 then there's likely a POWER or SPARC system in your price range.

The viability of that depends a lot on what you're doing. In a lot of cloud vendor cases, I can see why OpenPOWER (or any remnants of the days around fifteen years ago when there was an open source SPARC design) would be Of Interest. I suspect it has less to do with Intel specifically (although in the past few years that's probably become a bigger issue) and more with being in control of as much of the server design as possible. It's been well known, of course, for a long time that the biggest tech companies and cloud vendors aren't out there just buying the same Proliants and PowerEdges everyone else does.

But, point taken about OpenPOWER seeing some action at this piont.

In other datacenter curiosities ARM is realistically the big exciting new entrant to the datacenter (if only because by technicality POWER and SPARC never left)

$1000 is definitely closer to the curiosity range. I know Amiga enthusiasts happily pay around that much for PowerPC desktop systems and have for a while. It'll be interesting to see where you start seeing these things pop up, priced like that.

Predicting what will be valuable and desirable in 20 years' time is as much an art as it is a science.

Yeah, as time goes on and, importantly, as the values and priorities of people acquiring these machines changes, it'll be interesting to see how values change.

As I mention on a regular basis, we're kind of (I hope) at the tail end of people trying to treat late MacPPC machines as viable main modern computers, so today the value is mainly in the later and higher end machines: Power Macs, MDDs, The Quad, The 2.7, the 2005/PCIe versions, and the highest end PowerBooks anybody can find.

In a couple years as it loops back around to people having nostalgia for these machines, I suspect we'll see the values start to even out as people distribute themselves between "whatever I can get" and "what I had back in the day" (which is going to be eMacs, iMacs, iBooks).

Different individuals also try to collect at different points of interest, depending on how far ahead they can see, what interest them, or what deals they're looking for.

In terms of what will stand out though? Who knows. Apple's product line contracted a lot after the late '90s and really I have yet to see interest in anything specifically explode the way it did for the SE/30, Color Classic, and perhaps the TAM or a few other models everybody seems to want.

 

[nglevin]

In other datacenter curiosities ARM is realistically the big exciting new entrant to the datacenter (if only because by technicality POWER and SPARC never left)

The biggest problem with the ARM ISA has been, and will likely continue to be, that Apple has an impossible to acquire license to make their own variants on the design, and Apple appears to make the best ARM chips.

Which is a nicer way of saying that stock, unmodified ARM is not really that impressive compared to POWER and even SPARC. If Qualcomm or Samsung wanted to get ARM there, that would make the industry more interesting, but Samsung's working on POWER10 and Qualcomm's happier to stay with embedded devices.

My guess is that, in 2039, it will be the case that will be the attractive part of turn-of-the-millenium Mac OS X computing rather than the exact innards, assuming that we aren't too precoocupied with Basilisk III on our Apple Glasses.

I don't know. I would have thought that emulators would have knocked out the market for fast Quadras, but they ended up making a reasonable comeback on the second hand market as the supplies collapsed.

We are about less than a year out from 32 bit Mac OS X software being unusable on what will be the latest iteration of macOS, and maybe there will be a special Apple ARM laptop to go with it. It will be interesting to see where things land later this year, going into the next.

 

[Unknown_K]

Some 68K gear is worth a few bucks, even systems like a 950 that people didn't want to pay to ship are now being sold for a couple hundred a pop or more. 68K gear was expensive new and people held onto them for a long time so most didn't get recycled. PPC gear seems to end up at the recycler quickly since it was cheaper new. Capacitor rot has killed plenty of 68k machines and so has exploding batteries. I dread the day I need to recap early PPC boards since they have so many capacitors on them.

Is apple killing off 32 bit apps on OSX? People will be running 32 bit apps on Windows way after I am dead.

How is SUN Sparc gear selling these days?

 

[nglevin]

Is apple killing off 32 bit apps on OSX?

This support document does a good job of laying out the situation; https://support.apple.com/en-us/HT208436

Unless there's an unprecedented about face in the next iteration after Mojave, all signs point to "yes".

 

[Unknown_K]

Well I don't run anything after 10.6.x anyway so 32 bit apps are good enough for me.

 

[EvieSigma]

Some 68K gear is worth a few bucks, even systems like a 950 that people didn't want to pay to ship are now being sold for a couple hundred a pop or more. 68K gear was expensive new and people held onto them for a long time so most didn't get recycled. PPC gear seems to end up at the recycler quickly since it was cheaper new. Capacitor rot has killed plenty of 68k machines and so has exploding batteries. I dread the day I need to recap early PPC boards since they have so many capacitors on them.

Is apple killing off 32 bit apps on OSX? People will be running 32 bit apps on Windows way after I am dead.

How is SUN Sparc gear selling these days?

For a ton, it seems. Or at least people are trying to sell them for a ton.

 

[Cory5412]

There's new SPARC gear being sold, although I would imagine Oracle sells more x64 servers, and sells more software for non-Oracle servers anyway.

In terms of the used Sun market, it seems to have come up in the past couple years. You couldn't practically give away an Ultra series a couple years ago and now I see them fetching a little bit.

It's still not as bad as if you want to try to get an Alpha, but they seem to be holding their value.

Which is a nicer way of saying that stock, unmodified ARM is not really that impressive compared to POWER and even SPARC

It depends on the task. It seems like the use case is similar to the old SPARC T1/T2 types of designs and old AMD "magny-cours" (haha get it many cores) designs where you're optimizing for threads instead of raw speed. VPSes, processes that do the same thing over and over to small datasets (Sun's sales pitch for the Niagara/UltraSPARC T1 chips was essentially web serving and doing things like parallel image processing in web hosting situations, basically, using it for things like hosting facebook/flickr et al.)

Whether or not that market actually shapes up that way or it still ends up being cheaper to custom build some Xeon Gold/Platinum boxes or just rely on throughput over total number of cores.

Apple absolutely has the best ARM designs in their particular power envelope. Most of the server-based ARM designs aren't the same chips being used for other phones and devices, or if they're the same core, they're deployed to make use of the fact that they're in stationary computers with big power supplies and good cooling.

There are (well, were) a few real ARM servers that have shipped, because Apple isn't the only one that has a license to do that kind of design work, Calxeda is obviously dead, and it appears whatever it was IXSystems was using in one of their ARM-based storage boxes is dead.

Gigabyte appears to have some using chips that are now being sold by Marvell. (though, those particular server models might be old because they're labeled as being Cavium brand and I don't see that model listed on Marvell's site, so I don't know if those machines are actually available to buy.)

I believe the HP Moonshot system also had an ARM-based blade, the idea there being, basically same thing, lots of really small servers doing work lashed together, perhaps with kubernetes or whatever system automation tool was popular at that moment.

Though, it looks like not much exists in the rack-mountable or ARM-based blades realm right now. It'll be interesting to see if that comes back.

At any rate, it'll be interesting to see if anything can make a meaningful dent in the position x64 holds in the server world. Given that POWER isn't really new and that a couple generations of OpenPOWER have as far as I know have failed to make a big difference there, along with the fact that SPARC is still plugging along.

The next thing is risc-v, but I doubt that's in a place to be put into servers yet.

 

[nglevin]

I agree. RISC-V is about where ARM could have gone, if it wasn't for the IP issues and licensing. Which likely got more complex after Softbank swallowed ARM whole.

We'll see. It's early enough in its hype cycle that my enthusiasm is extremely tempered at the moment.  ;-)