You know what would be nice: Apple engineers and designers finding this thread and others like it and using it as inspiration to build another lovable Mac Pro.
The coffee can Mac Pro was a swing and a miss: an engineering achievement of sorts and an aesthetic bold move, but not loved by devs or scientists. I want an excellent case I can pack full of RAM, disks and GPUs, not external expansion with $70 cables that glitch and mess up the aesthetic in the process.
I don't know if Apple can make enough money on workstations to keep their attention, so maybe those days are gone.
> I want an excellent case I can pack full of RAM, disks and GPUs, not external expansion with $70 cables that glitch and mess up the aesthetic in the process.
I feel like Apple was onto something here, but they screwed up by trying to scale up the mobile paradigm (black boxes connected by cables) when there's a very similar paradigm much friendlier to professional use—the server rack-mount paradigm—that they could just scale down.
Imagine a small "compute core" just like the trashcan Mac Pro (CPUs + RAM + GPUs, replace it all at once when the time comes)... except it's in the form-factor of a server blade, and is sitting in a rack roughly the size of a workstation PC.
You'd buy a "Mac Pro", and in the box would be the (very elegant, Apple-y) rack, one "compute core module" to slot into it, and one "storage module" as well (which is your boot device, making the compute-core diskless and so a commodity.) You could customize your order to get a RAID storage module instead. You could buy more pre-populated modules from Apple, or other vendors, or buy empty modules to install whatever components you want. Third-party empty modules would probably cost about as much as a USB hard-drive enclosure costs today.
And, of course, the modules would all connect to a shared backplane... which would just be PCIe. So it's all the same Thunderbolt idea, just shaped differently.
So basically you are imagining something like Razer's never realised Project Christine [1], but in a blade form factor? I would like that. Also something like that being made by Apple (which has the market power to sustain an add-on ecosystem) would make more sense, whereas Razer would have never had enough traction (hence the project being cancelled).
I think enclosing each module like Razer did is defeating the purpose. Problem is, if you don't enclose I imagine it's hard to get this past the designers. Apple would need a way to make the internals themselves look good. That being said, the old Powermac / Mac Pro towers had very nice looking internals, I could see them exposing them like that....
Plus, it surely would have a nice looking front door.
Yes, but an Xserve would not fit under your desk; nor would its power requirements be compatible with the output of a 110V outlet; nor would it be bearable in sound or heat output to sit near one when it's running. And, on the other hand, an Xserve takes up 2U for a complete computer (a rack is for a cluster), whereas in the hypothetical system I'm talking about, one computer takes the whole ~10U by spreading its parts out over all those bays.
I'm talking, here, about the consumer electronics-ization of the concept of rackmount servers (or, more specifically, of blade-server chassis-es.) Even if Apple still had the folks who created the Xserve on staff, there really wouldn't be much in common with the resulting architectures. Totally different needs.
AFAICT this model never took off because the total cost came out really high. People have been clamoring for external GPUs for at least a decade, but even nowadays that it's technically feasible by the time you drop the cash for it you're a good chunk of the way to a new gaming rig.
Hell make it a Eurorack case. But all Apple styled. It's all the rage with hipster musicians. Imagine being able to intermix your analog modular synth modules with your Mac modules in the same cage.
If the backplane is PCIe, it'd be really hard to prevent third parties from just making their own modules. Or selling generic $10 PCIe-card enclosures.
But that's rather the point: if Apple made this thing, they'd know that, so they probably wouldn't be trying to prevent an open ecosystem of modules at all. They'd be encouraging it, since that's the whole pitch behind why it'd succeed as a "workstation" product in the first place.
Geneticist in a large academic lab here, I'd say 99% of all computers in use are Macs. Some instrument-tied PCs, a couple of Windows laptops, and that's it. Any data analysis (and these are predominantly sequencing scale datasets) that isn't running on the institutionally provided cluster is also being done on Macs.
In my ~20 years on the job in various places, OSX really changed everything; labs used to have a few Macs around for non-analytic tasks where the old MacOS was just easier. Post OSX, it's nearly all Macs unless it ships with an instrument of some kind (and then it's likely Windows). The analytic tool chain is/was massively unix based, so adapting to OSX was easy and came with a very nice GUI that also did the aforementioned easy non-analytic tasks.
The high-tide of Apple machines in genetic research actually has nothing to do with what most people might presume and has a very specific, very technical reason.
The AltiVec (“Velocity Engine”) present in G4 and even more so in 64-bit G5 machines was discovered to be extraordinarily efficient at performing the calculations required for the Basic Local Alignment Search Tool (commonly known as BLAST). At the time, the G4/G5 architectures had a massive lead on other chips present on the market at the time (i80x86-derived machines such as Pentium III, Pentium 4, and Pendium M; but also other exotic architectures such as MIPS, SPARC, and DEC’s Alpha.
Eventually other architectures caught up by adding vector units or vector instructions of their own, and Apple stripped any basis of fact out of this when they switched to Intel, but by a form of institutional inertia, Macs still rule in the genetics segment.
Yes, but can vary much from department to department. A decade back the immunology department I was in gradually switched almost entirely from Windows PCs to Apple Macs. Primarily by people seeing how much nicer and more productive a user-friendly Unix-based system was than Windows; even the die-hards eventually switched over. We also had a large number of Linux/Unix people as well, particularly for bioinformatics, who bought fully maxed out Mac Pros from the G5 onward to do work which required a lot of memory and disc. Back then, they were good value for money.
Today, Mac usage is declining. The cost:value isn't there and they don't manufacture any high-end kit. Given how much PC you can buy for the same price as a MacBook, like a 32- or 64-core Threadripper or EPYC with a massive GPU which blows any available Apple hardware away, it's not hard to see why. For computational biology, running Linux on it gives you a much more capable and productive system, and even Windows 10 with the Linux subsystem is perfectly adequate. Apple's mistake has been to focus upon only profitable consumer product lines, and drop the ball entirely on the high end. They didn't sell as many high-end units, but they were the driver for Mac adoption in many places, and without that Mac usage will continue to decline. A mediocre laptop isn't good for a lot of scientific uses.
Ignoring an entire market is how BlackBerry got clobbered by Apple. They didn't see a consumer device company as a threat. I don't know what vectors of attack there are on Apple from the business and research machine market, but someone is going to figure it out.
Seems there's no real attack vector; it's just a matter of waiting and doing small tests to see if Apple are completely blind yet, then doing some big thing.
but do "lab workstations" really equate to "workstations of people in the business" or "computers this school got a good deal on" ? I think the latter. I suspect that they're irrelevant to this discussion.
I can report that the climate scientists I knew mostly worked on macs (laptops, largely; a variety had trash-can Mac Pros as well). Fairly sizable motivator was the Unix CLI; they all had come up working on supercomputers with C/TC Shells, so having that locally was a big win.
AKA an expensive way of running telnet or xwindows sessions - when a basic leveano or dell would be cheaper and run full fat office and not the cut down mac versions
Research profs travel a lot, and thus enjoy battery life and light weight. Many CS and math profs live in LaTeX and shell and not Office. Mac is handy.
That's a matter of opinion and finder is a terrible UI for a file manager for any one who has used windows or Linux file managers.
And the way Apple has diverged from a more traditional UNIX like Linux is a big problem if you are developing to run on Linux as it adds a whole new set of risks.
More like Sun in those days (or maybe SGI or IBM rs6000, possibly HPUX though it was more for business) depending on whatever proprietary things you wanted to run.. but really Sun.
from there: linux PC's until OsX, then a mix of both.
It was certainly its ambition to be an academic workstation originally but lots of other things were. 'OS (or hardware) of choice in sciences' doesn't really reflect reality.
Huh? It’s a market. Not the biggest market, but a real market that spends billions, has extreme needs in some cases, and influences well-trained people who become decision makers in other markets.
Back in college (1998-2002), I had an old, white-bearded CS professor who kept (and actually did work on) an old NeXT Station in his office. He was a little... off.
I wonder if in a few years people will be fondly writing articles like this about the 2013-16 Retina MacBook Pros. Mine is getting a bit long in the tooth, but what to replace it with?
Ya, i'm holding on to my 2013 Macbook Air. Hoping there will be a new version of it soon. I'm going to wait one more year before jumping ship to a Linux ultrabook.
The note about the amount of fans, and how silent it is despite the relatively large power draw is interesting. I too ended up realising this when I purchased an Apple workstation - not the G5 but a Mac Pro, which was the first desktop I hadn't built myself since I had a Pentium 133 from Digital.
Until the Mac Pro I hadn't considered that a modern, high power computer could also be dead silent. It's stupid, but that was a revelation for me. I couldn't even tell it was turned on or not, except from the occasional hard drive seeking noise - which I fixed with an SSD. And 7 years later it still is as silent.
Unfortunately Apple is now focusing on keyboards with no depth and whatnot, so I'm back to building my own but I do keep in mind the lessons. My system now has 10 fans + 1 PSU fan, all running between 700-900rpm. I can hear myself think again.
http://www.silentpcreview.com has been a great resource in my pursuit of quieter computing (sadly not updated as often as it used to be). I think I currently have 7 or 8 fans, all high-quality and large, and all spinning as slowly as possible. I can't hear anything, and it's wonderful.
I meant the keyboard thing in general with regards to their apparent engineering direction. My current hackintosh does have a very loud keyboard otherwise :)
Other than the common correlation with electricity usage and possible environmental benefits, I have trouble sharing the concern for the loudness of a powerful PC. Unless room space is very limited, I always just kept my PCs far enough away from my desk that the sound of fans became little more than regular ambient noise, not very noticeable against the fans or air conditioning I already had.
Granted, I never had any insanely beastly machines, generally just sub-$1500 gaming rigs. The only sounds that I ever even noticed were shorter irregular ones, like spinning disks. Perhaps the problem is simply more noticeable to other people, especially those with more powerful and thus louder machines, or people with smaller spaces who have their machines much closer, like on their desk or around their feet.
I care about ambient noise much more for other applications, like laptops or home theater PCs. For both of those, I have a strong preference for fanless models if at all possible.
The trash can Mac Pro seemed to be designed to sit on a desk, so quietness makes sense, although I don’t understand what’s so important about having very powerful machines on your desk. For machines like iMacs, the problem of placement is obviously unavoidable, but I struggle to understand why that form factor is desirable for a very powerful machine.
These days it is actually pretty easy to casually use liquid cooling for CPU and GPU fans which ends up being significantly quieter.
I have routinely been building nearly silent 99% maxed out gaming/vr rigs with integrated off the shelf liquid cooling units with no tubing or scary liquid interactions needed.
No, as I said I didn't imagine it possible from what I was used to, so I never spent the effort to look it up. I thought the difference would be between loud and very loud.
All I would hope is that the design be pragmatic, rather than heroic. The Trashcan was pure heroic design, throwing massive resources at a problem that barely existed, pro desktops being too bulky. The G5 tower was almost the exact opposite, simply an elegant, thoughtfully-designed take on a proven, pragmatic form factor.
Agree with your sentiment. Also I really hope they include a low/mid range option for folks like me who want iMac-esque computing power, but don't need the display that comes with it.
The Mac mini is way too underpowered, but the next step up from there is $3,000.
They could even name it Centris or something like that - to indicate it's in the middle of their offering. On a serious note, try a hackintosh in the mean time. It's not difficult at all.
Assuming it’s a Core i5/i7 with the same tower case as the new Mac Pro, Apple should call it “Mac Tower”. The Xeon-based tower would then be “Mac Tower Pro”.
I was a teenager when the G5 came out and instantly fell in love with the design. Could not afford one of course, so built many computers with Lian-Li cases over the years all inspired more or less by the G5.
They still make the best PC aluminium chassis for any form factor in my opinion, and I have fond memories of assembling HTPC:s, desktop workstations and "bookshelf cubes" for my friends and relatives, all using Lian-Li cases.
Many times more or less dead silent (again, impressed by my first experience with the G5), I remember going to some lengths with replacing fans on graphics cards, CPUs and using passive power supplies.
Today I use a Macbook Pro and just can't be bothered assembling any PC computers anymore. Should I ever want to though, finding the right case should be easy.
If it does, I hope they use better quality/designed fans. As someone who's had both the PM-G5 and the MP-2010, I wasn't impressed with the acoustics of either of them, even at idle. [Picky? Guilty as charged]
The early to mid year Mac Pros were certainly good value for money though. No chance any new Apple workstation will be keenly priced!
It's beautiful, but carrying one around is a pain in the ass. I hope they make it out something lighter. Used to own a Intel Xeon Mac Pro, man that thing was the heaviest computer I've ever owned.
FWIW, we retired one of our CAD workstations at work a few years ago. It was a Dell Precision machine with similar specs as a 2007-2008 Mac Pro, and it was about as heavy, too.
I was really happy that the new workstations we got were much lighter.
I got one half a year ago just for fun (and because I wanted integers to be represented in memory the proper way round for the first time in my life. ;-))
I set it up at my workplace because my wife would kill me if yet another desktop PC appears at home, although the design might have appeased here enough. My colleagues just called me crazy for getting this useless old heat producing rubbish. I managed to install debian on it, which was a bit of a hassle (protip: If you only install one HDD, put it in the primary drive bay, otherwise openfirmware and debian will disagree about the disk naming/path and you'll have a bad time) but worked reasonably well in the end.
Also, when upgrading RAM I must have mixed ECC and non-ECC (or buffered and unbuffered?) because I just grabbed the first DDR sticks I could find at work. The machine still worked, but memory bandwidth was cut at least in half or so, notably slower. After I made sure I had matching sticks, everything was fine and I could enjoy the incredible speeds of DOUBLE data rate memory.
The first time I saw one of these was when my boss (a huge Apple fan boy at a time when that was a difficult lifestyle choice) managed to somehow convince the local Apple distributer to loan the office a demo machine for some reason.
You have to place the machine in the context of the time - the office was filled with crappy desktop machines from Dell, Compaq, and a bunch of white box OEMs. The G5 tower suddenly arrived like the monolith from 2001 with us all clustered around like filthy apes.
We never did anything useful with that G5 and after a few months it disappeared. But its job was done, over the next couple of years nearly everyone uplifted by that machine purchased at least one Mac for personal use.
I had this type of G5 machine for eight years. While I upgraded the disk and memory at some point, and decided to upgrade the GPU, the only thing that ever broke was that I had to send in one of the displays for a couple weeks because of a power supply issue (so the display failed, not the computer). And when I finally decided to get rid of the machine, it still worked perfectly; I gave it to a friend and his kids have been really happy with it. My reason for replacing the machine was purely due to needing a good Intel box for modern software development.
Apple really built things to last, and this is something I could never convince people of years ago. Everyone was so sure that some crappy PC costing a few hundred less was somehow so much better than these “expensive Macs” at the time, never understanding that the PCs were breaking all the time and the Macs basically never did.
> Everyone was so sure that some crappy PC costing a few hundred less was somehow so much better than these “expensive Macs” at the time, never understanding that the PCs were breaking all the time and the Macs basically never did.
Eh, I'm the proud owner of a crappy PC costing many times less than a G5, and it's been chugging along with since 2007 just fine. Some of the parts have been periodically upgraded where appropriate, so calling it the same PC is perhaps a bit of a Ship of Theseus dilemma, but there are certainly components in there which haven't changed at all (e.g. 3x 320GB Seagate drives, which are still going and I'm too lazy to chuck out).
Being able to build and upgrade it piecemeal has allowed me to go for good specials, and I've generally been buying at the price / performance sweet spot at every turn. It shouldn't be underestimated how much of a saving that gives. I probably still haven't spent nearly the price of a G5 on it, and I have a PC which is basically on the VR-capable performance line.
Bu basically, just don't buy the cheapest components and you should be fine.
They had great error-reporting features. If your kernel crashed, you'd know it if you were anywhere within 100 yards of it because the software-controlled fans would spin up to full blast.
I just checked and if you include the Power Usage of the Apple Cinema Display, my machine idles at just over 180W
It's a pity, because it runs quite well - always has. I have 10.4.11 and 10.5.8 on there and favor Tiger.
I have been trying to sell it, but obviously nobody wants it. On the other hand, i still have Ableton Live 7 and Logic installed with all my plugins, GB's of Samples and loops on it and I am having a hard time letting go the ability to open my old tracks...I know I will never need to, but I think I'll always keep it for that reason.
As a college freshmen in 2003, I used my once in a lifetime Apple Student Developer 20% discount on a stacked PowerMac G5 dual 2.0GHz. I loved that machine and couldn't believe how snappy OS X was on it. The way I tested UI speed then was clicking on a menu item (say the File menu), and moving my mouse back and forth on the other menu items as fast as I could — the G5 was the first computer I had that could keep up and show all of the sub menus in a flash without stuttering.
It’s funny how little tests like that can make such an impression. I experienced the same thing with iPhones, which had the bizarre feeling (now more normal and expected) of the content being attached to my finger when I scrolled at any speed, even back when the newly revealed content would just be a blank grid until it had time to paint the real content. I distinctly remember scrolling web pages as fast as I could on my friend’s original iPhone when it was first released, although I wonder how horribly sluggish that would feel now.
I had been trained by twitchy FPS video games from a young age to recognize even small frame skips in moving content, and the rarity of such frame skips is what enamored me about the early iPhones (although I didn’t “switch” to iPhones until the iphone 4) and made me dismiss Android phones for many years (Android flagships have been good in this department for a few years now).
I would bet there are still quite a few G5's to be found in recording studios around the world, not to mention G4's and even a few G3's.
Post-2004, did Apple follow up with a suitable replacement for the G5 in that context, i.e., audio engineering? What was it?
Addendum: By "audio engineering" I mean "DAW" use with software such as Pro Tools, DigitalPerformer, etc. and some non-Apple, dedicated hardware for line level inputs and outputs such as that made by Digidesign, MOTU, etc.
(The performance of "web browsing" on today's bloated web and other tasks not related to running the above software, which may include older versions, is irrelevant.)
Maybe Apple's acquisition of Logic had something to do with what happened?
I've a G3 Powerbook Pismo, it's not usable anymore for even basic task like web browsing (yay, 60 second at 100% CPU to load gmail, and with the most minimal setup I could come-up with).
For music, just no, for basic recording, it can do the job, barely, but for music editing (mixing tracks, applying effects, etc), you need a decent computer.
The modern web is an especially poor comparison. A big part of the reason why web works at all on modern computers is due to the intensive optimization efforts put into browsers, efforts that aren't backported to older operating systems or architectures. Chrome V8 will run on OS X 10.5 but not on PowerPC. So on a modern Macbook you'll get highly-optimized code from the JIT, and on an older G3 you'll get a much slower interpreter.
Native optimized apps that were made contemporaneously, like music studio software, are a different story. If it worked well on a G3 in 2000 it would still work well today. So maybe you have a bunch of plugins that you love using that you can't track down new copies of, or you don't want to pay for newer versions, or there's a computer in the corner with Gigasampler.
Music editing doesn't actually require that much processing power or I/O bandwidth, at least by modern standards. Quick back of the envelope math--your Pismo, if it runs at 400MHz, has a budget of ~200 cycles per byte to keep up with 16 tracks at 44.1kHz. That's easily doable, even if GMail is a mess.
Anecdotally, I remember looking at CPU utilization when I was working with DAWs back in 2000 or so, and it was never very high. You certainly can fill it with expensive plugins or huge stacks of oscillators, but even on a G3 there's a lot you can do.
This piece seems incredibly negative. My first ever Mac was the dual 2.5GHz G5 and I loved it - sure, it was huge, but it was also powerful (for the time, and for a good while thereafter) and looked amazing. Yes, mine did make a weird chirping noise periodically and I think it did eventually become a bit unreliable, but that was after quite a few years. I have nothing but fond memories of that machine.
I agree. It is very easy to judge it harshly in retrospect. I find the comments about heat output especially strange. Perhaps they are forgetting some of the hotblooded pentium 4s of the day.
This was not a serious Anandtech review! It was funny and full of hyperbole. From our nice position 14 years later it is amazing how much power draw we had to tolerate back then.
To me it read like an affectionate piece, not a harsh critique.
> From our nice position 14 years later it is amazing how much power draw we had to tolerate back then.
I find it incredible how indifferent and uneducated I was back then, about how much power it uses and what the consequences are on a larger scale. I remember keeping it powerd on 24/7 (of course with periods in which it slept). Nowadays I am (hyper)aware of the power consumption of pretty much every device in the house.
I still have a G5 for some specific software, and it has amazing power management. Nothing at the time even came close. It would put itself to sleep without fuss, and wake instantly and seamlessly. It certainly used less power than my newer linux x64 box over the course of a day.
I dunno, we had one of these at home for a while, as my wife worked at Apple and it was in the demo pool and we got to borrow it for a bit. It was my primary workstation for a while, and I was a big fan of anything not-x86 and a cheerleader for PowerPC at the time and this was the machine that made me realize what a dead end that architecture had become. I was sad when Apple went x86 but not in the least bit surprised, given my experience with that workstation.
Like others have mentioned here, its heat output was insane. I had to have the house air conditioner on on cool days because of that one room. In the winter, it was the warmest room in the house. It didn't even feel all that powerful to me, as I compared it to a newly build x86 system I had put together at the time.
Luckily Apple read the writing on the wall and corrected course accordingly.
> When Apple abandoned the PowerPC they temporarily took a step back into a predominantly 32-bit world with the Core Duo, only fully embracing 64 bits a few years later, with the Core 2 Duo and OS X 10.7.
As far as hardware is concerned, Macs with 32 bit Intel processors only shipped from January (iMac) to November (MacBook) 2006. I did wonder at the time whether gaining a few months was worth saddling software with i386 slices for several years (system frameworks had to be built for 4 different architectures, for instance). It's true that full 64 bit software support took a bit longer to emerge, but that may have been due to the need to support the 32 bit Intel Macs that were shipped.
The main cause of the Apple III's thermal issues was Job's instance that computers should be quiet with no fan and no vents.
And even then, those requirements should have been possible to meet, if only the Apple III hadn't been rushed to market without enough time for engineers to diagnose and fix these issues.
I've got a first-gen Intel Mac Pro. My first real computer splurge, August 2006 right when they came out. Currently with 9GB of RAM and the twin 2.6ghz processors it came with. The computer and the Dell 2407WFP still work great to this day. I did, however, buy a new 5k iMac last March because the Mac Pro was simply getting too slow to edit the 20MP photos coming out of my pocket camera.
Loved my G5. A huge beast of a machine but a lovely design nonetheless. Would be great to revisit the beautiful case and build something more modern in it sometime.
I am still happily using a stock(ish) Xeon-based 'cheesegrater' tower at home, but if you wanted to just reuse an old (PPC or Intel) Mac Pro case to build a new system in, there are some ATX adapter kits available[1].
I remember very clearly going into the original Geek Squad offices on Washington Ave and seeing a bunch of these lining the desks of the agents. I was in complete awe. I asked the guy what they were (I knew they were Apple hardware, but had never seen them before). He just quipped, "New Apple G5's, they're insanely fast and never go down. We love em'."
I purchased a ”dual-dual” (”quad”) Mac Pro G5 with two 970MP processors, 8 GB of ECC RAM, PCI Express system bus, and two SATA HDDs in November 2005 (?). Contrary to what many will tell you this machine was not water-cooled: that was the previous generation of G5s, which had multiple single-core processor chips clocked at higher frequency. This machine is air-cooled and varies between sounding like a thundering highway tuck and a jetliner on takeoff.
With a pair of CinemaDisplay 24” (?) it served as my main workstation well into 2009, and was a truly awesome machine and I keep it in working condition with OS X 10.7 for the purposes of nostalgia. It was by far the most performant-compared-to-prevailing-average machine I have ever owned (back at a time when such comparisons mattered).
I still use the Mac Pro (similar specs as mentioned - no trashcan) as a workstation today. Nothing too heavy, but it works for the basics. Crazy that a desktop from over 7 years ago can still upgrade to the same RAM as today's best 'pro' MacBook (16gig)
I run one of these at home as a server (NAS, Minecraft) to this day. Runs FreeBSD 11.1 very nicely.
I was going to email the author about it but couldn't find her email address anywhere - plenty of account details on the usual walled gardens, but no actual email address.
I kind of wish that Apple would just make a "barebones" x86 case + mb... take a stock intel design, add in a "copyright" chip (to cover OSX license/boot) and let you add your own CPU/GPU etc (GPU from supported list, or onboard to start).
Of course paired with an apple designed ATX/mATX case. I'm not sure full ATX is really necessary, but at least mATX would be nice. Then leave it to the tinkerers to put together their own... Most that would want a more powerful desktop mac would be perfectly fine putting their own together. Then just do a motherboard refresh each year with whatever the current chipset is.
Don't even try to be everything to everyone, and leave it to "developer" macs. Just imho.
You just described the "xMac", the code-name that all the fans and nerds used for the computer they wished for during the 'aughts.
And I think there's some possibility that a computer like this is on the horizon. Apple wouldn't announce a feature like first-class external GPU support unless they had really big plans for it. Maybe the new Mac Pro they promised will be more like what you're envisioning...
I have fond memories of the first PPC gen 8100/80. They really did amazing perf feats with the PPC CPUs. Apple managed to get lots of "faster than Pentium" ads out of those CPUs.
Also m68k IIci was really durable and upgradeable computer. They also lasted a lot longer than they were really needed <3
Been thinking of buys 6100/60 with PC Compatibility Card so I could have a retro gaming station with two different architectures.
Apple had bunch DOS/PC compatibility cards for handling best of the both worlds. The recent ones run even Windows 98...
I used to own one -- it was a fantastically designed machine, but holy cow was it a good space heater. I had a south-facing second story bedroom and during the summer, using my PowerMac, man, it was toasty in there.
I purchased a G5 tower for cheap, on a whim, from my local university surplus store. There was a strange sign that said "may leak". I was really confused as to why they were talking about memory leaks W.R.T. hardware. Only once I got home and opened up the case did I realize that they meant leaking of the liquid cooling system... I have yet to repair the cooling and boot up the system. Any tips/guides for doing so?
I ended up using another second-hand Mac Mini G4 for PowerPC testing. Much less hassle but also much slower. LLVM builds take ages.
I still use my 15 year old Dual-2.5GHz PowerMac G5 every now-and-then, as Adobe Creative Suite still works perfectly fine on it. Best investment ever, especially with the 30" Cinema display.
I still use my 2005 quad G5 with Mathematica. I don’t really understand all this hand-wringing about supposedly bank-breaking power consumption. Electricity (here, at least) is so cheap it is largely irrelevant as a proportional variable cost. (Or at the very least, I have never thought of it in those terms.)
I bought my G5 tower at the end of the PPC generation, and loved the thing. I think it was 2005 or 2006, and I ran the machine until only a couple of years ago. During it's working life, it never skipped a beat, and for a long time was still just as performant as anything being released new.
Then came the stage where I could no longer update OSX, couldn't really run the internet as no new browser versions were released... The machine was pretty much a brick.
"Leopard runs an obsolete version of iTunes, version 10, which is faster and easier to use than the latest version."
I still use iTunes 10.6.3 on Mac OS 10.9.5. It can sync my iPhone 4S over USB, so my contacts & calendars & music don't have to be in the cloud.
It's so much better than the newer versions of dumbed-down iTunes. I'm seriously considering rewriting my own version of iTunes 10 as a web app before jumping platforms to Linux about 5 years from now.
However, I remember mine being ridiculously loud. Ive promised me that more fans would mean less noise due to the "thermal zones" described in the above video, but, at least to my knowledge, that didn't pan out. It also had a habit of just blasting the fans in the middle of the night like an airplane if I remember correctly.
Apple moving away from PPC to x86 effectively ended any remnants of PPC in the consumer market.
All that's left now is the server grade POWER, its really hard now for anyone to get access to PPC without getting an expensive POWER setup.
You can see that in the ecosystem as well, ppcel builds exist from Ubuntu, but I don't think their performance is as well tuned as the x86 ones. This is another takeaway, if you don't have hardware in the consumer market, your ecosystem will die out. Nobody will know you.
I don’t think either Linux or MacOS X was ever tuned well for PPC, even when consumer PPC hardware was available.
MacOS X feels nice once you get above version 10.2 and above 1.5 GHz, but it never felt as snappy as MacOS 9 on 300 MHz. I always thought it was suspicious how snappy MacOS X was on the first Intel Macs, considering that NeXTstep was on x86 but never on PPC.
Linux on PPC always felt sluggish. It felt sluggish back when Power Macs were contemporary.
The graphics implementation was always inferior, just a basic framebuffer, which could never compare to the hand tuned assembler speed of Aqua. That made everything feel slower, but Linux was as fast for everything else.
This (a Dual 2GHz PowerMac G5) was my first Mac. I remember buying with cash from summer jobs before heading to college in 2005 – I named it Jupiter.
The thing was a beast and I loved it. Two years later, I bought it a companion, the first Macbook Pro – named Mars. Later that summer, I bought the iPhone the day it came out for a cool $599 and named it Mercury.
These were all great products, and they inspired me to start the Mac Users Group at my university :)
> "The G5's considerable weight is focused on these little pads. There were aftermarket cork pads, but I've used masking tape to wrap a pair of old cycle gloves around the handle-stands, which doesn't change the fact that in 1998 The Undertaker threw Mankind off Hell In A Cell, causing him to plummet sixteen feet through an announcer’s table."
I was just thinking about how I bought a G5 Thanksgiving weekend back in 2004.
I bought the base configuration. It came with only 256MB of RAM, which honestly was a criminally low amount for that size of machine. My wimpy 12" Powerbook felt faster, because it wasn't thrashing all over the place like the G5 was. I think I lasted three days before upgrading the memory.
I had one of these wind tunnels for a while in the 2000s. A dual 1.8ghz model. It looked ultra-cool, beautifully designed inside and out. But it ran really loud even by the standards back then...
There was actually a high end water-cooled model that was notorious for developing leaks - doubt there's many of those remaining in service these days!
I never got round to get rid of my watercooled dual G5, as I was hoping way to long to find some neat use case for it. Mostly because it's probably the most performant non-Intel computer readily available (i.e. not some SGI Onyx cluster).
Anyone has a better use case to either inspire me or get mine donated for the cost of shipping from Gernay?
> The Pentium M didn't just outperform the Pentium 4M mobile chip, it also benchmarked within a few percent of the desktop Pentium 4, while consuming less power and generating less heat. After a brief diversion with the Pentium D Intel essentially gave up on the Pentium 4 in favour of a multi-core development of the Pentium M, which was sold as the Core architecture.
I have a personal story: the application I worked on at that time needed some complicated calculations and decisions during them -- and to complete every run took a significant time, measured in minutes, or even hours, and there was a lot to calculate. But once somebody ran the application on a Pentium M notebook, nobody could believe at first: on the notebook it run faster than on the Pentiums 4 desktops in the office, even if the notebook had lower clock than the Pentiums 4 desktops!
The reasons are many, if I remember what I've understood then correctly, the Pentium 4 design was never meant to target (at least not much) the clock speeds at which it was eventually sold. Apparently there were marketing people in Intel who influenced the development of the new processors, who managed to press the engineers to target always "more GHz" -- that was simple and easiest to sell, having more GHz than competition. So apparently the Pentium 4 design was made to conceptually enable up to 10 GHz clock speeds! Today it sounds science fiction, but before that time the clock speed did rise a lot for quite a years nicely. Also apparently, some engineers warned the marketing that maybe it's not going to be easy to reach 10 GHz. "You just design the CPU, let others design the thermal and other solutions." Of course, everything but the CPU conceptual design wasn't successful: it's not just cooling the CPU off that's the problem, although that is a problem enough. There are some specific dependencies that simply don't go linearly. So in practice Pentiums 4 have been sold with the lower clock that they were designed for, most of the customers didn't notice much.
So what were the Pentium M advantages? That was a processor developed by another team (not even in the USA but in Israel), and with the specific goal of doing more in less clock counts. It worked marvelously for the real life demands, like the calculations present in our application. Compared to the Pentium 4, the M also had more CPU cache, obviously enough for our application to shine.
For our application Pentium M was not "within a few percent" it was really much better than Pentium 4.
The P4 NetBurst architecture used very deep pipelines, which allowed it to be clocked very fast. To fill that pipeline the scheduler had to predict a lot of branches. So if your code had an unrolled loop then it performed well. However, if the scheduler mispredicted branches then the time to flush all those wrong branches and refill the pipeline caused severe inefficiencies. The 'speed demon' design also suffered from power dissipation problems, excess leakage current at high frequencies.
The Core architecture did away with the super deep pipelines, and it sounds like that avoided a particularly pathological case in your application.
The G5 (and the G4) were from another time. Back in the early 2000's (and late 90's) it wasn't uncommon two have two machines at your desk. One for internet and email, the other for Microsoft Office. With OS X you could combine them both into one machine.
I would have, if I had been able to afford to. ;-)
I have seen people use to machines for Internet and regular desktop usage for security reasons. Back in the day, you sometimes could not even finish installing Windows on an Internet connected machine without catching one of those nasty RPC worms.
Having a dedicated Internet machine you could easily wipe and install from scratch without losing your applications, settings and files was not a bad idea, if you could afford it.
(Alternatively, one could just use a decent operating system, but that is a different story...)
»The G5 has three USB 2 ports, two FireWire 400 ports, and a FireWire 800 port. In my life I have never used FireWire to transfer data. I will probably go to my grave having never used FireWire.«
IIRC, there was an interesting article posted here on HN about the history of FireWire. It sounded like an interesting piece of technology, but various circumstances prevented it ever becoming the big success it could have been.
I feel the same thing is repeating with Thunderbolt.
It's kind of funny/sad because neither technology is strictly tied to Apple machines - I had a work laptop (Dell Latitude D630) with a Firewire port. One of our CAD workstations has Thunderbolt on board (but, sadly no actual port).
But there were almost no peripherals to support it. The only person I ever saw using Firewire used it to transfer digital video from their camera to their PC.
It came down to licencing. Apple wanted $1 per port in 1999, which was too expensive. Intel would rather develop their own bus (USB) than licence Firewire from Apple. Apple eventually caved to $0.25 but it was too late.
Firewire was heavily used for digital video, still used for heaps of industrial cameras. Still very prominent in digital audio. The multimaster bus design, guaranteed timing and bandwidth (isochronous transport), and speed, made Firewire a winner. A Firewire 400 drive is practically twice as fast as USB2. So anyone who had to do lots of transfers used it.
The last non-intel and non-x86 on desktop in wide use. I miss days of heterogenous computing at home. One could argue mobile brought that back in, but meh. Not the same.
What do you mean with heterogenous computing? Isn't that supposed to mean CPUs combining different types of cores like the Cell with SPUs or AMD APU with GPU? G5 was just a PPC implementation without any "sidekicks".
The coffee can Mac Pro was a swing and a miss: an engineering achievement of sorts and an aesthetic bold move, but not loved by devs or scientists. I want an excellent case I can pack full of RAM, disks and GPUs, not external expansion with $70 cables that glitch and mess up the aesthetic in the process.
I don't know if Apple can make enough money on workstations to keep their attention, so maybe those days are gone.