To add, again: One of my favorite linux shellboxes was a PIIIm laptop with a slow CPU and a low RAM ceiling. A G4 Mac mini might have a chance at outperforming that, and 1GB is more than sufficient for a text-only ssh console for things like irssi and alpine. It would likely be better suited to the task than an ultraportable laptop, too.
-
Hello MLAers! We've re-enabled auto-approval for accounts. If you are still waiting on account approval, please check this thread for more information.
Ideas for Use of Mac mini G4?
- Thread starter HFTaylor12
- Start date
I have the P4 Gateway laptop I got new for college in 2006 running XP and Firefox 52. Just in terms of web page rendering speed, my DLSD with Leopard and Leopard Webkit blows it away, although admittedly the Gateway is more usable on the web.
The benchmark I used was Cinebench. The ThinkPad T30 was mine, but the 1.67 PowerBook belonged to somebody else. It would be a fairer comparison for web browsing if you had a clean install, a fresh reboot, and contemporary versions of the same browser on both platforms. I forget which CPU the T30 had, but I'm sure I have it written somewhere. My guess is that it was a 1.8 or a 2.0GHz chip.
Do you have the detailed configuration information of the P4 based system and, like, JS rendering scores on something like Octane or one of the other js benchmarking tools?
The other thing is that Firefox has been, from time to time, the slowest of the modern browsers.
Do you have the detailed configuration information of the P4 based system and, like, JS rendering scores on something like Octane or one of the other js benchmarking tools?
The other thing is that Firefox has been, from time to time, the slowest of the modern browsers.
You're also comparing modern Webkit with an OS that's six years newer, so not a valid comparison.
I'm going to set up some side by side comparisons when I have a chance.
One of the things I have in mind is to run lens correction filters on some of my medium large format film scans(which are ~20mb and ~80mb in size). Most of the MF scans are of originals taken with Planar-type lenses which introduce some noticeable "mustache" distortion. The LF scans are simpler since all my LF lenses are Tessar types and tend to only have simple pincushion distortion. I
In any case, distortion corrections are among the most computationally intensive things you can do in Photoshop, so I think this is a decent stress test especially with the larger files.
Also, I'll mention that XP SP3 is roughly the same age as Leopard, and it's often thought that running an older OS with less computational overhead will give that computer an advantage. The P4s that were still shipping in ~2008 often still shipped with XP SP3.
One of the things I have in mind is to run lens correction filters on some of my medium large format film scans(which are ~20mb and ~80mb in size). Most of the MF scans are of originals taken with Planar-type lenses which introduce some noticeable "mustache" distortion. The LF scans are simpler since all my LF lenses are Tessar types and tend to only have simple pincushion distortion. I
In any case, distortion corrections are among the most computationally intensive things you can do in Photoshop, so I think this is a decent stress test especially with the larger files.
Also, I'll mention that XP SP3 is roughly the same age as Leopard, and it's often thought that running an older OS with less computational overhead will give that computer an advantage. The P4s that were still shipping in ~2008 often still shipped with XP SP3.
Cory makes the case for the various mobile Pentium 4 variants which, quite honestly, I can't say a single nice thing about, but the case for the G4 basically disintegrates when the Pentium M enters the picture. Here's the data sheet for the MPC7447A used in the last Powerbooks; at 1.42Ghz it sucks a maximum of 30 watts, 21 watts typical, whatever "Typical" means. 1.42gz is the maximum frequency in the datasheet, assuming it scales fairly linearly add maybe three or four watts to that for the 1.67mhz version. Intel set the design TDP for all Dothan Pentium Ms at 27 watts except for the 800mhz ULV version, which was set to 10.8 (By comparison there, the "typical" consumption of the 7447 at 1ghz is 16 watts, 23 watts max.)
Granted this *isn't* an apples-to-oranges comparison, because Intel bases TDP on some strenuous code mix but states it's not to be taken as the theoretical maximum the CPU might use, but even if we assume that Intel's "TDP" is directly comparable to Motorola's "Typical" the few watts theoretical win that gives the PowerPC is more than offset by the greater performance of the Intel CPU. That 27 watt TDP goes up to a 2.26ghz version, and according to GeekBench even the 1.6ghz Dothan handily outperforms the 1.67ghz PowerBook.
(Even the lowly 1.3ghz Banias Pentium M from 2003 scores around 870, which puts it right up with the very best Powerbooks and a good chunk faster than a Mini G4, which didn't come out until 2005. Banias actually drew less power than Dothan as well; Dothan doubled the size of the onboard cache and made a few other changes that significantly increased performance at the cost of increased power consumption. But still not enough to make it "worse" than where PowerPC was stuck at.)
I'm not entirely sure why Intel didn't embrace the Pentium M as a desktop solution once it became obvious just how much better it was than, well, basically everything in the consumer space. I think in part it was harder to produce/lower yield than the Pentium 4 at first because of the relatively huge amount of onboard cache it sported, and, well, there's also the fact that Intel's marketing had gone all in on the whole more-megahertz-is-better thing, it probably was wise of them to keep their counterexample quiet until the were ready to jump ship en-masse. (As great as the Pentium M was from an efficiency standpoint it didn't *quite* scale up high enough to match the very fastest, over-the-top desktop Netburst chips.)
Last edited by a moderator:
Core 2 basically was the desktop Pentium M you wanted - a Pentium M with an altered inner design, including adding all of NetBurst's go-faster features, which when added to a P6, makes it ludicrously fast.
Both Yonah (core 1) and Core 2 are, like Pentium M, in essence a heavily tweaked P6 core with a Pentium 4 bus controller, but yes, Merom/Conroe were the first of the family that could definitively crush the entirety of the Pentium 4 lineup. Pentium M's clock speed limit in 2-ish Ghz ballpark put beating the most ridiculous 3.4Ghz-plus single-core Netburst CPUs off the table, and in 2005 Intel rolled out the Pentium D which, likewise, ran at clock speeds high enough that Yonah (which pretty much was just two Dothan Pentium M's welded together) just couldn't *quite* beat it.
(Also, Yonah and Pentium M didn't support 64 bit, which really wasn't a big deal for almost anybody in 2006, but it was technically something NetBurst could do they couldn't.)
Nehalem (IE, the first CPUs sold with the "iNumber" branding) is technically the rev of the core architecture that really arguably incorporated all the best NetBurst go-fast tricks; it's where Hyperthreading came back, and its pipeline depth creeps back up into NetBurst territory. It's really sort of amazing just what a roll Intel was on from 2004-ish to around 2012. Not that they're exactly picking their noses today, but it does seem like the rate of improvement has stagnated somewhat.
(Also, Yonah and Pentium M didn't support 64 bit, which really wasn't a big deal for almost anybody in 2006, but it was technically something NetBurst could do they couldn't.)
Nehalem (IE, the first CPUs sold with the "iNumber" branding) is technically the rev of the core architecture that really arguably incorporated all the best NetBurst go-fast tricks; it's where Hyperthreading came back, and its pipeline depth creeps back up into NetBurst territory. It's really sort of amazing just what a roll Intel was on from 2004-ish to around 2012. Not that they're exactly picking their noses today, but it does seem like the rate of improvement has stagnated somewhat.
I suspect they held off on P6 being used everywhere again until Prescott was done, as that was more or less a total redesign of Netburst. Once it was obvious that Netburst couldn't be saved, they changed gears.
Intel's slowing pace of innovation, struggles with die shrinks (we're getting at least five revisions on 14nm) and recent "moar cores" philosophy makes me wonder if they haven't decided it's time for a clean sheet design.
There hasn't been a real IPC boost since Sandy Bridge in 2011 (insert i5-2500k joke here), and while 2013's Haswell brought amazing power savings (enabling machines like the 12+ hour battery life MBA13) newer chips have started to lose those power savings as they now ramp clocks to show any performance boost.
Intel's slowing pace of innovation, struggles with die shrinks (we're getting at least five revisions on 14nm) and recent "moar cores" philosophy makes me wonder if they haven't decided it's time for a clean sheet design.
There hasn't been a real IPC boost since Sandy Bridge in 2011 (insert i5-2500k joke here), and while 2013's Haswell brought amazing power savings (enabling machines like the 12+ hour battery life MBA13) newer chips have started to lose those power savings as they now ramp clocks to show any performance boost.
Last edited by a moderator:
I've been saying for a while that CPU performance has basically leveled off since about 2012 or so. As said, most of the improvements have been in performance per watt, and with the "Lake" series processors Intel has seemed to even stepped away from the "Tick Tock" design philosophy of earlier generations. That's a lot of the reason why I'm still using a 2012 MBP with a quad i7 Ivy Bridge processor, and at that the only real-world advantage I see over Sandy Bridge is that it has USB 3.0 native.
Granted Ryzen seems to really be shaking things up, although I have yet to do a lot of research on how it's doing in the real world. Perhaps, though, it will kick Intell into stepping things up.
I agree that Intel is seemingly "stuck" at 14nm, but truth be told I don't know how much smaller we can realistically get before wires don't actually behave as wires but rather you start seeing quantum mechanical effects. I spent a while discussing this with a colleague the other day. A copper atom is roughly 1.4Å, which means that a 14nm we're already laying down wires that are roughly 10 copper atoms wide. It may well be that 14nm is a quantum mechanical "brick wall" so to speak, but if it's not I suspect it's certainly close.
All of that aside, what HAS improved exponentially in the past 5 years is GPU performance, and I think that's where the makers are really concentrating their energy/resources now. GPUs excel at vector operations, and a lot of vector-heavy software now is smart enough to run vector operations on the GPU.
Granted Ryzen seems to really be shaking things up, although I have yet to do a lot of research on how it's doing in the real world. Perhaps, though, it will kick Intell into stepping things up.
I agree that Intel is seemingly "stuck" at 14nm, but truth be told I don't know how much smaller we can realistically get before wires don't actually behave as wires but rather you start seeing quantum mechanical effects. I spent a while discussing this with a colleague the other day. A copper atom is roughly 1.4Å, which means that a 14nm we're already laying down wires that are roughly 10 copper atoms wide. It may well be that 14nm is a quantum mechanical "brick wall" so to speak, but if it's not I suspect it's certainly close.
All of that aside, what HAS improved exponentially in the past 5 years is GPU performance, and I think that's where the makers are really concentrating their energy/resources now. GPUs excel at vector operations, and a lot of vector-heavy software now is smart enough to run vector operations on the GPU.
For CPU based benchmarks I am sure most P4 based T30's would be faster then G4 based Powerbooks. For web browsing don't forget the T30 has an anemic 16MB Radeon 7500 while the G4 1.67 models have either a 64 or 128MB Radeon 9700. Huge different in graphics speed and VRAM.
What Ryzen did for AMD is get their individual cores running almost as well as Intel designs and give you more of them at a lower price point. With current process you can't go much faster or much smaller so CPU speeds are going to hit a brick wall until somebody comes out with something radically different in materials. Being power efficient is great for laptops, but it is the only other real way of making better CPUs when you can't do much about core speeds. Same thing going on with GPUs needing less power then the they did 10 years ago.
As far as what to do with a G4 mini, no idea outside of a headless server. I would rather have a tower for better GPU choices and faster HDs for doing any OSX "work".
As far as what to do with a G4 mini, no idea outside of a headless server. I would rather have a tower for better GPU choices and faster HDs for doing any OSX "work".
Similar threads
