What is Digg?76 Comments
- organon, on 10/12/2007, -4/+114A processor from 2007 is faster than a processor from 2002? What are the odds ...
- plamoni, on 10/12/2007, -7/+107In other news... my cellphone has more computing power than my parents' 8-year-old desktop computer....
- SlvrEagle23, on 10/12/2007, -2/+81@ruz322: Sometimes I also like to throw Microsoft into conversations just to hate on them...
Friend: So, I got this new headunit with a DVD player and all that..it's awesome!
Me: Yeah, I guess so, but put Windows on it and it would be *****.
Friend: ...what? - UGM2099, on 10/12/2007, -0/+40Apple's G5's came out in 2003 and the XServe in '04, but you have a point.
A better comparison would be the Dual 2Ghz G5 I bought in 2005 and the 1.6 Ghz Intel Mini I bought one year later in 2006. The Mini was much faster in almost every aspect, even running Adobe apps in emulation. It cost 1/3 as much. - fkr3, on 10/12/2007, -5/+43This is hardly surprising. Mac Mini's are really just desktop computers in small cases. The hardware's newer, faster and more powerful than the old G5. It should be expected to perform better.
In other news, my celeron laptop is faster than the first server I managed - it had 300 users and was a dual p3 600! - Recode, on 10/12/2007, -0/+25Having a server is not just about performance. Reliability and longevity is just as important.
- cyberdork, on 10/12/2007, -1/+21And when are they finally going to release Core2Duo Minis???
I'm already waiting since last autumn. - UGM2099, on 10/12/2007, -5/+22Intel Minis start at $599 and are worth every penny, IMO.
- Philodox, on 10/12/2007, -0/+12Anybody remember this was on the Apple home page?
http://images.amazon.com/images/G/01/computer/detail-page/powermac_g5_performance_graph.jpg - cliffzdude, on 10/12/2007, -2/+14Very lame. If a HDD fails in one of our servers (HP DL series, hot swap HDDs) their led goes orange. Our non-tech admin assistant calls one of us who is technical, and with our approval yanks the bad drive and inserts a new drive. In a while the mirror is rebuilt via hardware raid. Bam, issue solved. No muss, no fuss... Try THAT with a MAC mini.
Same with power supply. Bad PSU? Ok, our DL ___ servers have dual PSUs. Yank bad psu, insert good psu. Done. Yes hot swap, yes other psu takes load, no need to power down server. Try that with a MAC mini.
Bottom line these are cool little consumer pcs, using them as servers is a cute trick but nothing more... - adidos, on 10/12/2007, -0/+11Direct link to original story: http://www.networkjack.info/blog/2007/03/29/intel-macminis-the-os-x-blade-server/
- tobsterius, on 10/12/2007, -1/+12Welcome, my friend, to something we call "Marketing"
- stmiller, on 10/12/2007, -1/+11There are several types of IBM cpus known by apple customers as 'G5'. While this mini intel chip is as fast/faster than the dual G5s, that is only one factor. Consider the notebook hard drive, ram limitations, etc. etc. of a mini compared to the powermac G5 or Xserve machines. 16GB ram, multiple (desktop) hard drives, fiber cards, ...
So I guess it all depends on your needs. - elvenseven, on 10/12/2007, -1/+10Captain Obvious, Emergency Call on Line 1!
- peterjhill, on 10/12/2007, -1/+10calling lame..
apples to oranges comparison. I hope no one is suggesting that a mac mini would make a better server than the xserve.. try swapping out a fan or power supply or disk in that mini. There is a company that specializes in hosting mini servers, but when one goes beyond price for comparison, a real server is designed to live in a datacenter, and a mini is not. - Kniggit, on 10/12/2007, -0/+7@Recode:
You're 100% right. Redundant hot-swappable hard drives and power supplies, rack mount hardware and standardized heights in multiples of 1.75", proper cooling, hardware management tools and network connectivity options are all reasons why nobody in their right mind would ever use one of these in a data center. - msgyrd, on 10/12/2007, -2/+8Yeah, you can buy the something equivilant to the Mini's specs for $300, but it won't fit it the Mini's form factor or look as good. Shuttle has a computer of about the same size and specs and starts at $999, although looks worse. The Mini is competitively priced for everything it offers.
Specs aren't everything for a computer. If you're using it as a media center in your living room, looks and size matter also. - cbreaker, on 10/12/2007, -0/+6It's true. Make sure when someone sells you "Mac Hosting" you're not paying for space on Mac Mini's =) I can see the benefit for a compute cluster or something, but for a web server? I'll take a properly cooled server unit with RAID and less performance any day.
- vagarach, on 10/12/2007, -1/+6Not the newer celerons, which are almost the same as their c2d counterparts, just without speed stepping and other nice things.
- UGM2099, on 10/12/2007, -4/+8True, crappy 3D graphics, and the hassle of opening up the Mini to add RAM, but it's still a fast computer and it fits equally well in your carry on luggage and next to your HDTV.
- jadenton, on 10/12/2007, -1/+5Not buying it. Maybe on integer code (which admittedly is the majority of code), but certainly not on floating point.
When I left my old employer, I left behind a dual core G5. It had a whopping 4GB of ram and a pair of whatever the highest clocked CPU was. It blazed. I used it to run research code (a pseudo-variant of ransac, expect much much faster), which was very very compute bound. Load up, suck in its modest amount of data, then churn through the data using both CPUs to do lots of floating point calculations. No altivec, because I needed more precision (double, rather than float). On one particularly "large" data set I used for speed tests, the G5 took about 28.5 seconds of compute time to finish the job (a little more than half that in wall clock time since it ran on 2 cpus).
When I went to my new job I picked up a top of the line Mac Mini. 1.83, Duo Core with a gig of ram as a home computer. For kicks, I used it to run the old code. Once all the CPU flags where set appropriately this same software came in at about 33.5 seconds of compute time. That's about a 17.5% speed difference between the two in favor of the G5. Note that the Intel mini was able to make use of SSE3, where as the G5 was NOT using altivec.
Again, the nature of this code and data is very compute bound. It can fit all required data into memory on a modern system. It makes no trips to the kernel for memory or I/O between start up and final output or during the time it is running on both processors. Relative I/O or disk speeds should not be a significant factor in the performance of this program. Given the relatively anemic bus and disks speeds for the mini, I would expect that more typical applications would do better on the G5. The Intel's general advantage on integer code MIGHT give it a boost in the server arena under low to moderate loads, but I would expect that the G5 would come out on top as the load and bandwidth requirements increased.
Of course, a new Mac Pro vs. the Old G5, no contest. - TroubleInMind, on 10/12/2007, -2/+5And that little box comes preloaded with most of the stuff you need to make a pretty sweet media center. If it only had native HDMI output...
- Aero1, on 10/12/2007, -4/+7in other news:
a pentium 3 desktop outperforms a pentium server - skinfitz, on 10/12/2007, -1/+4The Mini might number crunch better than the G4 XServe at some tasks but I know what box I'd rather trust my data to.
- ilgaz, on 10/12/2007, -1/+3Install Yellow Dog Linux to that XServe and see results. You will be surprised.
- bonexaw, on 10/12/2007, -0/+2"no matter how hard I tried I could not make that Mini use more than 0.37A of power."
Um... "A" = amps is a measure of current. Not power. amps times volts = power (in watts... like a lightbulb). - cbreaker, on 10/12/2007, -2/+4Yea, Rosetta stuff doesn't run faster on Intel. ANY current Intel CPU. It just doesn't. It's enough to run things allright, and it works, but the apps are faster on a PPC.
- UGM2099, on 10/12/2007, -1/+3@cacoe & cbreaker
Running Photoshop CS2 for the last 10 months on an Intel Mini has been a much snappier experience than the previous year I was working on the dual G5. Of course speed tests of Gaussian Blurs, etc, will favor the G5 but day-to-day I found Photoshop to be more pleasurable on an Intel Mac. I stand by my unscientific opinion. :) - dbr_onix, on 10/12/2007, -1/+3"I can see the benefit for a compute cluster or something, but for a web server? I'll take a properly cooled server unit with RAID and less performance any day."
That raises an interesting question : Would a single G5 (properly cooled and with RAID and such) really be more reliable than a bunch of Mac Mini's in a high-availability cluster? With a single machine, no matter how much more servery it is, a single piece of hardware dying (memory, CPU) will take down the entire server, but the chances of say 5 Mac Mini's all dying at the same time is near-zero. THe only thing that might be a problem is power-failures, but even then, a bunch of Mac Mini's could easily be distrusted to other data-centers (compared to sawing in half a XServe)
Perhaps a pile of white boxes might look less professonal than a nice rack-mounted server, but from a purely-stablilty/price view - for web-servers, it seems like the Mac Mini's would be more stable for less money (For the price of a £2200 XServe, you can get nearly 6 Mac Mini's).
(For the sake of the argument - I'm ignoring the fact you can't rack-mount the Mac Mini's, that they don't come with the OS X Server edition, and the fact that clustering doesn't always make sense) - drlha, on 10/12/2007, -0/+1@orangecrush: OS X on Intel can run 64 bit apps just as well as OS X on PPC, I personally have compiled some of my Unix tools for x86_64 platform on my MacBook Pro (Core 2 Duo). However the Mac mini only has a 32 bit Core Duo chip, so obviously won't run anything built for 64 bit.
- flickr, on 10/12/2007, -4/+5I'm glad it is so obvious to many of us, but I've got a (stupid?) friend saving up for a refurbished Dual-G5 which he swears is faster still than the Mac Pro (for games no less!)... oh well.
- inactive, on 10/12/2007, -0/+1Considering that you can get a 1U server for around 1300 bucks that will perform just as well as the mini they quoted in the article, I will take the better management features, higher quality drives, expansion slots, multiple gigabit ethernet, hot swap features, etc. of a real server over a mini any day. The power consumption of a good opteron or newer Xeon server will not be much more while the system will have been built for 24/7 use, and rack based cooling unlike a consumer product. Neat story, but this is something that you do in your basement and not a datacenter.
- cbreaker, on 10/12/2007, -0/+1If you have a data center with server farms, power simply isn't your biggest expense and not really a huge concern. Why bother saving a few amps on the computers when the AC unit cooling the room dwarfs that consumption?
- gettarat, on 06/11/2008, -0/+1awesome info
http://aphonesite.com
http://icases.biz - inactive, on 10/12/2007, -0/+1It's completely valid to compare apples to oranges. This is a dumb way to try to dismiss a comparison.
Both apples and oranges are medium-sized, spherical fruit of similar weight. But OMG, how could you compare them? Never the twain shall meet! Oh the retardedness! Woe is me! - jadenton, on 10/12/2007, -0/+1@orangecrush
Both where using whatever the most up to date version was at the time, running their respective versions of the Desktop OS (rather than the server). Although I can't be certain they should have had similar background processes running I expect they are similar. The number being reported here is CPU seconds, which is tracked separately from wall clock time and should not vary (much) under load, a fact which I observed to be true on various occasions (running iTunes or Handbrake in the background did not affect the CPU time reported, although it did affect wall clock time).
Now, the 64 bit nature of the G5 vs. the 32 bit nature of the mini is certainly a factor in the performance difference. It may even account for all of the G5 speed boost and them some. But a stock mini doesn't come with Core 2 yet, and so I think you have to count the 64 bit architecture in the G5s favor.
But again, I would expect a Core 2 processor to beat the G5, just not a current mini. - yoshitx, on 10/12/2007, -0/+1
http://www.macminicolo.net/cage.html - cbreaker, on 10/12/2007, -0/+1While the Intel Core CPU is better with Floating Point then previous Intel chips, they've been lacking in this area for some time. AMD chips have had excellent floating point performance since they introduced the Athlon- and apparently Intel is going to be bringing some better FPU's into the picture to compete with AMD in this area.
The PowerPC chip isn't a slouch, and it's not very surprising to me that you had a specific computation run much faster on a completely decked out dual G5 system with tons of RAM over a lower clocked Mac Mini on Intel.
I'm just not sure what you "don't buy" here. You yourself said that maybe on Integer operations the Core2 is faster then the older G5's - do you think that serving web pages uses floating point ops? - lilrabbit129, on 10/12/2007, -0/+1This can be explained by the difference in ram. Your G5 had 4Gb of ram, your mini had 1/4 of that. If your mini had to swap, that could account for the extra time (specially with a slower laptop HD in your mini).
- ilgaz, on 10/12/2007, -0/+1Unless you run a company doing work for Pixar using custom grids of Adobe apps, Xserve benchmark using PHOTOSHOP is completely insane.
Servers have their own PHP etc benchmarks, servers aren't optimised/intended for end user client apps! - msgyrd, on 10/12/2007, -0/+1I'd wager that the 32 to 64 bit capabilities of the OS shouldn't affect anything. If the code was written to exploit any gains in 64bit architecture, it probably can't be easily recompiled for 32 bit. Background processes also rarely use much CPU time, mostly they just fill up space in ram. If it doesn't have to swap out to the hard disk, his application shouldn't be affected by it.
- OrangeCrush, on 10/12/2007, -2/+2@jadenton
Which versions of OS X were you using on each machine? I was under the impression that PPC-OSX Tiger could run 32 or 64 bit code, but x86-OSX only ran 32 bit code, which may account for some of the performance difference. Also, did you have similar software configurations in the two machines? I ask simply because a home computer may be running additional background processes by default that would've been trimmed out of a workstation/server deployment which could also be affecting performance. - inactive, on 10/12/2007, -1/+1also consider the fact that the mini is running a 2.5" hard drive which top out at 200GB and aren't designed for running 24/7 with high demand like a server will face... The Xserve is built with 3 full size drives, and was able to handle 1.5TB of space with 3x500GB.. Plus although it didn't ever ship with 750GB drives, they be able to be used with no problem...
- inactive, on 04/05/2009, -0/+0Yo Music http://www.yomuzik.net Yo Music
- Derfus, on 10/12/2007, -2/+2that's awesome! I want a mini now
- JackAxe, on 10/12/2007, -1/+1@ugm2099
Get off your emulation high horse and come back down to reality. - cdcdark, on 10/12/2007, -1/+1Will a computer from 2007 be faster than a computer from 2002? Find out at 11:00
[http://www.youtube.com/watch?v=9j7o8ixbrLo] - lolo2007, on 10/10/2007, -0/+0For those of you who think computers should be priced like a commodity, without regard for the quality of the product—please, please go out and buy the cheapest computer you can find. IT'S EXACTLY WHAT YOU DESERVE! http://www.gwafi.com/news.html http://www.gwafi.com/story.html
- andyduncan, on 10/12/2007, -0/+0well, sure, but assuming the volts stayed constant, the amps are still relevant. it's not "power" but in this case it is a nearly perfect indicator of power consumption.
- unother, on 10/12/2007, -1/+1Eh?
I think the intent would be to swap out the entire mini in the case of a failure. The redundancy is not built into the unit, but is a product of the distributed design of having a "mini farm".
Keep in mind that they are suggesting usage as webservers, also. As the article makes clear, a modern XServe (or ilk) would be best utilized for a database server or the like.
Also, the article makes clear that one of the major reasons to use these is power consumption: clear and simple. As any server farm can eat up a lot of power (and throw off a lot of heat which shortens the lives of components) the low overall consumption that a mini-rack provides is a key selling point.
Remember, the writers are not suggesting that mini-farms are the "end-all-be-all" of servers, ready to take the place of all other hardware. They simply point out that a lot of the design factors that the mini has make it ideal for _certain_tasks_ in a high-utilization server situation. -
Show 51 - 73 of 73 discussions
Check out the new & improved 
© Digg Inc. 2009
— Content created and posted by Digg users is 