Slashdot Mirror
G4 Bug Keeps Them at 500MHz
Hal-9001 writes "I saw this link over at Ars Technica; apparently, the G4 has a bug that keeps it from running at 500 MHz or above. The story is over at MacWeek. "
← Back to Stories (view on slashdot.org)
← Back to Stories (view on slashdot.org)
1. Like I said, most. Sure, you can get Linux, NetBSD, etc. on a G4. However, you still rarely see them, because the G4 desktops generally are Macs, and Macs run the MacOS. Out of all the shops selling G4s, do you really believe most of its customer's run Linux or another non-Apple OS? And again, this is a Mac site, so they're talking mostly to MacOS users. If I'm only a BSD user, do I care if MS Windows crashes?
2. No, but as I said, its a side effect. The OS is poor, something I noted at. The new BSD derived OS should be far better. I'm sure Mac users would be happy to have the front end they love, and a back end that's worth they're love.
3. Sure it seems odd, but if we take them at their word and that it doesn't effect chips under 500mhz, than its ok. It sounds like it does, but they don't want to admit it.. or that they are giving an excuse for while the G4 500mhz chips will be delayed, but that its not a horrid problem that should scare customers away.
"Open Source?" - Press any key to continue
AS opposed to Linux where a stray pointer in kernel space or a power failure can lead to serious disk corruption? THe disk repair utilities for Linux are so crude that I have more problems with losing ext2fs partitions than I ever have had with losing HFS partitions.
Please God, get IBM to come onboard. The people at Motorola couldn't get up clock speed to save their lives. IBM is talking 500-700MHz G3s. If they start making G4s using SOI and using their amazing talent, Apple won't get screwed.
Steve Jobs should be on his knees begging IBM to make G4s. Motorola has made it abundantly clear that they can't meet deadlines, can't produce enough chips to meet even Apple's small demand, and can't keep up with Intel(much less AMD) in terms of clock speed and processing power
C'mon IBM, you can do it!
Statements like this worry me. Restarts are NOT ok, and i'm sickened that people think they are. Especially when they're frequent like it sounds like this bug would make them. The world is not windows, if it were then everyone including me would have no problem with an occasional restart. But more importantly, G4's are usually going to be used as servers. Reboots are NEVER acceptable in a server, it can mean 20 minutes of downtime on a website which can mean lost revenue. It can mean lost e-mails within a company which is a VERY bad thing. Get the picture? I'm glad they're delaying the release, of course, as this is not a good bug. OK my rant is done
Patrick Barrett
Yebyen@adelphia.net
Restating the obvious since nineteen aught five.
At least we know Motorola is thoroughly testing it's processors before letting them on the market.
Intel hasn't been *that* kind (read: professional) with the Pentium class processors... Remember the first series (60, 66, 90)??? They tried, but ultimately had to replace a helluva lot of them...
-- It's always darker before it goes pitch black.
I'd rather have a processor a month or two later than have a crappy one that causes me headaches and likely a few months to replace.
I mean, people can live without the 500 MHz version for a couple of months.
It would be far worse if they just decided to quietly go ahead and release it and hope no one notices. (Or if they didn't even do enough testing to detect it in the first place.)
The G4 suffers from analog noise, causing
cache corruption. This happens, to some extent,
at lower frequencies too!
That's right. G4 processors from Motorola crash.
IBM adds capacitors and other junk on top of
the chip to deal with this, but even they don't
really know the cause.
Thus, the G4 is a bad buy.
I'm serious, and have fairly direct inside
knowledge of this -- please fix the score on
this post, since I obviously must be an AC.
There are so many things wrong with this post I just don't know where to start.
:)
I'm really sick of Apple apologists...
I'm really sick of Apple critics that don't understand what they're offering advice on.
A hokey press release claiming that this might not affect MacOS users is just plain stupid
Who put out a press release? This was all from MacWeek's "sources." Apple and Motorola had absolutely nothing to do with MacWeek's analysis that "this issue might never evince itself in Macs."
Not to mention that atop all of this, Apple isn't offering any highly-clocked, discounted G3, nope
Pretty off-topic, but I'll address it. You can get a G3/400 for $1299 with DV stuff of in the form of an iMac. I'm sure there are still G3 towers around from various retailers. Selling G3s and G4s at the same time would cause unnecessary product line confusion.
just more G4 marketing hype... "Yeah, the AltiVec velocity engine will help AppleWorks with all of that vector math you'll be doing".
Right. Most people buy a G4 to use AppleWorks.
The least Apple could do would be to investigate the possibility of selling real workstations, based on MacOS X and powered by a Power3, Alpha, or maybe K7.
I agree. Apple should spend 6-12 months porting Mac OS X to Alpha, Power3 or K7, and concince developers to do the same. In the process they can alienate and confuse their customers, piss off wallstreet by radically altering their business strategy without warning, and lose the ease of administration Macs enjoy from tight software/hardware intregration, and ultimately elminate Apple's value proposition.
By why stop there? Why not just outsource the hardware manufacturing to Dell and just focus on software? Maybe Apple could work on some applications to compete head-on with MS Office, or maybe write a web browser. It should also open source everything and just sell the service contracts.
One more thing -- Linus should be the CEO.
- Scott
------
Scott Stevenson
Scott Stevenson
Tree House Ideas
Why do they even bother telling us there is a bug if it hasn't been released yet? Anyone with half a brain would quietly fix the problem and go on.
The best way to accelerate a windows box is at 9.8 meters per second square.
I would like to state for the record that although I am a Mac user/supporter, I don't put myself behind this post. :)
:P
"No need for Virtual Memory?" Help!
- Scott
------
Scott Stevenson
Scott Stevenson
Tree House Ideas
I'm really sick of Apple apologists...
I am really sick of PC bigots jumping the gun without reading the article.
first of all, the processor is soldered down
First of all you are wrong. These things are on Zif sockets.
This is not better than Intel
Secondly this bug is affecting nobody because Motorola is delaying release. Unlike Intel which releases stuff then does field replacements.
Although I believe the motherboards are jumperless in the Sawtooth 450's to begin with
It doesn't matter. The G3 upgrades don't require changing the jumpers - upgrade manufacturers came up with their own way to reclock the CPU. I am sure that the same techniques would work with G4's.
Ummm.... the general rule on the Mac is that you leave virtual memory on unless you really need the .5% of extra performance taken up by the overhead. As long as you're not pushing applications beyond your actual, built-in RAM, you really don't notice the speed hit, even in Photoshop.
On PowerPC Macs, having virtual memory on forces applications to leave their resource forks on disk rather than loading them into RAM. Resource forks can get mighty huge. Photoshop 5.5, for example, requires about 7000 K more RAM with VM off than it does on and I think Excel 98's RAM usage ballons by 11MB with VM off. (Then again, this is from the same company that brought us Microsoft RAM Gobbler 4.5, the web browser which, when left running for 6 hours, will suck every last bit of RAM for itself.)
Of course, I always turn VM off when doing a render :-)
----
Am I the only one who thinks Microsoft is a misnomer? Perhaps Macrosoft would be a better fit?
I believe that Alpha based Crays (T3E) use the Alpha in big-endian mode.
BTW, PPC is also bi-endian.
Don't get too worried about the comments. MacWeek attributed the comment to "another source."
This kind of attribution means the writer didn't talk to someone worth naming. Had the comment been from Apple or Motorola, it would have been attributed that way. Evidently, the "information" came from:
1) someone at the MacWeek watercooler
2) someone on IRC or comp.sys.mac.geek.at.home
3) a bum on the street (there's lots in SF)
4) the writer's kid
MacWeek frequently attributes quotes to nobody in such a way as to make it sound authoritarian. (e.g. "according to sources close to Apple"--what the hell does that mean? Nothing!)
Don't put too much stock into journalism. It's just an exercise in creative writing to fill pages.
article. Have you never noticed the button labeled "Log Out" on the "Post Comment" screen?
Simply selecting the "post anonymously" checkbox won't allow you to moderate and post in the same article. However, logging out does allow such a thing. I've done it.
Learn a bit about how Slashdot works and you will learn these things!
I guess the fellow does have a point, it's painful seeing all these "..." subjects, usually with no real content attached, always moderated up to 5. It reaffirms the dream of a Slashdot killfile!!
This 11 fellow is already violating at least one of the rules that he sees every time he posts: "use a clear subject that describes what your message is about." Hardly a role-model for the community.
Maybe someday, arrogant bastards won't be moderated up for being arrogant bastards, moderation won't be based on popularity, and moderation WILL be based on whether or not someone has something productive to say. Maybe someday, people will learn humility, people will learn to serve others before they serve themselves, people will learn that character is one's greatest asset, that hurting others is wrong, and that there's no "i" in "Slashdot Comment System"
Yeah, right!
Back when computers first started coming around,
:> This makes me
lockups and crashes were adamantly avoided at all
costs. Even today it's a vital role and is the way OS's should be designed.
This blatant statement:
Indeed, another source said, this issue might never evince itself in Macs, since the OS doesn't manipulate data rapidly enough to cause the problem -- the glitch
would more likely effect more-efficient embedded operating systems. Even if data corruption should occur, a source said, the result would be nothing more than a
system freeze, easily fixed with a restart.
Makes me sick to hear as it's read by people
who don't know better (having used buggy OS's for years)
and take it as a norm, thus compounding the problem
of getting companies to write apps and OS's that DON'T
crash every time you turn around.
I LIKE setting up a server, firewall, workstation
what have you and having it run for a year without
a re-boot.
I find it ridiculous that a popular OS must be rebooted
every other time a new program is installed.
Fortunately things are starting to turn around.
Linux (for me) is becoming more and more viable as
an alternative every day. Why.. it's been a week
since I had to boot into lose98.
quite happy.
OSS clone of QuickBooks/Quicken, anyone?
Please??!!
Friends don't let friends buy Compaq's. (Dell/Gateway... same same) You want a good computer? Build it yourself.
while i can appreciate that motorola actually admits when it has a problem, i don't think it's too much to ask that a company use conventional words in describing the problem...this is not an "outstanding performance issue", nor is it an "errata" (who dug that word out of its grave?) it's a bug, or at least an error. slowing a processor down is not a "workaround"...it's a defective product...just admit it and move on.
Heh. PowerPC chips are bi-endian too, actually support 3 modes... big endian, little endian, and a pseudo mode in-between. Pseudo mode simply reverses the bit order, rather than byte order.
Yes, it's changable via a simple flag switch (meant to allow emulation, so task switching can easily put the processor in the right mode).
Heh. Talking about 50W processors... I used to remember when the Pentium MMX came out, several manufacturers put the *real* chip in, not the "mobile" version.
I think it's time to optimize the x86 so they don't draw so much power, on the processing-power/electrical-power ratio.
For those interested in benchmarks, here are some SPEC numbers from motorola for a 450Mhz G4:
21.4 SPECint95
20.4 SPECfp95
For those who care they claim typical power dissipation of 5W @ 400Mhz.
For comparison a Pentium III @ 500Mhz:
20.6 SPECint95
15.1 SPECfp95
So the integer performance looks comparable and floating point is better for G4.
One of the big features of the G4 that doesn't show in these benchmarks is the altivec simd/vector unit. There are no standard benchmarks I am aware of for extensions like these (or MMX/SSE/3DNOW). Apple has posted some results here that compare Pentium III/SSE @ 600Mhz to G4/altivec @ 500Mhz and show the G4 to be 3X faster on this subset:
http://www.apple.com/powermac/processor.html
Obviously YMMV but for some algorithms altivec is much faster than the Intel SIMD unit.
Megahertz is a poor measurement of speed. It's not even valid when comparing Celeron to Pentium III, much less between chip families from very differnt designs, like the PPC 7400 and x86.
In addition, mindless Mhz comparision fails to take into consideration things like different operating systems, application quality and the design of the surrounding PC hardware system.
The fact the x86 processors have higher clock speeds isn't useful in itself. Apple has frequently advertised various other measurements that the idiot public hasn't paid much attention to. Besides the BYTEmarks the PC-lovers love to hate, you can take a look at:
a) independant testing that shows Mac users are more effecient that Windows users because the OS is more consistant and easier to use.
b) Intel benchmarks that Apple ran against the new G4s. Mostly Photoshop filters, which the G4's Velocity Engine is best at, that whomp the PIII. (Look at apples' G4 web pages).
Because the idiot public would rather rely on megahertz, Apple now is pushing PPC development to higher clock speeds rather than just higher performance (note they are not the same goal).
Of course, the Wintel group can boast faster speed in certain areas due to flaws in the MacOS. Or features: Windows is often faster at file system "Finder" work, because it isn't managing a desktop database of metadata, doesn't handle resource forks, and doesn't calculate the size of directories. This speed trade off means you lose much of the Mac feel and features of the Finder. But it's marginally faster.
Apple can charge more because they're doing more. They sell a whole product, not just clone a basic design and outsource the software to Microsoft. In turn, you have a richer system that's less crapped together. It's suprising that Macs aren't more; this is another result of the idiot public buying only by price, not by quality. Imagine if the BMW M3 was within 10% of the price of the Ford Escort. What would you drive?
I'd like to see BeOS on both x86 and G4 for comparison. Comparing two OS on the two chips is already doable, and pointless as a hardware comparision.
Compare Mac Word98 to Win Word2000. Then compare QuickTime4 on the PC and Mac. That's interesting.
Then compare FinalCut Pro on the Mac to... the PC just sitting there overclocked to 1000Mhz creating heat. Task: edit video. The Mac wins!
Forgetting about the Macs for a while (I can't even fit the damn things in my rack mount!), this delay will likely also delay the availability of really fast inexpensive PowerPC boxes based on the spec that IBM released and several manufacturers are (or expect to be) working on.
From a software development and architecture view, I much prefer the Motorola product line to Intel's. Its so much cleaner and IBM/Motorola have really taken a lot of the best ideas and put them into a single core without having to carry the baggage of the 8088 along with them.
The availability of Linux for the new platforms will make platform-independence a critical decision point from a marketing perspective. To translate for you Anti-BillGates types, this means that Windows will likely start losing serious marketshare as they're really tied into the Intel architecture and other proprietary PC technologies that Linux can pretty much ignore. If you're a commercial software developer and you have to decide what OS to write for, Linux starts looking all the more attractive if you see customers buying up those PowerPC boxes.
Typical Slashdot reaction -- anything remotely bad happens to Apple and it's immediately Apple's fault. Steve Jobs could have a cold and, by gawd, Apple is going to go down the crapper.
Let's look at this at face value, okay?
I'm not saying that this isn't a bad thing -- just saying that everyone is making it out to be far worse than it actually is.
----
Am I the only one who thinks Microsoft is a misnomer? Perhaps Macrosoft would be a better fit?
I'm one of the mac idiots. While other ppl in my groupt are bitching about i/o problems, I get to bitch about problems with linuxppc. I think that I'm winning the battle as linuxppc is pretty good if you have prior experience with linux. However, just mentioning that using a PowerPC chip based system makes ppl that I have to work with, want to puke.
Of course, the enemy of my enemy is my friend, may play a role in all of this.
What suprises me is that they haven't had more problems!
--
While I probably shouldn't respond, I will anyways.
A LOT more goes into selecting a microprocessor than mere performance benchmarks:
In this case, Apple was migrating from M68k's. A big-endian processor. Both x86 and Alpha are little endian and while Alpha (at least the newer ones) can mode switch to bid-endian. I doubt that DEC's chipset didn't support it. More cost there.
Also, DEC doesn't have the fab capacity to meet Apple's demand - not even close.
Thirdly, Apple kindof wants to have a notebook presence. Putting a 50watt chip in a notebook would likely yield several McDonald's-like lawsuits.
Tom
Even if data corruption should occur, a source said, the result would be nothing more than a system freeze, easily fixed with a restart.
Windows users could get very used to this.
if(!toilet_paper) roll.replace(new roll);
- A.P.
--
"One World, one Web, one Program" - Microsoft promotional ad
"Remember when the U.S. had a drug problem, and then we declared a War On Drugs, and now you can't buy drugs anymore?"