More Apple/970 Confirmation

[drastik]

Re: the G spot

Originally posted by Nepenthe
I'm not one to stop Apple from abandoning the G-series nomenclature...
...for good
IMHO it does "put everything into perspective", though. A G4 performs better than a G3. A G5 performs better than a G4, and so on.
But that system is just OMINOUSLY too simple....
*twitch*
-eyes glance about the room nervously-

it's all a conspiracy.

Maybe it's that I'm an audio pro. I am accustomed to products having about three letters, four numbers, and a couple more letters after the numbers. "The new 'Sony XYZ3450MKII 96'"
What would the world be like if digital mixers were called the "Yamaha B2"

The conspiracy of the technology world is obvious. They're out to brainwash us...to program us...
We MUST stand against this evil machine!
For the sake of humanity and all our bodily fluids!

Ok, so sorry, it's late and I just finished yet ANOTHER Dr.Strangelove marathon.

The O2-R is pretty close

 

[jethroted]

This is great news for all of us!

 

[StuPid QPid]

Re: the G spot

Originally posted by Nepenthe
I'm not one to stop Apple from abandoning the G-series nomenclature...
...for good
IMHO it does "put everything into perspective", though. A G4 performs better than a G3. A G5 performs better than a G4, and so on.
But that system is just OMINOUSLY too simple....
*twitch*
-eyes glance about the room nervously-

What about calling the new 970 chip the G64? That would nicely emphasize that it's 64bit, and it is (sort of) an advance on the name G4...

Although I don't really care what is called, as long as it's fast

edit: Added apostrophes to it's...darn, I always make that mistake

 

[StuPid QPid]

Re: Re: the G spot

Originally posted by StuPid QPid
What about calling the new 970 chip the G64? That would nicely emphasize that its 64bit, and it is (sort of) an advance on the name G4...

Although I don't really care what is called, as long as its fast

Actually thinking about it, G64 is uncomfortably similar to the old Commodore C64 I remember from a looooong while back in the 80s...
Mind you, it was a pretty good home computer for its time...

 

[JtheLemur]

Call it the PowerSmack G5. Done and done.

That is, provided the performance is at least on par with Intel and AMD...

 

[jettredmont]

Re: Re: No AltiVec for Linux?!

Originally posted by Awimoway
I don't think so. The 970 is coming sooner, and its initial speeds won't seem that impressive compared to x86.

As for the Super Bowl ad, it means one of two things: They're just planning ahead for a new ad campaign, but no earth-shattering new product.

Or they'll blow our socks off with something far more interesting than we often give them credit for when theorizing in forums like this. A really impressive DLD or--and I favor this--a revamp of the Orwellian ad, this time with Gates getting his image shattered, and the release of a new suite of office software that is fully compatible with MS Office formats.

Images being shattered ... I could think of three or four Ballmer soundbites that would go well intermixed with a still photo of Gates sneering that would be downright fascist ...

"dancing monkey" and "Developers! Developers! Developers!" top the list ...

 

[Vlade]

Why dont we just skip G5 through G16 and call it the G17?

 

[jettredmont]

Originally posted by synp
It's been done before. Remember Intel's 386sx? Remember the 486sx? The latter was a full-fledged 486 with the math co-processor disabled.

It is also remotely possible that disabling the altivec unit makes it possible to speed up the processor, which is good for other programs.

While it would be interesting if IBM shipped Altivec-disabled 970s (the 486SX was because there were many chips with defects only in the FPU ... the 386SX was just an odd beast designed to fit between the 286 and 386DX ...) I don't see that in the quote above. It sounds more like IBM doesn't want its initial Blade customers to be expecting AltiVec support in either the OS (Linux may well eventually support AltiVec, but that doesn't mean that the version of the kernel/packages that IBM ships with its blades will!) or in gcc (see above).

IBM is a conservative company. Directive #2 is that whenever a promise is made to a customer that promise must be kept regardless of cost. This typically leads to an extreme reticence to making promises of any kind unless their fullfillment is assured from the start. If IBM doesn't yet know what Linux seed they're going to ship their blades with (and I'm not even sure if AltiVec support is in the latest Linux builds ... anyone else know?), they're going to err on the side of lowered expectations and over-delivering rather than hype and under-delivery.

As I've said before, dealing with IBM is a bit different than dealing with most other tech industry companies.

But to the hardware issue ... doubtless there will be some 970s with defects only in the AltiVec regions ... the issue is if there will be enough of them to justify selling an Altivec-disabled chip, and if there is an easy way of reliably disabling the AltiVec code without making the chip unstable (Intel designs such switches into their processors, although I believe they "got lucky" with the 486 production ...) On the whole, though, I'd be surprised if either of these conditions were met. I don't expect 970"SX" chips to be marketable.

 

[Nepenthe]

yet more "confirmation" of the obvious

Originally posted by pyrotoaster
There's nothing as reassuring as confirmation of the obvious!

well here's some more
(If you call this "confirmation")
An article on Mac Daily news talking about apple's focus on narrowing the clock-speed-gap that it currently is victim to over Intel and AMD

"Sources inside Apple have told LoopRumors that increasing the clock speeds of its computers has become a major priority. The gap between the current G4 chips and its competitors such as Intel and AMD has strongly affected sales. Our sources have revealed that sales of Apple's pro line have been lagging for nearly two years as a result. The good news is that changes are 'well under way.' Our sources also hinted but wouldn't confirm that the current PowerMac form factor will be the last, as the new processor will reportedly bring a new design to the Pro line. When asked when we might see these changes, we were simply told: 'Before the Fall,'" reports LoopRumors.com."
http://www.macdailynews.com/comments.php?id=P765_0_1_0


So, yeah, there you go...

 

[Nepenthe]

Re: Re: the G spot

Originally posted by drastik
The O2-R is pretty close

yes, true....
However, now we have the "02R 96".... after that it's the
"O2R-24 96/192HD"
The DA-7 was realatively simple, but now it's the DA-7 MKII
of course, the mother of all would be the O2R2D2.... a-heh...

*crickets chirping*

 

[josepht]

Re: yet more "confirmation" of the obvious

Originally posted by Nepenthe
well here's some more
(If you call this "confirmation")
An article on Mac Daily news talking about apple's focus on narrowing the clock-speed-gap that it currently is victim to over Intel and AMD

"Sources inside Apple have told LoopRumors that increasing the clock speeds of its computers has become a major priority. The gap between the current G4 chips and its competitors such as Intel and AMD has strongly affected sales. Our sources have revealed that sales of Apple's pro line have been lagging for nearly two years as a result. The good news is that changes are 'well under way.' Our sources also hinted but wouldn't confirm that the current PowerMac form factor will be the last, as the new processor will reportedly bring a new design to the Pro line. When asked when we might see these changes, we were simply told: 'Before the Fall,'" reports LoopRumors.com."
http://www.macdailynews.com/comments.php?id=P765_0_1_0


So, yeah, there you go...

Great! That means revision #2 should be around January or February of 2004. That's when I plan to get mine.

 

[jethroted]

Originally posted by mac15
Cool, will this be called the g5?

I doubt it since moto is not making the chip, and they own the G4's and G3's.

 

[Macpoops]

Motorola does not own the G3 G4 naming convention. That is simply an Apple think. To moto it is the MPC 74xx and to ibm is 75x. Apple has their own name for just about everything. 802.11b= Airport g=Airport Extreme. Altivec= Velocity engine. IEEE1394= Firewire. Etc....

 

[ddtlm]

I don't know about how accurate the rest of this RUMOR is, but that part about AltiVec support not being in Linux is stupid. It took like 30 seconds or so to find this using Google:

http://maccentral.macworld.com/news/0203/11.redhat.php

Coming from Maccentral of all places, how hard can it be for someone else to find?

Anyway, I think this rumor is a fabrication and I find it surprising that almost everyone else swallowed it whole and speaks like it is fact.

 

[Sol]

That article mentions Motorola chips, not IBM. I think the rumour is credible. Remember that the 970 is still not out there yet so anything goes.

 

[ddtlm]

Sol:

I demonstrated Linux AltiVec support. The reason they call the vector units on both the PPC-970 and the 7455 "AltiVec" is because they support the same instructions. To support one AltiVec is to support both.

 

[jettredmont]

Originally posted by ddtlm
Sol:

I demonstrated Linux AltiVec support. The reason they call the vector units on both the PPC-970 and the 7455 "AltiVec" is because they support the same instructions. To support one AltiVec is to support both.

Umm, no.

You demonstrated that one year ago, Red Hat was committed to getting Altivec support into its Linux distribution.

A quick search for "linux altivec" on Google didn't turn up anything more recent except for a note from a developer working on Linux-PPC64's Altivec support late last year which indicated Altivec support was being developed but not quite "there" yet.

As I said before, IBM is not likely to be shipping the cutting-edge developer build of Linux on their blades. If Altivec support isn't time-proven yet, it likely will not make it into the IBM Linux-PPC distribution when it ships.

Do you have any more recent link indicating that Altivec support is in place and working?

 

[ddtlm]

jettredmont:

OK, if you insist on beliveing this silly rumor, then perhaps this link, also the first link on my first google search, will please you:

http://www.simdtech.org/apps/group_public/email/altivec/msg04822.html

To quote:

Has anyone used vector operations in (YDL-2.1) Linux kernel modules? I have user-space code that I need to port to kernel modules to run in real time under RTLinux.

So as you can see, one of the guys has a user-space i.e. normal program in Linux using AltiVec and he wants to make it run in kenel space (which is not as easy as user space). Conclusion: AltiVec programs run in Linux.

 

[macrumors12345]

Originally posted by ddtlm

So as you can see, one of the guys has a user-space i.e. normal program in Linux using AltiVec and he wants to make it run in kenel space (which is not as easy as user space). Conclusion: AltiVec programs run in Linux.

Wow, that is a VERY selective quote of that link that you provided. Maybe you should have included this part (the reply to the message you quoted) as well:

There are a number of challenges in doing AltiVec vector operations in
the Linux kernel. Among them are:

1) Saving/restoring registers - the kernel reserves these registers for
the user. Unless you KNOW that user-mode applications aren't using
AltiVec functions, you will have to manually save & restore the AltiVec
registers around your code. This consumes significant memory bandwidth.

2) The GCC compiler normally used for building the kernel isn't AltiVec
aware - you'll have to create your own build process.

Using AltiVec in the kernel is possible, but will require work.

Steve

That reply does not exactly make it sound like Altivec support in Linux is ready for regular use at the moment. I am not saying that I am sure IBM will not support it in their Linux distro at the release of the PPC 970, but certainly there does not appear to be any evidence out there that they WILL support it (while this rumor provides some evidence, albeit unsubstantiated, that they won't).

 

[dongmin]

Re: yet more "confirmation" of the obvious

Originally posted by Nepenthe
"Sources inside Apple have told LoopRumors that increasing the clock speeds of its computers has become a major priority. The gap between the current G4 chips and its competitors such as Intel and AMD has strongly affected sales. Our sources have revealed that sales of Apple's pro line have been lagging for nearly two years as a result. The good news is that changes are 'well under way.' Our sources also hinted but wouldn't confirm that the current PowerMac form factor will be the last, as the new processor will reportedly bring a new design to the Pro line. When asked when we might see these changes, we were simply told: 'Before the Fall,'" reports LoopRumors.com."
http://www.macdailynews.com/comments.php?id=P765_0_1_0

LoopRumors has a pretty shoddy record, if I recall. But it's still nice to hear of these rumors trickling in.

My fav part is the bit about a new case. I´m sorely waiting for one. The current design has only seen superficial changes since the Smurf cases. I´d like to see the two drive doors better integrated with the case. Also, I´d like to see audio, usb, and firewire jacks out on the front or sides.

 

[ddtlm]

macrumors12345:

There is a big difference between kernel-space and user-space. What more is there to say?

Anyway, here's another link from Google: http://www.synergymicro.com/Software/Linux.html

Quote from that link:

As with uniprocessor Linux, Linux SMP supports the Synergy's Math Library (SSSLª) for DSP and array processing, with AltiVec support.

And here's the whole Google search: http://www.google.com/search?hl=en&ie=UTF-8&oe=UTF-8&q=linux+kernel+altivec&btnG=Google+Search

 

[macrumors12345]

Originally posted by ddtlm

There is a big difference between kernel-space and user-space. What more is there to say?

According to the report the IBM rep said that "they have no plans to implement it in their Linux servers, or AIX versions of the blade servers." There are many ways to interpret that, ranging from they don't plan to implement it in the kernel to they won't let anyone program to it at all. Given that nobody here actually witnessed the conversation, none of us really know exactly what the rep meant when he said that they won't implement it. I agree that it would be unlikely that IBM would actively prevent anyone from ever using the Altivec unit, but that is a very strong interpretation of what he was saying, especially since we have no evidence to back it up.

Anyway, here's another link from Google: http://www.synergymicro.com/Software/Linux.html

Yes, I agree that it is possible to use Altivec within some Linux distributions. But that is hardly the same thing as saying it must be the case that IBM is going to implement it (at some unknown level) in the Linux (and AIX?) distributions that ship with the 970 blades!

 

[ddtlm]

macrumors12345:

It is inconcievable to me that IBM would in fact claim that they do not intend to support the AtiVec unit in Linux userspace (at least) when in fact they would have to go out of there way to not support it, for no apparent benefit. On certain tasks AltiVec is unmatched and I'm sure that IBM knows that, and I'm sure that many prospective customers know that.

In my opinion this whole AltiVec section of the rumor is a massive red flag. Like I said, I think this rumor is a fabrication, and I think the AlitVec chit-chat was thrown in to provide some meat and some excitement (purpose-built feature for Apple, etc).

 

[scem0]

Originally posted by arn
I imagine people don't realize that this sort of community exists... following apple's every move.

He may have been surprised simply because there hasn't been an announcement... how would a random joe know about this small part of his job?

arn

I can't imagine a life without MR .

This is definitely good news. I hope they change the whole
naming scheme. No more G'x's. Maybe they will call it the pMac
or something.....

 

[Sol]

G

I respectfully disagree. The G is a good way to put generations of CPUs into perspective. Apple has so far resisted re-branding any of Motorola's chips a G5, even with new technologies like Silicon On Insulator, more AltiVec units, etc. G5 is an appropriate name for a 970 or even an x86 that will take over from Motorola's CPUs.