Xserve G5 Updates?

[Ja Di ksw]

illumin8

Sorry for the confusion, but I wasn't comparing the sources, or reliability, or anything like that, of MacBidouille to MacWhispers. As I said, I'm not saying whether or not I believe this rumor is right or wrong, and I certainly never said ANYTHING about people being French or Iraqi or American or Russian or anything of that sort. All I said was, THEY (MacBidouille) said they were no longer going to do any rumors after WWDC, even if they were right. Then, after saying they were no longer going to post rumors, here they are doing it again. That is what I was comparing to MacWhispers, not anything else. Sorry for the confusion.

(Yes, I believe in being polite, and no, that's not an invitation to try to walk all over me or insult me, as others have done in here, though not in this thread. Just because I'm polite does not me I will back down from someone who can be 'louder'. Though, that being said, I'm not a stubborn fool who will keep arguing even when its shown I'm wrong)

edit: I guess this would be as good enough time as any to ask, how do you quote someone else with the bars and bold font? Thanks

 

[illumin8]

Re: illumin8

Originally posted by Ja Di ksw
All I said was, THEY (MacBidouille) said they were no longer going to do any rumors after WWDC, even if they were right.

I remember reading that comment as well, and I think there was something lost in the translation. The impression that I got from that comment that came out before WWDC was "if we are wrong we will withdraw from the Mac rumors community." I think they were very confident that they were right, and at the time were receiving a lot of flack for publishing benchmarks that seemed so much higher than even PC benchmarks.

I think time will prove that even though their benchmark numbers might have been a little off, the variances would seem normal considering they were run on pre-production hardware and a beta version of OS X 10.2.7.

 

[illumin8]

Re: Re: Re: wait a minute

Originally posted by arn
I don't think their benchmarks are yet confirmed.

You are correct, I guess nothing is really confirmed until the G5s have shipped and have been benchmarked by the major sites, but what I meant was their benchmarks seem to jive with the benchmarks that Apple has released at WWDC. Any variances could probably be explained by using beta versions of OS X 10.2.7 on pre-production hardware.

 

[fpnc]

Re: Re: wait a minute

About the MacBidouille G5 benchmarks:

Originally posted by illumin8
Their benchmarks were spot on...

I think that is unproven. In fact, I feel certain that those benchmarks were completely fabricated (i.e. fake). I'm not suggesting that MacBidouille tried to mislead us, just that they may have put too much faith in their source.

 

[Veldek]

Re: illumin8

Originally posted by Ja Di ksw
edit: I guess this would be as good enough time as any to ask, how do you quote someone else with the bars and bold font? Thanks

Look at the little "quote" button at the bottom right of each post.

 

[ddtlm]

Frobozz:

Well, Apple has said that within one year it will hit 3.0 Ghz. That puts it at next June. MacBidouille, and probably the public info on the 970 if I weren't too lazy to look it up, would peg the 970 at maxing out around 2.5 to 2.8 GHz. The 980 will be available in 2004. Assuming the 3.0 GHz by next June statement by Jobs is true, this will be a 980 chip.

I'm pretty torn between the two sides of the 970 vs 980 arguement:

View A: If we assume that the 980 is based on the Power5 it is not hard to believe that IBM will have designed it in paralell. IBM could very concievably be ready to ship 90nm PPC980's at the end of summer 2004.

View B: It would be very unusual (and expensive) for IBM to replace a processor design after only a year on the market because it would have a reduced money-making period to pay back R&D. While they could keep the 970 around in 90nm tech for lower-end Macs, I wonder if there is enough market to profitably support 980's and 970's at once. IBM might gain more from simply producing only one line of chips but with multiple L2 cache sizes to target different markets.

 

[ryan]

The 970/G5 isn't even shipping yet

Please somebody point to an article where IBM has even mentioned '980.'

I can't believe how people are getting into the same mind set they had when Motorola was rumored to be working on the G5, and look what happened, nothing! I'm sure we'll see faster/better processors from IBM down the line but any speculation as to when, speed, or even the model number is a waste of time.

 

[ddtlm]

ryan:

Unlike the Moto G5 rumorfest, Apple has publicly announced that the sucessor to the current IBM G5 is well on the way.

 

[AidenShaw]

two letters - 3U

Think bigger!

Most of these posts are about squeezing a 970 into a 1U Xserve chassis - what a limiting idea.

In the server world, 1U systems are the low end - what you get when you need to do many small problems.... You don't buy 1U for big problems. In the Intel world, Xeons and Itaniums do the heavy lifting, and the 1U systems are just now making the transition from Pentium III to Pentium 4 (Xeon).

I will bet that the G5 will show up in a much more "enterprise" system in a 3U or even 4U chassis. I'd expect it to have:

o support for 32GiB of RAM or more (maybe not initially, but with future denser DIMMs)

o 4 to 6 hot-swap PCI-X slots

o 6 or more internal disks, probably with embedded hardware RAID with battery backup

o redundant hot-swap power supplies, redundant hot-swap fans

o ECC memory, redundant (RAID) and hot swap DIMMs

o multiple GigE NICs on the mobo

o ILO (integrated lights out) capabilities

--------

Look at HP, IBM and Dell servers in the 3U/4U class - you'll see all this and more. Apple needs a system like this to be taken seriously.

Apple could keep the existing G4 1U Xserve as is, and add the 3U G5 as the high end server.

Bookmark this post - I'll bet that by a year from now this system will be in Apple's lineup. (1)

-as


(1) Disclaimer - I won't bet on ECC, Apple seems to be spectacularly stupid about not dealing with the issue of transient memory errors.

 

[ddtlm]

AidenShaw:

Thats a very narrow view of the rackmount computer market. 3U and 1U machines tend to be aimed at totally different customers. The 3U customer wants all that reliability stuff (for a server), the 1U customer often wants as much processing power per volume as possible (for a cluster).

the 1U systems are just now making the transition from Pentium III to Pentium 4 (Xeon)

Dual CPU Athlons and Xeons are not a new thing at all.

 

[daveL]

Re: two letters - 3U

Originally posted by AidenShaw
Think bigger!

Most of these posts are about squeezing a 970 into a 1U Xserve chassis - what a limiting idea.

In the server world, 1U systems are the low end - what you get when you need to do many small problems.... You don't buy 1U for big problems. In the Intel world, Xeons and Itaniums do the heavy lifting, and the 1U systems are just now making the transition from Pentium III to Pentium 4 (Xeon).

I will bet that the G5 will show up in a much more "enterprise" system in a 3U or even 4U chassis. I'd expect it to have:

o support for 32GiB of RAM or more (maybe not initially, but with future denser DIMMs)

o 4 to 6 hot-swap PCI-X slots

o 6 or more internal disks, probably with embedded hardware RAID with battery backup

o redundant hot-swap power supplies, redundant hot-swap fans

o ECC memory, redundant (RAID) and hot swap DIMMs

o multiple GigE NICs on the mobo

o ILO (integrated lights out) capabilities

--------

Look at HP, IBM and Dell servers in the 3U/4U class - you'll see all this and more. Apple needs a system like this to be taken seriously.

Apple could keep the existing G4 1U Xserve as is, and add the 3U G5 as the high end server.

Bookmark this post - I'll bet that by a year from now this system will be in Apple's lineup. (1)

-as


(1) Disclaimer - I won't bet on ECC, Apple seems to be spectacularly stupid about not dealing with the issue of transient memory errors.

Well ... I have to disagree. This post also addresses another post about Xserver and Sun. Apple is NOT attempting to compete with the mainline *nix server vendors. Why not just take a handful of asparins and slit your wrists? If you're in touch with the industry at all, you know that none of the enterprise *nix vendors are doing very well, not because of MS or itanium (which sucks so much power it's unreal), but because IT capital spending has dried up. Now you want me to believe the Apple is actively seeking out this market? I don't think so.

A typical IT server isn't concerned about FP or vector units. I think Apple's server strategy, using the G5 (NOT the Power4), is directed specificly at rendering and serving media, period. This is their sweet spot, this is what they know, and this is what, as a vertical market, has some margin in it. In short, Apple does not and will not compete with Sun/HP/IBM IT server platforms. They're too smart to do that.

To all those that think you can't build a 1U chasis with 1 -2 G5's - go do your homework. The G5 is a lot lower power than most CPUs out there *for the amount of real work they do*, particularly in the vertical space mentioned above (media). IBM plans to use this processor in blade configs, which have a *much* higher heat density then a 1-2 CPU 1U racked server.

BTW, Itanium doesn't do *any* "heavy lifting", yet. It consumes large amounts of power, cannot be packaged in dense enclosures, and has a *brand new*TM instruction set with very few applications.
What appications do exist are do, in large part, because HP and Intel have hung their hat on this CPU - and they are hurting badly because of the limited market acceptance, to date.

Hum, 1U is limiting ( I guess I'm taking the issues in reverse order, from the post). Interesting, given that, eventually, everything out there will be 1U. Look at what you can do in a laptop platform now, compared to 5 years ago. Remember the 8" floppy (don't take it personally). We're working toward higher densities - get it?

 

[rjwill246]

Re: Re: two letters - 3U

Originally posted by daveL
Well ... I have to disagree.

Thank you! Aiden has been very slow this week coming out to bash Apple since he messd up so badly about the G5s! Remember, Aiden said that there was no indication that Apple would opt for the 970s...or that IBM was working with them. And he had that from a highly respected sales rep of his who works for IBM. So much for inside info! BTW, Aiden, I wouldn't buy too much from this rep if I were you... he just might have a bridge..... well, you know.
So, Aiden, give it a bloody bone. For all the bashing, Apple is a stunning company that makes products in the Testarossa class, and yes, they do mess up once in a while, but those G5s will be shown (again) to be perfectly capable of equalling the furnace-like P4s...
And in the end, it is the OSX experience that really matters. It IS the crown jewel in a treasure trove of riches.
MS and Intel give us zircons... and even then, they are flawed.

 

[eric67]

IBM Power5 to come next early next year

Well for info here is the most recent news concerning the IBM Power5, directly from IBM :
http://www.theregister.co.uk/content/61/31467.html

of course this is clear that we will not see PPC980 or PPC970rev in the PowerMac at the sasme date (early 2004), one can easily imagine that PPC980 or PPC970rev call it as you want, let say PPC970 follow ip, will probably be developped let say in parallele to the Power5. I remember clearly have read on the web, but I can not find the link, that it is actually the fact that the follow up of PPC970 is indeed already being in development... which is in a way normal, knowing the time you need to do so... so I think it give us at least 1 year to speculate.

 

[eric67]

Re: two letters - 3U

Originally posted by AidenShaw
Think bigger!

Most of these posts are about squeezing a 970 into a 1U Xserve chassis - what a limiting idea.

In the server world, 1U systems are the low end - what you get when you need to do many small problems.... You don't buy 1U for big problems. In the Intel world, Xeons and Itaniums do the heavy lifting, and the 1U systems are just now making the transition from Pentium III to Pentium 4 (Xeon).

it could be named Xstation Enterprise server as I has already been speculated , so anyway the next place for the G5 is definetely the Xserve, way before the PowerBook.

 

[Rustus Maximus]

Originally posted by ddtlm
I'm pretty torn between the two sides of the 970 vs 980 arguement:

View A: If we assume that the 980 is based on the Power5 it is not hard to believe that IBM will have designed it in paralell. IBM could very concievably be ready to ship 90nm PPC980's at the end of summer 2004.

View B: It would be very unusual (and expensive) for IBM to replace a processor design after only a year on the market because it would have a reduced money-making period to pay back R&D. While they could keep the 970 around in 90nm tech for lower-end Macs, I wonder if there is enough market to profitably support 980's and 970's at once. IBM might gain more from simply producing only one line of chips but with multiple L2 cache sizes to target different markets.

It's surprising to me how many people I see doubting how quickly IBM and Apple could bring the 980 to market given how WAY ahead of schedule they were with the 970. I think ddtlm is right in that it isn't hard to believe they've been working on the 980 in paralell with the Power5, I would say it is a sure thing they have been and we will see 980 Pro Macs far sooner than even they are hinting at. Maybe it's just that we Mac users are so used to Motorola-induced dissapointment with our processor evolution. However I think we are all going to have to 'Think Different' (pardon the pun) about how quickly Apple ramps up it's processor family, now that we have IBM on our side. IBM is in this game to win, folks. This ain't your daddy's Mac.

I don't think it would be a bad thing to move the 970 to a consumer line of Macs and put the 980 in Pro Macs. By offering a consumer 64bit machine, people will still get the benefits of 64bit computing without having to pay pro prices.

But putting aside all of the 980/G6/G1,000,000 speculation...the G5 is here and is an awesome machine. It is the sign many in the creative industry have been waiting for, coupled with the Quirk release, to move to OSX AND new hardware. I think we will see a rather huge ramp up in Apple sales as a flood of loyal Mac users who have been in a holding pattern make the transition. It only spells great things for Apple which in turn spells great things for IBM which in turn will continue to develop better and faster chip technology for the Mac.

 

[AidenShaw]

Re: Re: Re: two letters - 3U

Originally posted by rjwill246
Aiden has been very slow this week coming out to bash Apple since he messd up so badly about the G5s!

Actually, Aiden's been on a business trip...

And, by the way, how is it that suggesting that a 3U server would be good is considered to be "Apple-bashing"?

Remember, Aiden said that there was no indication that Apple would opt for the 970s...or that IBM was working with them. And he had that from a highly respected sales rep of his who works for IBM.

Ummm - please show me any official announcement (before the keynote) that Apple would be using the IBM chip. Anything. Lots of rumours, and hints, and speculation - but no announcements. I even said that it was probable that Apple would use the chip - but there were no announcements.

My IBM rep said nothing about Apple, I didn't even ask. I certainly didn't say anything here about my IBM salesman saying anything about Apple.

What I *did* say was that my IBM rep was telling me that it would be late cyQ3 or to be safe cyQ4 before the 970-powered IBM blades would be available.

And guess what, Apple is saying that they'll start shipping the G5 in September - that's late Q3 on my calendar. Any delays, and it'll be even closer to the Q4 that my IBM salesman said was the safe bet.

In other words, I think that my statements were closer to the mark than the rumours that hundreds of thousands of PPC970 chips had already been shipped to Asian factories....


And about that 3U system....

If you look at any big server farm, most of the time you'll find lots of 1U servers being fed by a larger system. This bigger system would be a file server or network controller for the rest.

The system doesn't need more power (although quads and octos often appear in this space) as much as it needs more slots (more Fibre Channel and Gigabit Ethernet cards) and memory than you find in a 1U.

One is also a bit more concerned with reliability of the bigger server, since when it goes down everything is down. That's why these systems usually have redundant power standard, have at least ECC or some kind of redundant memory (RAID for RAM), and hot-swap PCI cards, hot-swap RAM, hot-swap fans, hot-swap power supplies and the like.

I'm not saying replace the 1U with a 3U, and I'm not saying that G5 in a 1U can't or shouldn't be done.

The problem is that if someone needs a big file server for an Xserve cluster, they are forced to get an x86 Linux box or other UNIX system. When that happens, it leads to the question "why not all Linux, rather than support 2 operating systems?".

If you want Apple to do better in the server space, you don't want people asking that question.

 

[ddtlm]

Rustus Maximus:

It's surprising to me how many people I see doubting how quickly IBM and Apple could bring the 980 to market given how WAY ahead of schedule they were with the 970.

Heh, they aren't way ahead of schedule with the G5. Doubters such as me have been predicting timeframes around September for some time, even when all the "immediate shipping" hysteria was sweeping the land. Unless Apple ships significantly earlier than they are currently claiming they will, the G5 shipping date will not be a victory for the optimists. (A long time ago I would have predicted a shipping date 3-4 months later, but few people have argued that recently.)

 

[Rustus Maximus]

Originally posted by ddtlm
Heh, they aren't way ahead of schedule with the G5. Doubters such as me have been predicting timeframes around September for some time, even when all the "immediate shipping" hysteria was sweeping the land. Unless Apple ships significantly earlier than they are currently claiming they will, the G5 shipping date will not be a victory for the optimists. (A long time ago I would have predicted a shipping date 3-4 months later, but few people have argued that recently.)

Sorry, my comment wasn't aimed at you, it was aimed more at the crowd I was hearing who were predicting January or later 2004 as the G5 release. Maybe I am overly optimistic but an August release date seems okay to me given what the G5 seems capable of. And I know we won't know it's true performance level until some production units are put through their 'real world' tests. But given the present state of the Mac processor family...the G5 is a tremendous leap, and if they hold true to Steve's 3Ghz in 12 months or less soundbite, then it's going to be a bright future for Apple. Sooo, yeah...I AM optimistic 😉

 

[ddtlm]

Rustus Maximus:

Heh, optimists, I'll crush you! 😀 But seriously, if Apple increases clock speed (or performance) by 50% every year they may still be behind the curve. If AMD and Intel go at it again like the "good ol days", 50% in 12 months won't look so good.

 

[ddtlm]

Just for curiousity's sake, I went to look how fast AMD scaled clock speeds when they introduced the Athlon in August 1999. In 7 months they went from 600mhz to 1000mhz... 66% jump in 7 months.

 

[Rustus Maximus]

Right. But aren't they and Intel nearing the end of what they can do with the x86 technology? They will have to move to Itanium, etc. and they aren't as far along in the 64-bit game as IBM or Apple, are they?

Plus they have their own marketing to contend with. They have spent the last few years telling everyone that MHz or GHz speed is what matters...however that scale is all changed with the move to 64bit. Sure they may have a 5GHz Pentium 4 next year...but it will be 32bit. As the developers for Apple gear up to take advantage of their 'ready-now' 64bit processor, the G5 will be a screamer, with 64bit, OSX Panther software and applications and the internal architecture to take full advantage of the latest technology (s-ata, huge memory addressing capabilities, etc.).

Ever the optimist 😀

Rustus

 

[ddtlm]

Rustus Maximus:

But aren't they and Intel nearing the end of what they can do with the x86 technology?

Yeah, just like Apple is always going to die "soon".

They have spent the last few years telling everyone that MHz or GHz speed is what matters...however that scale is all changed with the move to 64bit.

The 32/64 argument is pretty much hype and marketing for most people at this point, and I don't forsee it making all that big of a difference (in sales) just yet. Still, Intel would be all over 64-bit on the desktop if they weren't trying to avoid harming their Itanium. 64 bits are good, just maybe not important compared to other things Intel may be able to offer.

 

[frozenstar]

Originally posted by ddtlm
Still, Intel would be all over 64-bit on the desktop if they weren't trying to avoid harming their Itanium. 64 bits are good, just maybe not important compared to other things Intel may be able to offer.

Intel doesn't have a 64-bit chip that can run 32-bit code natively. That alone precludes them from entering the 64-bit desktop market.

 

[ddtlm]

frozenstar:

Intel could have and still can make a chip like that at any time, no sweat, in fact there was quite an internal debate about wether to make x86 64 bit or to do the Itanium. That debate didn't happen yesterday either, it was years ago. The only reason an Intel chip wasn't the first 64-bit "desktop" processor was because it contradicts their master (Itanium) plan.

 

[Rustus Maximus]

Originally posted by ddtl
The 32/64 argument is pretty much hype and marketing for most people at this point, and I don't forsee it making all that big of a difference (in sales) just yet. Still, Intel would be all over 64-bit on the desktop if they weren't trying to avoid harming their Itanium. 64 bits are good, just maybe not important compared to other things Intel may be able to offer.

I didn't say Intel was going to die (brief imaginary trip into a brighter, happier world...) I said they are reaching the limits of the x86 architecture and will need to move on if they want to compete in the 64bit market. The developers will react to demand combined with the ability of the technology...if 64bit is feasible (and it is now) then they will move that way because we will want to move that way.

Originally posted by ddtlm
Intel could have and still can make a chip like that at any time, no sweat, in fact there was quite an internal debate about wether to make x86 64 bit or to do the Itanium. That debate didn't happen yesterday either, it was years ago. The only reason an Intel chip wasn't the first 64-bit "desktop" processor was because it contradicts their master (Itanium) plan.

Their master plan...

('Does anyone else hear the Imperial March from Star Wars playing?')

The reason they haven't and won't go to 64bits yet is eactly what frozenstar said, The Microsoft Megasaurus that is holding them back at 32bits. That coupled with all of the Orwellian aspects of the Itanium and Longhorn (Paladium) will spell trouble for the Wintel beast.