satty said:To quote Rick in Casablanca:
We'll always have Paris
Apple to Intel
I think this is the beginning of a beautiful friendship.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Xserve Refresh?
- Thread starter MacRumors
- Start date
- Sort by reaction score
Apple to Intel
I think this is the beginning of a beautiful friendship.
Platform said:Well not really a problem because they can just put a billion fans running at 10.000rpm to cool it but still.........if they consume less power = less heat = possible to cram more chips inside the unit
don't forget less power = less cost = a major concern of, well, everyone.
Platform said:Not sure is Itanic another chip or Ithanium...???
If Ithanium.......I just read so many places that it seems to be lacking support for programs and stuff as well as Intel not so keen on it
Itanic was the industry tongue-in-cheek name for the Itanium chipset (think Titanic). Itanium originally tanked pretty badly due to lack of fast 32 bit compatibility, and due to Intel's overpromising on the technology. But while Itanium originally got a bad name, and Intel had to move 64 bit-ness into the Pentium 4 line, they have still not given up on it, and their big-iron servers are still based on newer versions of the technology.
I'm guessing Xserve will stay Power-based until 2007, when Intel will have new viable Itanium chips that can reliably outperform it.
Platform said:Not sure is Itanic another chip or Ithanium...???
If Ithanium.......I just read so many places that it seems to be lacking support for programs and stuff as well as Intel not so keen on it
Itanic is the pejorative nickname for the Itanium, which is not an x86 chip. Its native architecture is EPIC, and without software emulation it's support for x86 instructions performs very poorly. It's doubtful that apple would announce a switch to x86 if they were really moving to an EPIC chip. More likely is Xeon's in servers and perhaps high-end workstations, P4's in lower-end workstations if there is no Pentium M derivative with SSE3 support at the time, Pentium Ms or their successors in the mid/low-end desktops and in laptops.
-Lee
stingerman said:
Of course they intended to get the G5 across the line. The problem was power/heat issues something IBM never fixed until AFTER Jobs's WWDC announcement and lo and behold new low power chips are announced by IBM.
IBM threw a temper tantrum and played some maketing games to make it look like Apple was holding things back. Does anyone really think Apple would have held back a G5 PowerBook while they watched PB sales slide into a backhole? Not a chance in heck.As for the XServe what needed to be said has been said on the thread...hmmm..while in bed....by Ted...eating bread...*shakes his head* Hehe sorry.
Personally I'm guessing the XServe will go x86 at WWDC 2007 along with the PowerMac. The chips for the PowerMac and XServe aren't here yet and IT folk don't like being guinea pigs for new hardware or OS's. Once Apple proves itself in 2006 with a x86 Tiger and an x86 PowerBook and maybe iMac they will be less hesitant on the trans over to x86.
Platform said:Not sure is Itanic another chip or Ithanium...???
If Ithanium.......I just read so many places that it seems to be lacking support for programs and stuff as well as Intel not so keen on it
Itanium + Titanic = Itanic = Doomed chip = Won't be in a Mac = End of Post
Now how about those Intel Celery chips?
SiliconAddict said:Itanium + Titanic = Itanic = Doomed chip = Won't be in a Mac = End of Post
Now how about those Intel Celery chips?
Thanks people for making it clear.......and Itanium +Titanic makes sense....never heard anything reallt good about that chip
Toe said:But, more to the point, a machine with four 2.5 GHz processors is almost always going to be faster than one with two 2.3 GHz processors, no?
Given the same chips with the same chipset and same bus speed. I'd be worried if the 2.5GHz quad would be ever slower than the 2.3GHz dual. It would seem strange if it was, given how each core is faster, although there could be memory speed issues.
Hector said:
"IF" it happens then the Xserve will go to 2.5 to 2.7 GHZ PPC970MP
There is NOT a cooling issue with the 970MP
IBM will be offering 2 dual core 970MP's @2.5GHZ in their new JS40 blade servers starting in Sept.
JS40 blade servers are twice as dense as XServes. i.e 14(2 processor-dual cores)for a total of 56 cores in 7U.
The question remains whether Apple uses them or not.
dernhelm said:And just when it looked like they were abandoning my old buddy here. Seems like Xserve gets very little press from either the Mac press or even Apple itself. It would be great if they got a dual-core chip in these babies, but if the date is Aug 23, my guess is that you are only looking at a slightly faster version of the same ol' same ol'.
BTW - I don't think heat is that big a problem in the Xserves. Have you heard these things? I think they are 40% fan 60% computer parts.
Besides IBM can place a full Power5 in a 1U enclosure. It can't be that hard.
Nope IBM can't place a power 5 in a 1U.
A power 5 from IBM comes as a 2U and larger
backspinner said:
IMO the reasons stated by Jobs at WWDC where BS.
IMO it was sour grapes more than anything else.
IBM wasn't able to deliver the magic 3GHZ chip Jobs had publically said he would have.
He left Freescale/Motorola for almost the same reasons 2 years earlier.
IMO when IBM committed to deliver 10 Million 3GHZ Power-based chips to Microsoft for their XBox360 by OCT it was more than Job's ego could take so he dumped them.
2x 2.5GHz processors could equal a 5GHz Processor if:TrumanApple said:
A: The Bus Speed is the same as the processor
B: The processors don't need to communicate with each other.
C: The instructions don't rely on the result of a previous instruction
D: The OS is 64 bit and has pre-emptive simultaneous multi-tasking (meaning the operating system already knows how to divide it up and can send 4 instructions a cycle.)
Platform said:Why would it not qualify as a Mac......
It's neither called a PowerMAC, nor an iMAC, nor an eMAC nor a MACmini...(ok, the 'books don't have the 'mac' in their name either).
Moreover, searching for 'Macintosh' and 'Xserve' on Apple's page, I found this sentence in a pdf document: "... to adjust settings on a Macintosh or Xserve host."
(http://manuals.info.apple.com/en/XserveRAID1.2_AdminGuide.pdf), and no sentence explicitely calling the Xserve a Macintosh. However, software related info, like: "... this update is recommended for all Macintosh users..." seems to include the Xserve into the definition of a Macintosh, which makes sense since it is running Mac OS X.
I guess they don't want to call the Xserve a Macintosh, because they want the limit the term to end user computers.
Edit: I also found the following sentence in another pdf file: "The Xserve G5 is the Macintosh server platform..." (http://developer.apple.com/documentation/Hardware/Developer_Notes/Servers/XserveG5/XserveG5.pdf).
SMP systems are considered a fabulous success if they reach 80% efficiency. That's the reality. Intel is claiming 60-65% general performace increase with dual core P-M chips over single core processors. I think IBM's dual PPC 970s will fare well in this regard. You might expect that you will see something comparible to a 4GHz G5 AT BEST in a PPC 970MP running two cores at 2.5 GHz. That would be in a perfect world of perfectly parallel threaded code. (now, the other reality is.. a "4 GHz G5 would not scale linearly.. the memory bus would hold it back.. I'm talking about interpolating current single core performance up to a mythical 4GHz point)
They'll be fast though.
As for power.. G5s really don't run that hot. My dual 2.3 XServe runs CPU core temps in the office where we set it up of around 120-130 degrees F. That's pretty damn cool. That's under load. Idle, they run around 95-100F. In our chilled server room, where an XServe is SUPPOSED to reside, I'm running closer to 100F under load.
Now for comparison, my AthlonXP 3000+ upstairs runs at 159F at idle and it's got a copper/Aluminum heat sink with an 80mm fan on it... two 80mm fans in the front of the case, one 80mm fan in the back of the case behind the heat sink, and a power supply with a 120mm fan right above the heat sink pulling more dead air away from the processor area. That's actually a little high, I'm going to check the thermal compound, but 135F idle is normal for that chip and hitting 150F isn't unusual under load.
The problem isn't the ammount of heat the PPC 970 gives out, its that the chip is so damn small that 65W dissipation gives you twice the energy per area than a 100W Xeon or Opteron. It's a matter of getting a small.. comparatively small amount of energy out of the chip when you have much less surface area for conduction.
The PPC 970MP has two cores and twice the L2 cache (per core I believe). The surface area will be much greater and the revised process will make total waste heat much less per area. When Apple goes dual core (it's really a matter of when) they'll actually have a much easier time cooling the processors. The first run of duals is rated at 2.5 GHz, but from the typical thermal figures I've heard, they should be easy to cool in a loud 1U rack unit. They shouldn't even be that hard to silently cool in a big desktop with plenty of big fans. Apple will probably scrap the liquid cooling.
ffakr.
(now, the other reality is.. a "4 GHz G5 would not scale linearly.. the memory bus would hold it back.. I'm talking about interpolating current single core performance up to a mythical 4GHz point)They'll be fast though.
As for power.. G5s really don't run that hot. My dual 2.3 XServe runs CPU core temps in the office where we set it up of around 120-130 degrees F. That's pretty damn cool. That's under load. Idle, they run around 95-100F. In our chilled server room, where an XServe is SUPPOSED to reside, I'm running closer to 100F under load.
Now for comparison, my AthlonXP 3000+ upstairs runs at 159F at idle and it's got a copper/Aluminum heat sink with an 80mm fan on it... two 80mm fans in the front of the case, one 80mm fan in the back of the case behind the heat sink, and a power supply with a 120mm fan right above the heat sink pulling more dead air away from the processor area. That's actually a little high, I'm going to check the thermal compound, but 135F idle is normal for that chip and hitting 150F isn't unusual under load.
The problem isn't the ammount of heat the PPC 970 gives out, its that the chip is so damn small that 65W dissipation gives you twice the energy per area than a 100W Xeon or Opteron. It's a matter of getting a small.. comparatively small amount of energy out of the chip when you have much less surface area for conduction.
The PPC 970MP has two cores and twice the L2 cache (per core I believe). The surface area will be much greater and the revised process will make total waste heat much less per area. When Apple goes dual core (it's really a matter of when) they'll actually have a much easier time cooling the processors. The first run of duals is rated at 2.5 GHz, but from the typical thermal figures I've heard, they should be easy to cool in a loud 1U rack unit. They shouldn't even be that hard to silently cool in a big desktop with plenty of big fans. Apple will probably scrap the liquid cooling.
ffakr.
BTW, the other poster was correct.
The XServe is not a Macintosh. This could be an important distinction in the move to Intel.
.. though if you really parse words, Steve Jobs never said that Macs will never run PPC chips again. He said they are moving to X86. You can move a platform to X86 and yet still sell PPC machines though I'm not sure I see the advantage.
The way Apple is managing the transition though, it would be entirely possible and actually quite reasonable to keep both architectures alive for a long time. That's one reason why I'd still buy PPC.. though I expect I'll hold out for a dual core P-M based machine so I don't need to keep an Athlon for Gaming. Dual booting OS X and Windows will be a wonderful thing.
The XServe is not a Macintosh. This could be an important distinction in the move to Intel.
.. though if you really parse words, Steve Jobs never said that Macs will never run PPC chips again. He said they are moving to X86. You can move a platform to X86 and yet still sell PPC machines though I'm not sure I see the advantage.
The way Apple is managing the transition though, it would be entirely possible and actually quite reasonable to keep both architectures alive for a long time. That's one reason why I'd still buy PPC.. though I expect I'll hold out for a dual core P-M based machine so I don't need to keep an Athlon for Gaming. Dual booting OS X and Windows will be a wonderful thing.
manu chao said:
The 50 2.5GHZ nodes would beat out the 100 2.3 Ghz nodes in almost any scenario. Even considering that the memory bandwidth would likely be higher per processor on the dual 2.3s than on the quad 2.5s.
The main reason why the lower node count would probably win out is because its comm times will be lower than the 100 machines.
ffakr said:
Actually you dont have to parse his words at all. Steve said " we have several exciting PowerPC products still in the pipeline"
Hector said:
Ahh... But a 'Xeon' is just a multi-processor capable Pentium with extra cache. And as the Pentium-M is based on the older Pentium Pro/Pentium II/Pentium III core, it shouldn't be that difficult. (As the Pentium Pro was meant to be a high-end server chip in the first place, and the first 'Xeon' processors were just Pentium II or Pentium III chips with extra cache and capable of up to 4 processors per bus.)
As for the people suggesting or denying Itanium.... Steve just said 'Intel', he never once said 'x86' during his keynote. Yes, the fact that the transition machine is a Pentium 4 makes it about 99.9% likely that the PentiMacs will be x86, but that's still no guarantee.
And with the latest rumors of Intel's next core actually being roughly based on the same idea as the Itanium, (although not the same core instruction set,) it's possible that both the 'Pentium' and the 'Itanium' lines will merge into one basic core, with different translators deciding which line it will end up in. So it is entirely possible that Apple will release Itanium servers under this hypothesis.
I doubt it, though. I think Intel's new core will be a new x86 core that happens to have more in common with the Pentium-M than the Pentium 4.