Woodcrest June 26th. XServe Too?

[macaddict06]

Strangest part of this rumor...

People are still ordering Xserves?

Duh. Not to flame you...BUT, the XServe G5 is still a DAMN good machine. Plus, if a new one comes out in less than a month, Apple will either upgrade the ones that were bought or will give a discount on orders (Remember a year or two ago when the cost of a PMG5 and a 23" Cinema display together fell something like $700? People recieved rebates for the difference).
Further, just because new chips are out does not make the XServe G5 any less powerful in and of itself. Some network admins would MUCH rather run RISC over CISC, and they can get that in a G5, won't get it in an x86-based server.
So yes, people are still ordering, and with good reason.

 

[shawnce]

Provided that woodcrest clock starts at 1.6 ghz, would it be possible to put one of those new chips in a new iMac?

Would it represent an advantage over conroe or meron??

No. It would likely result in a slower more expensive and hotter system (nosier) then if Apple placed a Merom or Conroe in the system (note the first generation Meroms are pin compatible with current iMac mother boards... cheap for Apple to upgrade). Also Apple isn't going to make a 2 x Woodcrest (Quad core) iMac anytime soon and in many consumer situations such a system would perform more poorly then a higher clocked dual core system.

Apple could place a 2.3 GHz Merom in the top end iMac that would yield about a 40% advantage over the current top end iMac (Core Duo) while a dual core 1.6GHz Woodcrest would be likely be 10+% slower then the current top end iMac (Core Duo).

 

[Fabio_gsilva]

No. It would likely result in a slower more expensive and hotter system (nosier) then if Apple placed a Merom or Conroe in the system (note the first generation Meroms are pin compatible with current iMac mother boards... cheap for Apple to upgrade). Also Apple isn't going to make a 2 x Woodcrest (Quad core) iMac anytime soon and in many consumer situations such a system would perform more poorly then a higher clocked dual core system.

Apple could place a 2.3 GHz Merom in the top end iMac that would yield about a 40% advantage over the current top end iMac (Core Duo) while a dual core 1.6GHz Woodcrest would be likely be 10+% slower then the current top end iMac (Core Duo).

Thanks for the clarification...
Slowly I'm learning...

 

[Macnoviz]

hepta means 7. penta is 5.

I think Octo- en Hexacores are more likely

Or maybe Googoltriplexcores

 

[brianus]

Some network admins would MUCH rather run RISC over CISC

Why? Isn't that just a computer sciencey debate? What real world difference does it make?

 

[MacsRgr8]

HA! I feel your confusion. PowerPC fanboys I guess (or people that need a G5 server).

Nice to hear some real Intel fanboys around...

The G5 is still a very fast CPU. Great CPU for in the Xserve or Power Mac.
Nothing being PowerPC-fanboy about regarding an Xserve G5 as a good price/performance investment, same goes for the Quad G5, especially if you need power NOW, with real apps running natively NOW (which specifically goes for Mac OS X Server...)

 

[j_maddison]

For what purpose? You thinking 2 x 1.6GHz Woodcrest in an iMac? If not then why Woodcrest?

Why put a server targeted chip in a low end desktop when it will gain you near zero increase in performance (actually at 1.6GHz it will lose you performance) and only raise bill of materials costs.

• Woodcrest = server targeted (large thermal/power operating range, higher price, uses expensive chipset)
• Conroe = desktop targeted (average thermal/power operating range, low/medium price, uses low cost chipset)
• Merom = laptop / small form factor trageted (low thermal/power operating range, low/medium price, uses low/medium cost chipset)

Conroe, Woodcrest and Merom have basically identical cores and hence very nearly identical performance when operating at the same clock rate, thermal envelope and similar FSBs. The main differences is the socket supported, on die cache sizes and of course supported thermal/power envelope (reflected in clock rates supported and power throttling points).

Thanks, thats pretty much the explanation I was looking for. I was wondering if there was a performance advantage to be had from using a Woodcrest chip at a lower clock rate over a higher clocked Conroe. I didn't know, so I asked the question. thanks

Jay

 

[MacFan25863]

If it's true that Woodcrest is coming out both (i) ahead of schedule, and (ii) "over-powered," then I think we have really gotten our first dose of "this is what it's like to work with Intel."

Of course, just "ahead of schedule" is freakin' amazin'.

I agree! Its really amazing..we have had more speed increases and new technologies on the Mac in the past 6 months than we've had in the past 3 years!

 

[hulugu]

Why? Isn't that just a computer sciencey debate? What real world difference does it make?

Simply put, if you have code that only runs on RISC processors a CISC processor is going to be a problem. Not everything has been, or will be, translated to UB especially in the near-future.

I can't think of any examples, but I'm sure that a RISC processor is still necessary for some applications given that up until just this year, Apple was only using RISC with the PowerPC.

 

[danielwsmithee]

Simply put, if you have code that only runs on RISC processors a CISC processor is going to be a problem. Not everything has been, or will be, translated to UB especially in the near-future.

That is true no matter what processor type you are switching between. Whether you are using RISC or CISC has no impact on this. Code written for a G5 will not run a SPARC even though they are both RISC machines.

The concept of RISC and CISC is very outdated. Any modern processor you look at is in some ways RISC and in other ways CISC. Modern compilers hide the difference even when it does exist. x86 Machines use to be solely CISC now each CISC instruction is broken down into multiple RISC instructions that run on each of the processing units. The PowerPC's atlti-vec instructions would likely be considered CISC instructions.

There are reasons for wanting a G5 processor, but the fact that it is a RISC processor has little influence.

 

[andrewag]

I wouldn't mind a bit more of a variety in the specs. Maybe some lower end ones for apps like basic webservers etc.

With the Intel transition, I hope that Apple keep the hardware up to date. The Xserve slipped quite a bit next to its PowerMac sibling.

 

[andrewag]

I agree! Its really amazing..we have had more speed increases and new technologies on the Mac in the past 6 months than we've had in the past 3 years!

Apple have simply jumped over to Intel at the right time. A couple years ago they were going nowhere.

 

[Shintocam]

What is the typical xsere used for. Who are Apple's customers for this?

I assume these are used for more then just serving files to a small workgroup. A G4 would be enough for that. Do people actually run DBMSes on Apple hardware?

XServes are used in pretty much the same way as any Linux cluster. They make great webservers and database servers, but they are particularly well suited to multiprocessor data crunching. We have a small XServe cluster that we use for biology applications such as genome analysis etc....where you are running multithreaded applications etc...They scale well - similar to a linux cluster, but price for performance is extremely high even for the G5 XServes, particularly at academic prices. Companies like SUN have a hard time competing on price/performance and with the switch to the new chips, it's going to be even more difficult (even though ironically now the two companies are using the same architecture).

 

[IscariotJ]

Companies like SUN have a hard time competing on price/performance and with the switch to the new chips, it's going to be even more difficult (even though ironically now the two companies are using the same architecture).

Slightly OT, but Sun's Galaxy servers offer excellant price/performance. I installed a pair of X4100's ( dual dual-core Opteron's, 8GB RAM ) running Solaris10x86/Oracle for an ISP performing traffic shaping on 200,000 customers. The load very rarely went above 1.

I would be very interested to see how the new Xserves will compare running the same software.

 

[Core Trio]

No mention of the Powermac at the moment then. Could a 1.6ghz Woodcrest be used in an imac? I know its unlikely, but just curious to know if it could

Jay

I wouldnt really be efficient, the power consumption is high, the cost is higher than Merom or Conroe, and im sure the heat output of a server class chip is far greater than that of the chip currently housed inside the "laptop-on-a-stand" that is the current iMac.

 

[DharvaBinky]

What's the difference between a 1.6 Woodcrest and a 1.6 Conroe besides a dual processor configuration?

Hyperthreading? If so, 2xWoodcrests would appear to OS X as 8 cores. Too bad it's a shared FSB, no matter how fast it is.

I had actually heard that Apple was going to depart from 100% stock Intel components for the xServe and/or Mac Pro and that this departure was going to be a custom chipset that used Apple's point to point interconnect to make the FSB a switched fabric, but appear as a standard shared bus to the CPUs...

Did I dream that? I don't recall waking up with crusty pants...

 

[SPUY767]

My cluster suddenly looks underpowered. :-(

 

[ChrisA]

Simply put, if you have code that only runs on RISC processors a CISC processor is going to be a problem. Not everything has been, or will be, translated to UB especially in the near-future.

I can't think of any examples, but I'm sure that a RISC processor is still necessary for some applications given that up until just this year, Apple was only using RISC with the PowerPC.

I don't think it matters much. I have code here that runs interchangably on both i386 and SPARC. and either Linux or Solaris. From my point of view there is far more difference betwen the three operating systems (Linux, Solaris 10 and "older Solaris versions like 6") then between the SPARC and i386. The gcc compiler does it's job pretty much without me having to think mach about it. As for the differences between OSes. GNU "Auto tools" helps to make that resonably easy too.

That said, I KNEW I wanted my code to be portable and so I make it easy on myself by using only development tools that are present on all platforms. Porting between sets of development tools is hard work and in some cases parts need to be rewritten. I use X11 because that is common to all platforms but if I wanted to use Mac's native graphic it would be a re-write. But as is I'd bet my stuff would "just work" on either a G4/5 or Intel mac.

In summary it is the software environment in which the code is developed and runs that matters a LOT more then the type of CPU.

 

[ChrisA]

Too bad it's a shared FSB, no matter how fast it is.

The overall memory bandwidth depends on the speed of the bus times the probibility that the data is NOT is cache plus the speed of the cache times the probibility that the data item is in cache. If you can double the probibility of a cache hit then you have almost doubled the effective memory bandwidth. These new processors have huge caches and likely very high cache hit ratios. So you can't directly compare FSB speeds you need to multiply the FSB speed by the chance that the some data is on the other side of the bus.
Typically hit ratios are very high, on the order of 90%

Analysis id even harder because that while a cache miss does mean that one tread is "stalled" waiting for a cache line to be filled from RAM it may be that some other thread is not stalled and the processor can continue to do usfull work and so the over all prograss of work on the machine is not slowed. So sped depend on what you are doing that dy with your computer.

That said, of course faster is better.

 

[fervor]

Might be something to this...

I had a development server fry on me yesterday and wound up ordering an XServe as a replacement. I called a reseller friend of mine, had him check for availability - nothing until the week of the 26th. Wound up ordering an 'off the rack' configuration which his distributor had 4 of in the entire country. I would wait, but for what we're doing on that development server, even the G5 is way overkill.

I'm hoping we see an Intel XServe that week! That'd be friggin' awesome. I have big RAID/fiber channel to two XServes that I've been putting off until the Intel XServes hit.

By the looks of the XServe inventory channels, I'd say something is going on.

-bob

 

[bradc]

Exactly ^^^^ I got the same thing, I tried to order two Xserves & an Xserve RAID and everything was going to be late and coming from all over, so I said screw it and I figure I can wait.

 

[MacNemesis]

Strangest part of this rumor...

People are still ordering Xserves?

Just remember that OS X server isn't unversal yet.

 

[081440]

Just remember that OS X server isn't unversal yet.

It'll be released with the new Xserve, Why release it early? (They'll probably offer a cross-grade disk for people who just bought it, like they did for the pro applications)

 

[brianus]

It'll be released with the new Xserve, Why release it early? (They'll probably offer a cross-grade disk for people who just bought it, like they did for the pro applications)

Hmm.. splains why 10.4.7 still isn't out despite weeks of "release imminent" rumors and relatively bug free seeds. They'll do the same as they did with 10.4.4; wait for the introduction of a new product to release a new edition of the OS.

 

[DHagan4755]

Working under the assumption that the Mac Pro will use Woodcrest, is there any sign that the Power Macs are getting held up/drying up til the 26th?