2007 Harpertown Stoakley-Seaburg Motherboard With New Graphics Cards 8-Core Mac Pro

[MacsAttack]

Hmmm... Poll is missing the "I'll wait for Nephelium-based processors. Anything Intel is going to release before then will only provide incrimental improvements and I already have a Mac Pro thank you" option.

8 cores? Pifil! Intel have a demo of an 80 core CPU. Don't settle for anything less!

 

[Keebler]

No. Just clock speed.

thanks digitalbiker. i perused the site last night and figured that was the only diff. think i'll just get the 2.66 and add some ram and maybe the x1900.

cheers,
keebler

 

[Lycanthrope]

thanks digitalbiker. i perused the site last night and figured that was the only diff. think i'll just get the 2.66 and add some ram and maybe the x1900.keebler

Same here except I really don't need an X1900 - as I'm not a power user I can even live with 1GB RAM for a month or two and get a 23" ACD rather than 20"

 

[AidenShaw]

Though it would have been nice for Apple to add the other Quads, they were more expensive than they were worth apparently, not being much faster, if at all, in real world use than the dual 3s.

And the octo 3s, in real world use, won't be much faster than the quad 3s. Just about all arguments about not using the 2.66 apply to the 3.0.

The problem is software and tasks. Some things haven't been rewritten to be more parallel, and some things can't be rewritten to be more parallel.

If you have a parallel application, an octo 2.66 will clobber a quad 3.0. If you don't, you won't see a "real world" benefit.

Again, it would have been nice, but probably more of a hassle than it would have been worth for them.

What's Apple going to do when the 3.33 GHz dual core Xeons come out? Stop selling the octo - because it's slower?


I think that a big part of the equation is that OSX doesn't have processor affinity implemented, so that 10.4 will struggle on a 8-way system. Apple didn't want to see lots of benchmarks showing the octo as being significantly slower than the quad on many tasks.

Supposedly 10.5 has affinity, so now is a good time to introduce an octo so that the ADC developers will have an 8-way to fine tune applications for the 10.5 final release.

 

[BiikeMike]

I've waited and waited, and wasted so much time on chugging away at my little MacBook Pro (which is now broken, and about to be shipped home to see the doctor) that my cart is full and if nothing new is announced at NAB, the PLACE ORDER button will be clicked!


Waiting for new technology is an endless thing, and while I appreciate that some people don't really need a new computer, and can just wait and wait as long as they feel like, I do need a computer, and I can't wait!

3.0GHZ Octo-core, HERE I COME!

 

[digitalbiker]

I've waited and waited, and wasted so much time on chugging away at my little MacBook Pro (which is now broken, and about to be shipped home to see the doctor) that my cart is full and if nothing new is announced at NAB, the PLACE ORDER button will be clicked!


Waiting for new technology is an endless thing, and while I appreciate that some people don't really need a new computer, and can just wait and wait as long as they feel like, I do need a computer, and I can't wait!

3.0GHZ Octo-core, HERE I COME!

I feel your pain!

I have a PB 1.67 GHZ that I am using for my Filemaker Pro Client, Dreamweaver, PhotoShop, Acrobat, Illustrator, Painter IX, Starry Night Pro, FC Pro, mail, Keynote, ical, quicken and Xcode2 machine.

I have a 6 1/2 year old Dell Precision workstation, 1 Ghz P3 Xeon machine. I run Peachtree, CGM studio Pro, Microsoft Office, and Geographix Discovery in Windows on that machine. I also dual boot into RHEL3 to run Landmark Graphics Software and Oracle.

My Dell is fading fast and my PB is over taxed.

I have been holding out for the Mac Pro to go Octo to use virtualization for Vista, RHEL4, and run native in OS X.

I was hoping for that setup to replace my current work environment. But it has to last 4 -5 years at least and I am afraid that this current Mac Pro is not at the sweet spot for long-life technology yet.

Now I am waiting for NAB just to see if Leopard is demoed / announced and if that comes with any other internal goodies that indicate a switch to SS, BR, and additional GPU options for the Mac Pro by June.

However if NAB goes off without a mention of Leopard or any other significant announcement. Boom, I'm push'in the proceed to checkout button. I can't wait til Sept. / Oct. My Dell is fading fast.

 

[AidenShaw]

it's as sweet as it gets....

I was hoping for that setup to replace my current work environment. But it has to last 4 -5 years at least and I am afraid that this current Mac Pro is not at the sweet spot for long-life technology yet.

It should look sweet. If you're currently on a G4 PB and a 1 GHz P3 - a Core 2 Duo MiniMac would last you for 4 years!

No way that you need the dual-dual-dual just announced.

Stoakley will be a minor incremental improvement - unless you're running a relational database for thousands of users, or unless you really do need 60 odd lanes of PCI Express.

If I were you, I'd consider getting a dual-dual 2.0 GHz instead of the dual-dual-dual 3.0 GHz. You'll save a *lot* of money, and have something far faster than what you have now.

Put four 2 GiB FB-DIMMs in it, and leave four empty for the future.

In about 12 to 18 months, Intel will announce that the next chip won't use the socket 771 - and therefore no newer processors will come out to fit your motherboard.

When those new processors come out, pick up a pair of 4.2 GHz Nehalem true quads at closeout prices and upgrade. Fill out the four empty memory slots if it needed.

Doing the "big bang - this gotta last 4 years" upgrades in the WinTel world has two big drawbacks:

a: You pay top dollar for the very fastest parts (note that the quad core 2.0 GHz processors are 1/3 slower and 1/3 the cost of the 3.0 GHz quad)

b: Towards the tail end, you have a system that's far slower than state of the art


Some pundits recommend buying a new system each year, one that will get you through that year. Next year, replace or upgrade.

At the end of 4 years, you'll have a much faster system, and will have spent less money overall.

 

[Multimedia]

MultiCore Ready Branding Would Help A Lot

And the octo 3s, in real world use, won't be much faster than the quad 3s. Just about all arguments about not using the 2.66 apply to the 3.0.

The problem is software and tasks. Some things haven't been rewritten to be more parallel, and some things can't be rewritten to be more parallel.

If you have a parallel application, an octo 2.66 will clobber a quad 3.0. If you don't, you won't see a "real world" benefit.

What's Apple going to do when the 3.33 GHz dual core Xeons come out? Stop selling the octo - because it's slower?

I think that a big part of the equation is that OSX doesn't have processor affinity implemented, so that 10.4 will struggle on a 8-way system. Apple didn't want to see lots of benchmarks showing the octo as being significantly slower than the quad on many tasks.

Supposedly 10.5 has affinity, so now is a good time to introduce an octo so that the ADC developers will have an 8-way to fine tune applications for the 10.5 final release.

My questions to Apple PR:

Does Marketing plan on offering 8-core ready branding for developers? i.e. is there in development a packaging badge for developers of software that can see and use all 8 cores in an 8 core Mac Pro so they can let customers know their software will take advantage of all the power an 8-core Mac Pro can deliver?

This would be similar to the badge you offer for Universal Binaries and the Made for OS X Badge. Otherwise I'm confused how anyone will be able to easily tell who makes their software play nice with the 8 core and who doesn't. Perhaps even a 2 core, 4 core and 8 core set of badges would be in order. What do you think?

 

[digitalbiker]

It should look sweet. If you're currently on a G4 PB and a 1 GHz P3 - a Core 2 Duo MiniMac would last you for 4 years!

No way that you need the dual-dual-dual just announced.

Yeah, I'm sure your right and you make a lot of sense.

I didn't mean to imply that the 1Ghz P3 was adequate at all for my needs. It basically sucks right now and I am unable to update to the latest version of the Landmark Software that I normally run because the CPU & GPU are insufficient.

My main motivation for waiting for the Octo, was to be able to put a lot of memory on the system and run Vista, RHEL v4 with VMware, while in an OS X environment.

I figured that the performance hit from running virtual might be lessened if I had enough cores and memory to go around. I also think that this setup could save me a lot of time, instead of the constant, rebooting into a different OS when I needed to switch software.

I was hoping that when Leopard introduces the spaces feature that I could actually run full screen with the Virtual OS's in a different workspace. Then I could be in OS X, press a hot-key and switch space to RHELv4, press a hot key and switch to Vista, etc. etc.

So I was kind of waiting not only for the Octo but for Vista, VMWare, Leopard, Boot Camp, etc. to come together to make something like this possible.

All in all, I have been waiting too long! Money isn't so much the issue. The Landmark software that I run originally cost over $250,000.00 and has a $40,000.00 a year maintenance fee. It originally was designed to run on a Sun system that cost over $75,000.00.

So now that it runs on Red Hat Enterprise and I can use off the shelf PC's or Mac's, hardware is a minor business investment.

 

[Eidorian]

My questions to Apple PR:

Does Marketing plan on offering 8-core ready branding for developers? i.e. is there in development a packaging badge for developers of software that can see and use all 8 cores in an 8 core Mac Pro so they can let customers know their software will take advantage of all the power an 8-core Mac Pro can deliver?

This would be similar to the badge you offer for Universal Binaries and the Made for OS X Badge. Otherwise I'm confused how anyone will be able to easily tell who makes their software play nice with the 8 core and who doesn't. Perhaps even a 2 core, 4 core and 8 core set of badges would be in order. What do you think?

I have to agree that applications do need some sort of easily found label that will notify the buyer that an application is optimized for multiple core systems.

 

[Butthead]

Well the chips were made available to Apple and so they quietly added that expensive option for those who think they need it sooner, rather than later.

http://www.electronista.com/articles/07/04/05/mac.pro.special.xeon/

"For now, the product is in limited production and Apple has chosen to adopt it. We will introduce another 3.0GHz Xeon SKU later on as well...The model in question is the Xeon X5365, according to additional comments by Intel's Italian PR director Ruben Simpliciano."

Interesting, another 3.0GHz of a different SKU??? You know Apple loves to hype their video editing products capabilities, using the fastest available (or soon to ship) model they can come up with. All the video edit software is horsepower intensive. Perhaps they don't have a fast enough GPU option at this time, and the 8-core is what they needed for showing off the wow factor with all the vid ed software making it's debut at NAB. Which will kind of suck, because I want most, if not all of it; to be able to be run on an substantially improved Santa Rosa MBP 17in. w or w/o Leopard.

What happened to all those LED backlight screens, 2k res or higher that were supposedly coming soon?

Man this is hilarious. Almost 9 hrs after I post the link to the original story, Macrumors admin got around to reading MacNN's link, lol. 111 post on that yesterdays story, now picked up by Appleinsider...snooze.

MM, do consider what I have stated? Go to some FCP usergp meetings like that lafcpug.org. (see the heavy hitters like Walter Murch, guest speaker at NAB and frequently at lafcpug.org meetings). Apple loves to tout the latest, fastest desktop machine for pro users at NAB, using the latest beta FCP. It's almost a given that the release of the 8-core MP was done because it will be used to show all the wicked greatest of Apple's video editing solutions at NAB.

How dense are all of the macrumors readers to not figure this out?

Sure there will be better/faster/greater systems out before the end of the year...NAB starts soon, 8-core, but everything else the same...typical Apple, it's all about showmanship at NAB for the video editing market...like duh!

As always, everything introduced now is an incremental increase over last revsions (excepting the switch to Intel from lagging/stagnant PPC, which was grossly hyped as being 4x as fast a PPC, when in most instances it was hardly the case). 45nm process, new chipset/motherboard, etc does gain, and only those who absolutely need that minor but noticable increase in performance really feel the improvement. For the rest of us, it's all window dressing. the usual caveat, buy it if you need it now, wait if you can....same old, same 'ol; nothing has changed over the years.

 

[solvs]

And the octo 3s, in real world use, won't be much faster than the quad 3s. Just about all arguments about not using the 2.66 apply to the 3.0.

The Quad 2.66 can either be faster or slower than the Dual 3.0 depending on the app. The Quad 3.0 will either be faster or roughly the same speed, but not really slower. I would have been happy with the Quad 2.66. Even for the price, it would have been worth it for me. Apple didn't want to do it, and I don't blame them, but I didn't say I was happy about it.

I would've also liked to see a price drop and/or lower end machine as a stop gap, but frankly, I'll take what I can get, and the dev discount I get negates a big chunk of the overhead.

What's Apple going to do when the 3.33 GHz dual core Xeons come out?

They could wait for the Quad version, or go ahead and release it as they did with the dual 2.5 G5s vs. the single 2.7 depending on price. I don't know. Again, personally regardless of price, I think they should have them all. Including the Conroes. But they don't.

Supposedly 10.5 has affinity, so now is a good time to introduce an octo so that the ADC developers will have an 8-way to fine tune applications for the 10.5 final release.

It's supposed to be better. I haven't seen it in what I've been using, but I'm hoping for the best. Or at least better. 10.4 isn't bad, but it certainly has room for improvement. FCP 6 is supposed to be better too, and running multiple Adobe apps will be smoother, but yeah, most people are not going to see it in real world no matter how many cores they stick in there.

I will though, and I can afford it now, so

 

[uraniumwilly]

Waiting For The Logic 8 Thing

I'm guessing something is about to happen with Logic and Leopard w the 8-core. That's really the only reason I've changed my mind after waiting 6 years to buy a new Mac (upgrading from G4 with dual 1.8 accel.).

I had purchased 2 of the 150 Raptors in anticipation of the 8-core.

This SS motherboard information is very helpful, btw. I was very much in the dark about such arcane innerworkings.

 

[ATD]

The problem is software and tasks. Some things haven't been rewritten to be more parallel, and some things can't be rewritten to be more parallel.

I agree. I'm not ready to dismiss this computer at all, it looks good to me. I think everything has to do with the software you are running. Not the computer, not the OS. I have been using Mental Ray (Maya) for years and it has been a fully parallel app as long as I can remember. I have been rendering to 6 cpus (a quad + a dual) ever since the quad came out on OSX 10.2 (maybe 10.3, I don't remember). It handled it beautifully then as it does now in OSX 10.4. When I see a 260% increase in Mental Ray on a 8 core over a quad and disappointing numbers in other software I have to believe it's how the app is written, nothing else. Sure maybe Leopard will make it even better, but I don't see the downfall being the 8 core or the OS.

 

[daveL]

I agree. I'm not ready to dismiss this computer at all, it looks good to me. I think everything has to do with the software you are running. Not the computer, not the OS. I have been using Mental Ray (Maya) for years and it has been a fully parallel app as long as I can remember. I have been rendering to 6 cpus (a quad + a dual) ever since the quad came out on OSX 10.2 (maybe 10.3, I don't remember). It handled it beautifully then as it does now in OSX 10.4. When I see a 260% increase in Mental Ray on a 8 core over a quad and disappointing numbers in other software I have to believe it's how the app is written, nothing else. Sure maybe Leopard will make it even better, but I don't see the downfall being the 8 core or the OS.

The OS has a ton to do with scalability across multiple cpus / cores. It doesn't matter how well your app is written, if the OS has a poor threading model and /or poor kernel locking (mutex, semiphores) strategy. Try running Windows on an 8-way system. It doesn't matter what app(s) you run, it sucks. Maybe things have gotten better with Vista or recent updates to Server?

 

[ATD]

The OS has a ton to do with scalability across multiple cpus / cores. It doesn't matter how well your app is written, if the OS has a poor threading model and /or poor kernel locking (mutex, semiphores) strategy. Try running Windows on an 8-way system. It doesn't matter what app(s) you run, it sucks. Maybe things have gotten better with Vista or recent updates to Server?

I was just referring to OSX and that some believe it's going to take 10.5 to run a app on more than 4 cores. I don't believe I'll be using Windows for anything, thank you.

 

[AidenShaw]

Try running Windows on an 8-way system. It doesn't matter what app(s) you run, it sucks. Maybe things have gotten better with Vista or recent updates to Server?

I guess that's why the #6 system on the TPC benchmark page is a 64 core Windows system - because it sucks?

http://www.tpc.org/tpcc/results/tpcc_perf_results.asp?resulttype=noncluster&version=5&currencyID=0

 

[daveL]

I guess that's why the #6 system on the TPC benchmark page is a 64 core Windows system - because it sucks?

http://www.tpc.org/tpcc/results/tpcc_perf_results.asp?resulttype=noncluster&version=5&currencyID=0

Ah, AidenShaw ... I knew I'd get a reply from you. I tried, in my own lame way, to indicate my Windows knowledge was exactly current. Besides, your talking about a benchmark, not a real world app. However, it is interesting to note that the systems immediately ahead of your Windows-based result are running half as many cores, which speaks directly to scalability.

 

[daveL]

I was just referring to OSX and that some believe it's going to take 10.5 to run a app on more than 4 cores. I don't believe I'll be using Windows for anything, thank you.

Tiger has been accused of not scaling well on multiple cpus / cores. Leopard is supposed to improve in this regard. I sure hope so. We may see a fairly noticeable performance improvement on our 2-4-8 cpu / core Macs.

 

[AidenShaw]

However, it is interesting to note that the systems immediately ahead of your Windows-based result are running half as many cores, which speaks directly to scalability.

And they're also newer dual-core hyper-threaded chips with almost 3 times (24 MiB) the cache.

You should have stopped when you admitted that your opinions on Windows are based on out-of-date information.

 

[ATD]

Tiger has been accused of not scaling well on multiple cpus / cores. Leopard is supposed to improve in this regard. I sure hope so. We may see a fairly noticeable performance improvement on our 2-4-8 cpu / core Macs.

I have had no problem with Tiger or pre Tiger on 4+ cpus. That's Mental Ray, everything else is a different story. I have not seen anything else go over a 2 cpu usage. Maybe Leopard will help, I just don't think that the OS is fully the issue in blocking parallel apps from performing at top speed. Mental Ray scales very well, the numbers for an 8 core/Mental Ray render on the Apple site support that.

I'm waiting for CS3 and Leopard, at that point I'm going to consider a 8 core. If I get one I'm going to hook up the 8 core, a quad and a dual for 14 cpu renders. Should I post results?

 

[McEngineer]

My questions to Apple PR:

Does Marketing plan on offering 8-core ready branding for developers? i.e. is there in development a packaging badge for developers of software that can see and use all 8 cores in an 8 core Mac Pro so they can let customers know their software will take advantage of all the power an 8-core Mac Pro can deliver?

This would be similar to the badge you offer for Universal Binaries and the Made for OS X Badge. Otherwise I'm confused how anyone will be able to easily tell who makes their software play nice with the 8 core and who doesn't. Perhaps even a 2 core, 4 core and 8 core set of badges would be in order. What do you think?

It would have obvious value to consumers, but it sure opens a can of worms for developers. How ready is ready? Does the software's 8-core performance have to be convincingly close to twice the 4-core, or does a 20% improvement constitute 8-core ready? How do you certify readiness in a way that is independent of the user's amount of RAM? What about future chip designs that may still have 8 (or 4 or 2) cores but utilize different bus and caching architectures, thus changing the performance of previously certified code? Etc., etc.

...In my own little simulation niche, the emergence of "N-core" systems is a nontrivial problem; the optimization and scaling requirements are not nearly as understood (at least by me) as, say, using MPI on cluster-type systems. It also is not clear whether we should truly be looking ahead to "N-core" algorithms or stick to ones customized for a fixed number, or a selection of a few (e.g. 2,4,8). It may well be the case (I don't follow the roadmaps that closely) that the chip manufacturers will not rush to increase the number of cores until the software developers send clear signals that they know what to do with them.

 

[Lord Blackadder]

Give me Stoakley-Seaburg or give me death!

I'm more interested in what flavors of the GeForce 8 and R600 video cards show up.

 

[Multimedia]

Can You Set Up A Render Cluster of Macs Both PPC & Intel?

I have had no problem with Tiger or pre Tiger on 4+ cpus. That's Mental Ray, everything else is a different story. I have not seen anything else go over a 2 cpu usage. Maybe Leopard will help, I just don't think that the OS is fully the issue in blocking parallel apps from performing at top speed. Mental Ray scales very well, the numbers for an 8 core/Mental Ray render on the Apple site support that.

I'm waiting for CS3 and Leopard, at that point I'm going to consider a 8 core. If I get one I'm going to hook up the 8 core, a quad and a dual for 14 cpu renders. Should I post results?

Can you do this with an 8 Core Mac Pro + Quad G5 + Dual Core G5 set of Macs? Or do they all need to be Intel Macs?

 

[ATD]

Can you do this with an 8 Core Mac Pro + Quad G5 + Dual Core G5 set of Macs? Or do they all need to be Intel Macs?

You can do a mix of macs and windows PCs so I'm hoping I can do the same with PPC and Intel Mac. I hope to find that out before I start ordering.