A sliver of hope / Sandy Bridge Mac Pro??

[Kimmo]

From an interesting article that was posted here in another thread:

Kirk Skaugen, general manager of Intel's Data Center and Connected Systems group, confirmed last week that Intel is shipping Xeon E5 silicon right now for revenue to selected hyperscale database and supercomputing customers under non-disclosure agreements.

"Don't confuse volume shipments under non-disclosure going into the large cloud and HPC customers with when the best time to do a marketing launch is," Skaugen said.

http://www.theregister.co.uk/2011/09/19/intel_xeon_e5_follow_up/

 

[Marcush1286]

Nanofrog, one of many advisors on here.

nano?

 

[nanofrog]

Image

Nice. I'm downloading that.

Since Apple only really needs 3-4 PCI-e boards that are somewhat custom anyway (the video cards for BTO configs) to cover a very broad number of Mac Pro sales they should be able to launch on lower volumes of both cards and E5s. If they stuck close to the Intel reference designs the chips will work. ( Hand waving about how Intel could not round up enough chips to validate their own products because the yields are too low and I'll know folks are smoking powerful drugs. )

ES samples for testing hasn't been an issue from what I've seen. Not sure where this is coming from in terms of facts to support it (noticed it in this and another thread IIRC).

Apple has closely followed reference designs in the past, so I don't expect this to change as it's the most cost effective approach.

I think the majority of the other system vendors are dragging their feet on the E5 release because have broader set of risks and issues want to put under control and the holiday timing throws them all off. (e.g., won't have enough PCI-e v3.0 cards in place , field testing processes feedback loops closed , etc. ). This is exactly one of those context where Apple could move more swiftly because they have less baggage. Assuming that they prepared to take advantage of that difference.

I see this as less about the Holidays, and more about PCIe 3.0 (valid issue for DP systems, as those will make it into clusters/supercomputers/cloud configurations that need the additional bandwidth).

Hardware bugs that make it into final parts could be very costly as well as a PR nightmare, particularly as the information on Intel's slides indicated PCIe 3.0 design goals.

Apple could move the shipments into 2012 by announcing late in Nov that systems are shipping in January. But to blank 2011 entirely from Mac Pro news. That's a bad move. That may be enough to kill off the Mac Pro in terms of viable volume.

Maybe, but other vendors are in the same situation, which mitigates the issue IMO (comes down to when Intel ships their E5 orders, and if any particular vendors have a preferential deal in terms of shipment dates due to purchase volume).

 

[deconstruct60]

Maybe, but other vendors are in the same situation, which mitigates the issue IMO (comes down to when Intel ships their E5 orders, and if any particular vendors have a preferential deal in terms of shipment dates due to purchase volume).

Not really. The vast majority of the other vendors have 2-3x models that Apple has. There is an upside and downside to having a minimalistic line up. The upside is that Apple cuts way down on cannibalization and allows Apple to focus more attention on individual products. The downside is that if do not have overlapping coverage is that if they miss on product updates that whole sector is missed. The competitors with more overlapping coverage will scoop up more of the customers who have to move "now" (e.g., major projects to start, have hard windows within to spend the money, etc. ).

Competitors are going to come in and prune off potential Mac Pro customers with updated i7 base boxes for which Apple will have no answer for. It is already the case that the more price sensitive customers are moving that way. An extended delay on E5's will only add to the effect.

You are probably right for the even slower moving server market which wouldn't take the Core i7's anyway. However, that is not the Mac Pro's core target market.

Apple has to wait for the E5 and PCI-e v3.0 but with money to invest and fewer cards to test it shouldn't be hard to shake out the minor glitches earlier than the other vendors can. They can easily collect enough information to allow them to at least pre-announce in order to freeze customers (and perhaps collect some money that would expire on the quarter end: cash checks on 12/31 and ship very early in Jan. )

 

[WardC]

I hope Tim Cook would see some enlightenment in power computing, other than these consumer-based intel boxes...and give us a Power 7 machine with a Mac OS X build that would run on such a box.

Power795 -- http://www-03.ibm.com/systems/power/hardware/795/specs.html

 

[scottsjack]

I hope Tim Cook would see some enlightenment in power computing, other than these consumer-based intel boxes...and give us a Power 7 machine with a Mac OS X build that would run on such a box.

Power795 -- http://www-03.ibm.com/systems/power/hardware/795/specs.html

Amen brother, Tim needs to steer the company back to making powerful, professional level computers and an OS designed for them.

 

[nanofrog]

Not really. The vast majority of the other vendors have 2-3x models that Apple has. There is an upside and downside to having a minimalistic line up. The upside is that Apple cuts way down on cannibalization and allows Apple to focus more attention on individual products. The downside is that if do not have overlapping coverage is that if they miss on product updates that whole sector is missed. The competitors with more overlapping coverage will scoop up more of the customers who have to move "now" (e.g., major projects to start, have hard windows within to spend the money, etc. ).

I realize what you're getting at, but since Apple doesn't really compete in the server market, I don't include those (also think of board makers' product lines in terms of workstations, as they do tend to offer more than a single workstation board, such as ASUS's LGA1366 based offerings).

So I've been thinking in terms of (say in the case of Dell vs. Apple on workstations):

• SP MP vs. T3500
• DP MP vs. T5500

I realize that the MP's can be used as servers, but since they're not rack-mountable/blade solutions, and the slots cannot be configured for 4* 8x lane slots (highly useful in this segment as you well know), they're not really comparable IMO.

As per multiple workstation boards (same CPU family/socket, so only features to distinguish them) offered by system vendors, I'm not seeing it as much these days (sort of with HP's offerings, but I'm noticing that's mostly SP vs. DP boards and case formats, and there are consumer grade CPU's in some, such as the Z200). Primarily it seems, this is more the case with retail boards off the top of my head.

Seems just a matter of perspective as to how we're both looking at this I suspect.

Competitors are going to come in and prune off potential Mac Pro customers with updated i7 base boxes for which Apple will have no answer for. It is already the case that the more price sensitive customers are moving that way. An extended delay on E5's will only add to the effect.

Definitely true, as not every workstation user actually requires Xeons and ECC memory. Combine this with the current economy, purchasers are examining their needs more carefully, and purchase accordingly. Hence some are abandoning Xeons for consumer CPU based systems (implications for any vendor, but they at least have offerings to pick up those buyers that Apple doesn't).

But it's been Apple's choice to stay out of this market (they don't want cannibalization of the iMac and lower margins to compete with i7's on price).

Apple has to wait for the E5 and PCI-e v3.0 but with money to invest and fewer cards to test it shouldn't be hard to shake out the minor glitches earlier than the other vendors can. They can easily collect enough information to allow them to at least pre-announce in order to freeze customers (and perhaps collect some money that would expire on the quarter end: cash checks on 12/31 and ship very early in Jan. )

In terms of working out the kinks across both workstations and servers, PC vendors will have more work ahead of them.

But if they have the human resources to have concurrent validation teams, they could mitigate this to a good extent, and be able to meet a similar release date as Apple, assuming there's no preferential delivery schedule in place between vendors and Intel.

 

[Marcush1286]

I too want to see another PowerPC based Mac like before, except this time with Power 7 or Power 8 when it comes out. Who knows? Maybe Power might be an alternative for Tim to look into. As Ward mentioned, we need to steer away from these intel boxes.

Amen brother, Tim needs to steer the company back to making powerful, professional level computers and an OS designed for them.

 

[thunng8]

Power795 -- http://www-03.ibm.com/systems/power/hardware/795/specs.html

Its a beast alright .. way overkill for a mac though. Even if it could run Mac OS X, I don't believe any current mac application would take advantage of 256 cores.

 

[goMac]

I too want to see another PowerPC based Mac like before, except this time with Power 7 or Power 8 when it comes out. Who knows? Maybe Power might be an alternative for Tim to look into. As Ward mentioned, we need to steer away from these intel boxes.

Sigh.

Power 7 and Power 8 are way way more expensive than Xeons (they start around $7k) and they're unsuitable for desktop use. This would kill the Mac Pro.

Not to mention there are software issues. There aren't many universal apps left.

 

[Inconsequential]

I too want to see another PowerPC based Mac like before, except this time with Power 7 or Power 8 when it comes out. Who knows? Maybe Power might be an alternative for Tim to look into. As Ward mentioned, we need to steer away from these intel boxes.

No thanks.

Power7/8 is completely wrong and way way too expensive.

 

[xgman]

No thanks.

Power7/8 is completely wrong and way way too expensive.

Agreed, and it doesn't really fit Apple's current business model anyway. We will be lucky to hold on to the MP for a couple more years unless something changes.

 

[wallysb01]

No thanks.

Power7/8 is completely wrong and way way too expensive.

Seriously, at this point you're competing with clusters. I'd rather just buy hours at cluster that supports X11 forwarding or VNC, if you actually need a GUI and ~256 cores with ~1TB of RAM.

 

[philipma1957]

Agreed, and it doesn't really fit Apple's current business model anyway. We will be lucky to hold on to the MP for a couple more years unless something changes.

this is pretty much the truth. a top of the line iMac with 32gb ram can be built for under 3300. this machine will beat out many mac pros. not the dp but most of the sp models. this is a real problem for the mac pro.

 

[goMac]

this is pretty much the truth. a top of the line iMac with 32gb ram can be built for under 3300. this machine will beat out many mac pros. not the dp but most of the sp models. this is a real problem for the mac pro.

They need to start selling the SPs with Core i7 processors. They could really drive down the prices if they did that.

Because the monitor is still a separate component (making it more expensive), it wouldn't necessarily cannibalize the iMac, but it would make the Mac Pro more competitive.

We're not going to see $1k Mac Pros, but they need to nail down the high $1000/low $2000 range.

 

[Inconsequential]

They need to start selling the SPs with Core i7 processors. They could really drive down the prices if they did that.

Because the monitor is still a separate component (making it more expensive), it wouldn't necessarily cannibalize the iMac, but it would make the Mac Pro more competitive.

We're not going to see $1k Mac Pros, but they need to nail down the high $1000/low $2000 range.

Your forgetting that the i7 and the Xeon equivalent have been pretty much the same price, for like, ever.

On release the 980X and the W3680 were both... $999.
Currently you can pick the 980X and the W3680 new for £450 give or take.

Same applies to my i7 920/W3520, unit cost for them both were the same and are still now the same.

The SB-E chips:
E5-1660/3960X: $1088/999
E5-1650/3930X $583
E5-1620/3820: $294

So changing the CPU.. doesn't help.

The chances of a i7 Mac Pro, are zero IMO.

 

[goMac]

Your forgetting that the i7 and the Xeon equivalent have been pretty much the same price, for like, ever.

On release the 980X and the W3680 were both... $999.
Currently you can pick the 980X and the W3680 new for £450 give or take.

Same applies to my i7 920/W3520, unit cost for them both were the same and are still now the same.

The SB-E chips:
E5-1660/3960X: $1088/999
E5-1650/3930X $583
E5-1620/3820: $294

So changing the CPU.. doesn't help.

The chances of a i7 Mac Pro, are zero IMO.

Right, but those are i7 EEs. A more standard 3.4 ghz Sandy Bridge i7 is $299.00

http://www.newegg.com/Product/Product.aspx?Item=N82E16819115071

(Unless I missed something and only the EE parts work in a Mac Pro.)

The problem is really the Xeon loses it's value when it's not in a dual processor config. This put's the Mac Pro in this weird spot where the single processor models are a bad deal but the dual processor models are a good deal.

 

[nanofrog]

They need to start selling the SPs with Core i7 processors. They could really drive down the prices if they did that.

Because the monitor is still a separate component (making it more expensive), it wouldn't necessarily cannibalize the iMac, but it would make the Mac Pro more competitive.

We're not going to see $1k Mac Pros, but they need to nail down the high $1000/low $2000 range.

Concorde Rules is right (keeping the socket the same, the cost for the i7 vs. Xeon for the same core count and clock is the same).

Now they could reduce it by using a consumer grade socket (LGA1155), but there will be compromises involved.

1. Reduced max core count (up to 4x only; no DP models either).
2. Memory capacity would be an issue for some (16GB is the realistic limit due to costs; more is possible, but would be prohibitively expensive).
3. Reduced PCIe lanes.

There are other non-technical issues that could come into question as well, such as cannibalization of other systems (don't forget the Mini here).

Right, but those are i7 EEs. A more standard 3.4 ghz Sandy Bridge i7 is $299.00

http://www.newegg.com/Product/Product.aspx?Item=N82E16819115071

(Unless I missed something and only the EE parts work in a Mac Pro.)

The problem is really the Xeon loses it's value when it's not in a dual processor config. This put's the Mac Pro in this weird spot where the single processor models are a bad deal but the dual processor models are a good deal.

It's a different socket (LGA1155, not LGA1366), which is why it's cheaper (more per wafer).

Using such a part would cut costs, but it would also cut performance for those that need high speed I/O and larger than say 16GB of RAM in a workstation, even if ECC is useless to them (fewer PCIe lanes, past 16GB is possible, but very expensive <means using 8 or 16GB UDIMM's operating in non-ECC mode>).

 

[Eidorian]

Concorde Rules is right (keeping the socket the same, the cost for the i7 vs. Xeon for the same core count and clock is the same).

Now they could reduce it by using a consumer grade socket (LGA1155), but there will be compromises involved.

1. Reduced max core count (up to 4x only; no DP models either).
2. Memory capacity would be an issue for some (16GB is the realistic limit due to costs; more is possible, but would be prohibitively expensive).
3. Reduced PCIe lanes.

There are other non-technical issues that could come into question as well, such as cannibalization of other systems (don't forget the Mini here).

It's a different socket (LGA1155, not LGA1366), which is why it's cheaper (more per wafer).

Using such a part would cut costs, but it would also cut performance for those that need high speed I/O and larger than say 16GB of RAM in a workstation, even if ECC is useless to them (fewer PCIe lanes, past 16GB is possible, but very expensive <means using 8 or 16GB UDIMM's operating in non-ECC mode>).

On the personal enthusiast builder side of things, getting Xeon over Core i7 is a few dollars. Xeon UP matches up nearly identically to what Intel offers in the equivalent Core i7 socket. What I have noticed is the dramatic increase in motherboard pricing just for an appropriate chipset and/or BIOS microcode support. Though the ASUS's "Workstation" Series in the $200-300 range have made a wide feature set and Xeon support much more palatable to the masses.

I have not checked up on ECC DDR3 prices since I was looking at FB-DIMMs for a possible Mac Pro prices.

With the upcoming X79/LGA 2011 platform, I am hearing talk of RAM disks once more with the availability of cheap RAM and 4/8 slots to populate. $200 for 32 GB of RAM is not out of the picture.

Apple will more than likely continue their usage of a daughterboard for the processors, introduced with Nehalem, to simplify production. You will not have a free unused socket as in the LGA 711 days under Core 2 based Xeons.

 

[Inconsequential]

Right, but those are i7 EEs. A more standard 3.4 ghz Sandy Bridge i7 is $299.00

http://www.newegg.com/Product/Product.aspx?Item=N82E16819115071

(Unless I missed something and only the EE parts work in a Mac Pro.)

The problem is really the Xeon loses it's value when it's not in a dual processor config. This put's the Mac Pro in this weird spot where the single processor models are a bad deal but the dual processor models are a good deal.

The Core i7 920 was about $299 on launch give or take as far as I remember.

The i7-2600K is the wrong socket.

Personally I don't care that the SP mac is so much.

I need an expandable mac and the DP system is too rich for me, so the SP system it is.

 

[cragmr]

A "low end" MP based on LGA1155 would be great, but will not happen. Sad too, because I know a ton of people (me included!) who are interested in a powerful tower mac with a good price point (Mini isn't powerful enough, iMac has a screen attached to it), but if the refreshed MP still uses expensive "server" grade components a hackintosh is looking to be the solution.

 

[robertbrian23]

A "low end" MP based on LGA1155 would be great, but will not happen. Sad too, because I know a ton of people (me included!) who are interested in a powerful tower mac with a good price point (Mini isn't powerful enough, iMac has a screen attached to it), but if the refreshed MP still uses expensive "server" grade components a hackintosh is looking to be the solution.

in the near future, apple would do it.

i'm also on the same boat. want a Mac Pro without the server components. iMac specs with great GPU on a tower Mac would be awesome.

 

[thunng8]

Seriously, at this point you're competing with clusters. I'd rather just buy hours at cluster that supports X11 forwarding or VNC, if you actually need a GUI and ~256 cores with ~1TB of RAM.

That's the wrong answer. There are many applications that require a large SMP server.

----------

Sigh.

Power 7 and Power 8 are way way more expensive than Xeons (they start around $7k)

A few misconceptions here. You can buy an entire 4 core p710 Express server for 5.85k.

and they're unsuitable for desktop use. This would kill the Mac Pro.

No reason why they couldn't be used on the desktop apart from price. The lower end clocked Power7 chips (8 core 3.2Ghz or lower) don't use any more power than the Xeons used in the Mac Pro today and still provide higher performance.

It is all theoretical anyway. If IBM could sell the Power7 for desktop use with high volume, the price per CPU would go down dramatically. I don't see that happening. If anything, Apple will start use ARM CPUs on macs in the future

 

[zephonic]

No chips yet. Simple. We wait for Intel which was way better than waiting for Motorola

QFT

As an aside, I'm happy with my '09 W3520. I hardly ever run out of power often as it is, and with a CPU upgrade next year this machine should keep me going for quite a while.

 

[nanofrog]

On the personal enthusiast builder side of things, getting Xeon over Core i7 is a few dollars. Xeon UP matches up nearly identically to what Intel offers in the equivalent Core i7 socket. What I have noticed is the dramatic increase in motherboard pricing just for an appropriate chipset and/or BIOS microcode support. Though the ASUS's "Workstation" Series in the $200-300 range have made a wide feature set and Xeon support much more palatable to the masses.

In terms of retail pricing, the Xeons do go for a bit more due to a smaller supply (most Xeons are sold directly to system vendors).

In terms of SP boards, I don't really consider them expensive for what you get, particularly when you look at the feature sets vs. other boards.

BTW, I'm currently using an LGA1366 based ASUS workstation board.

I have not checked up on ECC DDR3 prices since I was looking at FB-DIMMs for a possible Mac Pro prices.

Last I checked, the difference between non-ECC and UDIMM for smaller capacities (1, 2, and 4GB sticks), wasn't that much (no where near the cost differences of DDR2 non-ECC vs. DDR2 FB-DIMM).

With the upcoming X79/LGA 2011 platform, I am hearing talk of RAM disks once more with the availability of cheap RAM and 4/8 slots to populate. $200 for 32 GB of RAM is not out of the picture. Maybe that's just me, but I don't mind not having to pay more for a Xeon workstation board than a gaming board.

Apple will more than likely continue their usage of a daughterboard for the processors, introduced with Nehalem, to simplify production. You will not have a free unused socket as in the LGA 711 days under Core 2 based Xeons.

Very likely IMO as well.

It's both cost effective, and adds additional PCB real estate to fit a pair of LGA2011 sockets.