Thunderbolt and MacPro....When?

[deconstruct60]

Surely it can't take that much effort to integrate one of these chips into a logic board design???

It does if you are already started and frozen the design for this product cycle. It isn't "hard" it is timing with the 10,000 other things that need to be aligned to get the board for that product out. Products appearing at the end of the year are well past the "freeze" date.

I also think vendors aren't too keen on shipping systems with zero peripherals to go with it. Reports are even the stuff that Intel showed isn't going to appear for months. Right now that is pragmatically a "display port" in terms of usefulness. Few vendors can sell the kool-aid Apple is pitching right now. "It a great port... look at that speed" .. "What can you plug in?.... well nothing.... but it is fast. " .

And the Intel kool-aid.

It is going to be optical ..... it isn't.
It is going to have 100m cable length ... it is 3m ( that 97m short of what talking about a year ago)
Only 7 devices ( what is this SCSI from 15 years ago ? SAS 10+K or USB with 100+ )
It is going to do a whole boatload of legacy protocols ... it only does two (PCI-e v2.0 and DisplayPort)
The connectors kind of look like USB ... they are display port
We have PCI-e cards working in lab in webvideo demos .... no PCI-e card in pipeline
There are going to be multiple optical transciever providers ... there is only one controller vendor.


Seriously would you jump on board with that? If I were I vendor I'd like to know what the freak it was really going to be before committing to putting it on board.

Apple doesn't really care. TB promotes the "you don't need slots" designs anyway. I think some higher end perhipheral vendors are willing to play along to because TB only socket devices are going to demand higher price points (than eSATA and/or USB 3.0 ). They are going to use the socket to segment the market. TB models will be below FibreChannel/Infinitband connectivity but above eSATA/USB 3.0 . It will take up the role/niche that FW had.

 

[deconstruct60]

But you can still get faster disk throughputs by using a PCI card (more than 8x lanes with a sufficient number of disks; whether it be DAS or SAN). NAS could exceed TB as well via port bonding (pair of 10G Ethernet ports for example), but cost wise, is nearing a SAN.

NAS and SAN ???? It is limited to 3m. Its got a shorter cable length than USB. It is a joke compared to Ethernet/FiberChannel/Infiniband right now for NAS/SAN applications. This is strictly a DAS solution right now.

Even if they get past the 3m limitation the 7 device limitation is also a huge blocker. I don't think optical is going to make a big dent in that problem if they make any attempt to keep up with the progressions of PCI-e (e.g., follow up to allow PCI-e v3.0 traffic )

 

[seek3r]

NAS and SAN ???? It is limited to 3m. Its got a shorter cable length than USB. It is a joke compared to Ethernet/FiberChannel/Infiniband right now for NAS/SAN applications. This is strictly a DAS solution right now.

Agreed, I speak from personal experience when I say it's difficult to saturate a QDR infiniband link for disk i/o alone

 

[goMac]

It is going to be optical ..... it isn't.

Optical is coming soon, and it's promised to work with the existing ports.

It is going to have 100m cable length ... it is 3m ( that 97m short of what talking about a year ago)

Which is a side effect of the switch from optical, but let's be honest... this was never supposed to be a replacement for fiber channel.

Only 7 devices ( what is this SCSI from 15 years ago ? SAS 10+K or USB with 100+ )

Non-issue, and again, never was supposed to be a replacement for fiber channel...

It is going to do a whole boatload of legacy protocols ... it only does two (PCI-e v2.0 and DisplayPort)

Do some research, it does do legacy protocols, in the exact same way as originally promised.

The connectors kind of look like USB ... they are display port

There were quite a few connector types demoed... Why exactly is this an issue?

We have PCI-e cards working in lab in webvideo demos .... no PCI-e card in pipeline

Valid complaint.

There are going to be multiple optical transciever providers ... there is only one controller vendor.

Errrr... The transceiver is a different component than the controller. We still have multiple transceiver vendors, and I'm pretty sure Intel hasn't closed the door to third party controller vendors.

I suspect that originally Apple was planning on using the USB ports, until the USB forum told them no. It probably required quickly moving to MDP, and as such, Intel may not have thought ahead for the issues of integrated video.

 

[Zeldain]

2012 when the new ones come out.

There is no way Apple is waiting until 2012 for a Mac Pro refresh. No way.... they may not have the SB yet, but we'll see something this year with Thunderbolt.

 

[Zeldain]

Even if a TB PCIe card was available tomorrow for the Mac Pro, would it matter?

It matters for those on the bubble of purchasing a Mac Pro system that don't want it to be a generation behind... it's not a big deal when it's a speed bump, or graphics upgrade, but when it's brand new tech it's a different story.

 

[philipma1957]

It matters for those on the bubble of purchasing a Mac Pro system that don't want it to be a generation behind... it's not a big deal when it's a speed bump, or graphics upgrade, but when it's brand new tech it's a different story.

You are spot on.

A lot of mac pro owners are like myself. An iMac with proper outboard drive connection would have been good enough.

A 16 gb ram iMac with a lacie raid0 T-Bolt as a boot drive and a 6 hdd promise T-bolt as storage raid5

with the interior hdd as backup boot would meet my needs. I own my mac pro because iMac does not offer this. So My next purchase may be the above or even a MBP with the 2 external hdds and a display from apple. Although the 8gb ram limit is not good enough for me as I do get past 10gb ram once in a while.

 

[Umbongo]

There is no way Apple is waiting until 2012 for a Mac Pro refresh. No way.... they may not have the SB yet, but we'll see something this year with Thunderbolt.

No we won't. They are not going to re-design their logic boards this year when new processors and chipsets are out in Q4. They have no reason to. All non-TB inventory would become redundant and no savvy buyer one would want one with Sandy Bridge Xeons around the corner.

 

[JesterJJZ]

Average Joes and Grandmas don't use MacPros...

 

[Zeldain]

Average Joes and Grandmas don't use MacPros...

Which is exactly why there needs to be something more often than 400 day refreshes on the Pro line.

 

[Zeldain]

No we won't. They are not going to re-design their logic boards this year when new processors and chipsets are out in Q4. They have no reason to. All non-TB inventory would become redundant and no savvy buyer one would want one with Sandy Bridge Xeons around the corner.

Nearly 12 more months is not "around the corner" and when you think about it, absolutely ludicrous that there aren't more mid-term refreshes for such a high-end, high-pricepoint machine.

 

[Hellhammer]

Average Joes and Grandmas don't use MacPros...

I didn't say a word about Mac Pro when I said what I said. Someone was asking when will we see PCs with TB and it was a reply to that, which then lead to some other discussion.

Why is it so damn hard to read the actual posts before posting?

 

[ozred]

Thunderclap is cursed.

 

[deconstruct60]

They are not going to re-design their logic boards this year when new processors and chipsets are out in Q4.

Q4 is this year. If the parts drop in Oct/Nov. they'll probably ship inside of this year. If the parts don't drop till Dec then probably not primarily wrong time of year to intro new products.

Xeon chips and supporting chipsets are shipped to system vendors long before they go to market. The testing on those systems is typically more rigorous then mainstream designs. Stuff like this SATA glitch in the recent Intel chipset get shaken out before they put it into production.


The two SandyBridge Xeons that aren't due till 2012 are a speciality one and the one targeted to 4 CPU package servers. The Mac Pro is workstation. The workstation versions are targeted in Q4.

http://www.cpu-world.com/news_2011/2011020803_Details_on_Intel_Xeon_E5_product_families.html

In 2008 and 2009 when Intel had new parts Apple dropped an update withing a month or so. Last year was an glitch on that timing due to some external factors ( busy with other product intros, including Mac ones) , need for some updates ( XServe being dropped), and Intel moving the schedule for this year. Unless Apple is waiting on some GPU card there shouldn't be a reason to wait. (and they shouldn't have to wait on that either. )

 

[Umbongo]

Nearly 12 more months is not "around the corner" and when you think about it, absolutely ludicrous that there aren't more mid-term refreshes for such a high-end, high-pricepoint machine.

Sure it is ludicrous, but we've been through all this with every release and nothing changes. Apple launch new Mac Pros (and pretty much every other system) when new processor/chipset combinations come from Intel.

Q4 is this year. If the parts drop in Oct/Nov. they'll probably ship inside of this year. If the parts don't drop till Dec then probably not primarily wrong time of year to intro new products.

I should have been clearer in my quote: We will not see Apple redesign their logic boards and release a Mac Pro based on LGA 1366 in order to include Thunderbolt this year. This is what the person I quoted seemed to be suggesting.

Xeon chips and supporting chipsets are shipped to system vendors long before they go to market. The testing on those systems is typically more rigorous then mainstream designs. Stuff like this SATA glitch in the recent Intel chipset get shaken out before they put it into production.

The two SandyBridge Xeons that aren't due till 2012 are a speciality one and the one targeted to 4 CPU package servers. The Mac Pro is workstation. The workstation versions are targeted in Q4.

http://www.cpu-world.com/news_2011/2011020803_Details_on_Intel_Xeon_E5_product_families.html

In 2008 and 2009 when Intel had new parts Apple dropped an update withing a month or so. Last year was an glitch on that timing due to some external factors ( busy with other product intros, including Mac ones) , need for some updates ( XServe being dropped), and Intel moving the schedule for this year. Unless Apple is waiting on some GPU card there shouldn't be a reason to wait. (and they shouldn't have to wait on that either. )

I hear what you are saying, but it just seems like being picky to me. I'm not going to deny that Apple could release a Mac Pro in 2011, but having followed all this for years and talked about it at length time after time I find it far more likely that it'll be an early 2012 release. There are so many factors that can push back a release, not least of all Apple being Apple, that it is safer to assume things won't come quickly and for anyone waiting to expect something in 2011 is likely to lead to disappointment.

 

[deconstruct60]

I think Thunderbolt is useful. However, it has been grossly oversold and overhyped.

Optical is coming soon, and it's promised to work with the existing ports.

This completely misses the point. If they were going to do two then announce going to do two. You don't do demos of stuff that isn't going to put into production and then try to get system vendors to jump on board. That's selling kool-aid.

The fact is this wasn't cooked and the economics didn't work out. So they are backtracking to something that has a better chance of working.

What Intel did was over promise and under deliver. If you are putting together a system and are trying to manage risk those are the kinds of people that don't make the cut.


Furthermore putting optical transceivers in both ends of the cable is only going to drive up costs. One factor why optical didn't arrive is costs. That isn't going to help. It is a ballon squeeze of the cost from one component to another (the devices into the cables).

Which is a side effect of the switch from optical, but let's be honest... this was never supposed to be a replacement for fiber channel.

You don't need to even compare to fiber channel. This is lower length limit than Firewire, USB, and SAS (longer lengths and more devices). About the only thing this does better than is eSATA.

The spin was that is was "better" than USB and Firewire. There are folks that wire up studio and put disk boxes at a distance from their computer. This currently solution doesn't work.

Do some research, it does do legacy protocols, in the exact same way as originally promised.

No it doesn't. It finally states the real plain truth that it only does PCI-e (and likely limited to v2.0) and Display Port data processing. There are two channels so in a sense doesn't even have to multiplex those two since there are 4 channels on the connector

The native controllers for the multitude of legacy protocols ( the FW controller , the USB controller, the SATA controller) handle the protocol and turn it into PCI-e . TB just carries the PCI-e. That's it.
Guess what those legacy controllers are coupled to legacy sockets. Which means this isn't the "one socket to rule them all".

Intel had slides like:

which only fed the notion that the technology "interoperated with everything". There is no handling of Ethernet. There is only handling after the Ethernet stream is turned into a PCI-e stream.

If they had simple slides that said what they say now, only does PCI-e and DisplayPort, then folks would know that it really is not a "connect to anything" socket. It is extend PCI-e and Display port socket and that "multiple" more precisely meant two.

There were quite a few connector types demoed... Why exactly is this an issue?

Because that is not representative of a standard. The functionality has to be designed into a system. You can't show up one week with a USB connector and then come back 6 weeks later with a ethernet connector and then 2 weeks after that with a display port connector. That is largely indicative that this tech isn't even fully baked.

Errrr... The transceiver is a different component than the controller.

I'm pointing out the misleading statements. Intel trotted out multiple transcievers when folks started questioning whether it is open and there would be multiple implementers. It was a fig leaf.

We still have multiple transceiver vendors,

Right now, no because there is no optical. Even those who were distracted by the fig leaf should be able to see the emperor wearing no clothes at this point.

Frankly, the "we got something working in the lab" doesn't carry lots of weight because that's exactly what they said last year..... and what do we have right now in terms of light energy... zip. So not buying it.

The other issue now is that the transceivers are going to get forced into the cables. Each cable needs two. I can see way those vendors probably still like the idea. I don't see how the peripheral vendors do since it isn't one but two they pragmatically have to buy. (unless switch to model were ship product with nothing to hook it up with. )

and I'm pretty sure Intel hasn't closed the door to third party controller vendors.

I'm sure they will let competitors who constantly run 6-12 months behind them to market to perpetrate that it is a standard. Adobe isn't blocking anyone from implementating Flash either and yet it got the "evil eye" in a Steve Jobs memo. At this point there is really no difference between the two.

It probably required quickly moving to MDP, and as such, Intel may not have thought ahead for the issues of integrated video.

I think you are giving them too much credit. Just about every demo revolved around video. There are lots of elements of TB now that the pulled back the curtain that were implemented in the current and previous versions of displayPort. I suspect this variation has been around for a while.

I think someone wanted to hijack USB. Whether it was Intel or Apple I'm not sure. Whoever is the "one socket to rule them all" advocate is probably behind that move. That plus it was cheap for Intel to trott out the old USB+fiber tech they already had.

 

[nanofrog]

NAS and SAN ???? It is limited to 3m. Its got a shorter cable length than USB. It is a joke compared to Ethernet/FiberChannel/Infiniband right now for NAS/SAN applications. This is strictly a DAS solution right now.

You misunderstood my post....

I was stating that a user can place a PCIe Fibre Channel, 10G Ethernet, or InfiniBand card in the system, and attach it to a NAS or SAN to blow Thunderbolt's throughputs out of the water. Distance too, but I didn't mention that (worst case is InfiniBand over copper <CX4 specification> = 15 meters). 10G Ethernet can do 100 meters over copper <6A>, and distances increase with optical of course (i.e. 10km for Single Mode Fiber).

Even for a DAS system, using a PCIe card can increase bandwidth and distance beyond what Thunderbolt is capable of (SAS entirely or FC/SAS configuration; SATA disks on SAS controllers can produce faster throughputs, but have a shorter distance <1.0 meters for passive signals>).

Hope this clears things up.

 

[deconstruct60]

Sure it is ludicrous, but we've been through all this with every release and nothing changes. Apple launch new Mac Pros (and pretty much every other system) when new processor/chipset combinations come from Intel.
......
I hear what you are saying, but it just seems like being picky to me. I'm not going to deny that Apple could release a Mac Pro in 2011,
.. but having followed all this for years and talked about it at length time after time I find it far more likely that it'll be an early 2012 release.

The logic here escapes me. Apple releases when Intel releases, but they won't for the Mac Pro.

The reason why Apple won't work on a redesigned LGA 1366 board is because they are already working toward a LGA 2011 (and perhaps LGA 1155 ) one. What was wrong with the guy's quote was that Apple doesn't have (or won't soon get) a Sandy Bridge Xeon to work with.


Yes is is more likely there will have been a new Mac Pro released by Feburary 2012 than by December 2011. However, is also more likely there will have been a new Mac Pro released by Feburary 2013 than the 2012 date.

that it is safer to assume things won't come quickly and for anyone waiting to expect something in 2011 is likely to lead to disappointment.

Not really. The Xeon release date used to be around early Q2. About a year ago Intel slipped the target to Q3. Now the estimates are Q4. It already has slipped. Twice!

Unless the other parts Apple targeted to that machine are also slipping it is more likely that two slips is enough so that the parts are around when they need them.

I expect there to be much wailing and moaning on these boards as late July comes and passes without a Mac Pro update. Then when Intel ships and it comes a chorus "I thought it was doom and would never get here".
[ Or a July/Aug update for the single package versions (E-3's because the E-5's cost too much) and a Oct/Nov update for the mulitple package version. That would mute the wailing and moaning. ]

 

[Umbongo]

The logic here escapes me. Apple releases when Intel releases, but they won't for the Mac Pro.

Point one: Apple don't release new Mac Pros before a new architecture is available to them. That is in regard to the idea of their being some interim Mac Pro update before Sandy Bridge ones come out.

Point two: I firmly believe, based on Intel, Apple and other system vendor history that an Intel Q4 processor release is unlikely to yield new Mac Pros before the end of the year.

 

[legreve]

There was a front page article in relations to thunderbolt. It stated what there well NO backwards compability hardware wise... no pci cards no converters. The only way you'd be able to use thunderbolt is to get a computer that has the tech build in....

 

[goMac]

This completely misses the point. If they were going to do two then announce going to do two.

...which is exactly what they did?

The fact is this wasn't cooked and the economics didn't work out. So they are backtracking to something that has a better chance of working.

So... you'd rather them release nothing at all? I don't get it. If optical isn't good for release right now, it's not good for release right now. It's either copper or nothing.

What Intel did was over promise and under deliver. If you are putting together a system and are trying to manage risk those are the kinds of people that don't make the cut.

Intel is horrible at engineering blah blah blah?

Furthermore putting optical transceivers in both ends of the cable is only going to drive up costs. One factor why optical didn't arrive is costs. That isn't going to help. It is a ballon squeeze of the cost from one component to another (the devices into the cables).

Optical transceivers in the cable was always the plan from day one... this isn't an engineering issue. Did you really think they were going to make an exposed optical connection?

You don't need to even compare to fiber channel. This is lower length limit than Firewire, USB, and SAS (longer lengths and more devices). About the only thing this does better than is eSATA.

Why do we care so much about the length again? It's three meters. This isn't meant for data centers or long runs, and never was...

The spin was that is was "better" than USB and Firewire. There are folks that wire up studio and put disk boxes at a distance from their computer. This currently solution doesn't work.

Pretty sure this is exactly why they announced long run optical?

Yes, you're going to have to wait for optical? I'm sorry?

No it doesn't. It finally states the real plain truth that it only does PCI-e (and likely limited to v2.0) and Display Port data processing. There are two channels so in a sense doesn't even have to multiplex those two since there are 4 channels on the connector

...I think you don't understand what you're talking about here at all. It's absolutely identical to what was originally promised.

The native controllers for the multitude of legacy protocols ( the FW controller , the USB controller, the SATA controller) handle the protocol and turn it into PCI-e . TB just carries the PCI-e. That's it.

This is exactly what Intel originally promised.

Guess what those legacy controllers are coupled to legacy sockets. Which means this isn't the "one socket to rule them all".

Intel was never ever planning on building in legacy controllers into TB. Not only that, but this would be a ridiculously dumb idea.

Intel had slides like:

which only fed the notion that the technology "interoperated with everything". There is no handling of Ethernet. There is only handling after the Ethernet stream is turned into a PCI-e stream.

I'm sorry there is no ethernet? But it doesn't matter because ethernet can be sent over PCI-E?

I'm really not sure what the deal is here.

If they had simple slides that said what they say now, only does PCI-e and DisplayPort, then folks would know that it really is not a "connect to anything" socket. It is extend PCI-e and Display port socket and that "multiple" more precisely meant two.

Again, I don't think you know what you're talking about... PCI-E makes perfect sense for a "talk to anything" protocol, and again, is exactly what was promised from day 1.

Because that is not representative of a standard. The functionality has to be designed into a system. You can't show up one week with a USB connector and then come back 6 weeks later with a ethernet connector and then 2 weeks after that with a display port connector. That is largely indicative that this tech isn't even fully baked.

No, that's engineering.

I'm pointing out the misleading statements. Intel trotted out multiple transcievers when folks started questioning whether it is open and there would be multiple implementers. It was a fig leaf.

I don't think you understand the difference between transceivers and the chipset. The TB transceiver market is totally open.

Frankly, the "we got something working in the lab" doesn't carry lots of weight because that's exactly what they said last year..... and what do we have right now in terms of light energy... zip. So not buying it.

...light energy?

The other issue now is that the transceivers are going to get forced into the cables.

Which is the way it's always been from day 1 when Intel announced Light Peak, but go on....

Each cable needs two.

Needs only one for copper if the device on the other end isn't TB...

I can see way those vendors probably still like the idea. I don't see how the peripheral vendors do since it isn't one but two they pragmatically have to buy. (unless switch to model were ship product with nothing to hook it up with. )

I'm not entirely sure what you're talking about. Tons of cables have transceivers in them.

I'm sure they will let competitors who constantly run 6-12 months behind them to market to perpetrate that it is a standard. Adobe isn't blocking anyone from implementating Flash either and yet it got the "evil eye" in a Steve Jobs memo. At this point there is really no difference between the two.

...ok?

I think you are giving them too much credit. Just about every demo revolved around video. There are lots of elements of TB now that the pulled back the curtain that were implemented in the current and previous versions of displayPort. I suspect this variation has been around for a while.

...yeah, the video demos were Light Peak talking to a video card, not directly to display hardware. That's a little different.

I think someone wanted to hijack USB. Whether it was Intel or Apple I'm not sure. Whoever is the "one socket to rule them all" advocate is probably behind that move. That plus it was cheap for Intel to trott out the old USB+fiber tech they already had.

Yes. Intel wanted to hijack the standard they invented. Woe is us.

I think you need to do some more research on TB. You have some fundamental misunderstands on it's design.