MP 7.1 why not PCIe4 or 5???

[DoFoT9]

there you go some time around 2021 PCIE 5 will be on intel so you can wait till 2022? maybe for PCIE 5 on apple

LGA4677 Socket Spotted: Future Server Socket with PCIe 5.0 Support

www.anandtech.com

can we end the topic now ?

Thanks for sharing. I'm on the market for a new machine, but personally really have no requirement for even PCIe4 let alone whatever 5 will be?

The only thing really of interest to me is having more ports that act faster overall.

 

[sirio76]

personally really have no requirement for even PCIe4 let alone whatever 5 will be?

It's not only you, 99.9% user doesn't need PCIe4 or 5, it's just a perfect argument for troll though.

 

[DoFoT9]

It's not only you, 99.9% user doesn't need PCIe4 or 5, it's just a perfect argument for troll though.

Sometimes, but I would like to think not. For example, my current rig is coming up to 10 years old. There is now technology that is faster than PCIe3.0 on my motherboard, so if I'm building for the long term, I prefer to have best in class available for me at the time, granted at a higher price upon initial release.

 

[throAU]

Apple really need to start making their own CPUs (or at least start shipping AMD if they don't have anything in the pipe for high end) ASAP. This is just another example of intel holding them back. Just like the later PowerPC days - they are stuck with a CPU vendor who is behind the times and having problems with keeping up both from a supply side and technology side. Intel are entirely outclassed in this space and have nothing in the pipe to fix that for the next couple of years.

And anyone claiming PCIe4 doesn't matter is just either blind or shilling for intel. Storage throughput matters in a lot of pro applications (and even on an end user desktop it improves responsiveness significantly), and PCIe 4 is much faster.

Existing PCIe 3.x based machines simply do not have enough lanes to dedicate sufficient throughput to m.2 any more.

 

[DoFoT9]

Apple really need to start making their own CPUs (or at least start shipping AMD if they don't have anything in the pipe for high end) ASAP. This is just another example of intel holding them back. Just like the later PowerPC days - they are stuck with a CPU vendor who is behind the times and having problems with keeping up both from a supply side and technology side. Intel are entirely outclassed in this space and have nothing in the pipe to fix that for the next couple of years.

I would like to think that they're already doing this, we have seen rumours of a MBA/MBP Apple chipset laptop for a few years now. It would take a lot longer to get competitive with a desktop "Pro" machine, and a lot of coin.

And anyone claiming PCIe4 doesn't matter is just either blind or shilling for intel. Storage throughput matters in a lot of pro applications (and even on an end user desktop it improves responsiveness significantly), and PCIe 4 is much faster.

Existing PCIe 3.x based machines simply do not have enough lanes to dedicate sufficient throughput to m.2 any more.

Not arguing the speed difference theoretically. Current PCIe4 SSDs are about 40% faster than their PCIe3 counterparts. Lanes is a good point, but a 16x PCIe3 lane can hit about 15,800MBs. This is far faster than any SSD can operate today. NVMe as I understand it requires 4 lanes to operate, so theoretical throughput should be doubled moving to PCIe4, but I think we're a ways away from maxing out ~7,880MBs 4x PCIe4.

I mostly feel it's other advancements such as reduced latency, higher throughput per lane, power reductions, etc that more so make the difference moving forward.

 

[throAU]

Lanes is a good point, but a 16x PCIe3 lane can hit about 15,800MBs. This is far faster than any SSD can operate today

Thats all well and good if you are willing to dedicate 16 of your (limited quantity of) lanes to a single SSD. A proper pro machine will contain multiple high speed SSDs, plus multiple graphics cards, 10, 40 or 100 gig ethernet, etc.

PCIe 3.0 is a limiting factor. PCIe 4.0 is less so.

 

[startergo]

In the same line of thoughts how the 7.1 will compare to the quantum computer which is already on the horizon?

 

[DoFoT9]

In the same line of thoughts how the 7.1 will compare to the quantum computer which is already on the horizon?

Hard no I’m afraid.

 

[crjackson2134]

In the same line of thoughts how the 7.1 will compare to the quantum computer which is already on the horizon?

Ha Ha... We could call it a Qmac...

There is so much work that needs to done before quantum computing becomes common place, it's not funny. But it's cool to ponder...

 

[t0mat0]

So what does Apple do now?
Say it takes at least 2 years for the design of a Mac Pro.
If they’re to release the Apple Silicon Mac Pro, it’s in design by now for release in next 2 years. So any update between now and then is already in the can?
Would they bother to change the socket knowing they’re going to something else for the next version in 2 years? To put it another way - did Apple know of their intentions and plans for the 2021 Mac Pro when they were making the current one, making the design future proof (this time...)

 

[Joe The Dragon]

So what does Apple do now?
Say it takes at least 2 years for the design of a Mac Pro.
If they’re to release the Apple Silicon Mac Pro, it’s in design by now for release in next 2 years. So any update between now and then is already in the can?
Would they bother to change the socket knowing they’re going to something else for the next version in 2 years? To put it another way - did Apple know of their intentions and plans for the 2021 Mac Pro when they were making the current one, making the design future proof (this time...)

CPU sockets change all the time. Maybe the 2018 plan was Intel with sockets / chipset changes as needed. Like how supermirco does it.

 

[deconstruct60]

So what does Apple do now?
Say it takes at least 2 years for the design of a Mac Pro.

Huge presumption that Apple "has to" start the next Mac Pro design when finish and ship whatever one was in flight.

Apple went from 2010-ish to 2013 . That is more than a 2 year break. Likewise. 2013 to 2019. That is a six year break. That is an even longer "take a pause before starting on the next one". Apple is on track to letting the iMac Pro languish for 3 years. At the high end of the Mac product line up Apple isn't on some "quick tempo" allocation of resources. Haven't been for very long time. I suspect people are going to be very deeply disappointed with the expectation that Apple Silicon is going to radically change that tempo. It probably is not going to do much at all.


If Apple had intended a late 2018 - early 2019 time frame ( and it slipped to late 2019). Then a "wait one year and spend two years" play would have put the next Mac Pro in the late 2021 - early 2022 time frame. If they were actually trying to shoot for 2H 2019 then even more solidly into 2022.
Part of that "wait one year" before starting is because there are higher priority Macs in the product development queue. And Apple gets to the top end ( much lower volume ) when they have "time".


Doing a Mac Pro appropriate "Apple SoC" is radically different from the I/O objectives that they are done so far. It would be quite prudent for Apple to add some 'pad' time wise to allow sufficient time to get something ready for this space. If Apple is committed to doing something for Mac Pro that is "competitive enough" there isn't a deep seated need for a stop gap. ( Again getting high tempo A-series chips out the door each year is a bigger priority for Apple. )

There are far more than just "CPU socket" upgrades Apple can do with the 2019 chassis. Apple was selling pragmatically 6 year old MP 2013 systems right up until the week or so before shipping the 2019 model. As long as there is a big enough trickle of folks still buying the 2019 units they can probably have a plan to keep selling them substantially into 2022.

If they’re to release the Apple Silicon Mac Pro, it’s in design by now for release in next 2 years. So any update between now and then is already in the can?

There doesn't have to be any CPU upgrades between now and then. Apple has done it before. From the consumer side it could be "nice" if there was an upgrade. But necessary for Apple given their Mac Pro track record for last decade. Nor necessary if look at the overall Mac revenues. ( it isn't like the Mac division is going to stop being profitable if the Mac Pro sales sag over next 1-2 years ).

Would they bother to change the socket knowing they’re going to something else for the next version in 2 years? To put it another way - did Apple know of their intentions and plans for the 2021 Mac Pro when they were making the current one, making the design future proof (this time...)

Again a substantive "leap" to presume that there was even a very long term plan for a 2021 Mac Pro.

Actually the current Mac Pro 2019 design gave Apple options if the early design feedback for the "Mac Pro SoC" didn't work out. Plan B could have been to move to another socket in 2021-2022 time period if they needed more time to work it out.

If Apple had been super-hyper confident about radically lowering TDP then might have stuck with something closer to the MP 2013 in size. They wouldn't have necessarily needed max wall socket power. The 64 (or more) lanes to provision 8 slots is a bit of opposite driver that Apple SoCs have. In many other cases, the Apple SoC is trying to get the system out of the corner that the industrial design put them in ( ever thinner case , more stuff with less battery volume , more complicated authentication (whole face as opposed to a fingerprint ) computations. etc. ). For the Mac Pro 2019 case there is more of a "Pull" of the Apple SoC that Apple has deliberately avoided. Lots of stuff outside of the SoC that requires connections. (relatively) Lots of high speed ports. Extremely high cross section bandwidth. etc.