Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

N19h7m4r3

macrumors 65816
Original poster
Dec 15, 2012
1,191
8
This is a rather interesting, and sad move from Intel. They'll be restricting their PCIe lanes to only their highest-end offerings, while offering cut down versions on the others.

Intel's $500-$750 Core i7-5930K and >$1,000 Core i7-4960X offer bigger 40-lane PCI-Express Gen 3.0 root complexes; the Core i7-5820K features a narrower 28-lane one.
On motherboards with, say, three PCI-Express 3.0 x16 slots, the i7-5930K and i7-5960X will let you run two slots at full x16 bandwidth, and a third slot at x8. On systems with the i7-5820K, the second slot won't go beyond x8, and the third one will cap out at x4. On boards with four slots, one of them will run out of bandwidth.
http://www.techpowerup.com/204387/i...atures-fewer-pci-express-lanes-after-all.html

I wonder if Intel will be doing something similar with their low-highmid Xeon range, and if so it'll certainly affect people wanting to either build a new Hackintosh, and the new Mac Pro that's now focussing on Dual GPU's, and using the other lanes for Thunderbolt.
 

repentix

macrumors regular
May 26, 2013
205
2
If I would build a socket 2011 workstation, I would build it with a six core 5830k nothing else!
 

deconstruct60

macrumors G5
Mar 10, 2009
12,360
3,935
This is a rather interesting, and sad move from Intel. They'll be restricting their PCIe lanes to only their highest-end offerings, while offering cut down versions on the others.
....
I wonder if Intel will be doing something similar with their low-highmid Xeon range,

If this leak is correct, then Intel is not.

http://www.chiploco.com/intel-haswell-ep-e5-1600-v3-35072/

The 1620 and 1630 v3 just have 4 cores. What Intel is doing is kneecapping a sub $300 6c CPU. If it has stuck with 4c they wouldn't want to offset it from the overlap area with the desktop-mobile offerings in the 4c (E3 and others). The "kneecapping" is more so on limiting the turbo cap (which probably could be higher without much problem. The 1620 v2 appears to be have shaved off base clock to drop the price a bit lower too.

E5 2400 versus E5 2600 summary
http://www.supermicro.co.nz/content/view/262/39/
)

In the Desktop zone, this 28 lane limit is a step up from 16. ( 28 is a bit odd. the E5 2400 series ( -EN) models so far have differed from the E5 2600 series (-EP) models with only x24 and just 3 memory channels ). Perhaps in Haswell that gets bumped to 28. I would be surprised is Intel came up with a custom market segmenting mask solely for this specific low end i7. If they coupled it to the E5 2400 series constraints. ( IMHO the need for the 2400 series is a bit questionable. )


and the third one will cap out at x4. On boards with four slots, one of them will run out of bandwidth.

On most mainstream boards there are already slots that "run out of bandwidth". With mainstream processors at x16 with four slots there is some switch watering down the bandwidth.

and if so it'll certainly affect people wanting to either build a new Hackintosh, a

Most folks on the Hackintosh track have cheaper as a high objective. The single core drag racing will probably come out cheaper if just use something overclocked from the mainstream desktop range if want to "pass by" the 4c entry Mac Pro.

and the new Mac Pro that's now
focussing on Dual GPU's, and using the other lanes for Thunderbolt.

The E5 1400/2400 (-EN) never were good options for the Mac Pro (or most other workstations ). The seem to targeted at marginally cheaper servers with relatively modest I/O that want to deliver "utility" services through (e.g., embedded at a telco branch data center, etc. )
 

chrfr

macrumors G5
Jul 11, 2009
13,546
7,071
I wonder if Intel will be doing something similar with their low-highmid Xeon range, and if so it'll certainly affect people wanting to either build a new Hackintosh, and the new Mac Pro that's now focussing on Dual GPU's, and using the other lanes for Thunderbolt.

I'm sure whatever Intel is doing to the Xeons going into the Mac Pro is being done with approval from Apple. This is an obvious product differentiation strategy that won't affect Apple.
 

deconstruct60

macrumors G5
Mar 10, 2009
12,360
3,935
If I would build a socket 2011 workstation, I would build it with a six core 5830k nothing else!

the i7-5820K is aimed more so at gamers with larger than average, but limited budgets, than at workstation workloads.

x16 optionally split into 2 x8 PCI slots
x8
x4

total x28.

A three way x8 SLI/Crossfire set up with a optionally a high I/O storage card. for the x4 slot. Or two way x8 SLI/Crossfire and two high I/O storage card (e.g., PCIe SSD). The key issue were only pragmatically looking for x8 (used) cards any. ( three x8 cards is just x24 worth so x28 provides a bit of headroom still)

The approximately $150-200 gap between the the next step up will be thrown at cooling and overclocking infrastructure over time. As long as non linear pricing (bang for buck) between top end single x16 GPU cards and multiple GPUs then going "wider' with x8 effective cards will probably pay and some gap over desktop limited single x16 CPU package pay off for Intel.


Modern Workstations are generally in a different zone in that what multiple x16 paths available for max GPGPU and higher data transfer GPU loads.
 

deconstruct60

macrumors G5
Mar 10, 2009
12,360
3,935
I'm sure whatever Intel is doing to the Xeons going into the Mac Pro is being done with approval from Apple. This is an obvious product differentiation strategy that won't affect Apple.

You are correct that Apple would probably strongly object to loosing 12 lanes of PCIe v3. The Mac Pro design is based on a minimum of 40. But if Intel drew a hard line at 6c+40 lanes in the pricing structure, Apple doesn't have much of a choice. It wouldn't impact Apple deeply since they could just move up the price curve and pass along the increased costs to customers.

If the 1620 v3 is going cheaper (~$220 from ~$289) then perhaps at v4 (or beyond ) Intel will "chop off" the PCIe lanes ( and maybe get rid of -EN options by just doing a much smaller set of 'more limited' -EP ones. That would get rid of supporting yet another socket which with rise of their own Atom server line up and upcoming ARM competitors probably isn't buying much with. ). As long as there was a E5 1600 40 lane product in the $280-330 range Apple could probably still make it work with no major price changes.

Apple isn't the only customer though. If this i7 5820k is far more successful than the 1620 and 1630 v3 (and the -EN line deep erodes ) then there will be some tension that offset Apple being unhappy.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.