Why doesn't Intel and AMD Integrate Gigabit Ethernet?

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

SocketF

Senior member
Jun 2, 2006
236
0
71

Lol ...
Recommended Customer Price TRAY: $149.79
I guess it is doable if you can sell products for a high price like that. But for the normal low-cost market, I strongly doubt it.

But there is still the chance, that they didnt shrunk the PHY-part and just use 65nm sizes there. That would be something similar that GF and TSMC will do officially with their 14XM and 16nm processes, where the back and will be identical to 20mn.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
Lol ...
I guess it is doable if you can sell products for a high price like that. But for the normal low-cost market, I strongly doubt it.

But there is still the chance, that they didnt shrunk the PHY-part and just use 65nm sizes there. That would be something similar that GF and TSMC will do officially with their 14XM and 16nm processes, where the back and will be identical to 20mn.

You are aware thats 10Gbit controllers?
 

JimKiler

Diamond Member
Oct 10, 2002
3,559
205
106
The mobo makers are doing this integration already and it does not make sense to integrate with the CPU since you need a port on the mobo not on the CPU to plug your Cat 5 or Cat 6 cable into.

I bet i still have a couple old Intel NIC where you could offload duties to the NIC instead of the CPU. Newfangled tech back in the day.
 

tynopik

Diamond Member
Aug 10, 2004
5,245
500
126
Maybe I was slightly confused that day, but I thought I read that gigabit ethernet is already included in (nearly) every Intel chipset.

The thing is board makers have to pay to license it to actually use it, so even though it's already there, it's cheaper to add a 3rd party controller chip and not pay Intel.
 

IntelEnthusiast

Intel Representative
Feb 10, 2011
582
2
0
I thought we got enough complaints about changing our sockets already but if we were to intergate the NIC on to the processor we would have to change the socket again.
 

tynopik

Diamond Member
Aug 10, 2004
5,245
500
126
I thought we got enough complaints about changing our sockets already but if we were to intergate the NIC on to the processor we would have to change the socket again.

1. As you said, you change socket frequently enough that it shouldn't be an issue to slip into upcoming socket change.

2. The original question asked about cpu OR chipset. Changing the chipset wouldn't require a new socket.

3. My understanding is that it already is included in the chipset, it's just disabled by many manufacturers so they don't have to pay Intel a licensing fee for it.
 

sm625

Diamond Member
May 6, 2011
8,172
137
106
I bet the trace length from the RJ45 to the PHY can only be a certain distance. So there would need to be some sort of IC between the RJ45 and the cpu to hold the signal to proper specs. So it is one of those cases where it is cheaper and simpler to just put the PHY out by the RJ45 and then have it talk to the cpu over a bus that is designed to be laid out on a pcb.
 

VirtualLarry

No Lifer
Aug 25, 2001
56,452
10,120
126
I bet the trace length from the RJ45 to the PHY can only be a certain distance. So there would need to be some sort of IC between the RJ45 and the cpu to hold the signal to proper specs. So it is one of those cases where it is cheaper and simpler to just put the PHY out by the RJ45 and then have it talk to the cpu over a bus that is designed to be laid out on a pcb.

There's probably some truth to that, but were people in this thread actually suggesting to integrate the analog PHY chip with the CPU, or just the controller? It would make sense to integrate the controller into the PCH chip. The PHY should be a seperate chip like you suggest.

NV integrated an ethernet controller into their NFORCE platforms, did they not?
 

Kougar

Senior member
Apr 25, 2002
398
1
76
Maybe it's not worth the silicon real-estate? Or the cost/complexity of micro-managing additional CPU models / fabbing silicon with and without it built in.

How many 1Gbit ports does a 1P board need? 2 maybe? So looking at this from purely the server market... then what about a 2P or 4P server? That'd mean 8 1Gbit ports on a single 4P server board, meaning most would just be completely wasted.

With that in mind I'm guessing it's probably just not worth the silicon space + power consumption to add it when it's so much simpler all around to keep it built into the chipset. It'd only be in scenarios like the Atom where it's always a 1P system and only one port would ever be needed that it'd start making sense to do it.
 

Rifter

Lifer
Oct 9, 1999
11,522
751
126
If intel built in a Intel NIC into the CPU or NB I would be sold on Intel for life and never consider another AMD system, I think their NIC's are that good.
 

SocketF

Senior member
Jun 2, 2006
236
0
71
There's probably some truth to that, but were people in this thread actually suggesting to integrate the analog PHY chip with the CPU, or just the controller? It would make sense to integrate the controller into the PCH chip. The PHY should be a seperate chip like you suggest.
Yes but where is the benefit of it, if you have to connect MAC and PHY anyways with something similar to PCIe x1? Then you could also already connect an integrated MAC+PHY networkchip with PCIe x1.
NV integrated an ethernet controller into their NFORCE platforms, did they not?
Only the MAC, not the PHY:
Gigabit Ethernet — The price of Gigabit networking hardware is falling at a remarkable rate, making 1000Mbps networks an affordable reality for businesses and even bandwidth-hungry enthusiasts. Gigabit Ethernet is nothing new in the motherboard world, where GigE chips have been riding the PCI bus and Intel's CSA north bridge link for some time. However, the nForce3 250GB's Gigabit MAC resides directly on the chip where it has unfettered access to the chipset's HyperTransport bus. Although the nForce3 250Gb's Gigabit MAC is integrated into the chipset, it uses a PHY chip to act as an Ethernet transceiver. The 250Gb's GigE implementation supports TCP and checksum offloads to reduce CPU utilization, plus jumbo frames to improve throughput by reducing frame overhead.
http://techreport.com/review/6547/nvidia-nforce3-250gb-chipset
 

Ayah

Platinum Member
Jan 1, 2006
2,512
1
81
That guy went through a herculean effort to debug and characterize that bug. I bet he will be scarred for life from the mental trauma, I would be!

It's already proven to be a vendor EEPROM programming issue.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
Yes but that is of no importance to the thread's topic.

Yet you made a fool of the price thinking it was a 1Gbit PHY, without checking specs first that says its a 10Gbit. :awe:

The 40nm 1Gbit PHYs are dirt cheap.
 

coffeejunkee

Golden Member
Jul 31, 2010
1,153
0
0
Maybe I was slightly confused that day, but I thought I read that gigabit ethernet is already included in (nearly) every Intel chipset.

The thing is board makers have to pay to license it to actually use it, so even though it's already there, it's cheaper to add a 3rd party controller chip and not pay Intel.

That's what I thought too, at least the chipset diagrams always mention something about gigabit lan in the pch.
 

SocketF

Senior member
Jun 2, 2006
236
0
71
Yet you made a fool of the price thinking it was a 1Gbit PHY, without checking specs first that says its a 10Gbit. :awe:
No .. I told you already that I saw it. It is just not important.
In the chip industry, "impossible" often means just "too expensive" to do, without any benefits.

Now, if you can sell a network-chip for ~$150 the effort for an expensive and complex 40nm design could probably pay off. If that is a 10GbE Chip or a 10Mbit is of no concern, important is the price.
The 40nm 1Gbit PHYs are dirt cheap.
Do you have their prices, or are you just guessing?

I googled for it and was not successful. It doesnt matter anyways, these 1 GbE parts are very likely only some chips that didn't qualify as 10 GbE, i.e. the scrap parts. These are always cheaper. However, without the expensive 10GbE part the 40nm design wouldn't be profitable.

Edit:
Checked for NICs instead for the chips. NICs with i210 chips are not "cheap" either. They cost around US-$75:
http://www.compsource.com/ttechnote.asp?part_no=I210T1&vid=211&src=PW

At the same retailer, another intel GbE-NIC costs only $30:
http://www.compsource.com/pn/SDPEX2...-Ethernet-PciExpress-Network-Card-SDPEX24009/

And an realtek is again only half of that for $14:
http://www.compsource.com/pn/SDPEX2...-Ethernet-PciExpress-Network-Card-SDPEX24009/

It may be that the i210 has some special server features, but again, only the (high) retail price is of concern.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
No .. I told you already that I saw it. It is just not important.
In the chip industry, "impossible" often means just "too expensive" to do, without any benefits.

Now, if you can sell a network-chip for ~$150 the effort for an expensive and complex 40nm design could probably pay off. If that is a 10GbE Chip or a 10Mbit is of no concern, important is the price.
Do you have their prices, or are you just guessing?

I googled for it and was not successful. It doesnt matter anyways, these 1 GbE parts are very likely only some chips that didn't qualify as 10 GbE, i.e. the scrap parts. These are always cheaper. However, without the expensive 10GbE part the 40nm design wouldn't be profitable.

Edit:
Checked for NICs instead for the chips. NICs with i210 chips are not "cheap" either. They cost around US-$75:
http://www.compsource.com/ttechnote.asp?part_no=I210T1&vid=211&src=PW

At the same retailer, another intel GbE-NIC costs only $30:
http://www.compsource.com/pn/SDPEX2...-Ethernet-PciExpress-Network-Card-SDPEX24009/

And an realtek is again only half of that for $14:
http://www.compsource.com/pn/SDPEX2...-Ethernet-PciExpress-Network-Card-SDPEX24009/

It may be that the i210 has some special server features, but again, only the (high) retail price is of concern.

They cost between 2 and 4$. And no, they are not failed 10Gbit chips as you somehow make up from imagination. The 10Gbit chips are packaged in a 25x25mm package, the 1Gbit in 6x6mm.

Your links also show different products. Not to mention the Intel one is a server NIC and not a desktop one...

40nm 1Gbit PHY chips are also compatible with the 82579 that goes on mobos.
 
Last edited:
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |