News Realtek's $10 tiny 10GbE network adapter is coming to motherboards later this year

It kills me that this is such a chicken and an egg problem. Motherboard and AIC manufacturers complain that nobody is seeking out the higher end network options, and the reason nobody is doing so is because the rest of the infrastructure is so wildly overpriced as to be a non-starter.

So many people would gladly move up to 2.5GbE or 5GbE if not full 10GbE quickly if there was something closer to reasonable price parity to the current crop of hardware. But when a 5 port switch has such wild disparity, nobody can be blamed for the lack of "excitement" around faster networking technologies.

5 port switch costs:
1GbE switch is around $15
2.5GbE switch is $80-$140
10GbE is $300

Even taking the argument that 10GbE should be 10x the price of 1GbE, that should put the price at $150 for a 10GbE 5-port switch. Prices are so wildly skewed on switching gear as to make any home network speed upgrades unfeasible for the large majority of the populace.
 
If a 10GBe NIC is going to cost just $10 or even $20-$25 then there should be plenty of money left for Cat6 or 6a cabling. People seeking out 10GB are not going to mind the cost so much and in fact these low 10GB NIC costs may justify finally spending on a 10GB switch to upgrade the home or home lab.
 
  • Like
Reactions: MosephV
5 port switch costs:
1GbE switch is around $15
2.5GbE switch is $80-$140
10GbE is $300
A couple months ago, I got this 4x 2.5G + 2x 10G (SFP+) switch for $34 on Amazon. Works great!

Here's a version of the same switch, but with 2x 10GBaseT ports (instead of SFP+), currently $100:

ServeTheHome reviewed a 6x 10GBaseT switch they claim cost $195.

 
9mm x 9mm ?! Can't be correct. 9mm is 3/8" =maybe the size of the chip but not the board, for sure. the slot is bigger...
It would be better if y'all got rid of inches all together, archaic measurement based off a british king's finger segment length, because he was king=god, come on, get over it.
 
  • Like
Reactions: Mr Majestyk
9mm x 9mm ?! Can't be correct. 9mm is 3/8" =maybe the size of the chip but not the board, for sure. the slot is bigger...
It would be better if y'all got rid of inches all together, archaic measurement based off a british king's finger segment length, because he was king=god, come on, get over it.
Yes, the size of the chip. This is a motherboard targeted NIC. The 'card' is the whole motherboard.
 
  • Like
Reactions: gschoen and MosephV
Well, I suppose that depends on what your WAN speed is :) In our area there are already multi-gig home Internet options.
Even when you nominally have > 1 Gbps internet service, I think it's rare to see downloads faster than 1 Gbps, in actual practice.

And no, you can't trust those "speed test" sites, because I'm sure ISPs give them prioritized access, so that their customers think their internet speeds are better than they really are. I definitely get speeds from one of those sites that I never see, for anything else on the internet.
 
Regarding the edit: CAT6 has better shielding and signal integrity, but it and CAT5e are both using the same number of wires and the same connectors, and there’s no way for the cable to signal its capabilities to the connected devices. I’m almost certain I’ve heard of 10GBE over CAT5e in short runs/favorable environments before, even if the CAT5e spec can’t guarantee it over the maximum cable length and in hostile environments.
 
I’m almost certain I’ve heard of 10GBE over CAT5e in short runs/favorable environments before, even if the CAT5e spec can’t guarantee it over the maximum cable length and in hostile environments.
Reportedly, it works up to about 45m, though I wouldn't count on getting anywhere close to that range. I'd also keep an eye on the Ethernet-level stats, like number of CRC errors and TCP retransmits.

At home, every twisted-pair copper cable I've bought for probably at least a decade has been Cat 6 or 6A. For shorter lengths, the cost hasn't been too off-putting. Right now, I only have a single switch with a 10GBaseT port.

I also have two SFP+ NICs, I got cheap off ebay (because they require PCIe x8 slots). That cable is more expensive and limited to just 5m, if you opt for copper instead of fiber + transceivers.
 
And no, you can't trust those "speed test" sites, because I'm sure ISPs give them prioritized access, so that their customers think their internet speeds are better than they really are. I definitely get speeds from one of those sites that I never see, for anything else on the internet.
I think of those as a synthetic benchmark. If you are getting slow speeds downloading something they can be a quick sanity check to see if you have something wrong on your end or the other site is just slow.
 
Reportedly, it works up to about 45m, though I wouldn't count on getting anywhere close to that range. I'd also keep an eye on the Ethernet-level stats, like number of CRC errors and TCP retransmits.

At home, every twisted-pair copper cable I've bought for probably at least a decade has been Cat 6 or 6A. For shorter lengths, the cost hasn't been too off-putting. Right now, I only have a single switch with a 10GBaseT port.

I also have two SFP+ NICs, I got cheap off ebay (because they require PCIe x8 slots). That cable is more expensive and limited to just 5m, if you opt for copper instead of fiber + transceivers.
I always heard that about 33m was the max you could get Cat 5e to work at effectively for 10GbE. Technically Cat 6 isn't certified for 10GbE either as it cannot do it at 100m as is required for the specification.

For your DAC I find it interesting that you can only go 5m on a 10Gb DAC. You can get passive cables to 7m and active to 10m. Granted those are $40 and $100+ respectively.
 
I’m almost certain I’ve heard of 10GBE over CAT5e in short runs/favorable environments before, even if the CAT5e spec can’t guarantee it over the maximum cable length and in hostile environments.
This all comes down to what people using actual real world experience know works...well most the time... and what the official standards say. Everyone has known for a very long time normal Cat6 cable will pass 10gbit at distance close to 50 meters. It is just not "certified" to work.

What has happened is cat6a cable which is certified at 10gbit at 100meters has become cheaper. This though is more of a illusion. The key cost to ethernet cables is the copper metal used to make the cables. This has gone up massively more than the extra manufacturing costs to make cat6a. You just don't notice this extra manufacturing costs for cat6a cables when the costs of all cable has increased so much.

You might as well just use cables that are certified.
 
  • Like
Reactions: jlake3
This all comes down to what people using actual real world experience know works...well most the time... and what the official standards say. Everyone has known for a very long time normal Cat6 cable will pass 10gbit at distance close to 50 meters. It is just not "certified" to work.

What has happened is cat6a cable which is certified at 10gbit at 100meters has become cheaper. This though is more of a illusion. The key cost to ethernet cables is the copper metal used to make the cables. This has gone up massively more than the extra manufacturing costs to make cat6a. You just don't notice this extra manufacturing costs for cat6a cables when the costs of all cable has increased so much.

You might as well just use cables that are certified.
My point was mostly just that while “We got it working on CAT5e!” sounds impressive, RTL8127 hasn’t really been demonstrated to be doing anything special with that. If their hardware error correction could make 10G work and a NIC from a competitor had to negotiate down to something lower on the exact same cable that may be a demonstration that their chip is superior, but they didn’t do anything like that.