The automotive market is less about raw bandwidth and more about determinism in such a scenario. Being able to truly dictate bandwidth usage with predictable latency is a key capability of the Aquantia chips. For automotive this is critical considering the number of switches just in the topology outlined in the article.
Ideally you'd want to have all the devices going to a pair of redundant switches with each having their own dedicated runs. I presume that the automotive industry has some unique requirements for redundancy due to things like crashing and still keeping the maximum number of sensors operational.
Shame it wasn't bought by AMD. nVidia just acquired Mellanox.
Would have been cool to have 10g or 5g in the chipsets or have AMD push 5g/10g by discounting ethernet chips when bundled with chipsets or something like that.
Considering the changes to bring 2.5 Gbit to market, I see it becoming widespread rather quickly. 10 Gbit is already a well understood and deployed widely at the highend (servers, workstations etc.).
The odd ball is the 5 Gbit rate which recommends CAT6 over CAT5e cabling for any significant distances and increases the line rate over 1/2.5 Gbit. At this juncture, it'd make sense to simply add 10 Gbit support alongside the 5 Gbit rates from a NIC standpoint.
The real question is when we'll see inexpensive 2.5/5 Gbit capable switches for consumers. 10 Gbit units are just now trickling into the highend consumer side after being common place in the datacenter but carry a price premium. The hope is that 2.5 Gbit will quickly supplant 1 Gbit at the same price.
You don't really need anything special to do EtherCAT-- it's just 802.3 frames. Virtually all Ethernet NICs will do EtherCAT just fine. (Though Realtek's tend to have issues, but that's mostly a result of horrible drivers not the actual hardware.)
It puts an end to an era of 10GBase-T startup wars: Plato Networks, Solarflare (bought by Marvell too), Teranetics (bought by PLX and Aquantia), and Aquantia. I used to work for Teranetics/PLX/Aquantia... I was hoping they'd be the next Marvell/Broadcom with all the new ethernet money- this is what was always the goal but 10GBase-T was just too fast- hence going SLOWER to 5G and 2.5G.
Does anyone know if Aquantia ever licensed their 2.5/5G IP out to the likes of Marvell and Broadcom?
Saw this and instantly bought an Aquantia 10gb nic that I was hesitant on buying. I don't trust Marvell not to screw this up and do away affordable 10gb nics
I don't really feel automotive is really a trend to invest in. I mean sure i get that devices need to connect efficiently with all the electronics communicating. I really think these companies don't know the average human adult does not go out and buy a car that has a lot of electronics to begin with. Despite what it my seem online, most people are buying cheap cars from each other. lol
...like someone did a average price a person spends on a car and it was around $8k mark. I'm pretty sure not many of those have more than a car radio and maybe off the shelf add ons put in. lol
Autonomous cars have lots of sensors that need to communicate with low latency. That means fast in-car networking. I think the industry has decided that autonomous cars will eventually rule the world.
As for the pricing question, I think another industry assumption is that many people will not own vehicles, in the future. You'll simply summon a robo-Uber, anytime you want to go somewhere. The bonus is that you can transport an entire family to different locations, simultaneously. You can even have a robo-Uber pick up the kids from soccer practice, instead of having to go there and get them yourself. The downside will be the wait times at peak demand (but today, you'd just be sitting in traffic, which autonomous cars will greatly alleviate, once human drivers are eliminated).
I'm not saying it's a good future, but it appears to be a fait acompli. There will certainly be a lot more driving, so electricity generation must be from renewables or we're doubly-screwed.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
21 Comments
Back to Article
Kevin G - Monday, May 6, 2019 - link
The automotive market is less about raw bandwidth and more about determinism in such a scenario. Being able to truly dictate bandwidth usage with predictable latency is a key capability of the Aquantia chips. For automotive this is critical considering the number of switches just in the topology outlined in the article.Ideally you'd want to have all the devices going to a pair of redundant switches with each having their own dedicated runs. I presume that the automotive industry has some unique requirements for redundancy due to things like crashing and still keeping the maximum number of sensors operational.
Lord of the Bored - Monday, May 6, 2019 - link
Aaaaaand cheap multi-gigabit ethernet just died. Yay.Ian Cutress - Monday, May 6, 2019 - link
https://www.anandtech.com/show/13428/realtek-has-2...DigitalFreak - Monday, May 6, 2019 - link
No 5 or 10gig support.Lord of the Bored - Tuesday, May 7, 2019 - link
"Things we missed" is an appropriate headline, it seems.I stand pleasantly corrected.
Kevin G - Tuesday, May 7, 2019 - link
2.5 Gbit speeds are pretty much a freebie since the change between 2.5 and 1 Gbit is expanded symbol space at the same line rate.mariush - Monday, May 6, 2019 - link
Shame it wasn't bought by AMD. nVidia just acquired Mellanox.Would have been cool to have 10g or 5g in the chipsets or have AMD push 5g/10g by discounting ethernet chips when bundled with chipsets or something like that.
Kevin G - Tuesday, May 7, 2019 - link
Considering the changes to bring 2.5 Gbit to market, I see it becoming widespread rather quickly. 10 Gbit is already a well understood and deployed widely at the highend (servers, workstations etc.).The odd ball is the 5 Gbit rate which recommends CAT6 over CAT5e cabling for any significant distances and increases the line rate over 1/2.5 Gbit. At this juncture, it'd make sense to simply add 10 Gbit support alongside the 5 Gbit rates from a NIC standpoint.
The real question is when we'll see inexpensive 2.5/5 Gbit capable switches for consumers. 10 Gbit units are just now trickling into the highend consumer side after being common place in the datacenter but carry a price premium. The hope is that 2.5 Gbit will quickly supplant 1 Gbit at the same price.
tezcan - Saturday, May 11, 2019 - link
Mellanox is a system manufacturer- they buy chips from Aquantia (an IC manufacturer).basroil - Monday, May 6, 2019 - link
Ethernet? Count me out, they need ethercat at minimumbenzosaurus - Saturday, May 11, 2019 - link
You don't really need anything special to do EtherCAT-- it's just 802.3 frames. Virtually all Ethernet NICs will do EtherCAT just fine. (Though Realtek's tend to have issues, but that's mostly a result of horrible drivers not the actual hardware.)mryamaguchi - Monday, May 6, 2019 - link
Is this Marvell's end game?HardwareDufus - Monday, May 6, 2019 - link
I see what you just did there.ksec - Tuesday, May 7, 2019 - link
Nah, just the first part of Marvell Vs Broadcom.Kevin G - Tuesday, May 7, 2019 - link
Infintite Loopback Editiontezcan - Saturday, May 11, 2019 - link
It puts an end to an era of 10GBase-T startup wars: Plato Networks, Solarflare (bought by Marvell too), Teranetics (bought by PLX and Aquantia), and Aquantia. I used to work for Teranetics/PLX/Aquantia... I was hoping they'd be the next Marvell/Broadcom with all the new ethernet money- this is what was always the goal but 10GBase-T was just too fast- hence going SLOWER to 5G and 2.5G.Does anyone know if Aquantia ever licensed their 2.5/5G IP out to the likes of Marvell and Broadcom?
Ej24 - Tuesday, May 7, 2019 - link
Saw this and instantly bought an Aquantia 10gb nic that I was hesitant on buying. I don't trust Marvell not to screw this up and do away affordable 10gb nicsGunbuster - Tuesday, May 7, 2019 - link
Maybe they can use some of the Aquantia IP to make the Marvell Avastar not the worst WiFi/Blutooth on the face of the planet.imaheadcase - Tuesday, May 7, 2019 - link
I don't really feel automotive is really a trend to invest in. I mean sure i get that devices need to connect efficiently with all the electronics communicating. I really think these companies don't know the average human adult does not go out and buy a car that has a lot of electronics to begin with. Despite what it my seem online, most people are buying cheap cars from each other. lolimaheadcase - Tuesday, May 7, 2019 - link
...like someone did a average price a person spends on a car and it was around $8k mark. I'm pretty sure not many of those have more than a car radio and maybe off the shelf add ons put in. lolmode_13h - Tuesday, May 7, 2019 - link
Autonomous cars have lots of sensors that need to communicate with low latency. That means fast in-car networking. I think the industry has decided that autonomous cars will eventually rule the world.As for the pricing question, I think another industry assumption is that many people will not own vehicles, in the future. You'll simply summon a robo-Uber, anytime you want to go somewhere. The bonus is that you can transport an entire family to different locations, simultaneously. You can even have a robo-Uber pick up the kids from soccer practice, instead of having to go there and get them yourself. The downside will be the wait times at peak demand (but today, you'd just be sitting in traffic, which autonomous cars will greatly alleviate, once human drivers are eliminated).
I'm not saying it's a good future, but it appears to be a fait acompli. There will certainly be a lot more driving, so electricity generation must be from renewables or we're doubly-screwed.