AMD Targeting Avionic systems

[Prysin]

So it seems like AMD has found themselves a new market they want to get a piece of.

 

Quote

In what promises to be AMD's most ambitious (and lucrative) semi-custom hardware deal to date, the company partnered with CoreAVI, a leading manufacturer of commercial jet avionics, to develop high-performance cockpit display modules that perform on-the-fly terrain mapping. The company is designing special variants of its Radeon embedded GPUs and APUs (as in accelerated processing units, not auxiliary power); which drive high-resolution cockpit displays with reliable levels of performance.

The company is tapping into its most proven GPU IP and the ubiquitous x86 CPU architecture to drive PFD (primary flight display) modules. AMD hardware's surplus-to-requirement processing power comes in handy in accelerating terrain mapping software on-the-fly. On the software side of things, AMD is tapping into open industry standards such as OpenGL, custom-built for mission-critical (safety-critical) applications. AMD isn't new to the CoreAVI product stack. The company has been supplying extended temperature-range variants of its Radeon embedded GPUs to the company for some time now.

 

 

 

 

Source:

https://www.techpowerup.com/226747/amd-to-enter-the-aerospace-industry-with-coreavi-avionics-technology

 

 

 

I find this rather neat. There are gigantic contracts to be gotten if they manage to pull through and actually make this venture work.

[Doobeedoo]

They're getting everywhere they can, that's good when they can expand on many fronts.

[226477_1454181668]

4 minutes ago, Doobeedoo said:

They're getting everywhere they can, that's good when they can expand on many fronts.

Or spread themselves too thin

[GoodBytes]

I am actually surprised they are going in this market. In order to have equipment be FAA approved, FULL and DETAILED documentation needs to be giving, and that is public info. Also requires extensive multi-year testing by a third company costing a fortune. This is why GPUs and CPUs in aircraft are AGED old. Like we are talking about 3Dfx days with buggy z-buffer, and partial OpenGL support.  AMD will definitely be the forefront of this market, if indeed their stuff are FAA approved. I mean, any junk old (for us) GPU will be miles better than what is on the market.

 

[Notional]

20 minutes ago, GoodBytes said:

I am actually surprised they are going in this market. In order to have equipment be FAA approved, FULL and DETAILED documentation needs to be giving, and that is public info. Also requires extensive multi-year testing by a third company costing a fortune. This is why GPUs and CPUs in aircraft are AGED old. Like we are talking about 3Dfx days with buggy z-buffer, and partial OpenGL support.  AMD will definitely be the forefront of this market, if indeed their stuff are FAA approved. I mean, any junk old (for us) GPU will be miles better than what is on the market.

None the less, AMD has the CPU IP, the GPU IP, fully functional graphical API IP, etc. They can just remove all the gaming/computing stuff, that gives them a competitive edge, and focus on older/simpler architecture in a new node. 

Either way, interesting to see where this is going.

[GoodBytes]

4 minutes ago, Notional said:

None the less, AMD has the CPU IP, the GPU IP, fully functional graphical API IP, etc. They can just remove all the gaming/computing stuff, that gives them a competitive edge, and focus on older/simpler architecture in a new node. 

Either way, interesting to see where this is going.

Even if they did, and you are probably right... they would still have a large competitive edge. Let me put this way: The guys I know working on this, will DREAM for a GeForce 4 MX chip... yea that bad. Of course, ideally a GeForce 8000 series as it is programmable (by this age, it has been the case long time ago), and more importantly the GPU was fully and deeply tested by Nvidia. Has the most bug fixes, which put Nvidia in a good path for future chips. So at least it is a solid chip. But anyway.. now with AMD, I think their chip will sell like hotcakes in the sector once new instruments for new planes comes.

[Prysin]

3 hours ago, GoodBytes said:

I am actually surprised they are going in this market. In order to have equipment be FAA approved, FULL and DETAILED documentation needs to be giving, and that is public info. Also requires extensive multi-year testing by a third company costing a fortune. This is why GPUs and CPUs in aircraft are AGED old. Like we are talking about 3Dfx days with buggy z-buffer, and partial OpenGL support.  AMD will definitely be the forefront of this market, if indeed their stuff are FAA approved. I mean, any junk old (for us) GPU will be miles better than what is on the market.

 

they already have sold APUs for powering the "seat mounted entertainment systems". Was a article about them getting a contract with Boeing about a year or so ago

[Space Reptile]

i mean they have the tech and exp gained from making console chips , so why not 

[Carclis]

I wonder how profitable this will be compared to Nvidia's automotive venture.

[Prysin]

28 minutes ago, Carclis said:

I wonder how profitable this will be compared to Nvidia's automotive venture.

considering how much planes cost, and how much just a dashboard costs (we're talking a couple of millions), probably very.

[LAwLz]

25 minutes ago, Prysin said:

considering how much planes cost, and how much just a dashboard costs (we're talking a couple of millions), probably very.

But you also have to take into consideration the amount of units sold. Selfdriving cars are by no means common right now, but in the long term I am fairly certain than there will be more self driving cars sold than airplanes sold.

 

It's two completely different strategies with different time frames. In the short term (not sure when this AMD system goes on sale but let's assume like 1 year) AMD will make more for sure, but 5 years down the line? I am not so sure.

[Prysin]

1 hour ago, LAwLz said:

But you also have to take into consideration the amount of units sold. Selfdriving cars are by no means common right now, but in the long term I am fairly certain than there will be more self driving cars sold than airplanes sold.

 

It's two completely different strategies with different time frames. In the short term (not sure when this AMD system goes on sale but let's assume like 1 year) AMD will make more for sure, but 5 years down the line? I am not so sure.

plane sales are going up. Maybe not the huge ones, but smaller private jets like Bombardier, Learjet, Cessna and Bell are on the rise. They are often bought by larger companies that realize that buying the plane and storing it in a hangar can save the company money on travel expenses rather then buying business class tickets from the big airlines. (although they could just use Economy tickets, but their CEO asses are probably too delicate for such reduced comfort).

 

The danger is if one of these systems fail. The rammifications are absolutely devastating, and can cost them all future endeavors into the aviation business. Doesn't matter if the product they deliver are good. One failure that causes a crash or near crash can cost them the entire investment just like that.

 

In that regard, self driving cars are much safer, as the life-toll in a potential fatal accident can be measured in the dozens at worst, rather then thousands (IE: plane crashing into buildings or mid-air collisions between two super-jumbos)

[LAwLz]

15 minutes ago, Prysin said:

plane sales are going up. Maybe not the huge ones, but smaller private jets like Bombardier, Learjet, Cessna and Bell are on the rise. They are often bought by larger companies that realize that buying the plane and storing it in a hangar can save the company money on travel expenses rather then buying business class tickets from the big airlines. (although they could just use Economy tickets, but their CEO asses are probably too delicate for such reduced comfort).

 

The danger is if one of these systems fail. The rammifications are absolutely devastating, and can cost them all future endeavors into the aviation business. Doesn't matter if the product they deliver are good. One failure that causes a crash or near crash can cost them the entire investment just like that.

 

In that regard, self driving cars are much safer, as the life-toll in a potential fatal accident can be measured in the dozens at worst, rather then thousands (IE: plane crashing into buildings or mid-air collisions between two super-jumbos)

Again, even if plane sales are going up they are not selling as much as self driving cars will sell.

 

It's few, high profit sales vs lots of lower profit sales.

 

I mean come on, you can't seriously mean that you think there are more planes sold than cars. Most ordinary families has one or more cars. How many ordinary families do you think has a private jet?

[Miguel Batista]

The tech is not old just because of Regulations. It also has to be able to withstand higher levels of radiation. In space, the first quad core was developed recently which went from a 192 nm single core architecture to 68nm quad core. Needless to say, everyone is excited

[Prysin]

1 hour ago, LAwLz said:

Again, even if plane sales are going up they are not selling as much as self driving cars will sell.

 

It's few, high profit sales vs lots of lower profit sales.

 

I mean come on, you can't seriously mean that you think there are more planes sold than cars. Most ordinary families has one or more cars. How many ordinary families do you think has a private jet?

oh, there is definetively more cars sold then planes, but at the same time, this is like comparing consumer vs enterprise hardware.

 

A car, even a ridiculously expensive one is only going to cost what? 200k $/€ for a really high end sports car. A plane costs 30m $/€ or much much much more... A A380-200 is like 200m $+..... So this will be like comparing the profit margin of a A6 APU vs a W9150 Firepro. The margins are nowhere near the same (yes validation and 24/7 support is expensive, but we both know the margins for FirePros/Quadros are through the roof).

 

For Nvidia to make real money, they need to sell massive quantities, and i don't see "autonomy" as a feature in your VW Golf in the coming 10 years even. Simply because it will cost shitloads for the R&D to get it made, likewise it will cost shitloads for the car manufacturer to get it homologated. Thus i do not see it making it into high volume production units anytime soon. 

 

Nvidia isnt known to "sell cheap" in order to get into a market. They want to dazzle you with performance numbers and features in the hopes that you will pay the full price without questioning the price itself. That approach isn't going to get them into volume production. If it costs the consumer 10k€ extra to get that "autonomy" package, then that isnt enticing for the wast majority. Because in that case, the 5k€ carbon ceramic disc brakes and 4.5k € alcantara and carbon fibre trim package is much more alluring.

[LAwLz]

42 minutes ago, Prysin said:

oh, there is definetively more cars sold then planes, but at the same time, this is like comparing consumer vs enterprise hardware.

 

A car, even a ridiculously expensive one is only going to cost what? 200k $/€ for a really high end sports car. A plane costs 30m $/€ or much much much more... A A380-200 is like 200m $+..... So this will be like comparing the profit margin of a A6 APU vs a W9150 Firepro. The margins are nowhere near the same (yes validation and 24/7 support is expensive, but we both know the margins for FirePros/Quadros are through the roof).

 

For Nvidia to make real money, they need to sell massive quantities, and i don't see "autonomy" as a feature in your VW Golf in the coming 10 years even. Simply because it will cost shitloads for the R&D to get it made, likewise it will cost shitloads for the car manufacturer to get it homologated. Thus i do not see it making it into high volume production units anytime soon. 

 

Nvidia isnt known to "sell cheap" in order to get into a market. They want to dazzle you with performance numbers and features in the hopes that you will pay the full price without questioning the price itself. That approach isn't going to get them into volume production. If it costs the consumer 10k€ extra to get that "autonomy" package, then that isnt enticing for the wast majority. Because in that case, the 5k€ carbon ceramic disc brakes and 4.5k € alcantara and carbon fibre trim package is much more alluring.

We will see. I think it is foolish to claim to be able to predict the growth and margins of selfdriving car tech right now.

 

At least we agree that they are two completely different markets and strategies.