5 hours ago, Mike T said:

What confounds me is that CPUs are cheap

Do we know what CPUs the MSFS and specifically the Azure servers use? Because yeah, an AMD Ryzen is cheap. An Nvidia B100, which is probably what's running the Azure systems (I know they're Nvidia, but not exactly what chip) is $35,000. Its replacement, the B200 will be $50,000 and will consume 700 watts all by itself. String a whole pile of those together and the upfront costs are anything but cheap, and the electricity bill would make my "the thermostat stays on 65!" dad faint.

 

1 minute ago, eslader said:

Do we know what CPUs the MSFS and specifically the Azure servers use? Because yeah, an AMD Ryzen is cheap. An Nvidia B100, which is probably what's running the Azure systems (I know they're Nvidia, but not exactly what chip) is $35,000. Its replacement, the B200 will be $50,000 and will consume 700 watts all by itself. String a whole pile of those together and the upfront costs are anything but cheap, and the electricity bill would make my "the thermostat stays on 65!" dad faint.

Why would Azure run game servers on AI/ML silicon? That seems way too expensive for no benefit, even if it was possible. Seems like a basic linux box would do the trick and be very cheap comparatively speaking. You don't need AI/ML functionality to stream scenery to users, even if the scenery was generated with ML to begin with. None of this really makes any sense to me. Maybe I am wrong- I only work on a team that does ML work, I am not a data scientist myself.

18 minutes ago, eslader said:

Its replacement, the B200 will be $50,000 and will consume 700 watts all by itself. String a whole pile of those together and the upfront costs are anything but cheap, and the electricity bill would make my "the thermostat stays on 65!" dad faint.

And this is in part why MS is buying Unit 1 at Three Mile Island!

2 hours ago, Mike T said:

We would all like to pretend there's some mean scrooge in a suit sitting in an office somewhere messing everything up, but in truth it's not the bean counters, its often the beans.

Spoken like a bean counter 😁 But isn't it the bean counter's job to make sure there are enough beans to begin with? Then again, you can't really adequately fund every project, that will cut too deep into the bottom line and won't keep shareholders happy with nonstop record profits. So, you overpromise, disillusioned that your team will be able to "step up" with limited time and resources. And that actually worked sometimes in the days of "crunch", but these days it's not worth the bad PR when the public learns about your dev team that has slept under their desks for the past month in a desperate attempt to release a finished game.

In the end it's not a mean scrooge sitting in the office... it's an exec who is completely clueless because they're surrounded by "yes" people who also have very little knowledge of what is happening on the ground floor because the reality is filtered through a quagmire of middle management. Meanwhile, the engineer that actually spoke up during planning sessions was fired... for being too much of a cowboy.

I won't deny the value in project management, but frameworks like agile have been abused to the point that stable software these days is the exception instead of the rule. And in the end, it is all to keep the bean counters happy: getting to market first, getting the revenue sooner, keeping those shareholders happy.

8 hours ago, Mike T said:

One last comment. Multifaceted, complex, and technically forward projects like MSFS 2024,with lots of moving parts, require a strong OPM (office of project management) to be successful. If there was an OPM for this project - they failed and they failed miserably. If there was no OPM, well, then there's your problem.

Firstly, kudos for putting your theses on here for viva by Avsim!

OTH, (and not to teach a doctor how to suck eggs) I'd like to see any claims that the management of this project failed structured very tightly with clear criteria for what success should reasonably look like for such an innovative and unique project, and robust data laying out to what degree those criteria were missed. There are clearly many people very vocally saying it's a disaster, but there are also a lot, and increasing number of us saying 'hey, this is actually really good'. You seem to be starting from the position that it's already a failure?

Edit: and would you happen to be the same Mike Toussaint who makes a living as "a Senior Director Analyst covering enterprise networking technologies"?

Edited December 4, 20241 yr by scotchegg

Interesting thread. As someone with similar high level project experience (15+ years in enterprise tech, including at Microsoft) I will say that Microsoft does know how to win. They also know how to lose (Windows Phone, anyone?).. but when they align both the business and the engineers behind something, they win. There's a reason they are a $2T company.  

MSFS is a very important part of the company's history, and they will not allow it to fail. They may fumble and stumble around a bit, but they will make sure they throw enough at it to make it work long term. As an example, they launched MS Flight perhaps expecting it to fail.. they didn't have the MSFS label on it. Conversely, MSFS 2020 was celebrated throughout the company.

So I'm not too worried about they long term. Asobo and MS have demonstrated (the sheer number of World Updates.. for free) that they will continue working hard at this, this isn't a cash grab. The launch was an omnishambles.. but in six months I bet this will all be forgotten and we'll be complaining about other things.. not the very continued existence of the franchise. 

55 minutes ago, scotchegg said:

Firstly, kudos for putting your theses on here for viva by Avsim!

OTH, (and not to teach a doctor how to suck eggs) I'd like to see any claims that the management of this project failed structured very tightly with clear criteria for what success should reasonably look like for such an innovative and unique project, and robust data laying out to what degree those criteria were missed. There are clearly many people very vocally saying it's a disaster, but there are also a lot, and increasing number of us saying 'hey, this is actually really good'. You seem to be starting from the position that it's already a failure?

Edit: and would you happen to be the same Mike Toussaint who makes a living as "a Senior Director Analyst covering enterprise networking technologies"?

And it is only right to teach a Dr. to suck eggs. Socrates would have it no other way!! And Drat...my cover's blown!

9 hours ago, Mike T said:

What confounds me is that CPUs are cheap, and bandwidth is cheap.

I agree with your thesis, but the above is simply not correct, even if you're Microsoft and you own Azure.

You're missing a significant number of key conditions that affect very large systems like this.

Quote

If there was an OPM for this project - they failed and they failed miserably. If there was no OPM, well, then there's your problem.

And this statement just reinforces it.

///

You've got some nice ideas cruising over the scene at FL350, and youre not close enough to understand what's really going on.

8 hours ago, Mike T said:

To be fair, Asobo did attempt a limited traffic study with the short alpha test. But then it was missing most of the content that had to be streamed, no store, and only a limited subset of the population. That's like doing a traffic study for a new major highway by observing traffic at 3 a.m. and not considering what happens at rush hour. So you build a two lane highway when you actually needed an 8-lane highway (um, hello, Atlanta?) And the materials you used to build the 2-lane highway don't support the weight of 18-wheelers! So now you've got to widen the highway AND fix all the potholes and collapsed bridges on the existing ones. That says nothing of the budget required.

Dontcha think they used the LAST FOUR YEARS as a traffic study...?

There are some problems that horsepower can't solve.

4 hours ago, flyforever said:

some of us come from businesses that required a very strong PM experience to complex projects.

But then.... Enter... people like Elon Musk and Steve Jobs who both didn't think much of consultants (sorry) and marketing studies, and people like one old boss I had (owner of a company. He could sell sand to Arabs and ice to Eskimos yesterday, today, and tomorrow now!).

Musk: the only way to succeed is by experiencing and learning from failure ("when it's important enough, you do it even if the odds are not in your favor" ie. risk taking). Jobs: hated power point presentations ("stay hungry, stay foolish" ie. risk  taking). 

 

17 minutes ago, UrgentSiesta said:

Dontcha think they used the LAST FOUR YEARS as a traffic study...?

There are some problems that horsepower can't solve.

Exactly.  MSFS 2024 is arguably a different business model (using streaming and thin clients) compared to the old business model that MSFS 2020 is or was. It's like breeding race horses (local storage of scenery eg. Orbx's TrueEarth) only to find out later that no amount of studying racehorse performance, will horses ever win races against a car (cloud streaming of scenery eg. MSFS 2024).

9 minutes ago, bofhlusr said:

Exactly.  MSFS 2024 is arguably a different business model (using streaming and thin clients) compared to the old business model that MSFS 2020 is or was. It's like breeding race horses (local storage of scenery eg. Orbx's TrueEarth) only to find out later that no amount of studying racehorse performance, will horses ever win races against a car (cloud streaming of scenery eg. MSFS 2024).

The problem with that analogy is that it is well known that a racehorse cannot beat a car and anyone breading race horses to attempt to beat a car is a fool. I don't think that's what happened here at all.

FS2024 is not a new business model - it is a new technical model.The business model remains the same - pay a fixed fee, buy items from the market place. The technical model of a thin client and streaming everything from the cloud is very aggressive. BUT it is in line Microsoft's direction. 

I think that MSFS 2024 is a test bed for Windows 365 which is indeed the future of windows. And as we know, MSFS has always been a test bed for Windows. The future of Windows is a thin client and everything in the cloud. It's already available or business and it will be available or personal use soon.

59 minutes ago, Mike T said:

The problem with that analogy is that it is well known that a racehorse cannot beat a car and anyone breading race horses to attempt to beat a car is a fool. I don't think that's what happened here at all.

FS2024 is not a new business model - it is a new technical model.The business model remains the same - pay a fixed fee, buy items from the market place. The technical model of a thin client and streaming everything from the cloud is very aggressive. BUT it is in line Microsoft's direction. 

I think that MSFS 2024 is a test bed for Windows 365 which is indeed the future of windows. And as we know, MSFS has always been a test bed for Windows. The future of Windows is a thin client and everything in the cloud. It's already available or business and it will be available or personal use soon.

Well, first, I can beat a Ferrari with a race horse easily. Just race it on the horse track after it's been raining. I'm not being pedantic there. Well, ok, maybe a little, but I'm making the point that sometimes a shift in assumptions can reveal problems with a premise.

 

I think you're right that MS thinks this is the direction they want to go, but where I get a little lost is in whether or not it will actually be profitable. Lots and lots of things, both software and hardware, have been sold for a flat up-front fee, but which require some sort of ongoing service from the vendor. IoT is a great example. I buy 20 switches for $300 and my house is smart. Until one day the manufacturer realizes that everyone's already bought all the switches they plan to make, no more money is coming in, and that cloud service which powers those switches is costing them money. Now they either have to go to a subscription to use the cloud service, which will infuriate early adopters who were not told there would be ongoing costs, or design a new switch architecture that is incompatible with the old one but which offers something the old one can't do, which will get the ubergeeks to rebuy everything even though they don't need to. Or the third option, which is to shut down the server, which will infuriate everyone.

I know they're selling stuff on the marketplace, but I'm a little doubtful that they're able to sell enough to cover ongoing server costs. I thought it was bad enough in 2020 with just streaming the scenery, but now they're streaming almost everything and... How? How will that end up being profitable after 10 years?

It wouldn't surprise me if 2024 really is the planned iteration for 2020, but they needed that cash injection to justify the continued server expenses. This suggests there will be an MSFS 2028 or, given that they're using the servers even more now, a 2026. It also suggests we're all playing the game of hoping that by the time the server bill kills MSFS, someone else has come up with a way to make a flight sim look this good using local storage.

 

At any rate, I remain highly skeptical of this mad dash toward the cloud. Look around and it's not hard to find FinOps whitepapers talking about how businesses decided they were gonna save tons of money by moving everything to the cloud, only now they're still spending like crazy and don't know why. In some cases it's because they're not properly auditing their cloud spend, but in others it's because the cloud architecture is inherently more expensive on an ongoing basis. It's like leasing a car. Sure, the barrier to entry is cheaper than buying, but at the end of 3 years you've spent tens of thousands of dollars and have no car.

Moving things to the cloud creates an instant addiction. You're now addicted to the cloud because your service won't work without it. If you're charging an annual subscription, that's maybe doable and it enables things that would be more difficult to accomplish on customers' individual computers such as making a game that takes up a petabyte or two. If you're not, then you have to seriously think about how you're going to keep paying the tab once the purchase money stops flowing.

 

1 hour ago, Mike T said:

The problem with that analogy is that it is well known that a racehorse cannot beat a car and anyone breading race horses to attempt to beat a car is a fool.

Yes. Now, it's well known but not when horses were the only mode of transportation before the car.  And so it is too with local vs cloud-based scenery.

 

1 hour ago, Mike T said:

think that MSFS 2024 is a test bed for Windows 365 which is indeed the future of windows.

Competiton for ChromeOS.

29 minutes ago, eslader said:

once the purchase money stops flowing

I don't think it ever will.  For the majority, the digital twin is alive and not a static model.  Change is inevitable.

Edited December 4, 20241 yr by bofhlusr