Intel Finalizes 'Intel on Demand' Pay-As-You-Go Mechanism for CPUs

If it's not a subscription already then I'm sure that's what they have planned later down the road. Everything is heading towards subscription based
It's not likely, Intel would have no means of enforcing the license as the vast majority of these systems are closed off and can't report back to a check-in server if they wanted to. And there is a somewhat extensive process needed to write to the FPGA on the Xeons, it's not something that gets automated.
Under Intel's existing chip structure you are already doing this, You have the base processor for X, but then you have a big long sheet of suffixes so say you need the H suffix for a multi-media server, well they make that Xeon with an H option it just costs a few hundred more, say I needed that same core count with the S suffix, well that is an option too it just again costs a few hundred extra over the base chip.
This new process lets Intel continue doing the exact same thing they are doing but it lets them just keep the base chip, and charge for a software unlock and a small write to an FPGA instead of having to produce the base models and quantities of each of the suffixed models while they estimate demand, this lets demand be met with an email and a login to a website.
 
It's not likely, Intel would have no means of enforcing the license as the vast majority of these systems are closed off and can't report back to a check-in server if they wanted to. And there is a somewhat extensive process needed to write to the FPGA on the Xeons, it's not something that gets automated.
Under Intel's existing chip structure you are already doing this, You have the base processor for X, but then you have a big long sheet of suffixes so say you need the H suffix for a multi-media server, well they make that Xeon with an H option it just costs a few hundred more, say I needed that same core count with the S suffix, well that is an option too it just again costs a few hundred extra over the base chip.
This new process lets Intel continue doing the exact same thing they are doing but it lets them just keep the base chip, and charge for a software unlock and a small write to an FPGA instead of having to produce the base models and quantities of each of the suffixed models while they estimate demand, this lets demand be met with an email and a login to a website.
Where there's a will there's a way. If companies can do it for car features then they can do it for anything. Also IoT means that anything and everything will be connected eventually.
 
  • Like
Reactions: erek
like this
Re-reading the post I'd replied to a bit flippantly, you're 100% right. My mind was elsewhere when I wrote it - namely the idea that "AMD would not follow Intel". Because what immediately came to mind was how they changed after the success of Zen2 Threadripper, and for Zen3 TR Pro abandoned HEDT/WS/Enthusiast and with really no competition in that market began to act in many ways exactly like Intel would, with the timed-exclusivity lockins (and lockouts), i.e. Lenovo, and leaving AIB partners to bleed out in the snow.

And I don't fault AMD for acting like a business is supposed to - no kidding you sell your silicon for $110 per core to server/datacenter instead of to the market that expects it for $35 per core for the same piece of silicon, even though the latter market helped get them to the favorable market position they found themselves in. It's just that the narrative about AMD being the good guys that would never act like Intel is becoming a little worn, if not detached from reality.

AMD will act every bit like Intel - and Nvidia for that matter - the moment they're in the market position to be. IMO.
No problem. But that's why I said earlier that it still is a business. However, it's not doing the things (yet) that really piss people off. AMD for me is still in the good camp for now. Mainly because of their previous commitments to open source, ECC on their desktop parts, and as long as AM5 lasts at least 3 generations they are fine with me. The fact that I can take this old ass board I have flash the bios and go from a 2400g to a 5800X3D is just straight up value. Granted I like what Intel did with the dual memory controllers (DDR4/DDR5) but I've got to see it do that for more than a few generations.
 
Yes actually, that's the long term goal of it.
In Enterprise IoT is locked down to a degree where it doesn't get the I, it's more like a NoT because local Network is the most it's getting assuming it's even allowed off its specified VLAN, and almost certainly their internetwork traffic is limited to just the bare required services and applications with access strictly controlled by user ID credentials and Source IP.
I mean you want to change a flow valve on the boiler system you need to be logged in to a specified terminal, as a member of a specified user group, then also have the credentials to the control interface that said valve reports to. That valve itself doesn't have the ability to see anything it doesn't specifically need to see, and any and all traffic coming to or going from it must match a pre-approved application ID and port structure otherwise it's not even getting past the switch to be delivered to it in the first place.
Say what you will for all the data breaches over the past decade, but they have forced us to go hard on things and get draconian because we have to assume they are already inside the building and it's our job to keep them from getting out with anything.
 
In Enterprise IoT is locked down to a degree where it doesn't get the I, it's more like a NoT because local Network is the most it's getting assuming it's even allowed off its specified VLAN, and almost certainly their internetwork traffic is limited to just the bare required services and applications with access strictly controlled by user ID credentials and Source IP.
I mean you want to change a flow valve on the boiler system you need to be logged in to a specified terminal, as a member of a specified user group, then also have the credentials to the control interface that said valve reports to. That valve itself doesn't have the ability to see anything it doesn't specifically need to see, and any and all traffic coming to or going from it must match a pre-approved application ID and port structure otherwise it's not even getting past the switch to be delivered to it in the first place.
Say what you will for all the data breaches over the past decade, but they have forced us to go hard on things and get draconian because we have to assume they are already inside the building and it's our job to keep them from getting out with anything.
That's great, I'm sure your place has good security. Doesn't really change anything I said though.
Unless you're an engineer who designed those systems, you don't really know if remote access is possible or not.
 
Last edited:
  • Like
Reactions: erek
like this
In Enterprise IoT is locked down to a degree where it doesn't get the I, it's more like a NoT because local Network is the most it's getting assuming it's even allowed off its specified VLAN, and almost certainly their internetwork traffic is limited to just the bare required services and applications with access strictly controlled by user ID credentials and Source IP.
I mean you want to change a flow valve on the boiler system you need to be logged in to a specified terminal, as a member of a specified user group, then also have the credentials to the control interface that said valve reports to. That valve itself doesn't have the ability to see anything it doesn't specifically need to see, and any and all traffic coming to or going from it must match a pre-approved application ID and port structure otherwise it's not even getting past the switch to be delivered to it in the first place.
Say what you will for all the data breaches over the past decade, but they have forced us to go hard on things and get draconian because we have to assume they are already inside the building and it's our job to keep them from getting out with anything.
Oh I wouldn't go that far with Enterprise IoT devices and configurations.
While you are correct that this is how they should be configured, lets not forget what happened to a university campus' enterprise IoT environment back in 2017...

University Hackers Attacked 5,000 IoT Devices on Campus

How IoT hackers turned a university's network against itself

I remember this making international news back then.

trip-Ampoule-connect%C3%A9e-pirat%C3%A9e-englishV2.jpg
 
Last edited:
Oh I wouldn't go that far with Enterprise IoT devices and configurations.
While you are correct that this is how they should be configured, lets not forget what happened to a university campus' enterprise IoT environment back in 2017...

University Hackers Attacked 5,000 IoT Devices on Campus

How IoT hackers turned a university's network against itself

I remember this making international news back then.

View attachment 527009
Yeah WiFi makes security hard, such a PITA, but that's why you use multiple SSID for different tasks, if you have an "IoT Devices" wifi network you better make sure it follows the same rules as the hardwired IoT one, but yeah, the lighting ballasts are on the network, or more their controller is on the network and the controller then handles the responses from their zone, but that follows the same rules as the HVAC system. Industrial controllers have come a long way in the past few years to prevent these sorts of things, but if you cut corners it's gonna happen.
 
How quickly did AMD stop providing the Wraith Prism (good Quality) stock cooler in the box with their high-end processors? About as soon as they had the performance crown.

AMD only plays nice when they are the underdog. Once they're the top-dog, its a dog-eat-dog world.

dogs.
Let's not forget the time when they charged a good $1,000 for the Athlon 64 FX-51 back in the day, when they just utterly dropped the Hammer on Intel's NetBurst crap at the time (and probably Itanium, SPARC and MIPS too with the Opteron lineup), and no gamer or enthusiast with sense built anything other than AMD. They had the lead, they knew it, and they priced to match because they knew some people are willing to pay up for the very best.

Same deal with Threadripper much more recently, especially now that Intel isn't even trying to compete in the HEDT segment any more (Sapphire Rapids is increasingly feeling like Larrabee-tier vaporware) and AMD is redirecting all of those CPU designs toward the much more profitable EPYC server/HPC market, while mainstream Ryzen has significantly closed the gap in core count if not I/O (PCIe lanes in particular).

With that said, I don't even recall the last time I ever used a stock HSF with a CPU. If it's one of my own builds, it's getting the trusty ol' XSPC Raystorm slapped onto it if mounting hardware permits. (I actually had to hold off on my 12700K build for a few days due to the need for updated hardware with LGA1700, and should check if I can just buy AM5 mounting hardware for it in case I switch.)
 
Sapphire Rapids is increasingly feeling like Larrabee-tier vaporware
You know it’s shipping to their government contracts and the and OEMs and they have product ready to go for Jan 10? The E4 stepping had more bugs but E5 had been cleared. I’m looking forward to it I’ll have some options for my August server refresh. I delayed and held off the refresh this year because AMD availability was bunk and the Xeons were meh…
So I have some calls with Dell in March to get ready for that.
 
You know it’s shipping to their government contracts and the and OEMs and they have product ready to go for Jan 10? The E4 stepping had more bugs but E5 had been cleared. I’m looking forward to it I’ll have some options for my August server refresh. I delayed and held off the refresh this year because AMD availability was bunk and the Xeons were meh…
So I have some calls with Dell in March to get ready for that.
I admittedly wasn't aware of that, but it does go to show how awfully late to market it is when we're already on Raptor Lake for mainstream desktops and laptops, and Alder Lake's Golden Cove cores completely steamroll their earlier architectures, especially given how their HEDT/server CPUs are all still stuck in Eternal Skylake hell while AMD's keeping EPYC current with the latest Zen microarchitectures.

I'm itching to see a new generation of HPE and Dell EMC servers with Sapphire Rapids as it currently stands. Wouldn't want AMD getting too complacent there for reasons stated earlier, and my workplace's full of HPE Gen8 servers that have not aged well compared to today's hardware.
 
I admittedly wasn't aware of that, but it does go to show how awfully late to market it is when we're already on Raptor Lake for mainstream desktops and laptops, and Alder Lake's Golden Cove cores completely steamroll their earlier architectures, especially given how their HEDT/server CPUs are all still stuck in Eternal Skylake hell while AMD's keeping EPYC current with the latest Zen microarchitectures.

I'm itching to see a new generation of HPE and Dell EMC servers with Sapphire Rapids as it currently stands. Wouldn't want AMD getting too complacent there for reasons stated earlier, and my workplace's full of HPE Gen8 servers that have not aged well compared to today's hardware.
My Dell AMD servers have been OK… as long as I don’t stray from the beaten path. My Xeons though, they might be older and slower but I’ve never had a “well why the fuck is this doing that” moment. They are very cut and dry.

edit:
Intel claims that Emerald Rapids is still on track for a "Late 2023" release.
 
Last edited:
Yeah WiFi makes security hard, such a PITA, but that's why you use multiple SSID for different tasks, if you have an "IoT Devices" wifi network you better make sure it follows the same rules as the hardwired IoT one, but yeah, the lighting ballasts are on the network, or more their controller is on the network and the controller then handles the responses from their zone, but that follows the same rules as the HVAC system. Industrial controllers have come a long way in the past few years to prevent these sorts of things, but if you cut corners it's gonna happen.
Network Slicing and Private LTE/5G, or depending what it is , LoRaWAN
 
Network Slicing and Private LTE/5G, or depending what it is , LoRaWAN
Yeah, so many options for IoT out there, and more popping up every year it seems. PITA but you either need manpower to manually inspect valves, fans, filters, power usage, gas levels, etc. which costs money. Or you get it so all of that I just emailed in a morning report to somebody so they can check for anything out of spec in a quick glance over their morning coffee.
 
It's not likely, Intel would have no means of enforcing the license as the vast majority of these systems are closed off and can't report back to a check-in server if they wanted to. And there is a somewhat extensive process needed to write to the FPGA on the Xeons, it's not something that gets automated.
Under Intel's existing chip structure you are already doing this, You have the base processor for X, but then you have a big long sheet of suffixes so say you need the H suffix for a multi-media server, well they make that Xeon with an H option it just costs a few hundred more, say I needed that same core count with the S suffix, well that is an option too it just again costs a few hundred extra over the base chip.
This new process lets Intel continue doing the exact same thing they are doing but it lets them just keep the base chip, and charge for a software unlock and a small write to an FPGA instead of having to produce the base models and quantities of each of the suffixed models while they estimate demand, this lets demand be met with an email and a login to a website.
Easily do it with time based license keys, you generate an ID based code from your side - take that apply it to Intel, they give you a unique license key that is now tied to that CPU and has a time limit on it, once it expires, rinse repeat. This also means the license can be locked to that CPU as well. No online connection needed. Or business would just have to run the equivalent to a KMS server for CPUs.
 
  • Like
Reactions: erek
like this
Outside of very niche (likely enterprise) applications, I can't see this being anything good for openness, sovereignty over hardware you buy, or the enthusiast user overall. As far as AMD not being "as bad", they're certainly not perfect but I think they've found out that even when they're competing well with Intel or Nvidia that a degree of better behavior (openness etc) is good for their brand. We should encourage them to continue with this as much as possible, both vocally and with our wallets, as there is still a ways to go . I'm eager to see if the X670E boards and Zen4 procs allow easy disabling of PSP functions in the BIOS/UEFI as was done in previous generations ever since both Intel and AMD realized there was a big fucking oopsie of a vulnerability a few years back. Of course, you have to trust that those disabling options actually function, but given the microcode etc..updates some tests have shown that there have been differences between the two For Intel there are a few projects like coreboot/libreboot that neuter the ME as much as possible, but with the PSP having fewer components with less reach, it seems to be more difficult. Still, any degree of blackbox at the heart of your CPU's security isn't acceptable and perhaps with time it would be great to verifiable deliver a fully open solution; a portion of the infosec, privacy etc..community would throw money if AMD was able to do so.

In any case, I don't want to see this proliferate in user-unfriendly, profiteering manners used to lock off more and more features unless you pay or worse, subscribe. I can remember vaguely a few years back that, when I was looking at getting an X299 based Intel HEDT system that, aside from the expense of these high end boards themselves, there were certain features that you needed a particular "key" to enable on the chipset that was an additional purchase? I never ended up getting that platform, but I can remember thinking that itwas pretty asinine to buy a HEDT platform and an expensive proc, and then be asked to pay again to unlock its full chipset potential and as I recall it wasn't a trivial amount. Edit : It was VROC, to enable RAID on SSDs and the like back then! Sure you could have free passthrough JBOD, but if you wanted RAID you bought a little $200-ish physical plug that (assuming your board supported it) would give the license to the RAID features. There was a cheaper model that gave you only RAID 0 and 1, the "premium" added the full suite including 5 , and there was apparently even a version that gave you the full premium features but ONLY if you were using Intel SSD! Looking at the board I would have likely purchased now, the Rampage VI Extreme , it lists having a VROC hardware slot onboard and the overall features talk about VROC, but it doesn't list the key as an "accessory" included so I'm unsure if even after buying a top of the line board like that you'd need to buy the damn key to enable it, or if the slightly more reasonable situation was that certain mobos shipped with the key included. In any case, still a frustrating set of features being locked away.

Lets hope such practices don't become common outside of a very specific niche; between all the varying forms of lockdown, invasive software or firmware, streaming or remote etc... in many ways we're losing control over our computing experience. There are more and more things that require either cumbersome workarounds or the willingness to go without ease or modern features if you want control, privacy and the like. We don't need one more intrusion.
 
Back
Top