I LOVE Paywalls. Thanks Intel!

[Plouffe]

12th Gen Intel supports ECC memory, but you're going to need a specific chipset to utilize it. A chipset only available on expensive workstation motherboards that lack other features you might want... So just how badly do you need Error Correction Code memory in the first place?

 

 

 

 

[echos]

I think this video made a big oof, by not talking about features that are enabled on a CPU via FPGA/CPLD. While the ECC memory is true to some extent as it is built into the silicon, the video misses "soft" features.

https://www.tomshardware.com/news/intel-software-defined-cpu-support-coming-to-linux-518

 

With "software defined" CPU support it leverages FPGA/CPLD blocks on the CPU to enable special features. These could be anything from new hardware crypto features or media encoders. As such these features are a legitimate reason for a "paywall", since you're buying new features that never shipped with the processor. This is also a big boon to data centers as you can extend the life span of a process, by simply buying new IP cores for newer features.

[ToboRobot]

I found out my '96 Dodge Neon had a governor that limited the speed to 196km/h.

[BaidDSB]

I feel like im the only one who likes the idea of the paywall thing if it makes sense for environment.

 

for example, say if my 5600x was capable of becoming a 5700x for the extra fee, what would be better for the environment?

 

throwing away my already existing chip and creating waste to buy a new 5700x? or just upgrading and using the unlocked cores and features and NOT creating more waste?

 

i prefer #2. As long as companies dont abuse this [they probably will but thats another problem], its a GREAT idea to reduce e and other types of waste

[Ubersonic]

If you think it's difficult with Intel then oh boy you should see how hard it is with AMD! I had to buy an X370 motherboard AND plug ECC RAM into it, then it just worked wiht no BIOS changes, it was a nightmare!

[Oshino Shinobu]

7 minutes ago, BaidDSB said:

I feel like im the only one who likes the idea of the paywall thing if it makes sense for environment.

 

for example, say if my 5600x was capable of becoming a 5700x for the extra fee, what would be better for the environment?

 

throwing away my already existing chip and creating waste to buy a new 5700x? or just upgrading and using the unlocked cores and features and NOT creating more waste?

 

i prefer #2. As long as companies dont abuse this [they probably will but thats another problem], its a GREAT idea to reduce e and other types of waste

Or they could release the full version of the CPU for the price of the cut down one as clearly they're still making a profit on it. If it can be unlocked and run at the higher spec fine, then all they're doing is artificially cutting it down so they can charge more for the full version (not so they can charge less for the cut down version). It seems like a good deal being able to upgrade your CPU via a software unlock, but really they're just selling you the same product at a higher price.

 

It's just a way to make more money, it's not good for the consumer or the environment.

[bmx6454]

22 minutes ago, ToboRobot said:

I found out my '96 Dodge Neon had a governor that limited the speed to 196km/h.

can it even go that fast?? at least without blowing up the head gasket? lol

[Kilrah]

18 minutes ago, BaidDSB said:

for example, say if my 5600x was capable of becoming a 5700x for the extra fee, what would be better for the environment?

 

throwing away my already existing chip and creating waste to buy a new 5700x?

You wouldn't throw the 5600x away, you'd sell it for someone else to use so where's the impact for the environment?

[ToboRobot]

29 minutes ago, bmx6454 said:

can it even go that fast?? at least without blowing up the head gasket? lol

It totally did.  I was pissed I couldn't go 200KM/H.  I was also I stupid 19 year old at the time.

[Nystemy]

Just now, echos said:

I think this video made a big oof, by not talking about features that are enabled on a CPU via FPGA/CPLD. While the ECC memory is true to some extent as it is built into the silicon, the video misses "soft" features.

https://www.tomshardware.com/news/intel-software-defined-cpu-support-coming-to-linux-518

 

With "software defined" CPU support it leverages FPGA/CPLD blocks on the CPU to enable special features. These could be anything from new hardware crypto features or media encoders. As such these features are a legitimate reason for a "paywall", since you're buying new features that never shipped with the processor. This is also a big boon to data centers as you can extend the life span of a process, by simply buying new IP cores for newer features.

Except this is really not the case.

 

A lot of the reconfigurability isn't usually about hardware features. But rather just microcode. And a firmware update usually has these coming along with a new BIOS version regardless, something the motherboard vendor provides for free.

 

Using programmable logic for actual computing has long since been proven inept. _{(with an exception to real time control or a few other rather niche situations where it usually is more about avoiding court than having good power efficiency or performance.)}

Just now, BaidDSB said:

for example, say if my 5600x was capable of becoming a 5700x for the extra fee, what would be better for the environment?

 

throwing away my already existing chip and creating waste to buy a new 5700x? or just upgrading and using the unlocked cores and features and NOT creating more waste?

There is an argument to be made that "if" one could upgrade it, then it would be logical.
But the "if" is usually followed with "your chip is junk. And can't run at 5700x specs."  So most 5600x just aren't a 5700x.

Even if two products uses the same chip doesn't mean that both products can perform to the same spec.
Usually a lower tier product using the same chip as a higher end product, then the chip in the lower end version either has a manufacturing defect, or signs of contamination limiting peak performance.

To give some perspective, the semiconductor manufacturing field considers a 60% yield as "good", 80-90% is phenomenal. So there is a ton of chips that don't come close to working properly. Manufacturing defects are common. Now depending on the chip, the defects can be more or less debilitating. A lot of chips do though have yield optimizing in mind, ensuring that the common defects will result in still useable chips, but with lesser specs. _{(One common issue is just contamination in the production line, resulting in higher leakage and lower power efficiency. And less power efficiency on an already thermally limited chip isn't all that ideal.)}
 

However, yes sometimes people want more of the lower end skew than manufacturing defects alone will produce. So fusing down a chip isn't unheard off. But most of the time, the lower end chip can't be upgraded, regardless of how much one desires it to be, since most of the lower end chips have defects. _{(usually all of them if the manufacturer has set the price just right and haven't made a too large supply agreements with OEMs for the specific skew. However, there is also those situations where a manufacturer comes out with some odd new skew that is defects that the manufacturer didn't have an appropriate skew to ship them as, ie they can be too good in some regards, but too bad in other regards. These are though rare to see, often goes to OEMs or the cloud, and when they finally trickle out into the second hand market they become these weird chips that officially doesn't exist.)}

4 minutes ago, Oshino Shinobu said:

Or they could release the full version of the CPU for the price of the cut down one as clearly they're still making a profit on it.

Here is the thing with manufacturing defects, throwing the chip in the trash gives you 0 back, selling it for something makes the losses on the chip less severe.

Deducting away manufacturing, packaging and shipping costs from the selling price is indeed potentially giving profits. But that economic assessment lacks consideration of development costs, among other associated costs with having a product on the market, like warranty and general customer support, not to forget marketing so that people know the product exists.

 

29 minutes ago, Kilrah said:

You wouldn't throw the 5600x away, you'd sell it for someone else to use so where's the impact for the environment?

And yes, selling one's old computer parts on the second hand market is a way better way to give these products a second life. Instead of trying to magically remove defects from a chip by throwing money at it.

Though, when selling on the second hand market, it is important to keep in mind that hardware depreciates in value far faster than cash does. So better sell it quick than having it at an unrealistically high price where no one wants it until it literally is a worthless paperweight. _{(Ebay at least gives one the tool to see what the buyers currently are willing to accept, just check out the "sold listings" feature. Scroll a bit there, and place oneself at a reasonable price against those and the thing is gone in days, not years. Yes, one could have been lucky and squeezed out 5% more, but the risk is larger that it ends up collecting dust or one has to drop the price way bellow what one would have been able to get initially. Hardware isn't a good economic investment.)}

[BaidDSB]

50 minutes ago, Kilrah said:

You wouldn't throw the 5600x away, you'd sell it for someone else to use so where's the impact for the environment?

where i live no one buys high end hardware like this second hand. this isnt the west. everyone wants  new and warranties

[Oshino Shinobu]

17 minutes ago, Nystemy said:

Here is the thing with manufacturing defects, throwing the chip in the trash gives you 0 back, selling it for something makes the losses on the chip less severe.

Deducting away manufacturing, packaging and shipping costs from the selling price is indeed potentially giving profits. But that economic assessment lacks consideration of development costs, among other associated costs with having a product on the market, like warranty and general customer support, not to forget marketing so that people know the product exists.

Not talking about manufacturing defects here though. I said if it can be unlocked and work fine, where they've just disabled features that would otherwise work fine and then sell them to you later. At that point, they're just artificially cutting down a product so they can sell the full version to you for more money.

 

Making lower end SKUs from the same silicon where it can't match the higher end SKUs is fine and good practice to not waste a lot of perfectly usable product.

[Kilrah]

1 minute ago, BaidDSB said:

where i live no one buys high end hardware like this second hand. this isnt the west. everyone wants  new and warranties

There will most likely be someone willing to unload it from you if the price is low enough, and you'll be making more than the 0 if you threw it away.

[Nystemy]

2 minutes ago, Oshino Shinobu said:

Not talking about manufacturing defects here though. I said if it can be unlocked and work fine, where they've just disabled features that would otherwise work fine and then sell them to you later. At that point, they're just artificially cutting down a product so they can sell the full version to you for more money.

 

Making lower end SKUs from the same silicon where it can't match the higher end SKUs is fine and good practice to not waste a lot of perfectly usable product.

Artificially limiting a product as to ask for more money later is a practice I think is fairly bad for everyone at large. Since it generally sets a precedence for, "most people don't use this, lets put an additional price tag on it.", rarely does it actually end up with the base product getting cheaper.

Turning slightly defected products into a lower end version is beneficial for everyone however. Since the defected version is obviously not the full thing, but if it still is a decent working product, then someone likely wants it.

However, in regards to extra features that is technically there but normally fused off or otherwise disabled, this is more debatable.

• If it is off due to not at all working, fine.
• If it is off due to not being part of the spec. Well does it have to be off? _{(for software stability reasons yes, especially on a flaky architecture like x86.)} The feature were never promised to work, so it logically wouldn't be validated. _{(Or more likely failed validation, so likely falls into "don't actually work.")}
• If it is off since one didn't cough up enough money. Well, I say the manufacturer should have said, "it is out of stock currently." and simply waited for more correctly defected chips to appear in production. Or make a lower end chip. _{(Manufacturers obviously do have lower end chips, usually there is a few different chips in production in parallel targeting different markets.)}

In regards to ECC memory and the Intel CPUs.
To my knowledge, the data traces responsible for ECC are optional. Ie, the motherboard vendor can just not put them onto the board, as to both save development time, and also tighten up the traces a bit, reducing the effects of EMI on the signal and slightly reducing the need for ECC. So yes, it does depend on the motherboard if ECC is a feature or not.

A similar story applies to a lot of other PC features.
A lot of stuff is just optional and fully dependent on the motherboard having it or not. The CPU/chipset often does support all sorts of things, doesn't mean that the motherboard does as well. Yes, it becomes a fee of admission for these features, but at least a logical one. _{(As an example, a lot of chipsets offers a handful of SATA ports, USB, and even a serial port. Yet, most motherboards rarely actually brings all these ports out to the user. The chipset still has them, the BIOS typically disables them for obvious reasons. One can technically modify the board and hook them up, but the BIOS will still have them disabled. Is this immoral and wrong of the motherboard vendor? No, not really. To a similar degree I can accept a CPU/GPU vendor not including every feature that technically is there.)}

[TheDragonLord]

As someone who got fed up with system crashes and corrupted data, and bought himself an Opteron 3680 "flowered" (duel chip) motherboard back in late 2012/early 2013, I can say from my own tests and experiences that while compatibility on the desktop OS [Damn Microsoft for shuttering your Hardware Testing Department to save a few bucks, you BASTARDS! You hath screwed us all! #OneOfTheRealReasonsWindows11HasACutoff] at times has been glitchy as frack [DAMNIT WINDOWS! STOP ETERNALLY BREAKING WHEN YOU GO INTO ENERGY SAVE MODE! I KNOW OPTERON IS A DAMN WORKSTATION/SERVER PROC DESIGNED TO BE ON 24/7/365 with minimal downtime, BUT FOR FUCK'S SAKE—I SHOULD *STILL* BE ABLE TO run the desktop OS without it nuking every time it goes into energy save mode and not be required to run Windows Server just cuz you want more money! That shit can't game for fuck! But, I digress…], but outside of weird compatibility issues or games simply being too much for my then, top of the line, GTX 690. , the overall system stability has been insanely better when compared with my other systems. Other than *my* ADHD fuckups, games don't crash ANYWHERE near as much, they run more consistent frames per second, the whole system generally runs better, and despite a theoretical lag associated with Registered ECC memory, I found it made a world of difference — frankly, it has made me a believer that the only reason ECC isn't standard in all devices is corporate capitalist greed on the whole.


I will likely never buy a desktop or laptop system that isn't capable of ECC memory as a result. It's just not worth having to reinstall my OS every 3–6 months because of the amount of baked in errors the OS alone creates in addition to all the other potential crap that goes on (same reason why I fell for RAID1 and SCRUB checks, and those extra bits of data that assures data integrity—now, if only I could get that two-digit error hammering code from Raid 2/3 combined with RAID 1 mirroring and I'd be at bliss. I may end up with less overall storage and older drives as a result, but reducing errors & the subsequent increase in overall system stability, and the resulting reduction in forced system reinstalls is just worth it. Even after God only knows how many ADHD brain fart induced system nukings I've done, or the countless other outside induced code fuckups, ALL of my data has been easily recoverable and the system has been so much smoother & more resilient.


It reminds me of this review of Maxon's drivers—if other graphics manufacturers (and other manufacturers) put in the level of quality control that Maxon does with its graphics cards, the lifespan, speed, and general usability of our systems would exponentially improve! I know that MS is a *big* part of the failure with the death of their Hardware Testing Department (and the lack of incentivizing sending *FULL SYSTEM DUMPS*, which would be in the gigabytes at minimum) secondary hardware manufacturers & their driver divisions are left with far less ability to rigorously test their driver design. That said, y'all are weak sauce & need to up your game!



Open-Hardware access is indeed a need for consumers. If manufacturers want to "save money" by only using one fab run for multiple chips and sort components based upon more than reliability/speed/etc. testing, where they can physically turn a whole batch of "blanks" into "gaming grade" vs "workstation grade" cards by a mere stroke of a key or selecting a particular set of pins to connect on the chip die, or other nefarious practices that ultimately prevent consumers from owning their own hardware — we don't want to pay for a damn license! We are paying to buy this shit, DAMNIT!


Firms "guaranteeing" a product is one thing. For example, if AMD say used the same chip for both products, but would only guarantee one product to reliably work as stated given the increases quality standards, companies would still be likely to pay given they aren't just paying for the hardware — if a graphics card fails and destroys work being done on a film, the cost in terms of losses time/effort/capital investiture would likely well exceed the added costs of the guarantee. Whereas, someone who is learning about this stuff or who simply doesn't have the resources could well afford to take the risks, even if they lose their data if only because they aren't putting in the level of resources that someone who is doing this for a living might be (for example). In such a case, software unlocks should be made available, even if we risk voiding our warrantees for running hardware at specs not guaranteed by the QC department at the manufacturers. Those risks SHOULD be up to consumers to decide.

INSTEAD of Samsung bricking my device when I want to root & unlock its bootloader (cuz I want to purge my old phone of all the bloat ware that the manufacturers or secondary distributor gets paid bank to include & hardware lock), how about instead you blow the warranty fuse? Still sketchy as fuck because of the risks of abuse/failure, but it would be ORDERS OF MAGNITUDE more ethical/less sketch than what these firms have been doing for DECADES now…

#My4¢ [2¢ ain't enough to cover it.]

[TheDragonLord]

Does anyone have a link to the ECC Memory manufacturer that Linus mentioned? Was it Munkin? I couldn't entirely tell & wondered if anyone knew/had a link/etc.

 

Many many thanks!

[jed351]

I only created the account now because I am surpised nobody realized this video is nonsense. Please don't be offended by that fact that this is my first comment. 

 

Linus already pointed out Intel's argument is that consumer CPUs do not need ECC. If you ask the question whether you need ECC, the answer is always no. Those who need ECC don't ask this question because they know they need it. The GUI of any OS is more likely to crash before the system memory is altered by an accident (even my Mac's GUI encountered a bug and forced me to restart it 2 hours ago). This is just not the same as car manufactures paywalling functions that more than 0.001% users are actually going to use. 

 

ECC memory requires addtional time and resources to validate the CPU and motherboards. Linus even pointed out in the video that AMD's consumer CPU's ECC function is not validated by AMD. There will be two cases if Intel would like to "support" ECC memory on their consumer platform. First, if Intel decides to spend resources on validating ECC memory on consumer's platforms, Linus will make another video critcising Intel asking consumers to pay for the validation of a feature which 99.999% don't use (for a consumer CPU). Secondly, if Intel is now unofficially supporting ECC on their consumer CPU but it didn't work as intended, Linus will also make a video criticising Intel. (If anyone has the courage to take the gamble that ECC might not work, he/she doesn't need ECC)

 

Regarding whether you need ECC for your home server, the answer is no if the server is idle for most of the time. (which is extremely likely the case if you are using an old PC as the server) The Google's study has nothing to do with a home NAS. Google's servers are likely fully plugged with memory modules and run at near full load 24*7.

 

 

[wanderingfool2]

The issue I have with the video is the whole talk regarding things like already making a profit and such.

 

The fact is for most people, it's not a feature that really matters.  The people who it might matter for are people who are purchasing it for more business/server level of stuff.

 

Acting like the a decision makes "business sense" and you shouldn't "lie down and take it" is a completely backwards way of looking at it.  To use a Tesla as the example, as they did.  The heated seat argument.  Does it reduce the cost to manufacture?  Yes.  Does purchasing it help cover the cost of having it installed in all cars? Yes.  Does it make business sense to do so?  Yes.  It's foolish to talk about it being a negligible cost, as from a business sense it's negligible cost (but giving it away free is not negligible...logic below)

 

But that ignores looking at the list of consequences of why lying down and taking it makes sense.

Split into more SKU's.  You now have wasted downtime and higher complexity.  Higher base model cost

Keep single SKU and give it to everyone.  That added cost now goes into the base model.  Say 1/4 of people ended up purchasing heated seats(@$400)...but extra cost to add in was $50 (when considering all SKU's and such)...so making on average $200 for every 4 cars sold.  Eliminating that option and providing it to everyone (while making the same profit) is now $100 addition to the base model...effectively 3 people would pay for that additional person to have it.

 

Guess what, they also did start providing the rear heated seats to everyone...when they increased the base level price.  So good for you, that shouldn't lie down and take it strategy is really working well, just gave a justification to increase the base level price for those who don't want rear heated seats.

 

Also I hate the blindness of claiming $40k car and expecting it to have features of a $40k ICE car...the battery cost literally is like $20k (their service/parts doesn't make money...as per their balance sheet...and it's $20k to replace batteries so that's how I get that number).  The Mach-E costs that, and Ford has admitted they are losing money on each vehicle they sold.

 

 

The tl;dr though is, I don't really care too much about Intel putting resources into making it work with EEC.  If lets say there is some evidence of them spending extra money to intentionally disable it, then I have a problem.

[WhiteAirLock]

When it comes to cars and their features being locked behind paywall, it's kinda like DRM in Games. This stuff can be bypassed digitally and in the old fashion way.

I didn't wan't to comment this on YT since I already have a bit of trouble after exposing a German politician and getting sued by him.. luckily they don't even had my real address or my real name since I didn't want to feed uncle Google with that.

 

Anyway, on the BMW you can bypass the heated seats with a potentiometer that's used for the heated seats in a Golf 4 (1999) 1.4 and you can basically hot wire that into the fuse box. Don't mind the error messages that will appear after you disconnect the heating element from the connector.

When it comes to other features, my friend has some 280Euro OBD2 Programmer from Aliexpress with the decryption software for the unique 4-8 digit password that's generated from the VIN or "key number" (if you buy the BMW from German dealership) and it's as simple as enabling the expert mode in the software and searching for the right keyword like "Sitz_VR_Heizelement" (seat heater front/right)  or "Sitz_VL_Heizelement" (seat heater front/left).

In my opinion, the manufacturers are stupid for locking things behind a paywall since cars became something like a ARM64/x86 computers on wheels. This just begs to be hacked or tempered with. 

It's not like that 1997 Renault Clio electric car which had it's ECU and BMS equipped with one time flash/read only storage modules which didn't allow you to get more than 21hp out of the electric engine or put higher capacity battery into the car.

 

Before anyone asks, I won't provide any tutorials on how to modify things on your "It's plastic, It's fantastic" car and neither will I provide any purchase links.

Just do a little bit of research and you'll eventually reach your destination. Also, try to search things in other languages like Polish, Russian or Español.

[Senzelian]

13 hours ago, TheDragonLord said:

Does anyone have a link to the ECC Memory manufacturer that Linus mentioned? Was it Munkin? I couldn't entirely tell & wondered if anyone knew/had a link/etc.

 

Many many thanks!

Mushkin

http://www.poweredbymushkin.com/Home/