Despite Initial Claims, AMD Confirms Ryzen 8000G APUs Don't Support ECC RAM - Sl...
source link: https://slashdot.org/story/24/02/17/041200/despite-initial-claims-amd-confirms-ryzen-8000g-apus-dont-support-ecc-ram
Go to the source link to view the article. You can view the picture content, updated content and better typesetting reading experience. If the link is broken, please click the button below to view the snapshot at that time.
Despite Initial Claims, AMD Confirms Ryzen 8000G APUs Don't Support ECC RAM
I can't make much of it excepted that maybe they want to force customers to buy the pro version. ECC memory isn't rocket science after all.
Screwing consumers for more money. Paying more and getting less. What part of the re-definition of “progress” is unclear to you?
Old people used to say they don’t make ‘em like they used to. Now everyone says it. Because it’s true.
BS.I would argue they would force people to pay more and get less if they actually introduced ECC support in those CPUs.Most people in the target group wouldn't need it, therefore they would pay more for a feature that's useless to them. And if some of the others wanted it, they would have had to also get ECC RAM, at which point they would be better off getting a Pro CPU.
BS. I would argue they would force people to pay more and get less if they actually introduced ECC support in those CPUs. Most people in the target group wouldn't need it, therefore they would pay more for a feature that's useless to them. And if some of the others wanted it, they would have had to also get ECC RAM, at which point they would be better off getting a Pro CPU.
I would argue that everyone needs ECC RAM. Sure, a random bit flip causing a crash or hard drive corruption every so often might not be the end of the world, but it still sucks. ECC should be the default, not the exception. The tiny extra cost in the overall price of computers caused by eliminating non-ECC RAM across the board would be well worth it in terms of overall stability.
-
And just to add to that, I would note that nearly every time I've helped someone with a computer that was crashing constantly, bad RAM was at fault. ECC would have made it possible for the OS to detect those problems and alert the user that their RAM was defective. Instead, some of them had suffered for years, thinking that their computer's behavior was normal.
That's why ECC should really be an "everyone" thing, not a "server" thing.
-
I am curious about what kind of daemon would notify the user of errors reported by the ECC code. I had to look at the system logs to see some ECC reported errors that the layer managed to fix. Sometimes even filling up the logs. There must be some kind of daemon for that right?
-
On some Dell tower PC's from around 2008, the notification was that at the next boot, the BIOS firmware showed a message saying something like "A single bit memory error was detected and corrected, press F1 to continue."
I have some sticks of RAM (in a box of junk somewhere) that caused the message on the hottest days of the year.
-
I think you might be confusing boot time or off-line write/read memory check which require no ECC at all and run time memory check which can only be performed realistically with ECC. Of course, you could write ECC data to disk but this would slow down the computer so much that it isn't a practical solution.
-
-
-
People thinking a computer crashing is normal would have suffered through the OS throwing up errors they didn't understand thinking it was normal as well. You're not building a case for ECC, you're building a case for basic computer educations.
If a computer crashes once, shit happens. If it crashes twice it something to investigate, and it's insanely trivial to test memory.
-
it's insanely trivial to test memory
True - if you don't expect your test to identify 100% of the problems - esp. those that may be temperature sensitive.
And it's definitely not "insanely fast" even to do a "kind of intense test" which actually belies the "insanely trivial" claim. I've had memory failures that would show up on, perhaps, once every 20 hours of running MemTest+ - and go through many passes which showed no errors.
-
-
-
I would argue that everyone shouldn't need to pay extra for suffering one extra crash per maybe thousand people per lifetime of their PCs.
Bonus points for not having to suffer a couple of extra deaths per those maybe thousand gamers for hardcore gamers because ECC memory latency caused a hitch at just the right moment.
-
If it was only that, I could go for it but the worst is data corruption, your computer keeps running most of the time and data on disk becomes corrupted slowly but surely.
-
The probability of one-bit errors is proportional to the amount of RAM. When computers had 64k of RAM, that might have been a realistic number. With the average computer having five orders of magnitude more RAM, the probability is five orders of magnitude greater.
Some studies have shown as much as 1 bit error per gigabyte of RAM per 1.8 hours. That means over a person's lifetime, a computer with 16 GB of RAM, the average person would experience over 6 *million* one-bit errors.
-
-
I agree, on anything but bargain basement models. That category is so orice sensetive that ubkess thetevis a maker wide coordination to such a level that it might be ilegal in some places lest the first mover having a quarter where the costumers and hance ther stock market wil punish them for being slightly more exspensive and no poard of directors wants that.
-
No thanks. ECC comes with a massive performance limitation. I'll happily trade off the insanely rare chance that one of the bits in my memory is going to be flipped and may lead to a crash (odds are it won't) for the added performance benefit of not being limited to slow RAM speeds. And that's before we start discussing the general availability of DDR5 ECC in the first place.
If you're doing something important than ECC that shit. Most of the world's computers are not and it shouldn't be the default.
Recommend
About Joyk
Aggregate valuable and interesting links.
Joyk means Joy of geeK