Ryzen problems - AGAIN!

[noobpwnftw]

Previously I had 100 Zen1(Ryzen) chips and they all had the segfault bug, took me more than 3 months to "reproduce" the bug, then 2 attempts before an RMA: the first attempt failed when I told them I had 100 of them, they kick me back and forth between the distributor and AMD customer service, the second time I told them confirm the faulty batch number first, issued an RMA ticket to the distributor and only after that I sent them my full stock.

At that time I thought that I'd never ever again buy any AMD stuff, but then I cannot resist the CP value of Zen2 released recently, so I bought another 50 of them, guess what?

They have this weird idle state bug that randomly freezes! Even better is that virtually every fix available are basically disabling C6 entirely(either disable it in BIOS or via BIOS update which essentially does the same), or stuff like zenstates, so there goes turbo boost and power savings.

Seriously do those guys do chips in "see if it works" mode? No wonder why there is no public cloud that runs on AMD today: the ones who did went bankrupt paying SLA penalties.

I've seen a few benchmark threads about their performance, they are good for as long as they last, you've been warned.

[Ozymandias]

they are good for as long as they last, you've been warned.

I've had one for two years, two for one year. So far no problems, although I have no use for turbo boost and the like. I hope they last, they're the most power efficient CPUs I've bought, cheapest too.

[Werewolf]

Previously I had 100 Zen1(Ryzen) chips and they all had the segfault bug,

then I cannot resist the CP value of Zen2 released recently, so I bought another 50 of them, guess what?

They have this weird idle state bug

You bought 150 CPUs and they all had bugs? Really?

[Zenmastur]

Previously I had 100 Zen1(Ryzen) chips and they all had the segfault bug, took me more than 3 months to "reproduce" the bug, then 2 attempts before an RMA: the first attempt failed when I told them I had 100 of them, they kick me back and forth between the distributor and AMD customer service, the second time I told them confirm the faulty batch number first, issued an RMA ticket to the distributor and only after that I sent them my full stock.

At that time I thought that I'd never ever again buy any AMD stuff, but then I cannot resist the CP value of Zen2 released recently, so I bought another 50 of them, guess what?

They have this weird idle state bug that randomly freezes! Even better is that virtually every fix available are basically disabling C6 entirely(either disable it in BIOS or via BIOS update which essentially does the same), or stuff like zenstates, so there goes turbo boost and power savings.

Seriously do those guys do chips in "see if it works" mode? No wonder why there is no public cloud that runs on AMD today: the ones who did went bankrupt paying SLA penalties.

I've seen a few benchmark threads about their performance, they are good for as long as they last, you've been warned.

You should report this and the methods to reproduce it to one of the big bloggers. Most of them have access to people inside AMD AND if it's a real issue then they can make people in "general" aware of it and put pressure on AMD to resolve it. This is how the turbo-boosting bug got fixed/mitigated. If it becomes a public relations problem for AMD they will take action. You should do this soon as this type of publicity BEFORE the Ryzen 9 3950X and Threadripper launch will make AMD take action. Keeping it a secret or publishing it here will NOT get the desired result. I would suggest 3 bloggers, Steve @ Gamers Nexus, Roman @ Der8aurer, and Linus @ Linus Tech Tips. They all have good reputations, have direct contacts at AMD, and can give the problem the publicity it deserves assuming it can be reliably reproduced.

Also note that a new Agesa version will launch with the new chips and I understand that about 100 bug fixes plus some new features will be in the new release. So a fix may already be in the works. But just in case I would talk one or all of the bloggers mentioned above as they are likely to be able to determine if this is one of the fixes in the new patches.

Regards,

Zenmastur

[noobpwnftw]

Well the problem is already well-known:
https://bugzilla.kernel.org/show_bug.cgi?id=196683

Basically what it takes to be safe is to either use a crippled kernel or a crippled BIOS, which disables C-STATES and turbo boost.

What really frustrates me is that for stuff like CPUs I would assume them to be stable, just like for drinking water, I'd assume it is not toxic.

Now it seems for them it is launch first, then everyone else(OSes, motherboards, users) have to fix their mess and it is not like those are some corner cases that only happen once out of a million.

[Zenmastur]

Well the problem is already well-known:
https://bugzilla.kernel.org/show_bug.cgi?id=196683

Basically what it takes to be safe is to either use a crippled kernel or a crippled BIOS, which disables C-STATES and turbo boost.

OK, so you have to see if the new Agesa version fixes it. If you have purchased a large quantity of MB's I would talk directly to your MB manufacturer as they may have a partial list of fix the new version will address and be able to give you a firm date when they will release new bios with the bug fixes in it.

Regards,

Zenmastur

[noobpwnftw]

OK, so you have to see if the new Agesa version fixes it. If you have purchased a large quantity of MB's I would talk directly to your MB manufacturer as they may have a partial list of fix the new version will address and be able to give you a firm date when they will release new bios with the bug fixes in it.

Thanks, I will. You now may understand that for me it takes some labor work to update BIOS on every one of them, also for the last time I have to remove and reinstall every chip on the motherboard. I'd be very thankful next time when I buy some brand-new AMD chips they'd work out-of-box.

I'm not saying Intel chips don't have problems, they do too, but most of the time problems are fixed before launch, not after launch, also they had pre-launch QS/ES samples(while too bad there are people making a fortune reselling those), so after the launch the remaining problems are usually more "gentle", they don't bother me much and I'm not forced to apply a fix to get the computer running without any crashes.

[Zenmastur]

OK, so you have to see if the new Agesa version fixes it. If you have purchased a large quantity of MB's I would talk directly to your MB manufacturer as they may have a partial list of fix the new version will address and be able to give you a firm date when they will release new bios with the bug fixes in it.

Thanks, I will. You now may understand that for me it takes some labor work to update BIOS on every one of them, also for the last time I have to remove and reinstall every chip on the motherboard. I'd be very thankful next time when I buy some brand-new AMD chips they'd work out-of-box.

I used to be a sysadmin and I had over a 1,000 devices to take care of, so I know what a pain in the ass it can be when something isn't right.

one time I spent 6-weeks every day, 7 days a week, 18-hours a day, trouble-shooting a data base issue we were having with a group of servers. I had to give briefings about every two hours to the executives on what I had done in the last two hour to fix the problem. Every vendor that had sold us equipment had a rep in my office for ALL 6-weeks to help trouble shoot the issue. Come to find out, corporate had upgraded software in our main router AND had changed the bandwidth allocations on some of our WAN lines. Once that was determined, the router software was rolled back and bandwidth was set back to what it should have been all the DB problems went away. Imagine that. I got almost no sleep for 6 weeks! So, I can relate!

Regards,

Zenmastur

[Dann Corbit]

Well the problem is already well-known:
https://bugzilla.kernel.org/show_bug.cgi?id=196683

Basically what it takes to be safe is to either use a crippled kernel or a crippled BIOS, which disables C-STATES and turbo boost.

What really frustrates me is that for stuff like CPUs I would assume them to be stable, just like for drinking water, I'd assume it is not toxic.

Now it seems for them it is launch first, then everyone else(OSes, motherboards, users) have to fix their mess and it is not like those are some corner cases that only happen once out of a million.

Even though it is well known, it is still a good idea to raise a stink about it. The squeaky wheel gets the grease.
Your statements put some fear in my heart about buying the 64 core threadripper.

[noobpwnftw]

Well the problem is already well-known:
https://bugzilla.kernel.org/show_bug.cgi?id=196683

Basically what it takes to be safe is to either use a crippled kernel or a crippled BIOS, which disables C-STATES and turbo boost.

What really frustrates me is that for stuff like CPUs I would assume them to be stable, just like for drinking water, I'd assume it is not toxic.

Now it seems for them it is launch first, then everyone else(OSes, motherboards, users) have to fix their mess and it is not like those are some corner cases that only happen once out of a million.

Even though it is well known, it is still a good idea to raise a stink about it. The squeaky wheel gets the grease.
Your statements put some fear in my heart about buying the 64 core threadripper.

If it is for just one desktop I think you'd be fine, it is a lot easier to apply a fix, there however does not exist a "wall-of-shame" list of most common problems you will get with a Ryzen, so I'm only discovering them when I hit these particular bugs, go on search only to find people here and there ranting about it, including my threads here.