New Ryzen Is Running Solid Under Linux, No Compiler Segmentation Fault Issue

[SemiLucid]

Michael does this count as an official statement from AMD, and all week 25 onwards Ryzens are fixed?

OR

Have the week 25 - 30 Ryzens just been what was in AMD's inventory at the time, and have received extra testing before being sent out as replacement parts.

[SavageX]

It is fixed for CPUs that were produced after week 28 (or so?). But you can't order specifically those anywhere. Unless you show up in person in a shop, you risk getting a broken Ryzen and have to go through RMA, which can take several weeks.

Until potentially affected CPUs have been purged from the channel, beware of buying Ryzens.

This is exactly my problem: I don't want to go through a RMA procedure, I just want to buy a CPU that works according to spec, as can be expected.

I recently had to replace my main rig (which sadly fell victim to a water-damage incident) and I planned on getting a Ryzen 1700. For now, I went with a A12 9800 as interim solution (which has the nice side-effect of coming with nice-enough graphics so I can wait for the GPU market to get back to sanity before getting a dedicated card) as I currently have no reliable way to source a known-good Ryzen.

All I need from AMD is a statement that they identified the issue and some sort of guidance on how to tell the potentially faulty CPUs apart from the guaranteed-okay ones. For now, it's unclear if 1725+ chips are okay on a regular basis or if the RMA-chips are just hand-picked.

[Cres]

Has anyone happened to confirm whether op cache is actually still enabled on their replacement CPUs or ones produced after week 25 for that matter? As the corresponding feature control MSR isn't known, the best way to verify it would probably be to find a sufficiently sensitive benchmark with a known baseline or sample for PMCx28A[0:1] events.

[qsmcomp]

My Ryzen 7 1700 got RMAed, and there is no segfault problem on the new chip.

[nomadewolf]

It's already been proven that it's not microcode. The RMA'd chips have the same same microcode which is what's embedded in AGESA 1.0.0.6.a - if it was as simple as that, AMD would have dropped the microcode like it's hot and fixed this a long time ago.

Instead, it's still being brought up in the news because it's still broken (yet, they still keep mum about this issue).

Can't argue with that.
It's not microcode then.
But am i alone in thinking that a silicon fix would take more time to get out than the time they had so far?

[RussianNeuroMancer]

But am i alone in thinking that a silicon fix would take more time to get out than the time they had so far?

Maybe they just test hardware before sending it?

[x3sphere]

Retailers still have pretty old stock. I just got a new 1700X from Newegg last week that was week 9. I was having other issues with Linux in addition to the segfault bug too... system would randomly lock up every few days.

Fortunately Newegg let me return it as I didn't want to wait around for an RMA. Decided to just go with Threadripper instead.

[nomadewolf]

Maybe they just test hardware before sending it?

That may very well be the case.
But it would mean AMD might actually be losing money with Ryzen... Ouch!


EDIT:
Wait. Is this also confirmed in Windows? Or is it just Linux?

[Funks]

Maybe they just test hardware before sending it?

Per the reports on the amd community forums, looks like they are testing the RMA'd chips one at a time as some of us received them with testing notes. The UA1728SUS R7-1700X CPU I received yesterday has passed the Linux GCC 7.1 make -j16 test (15 hours so far), I believe it was tested as well as the box looks like it was opened from the bottom (no sticker there). The new CPU looks like it's running at a lower default voltage.

Bad news is that it's taking them AMD a while to send RMA's back. Still waiting for my R7-1700 back from them (both my two chips affected), original interaction regarding the RMA was August 11, the CPU was sent back to them on August 18 - it's now the 29th and they still haven't shipped back an R7-1700. They may be out of stock and waiting to test one before shipping it back to me.

[vw_fan17]

Per the reports on the amd community forums, looks like they are testing the RMA'd chips one at a time as some of us received them with testing notes. The UA1728SUS R7-1700X CPU I received yesterday has passed the Linux GCC 7.1 make -j16 test (15 hours so far), I believe it was tested as well as the box looks like it was opened from the bottom (no sticker there). The new CPU looks like it's running at a lower default voltage.

Bad news is that it's taking them AMD a while to send RMA's back. Still waiting for my R7-1700 back from them (both my two chips affected), original interaction regarding the RMA was August 11, the CPU was sent back to them on August 18 - it's now the 29th and they still haven't shipped back an R7-1700. They may be out of stock and waiting to test one before shipping it back to me.

Almost sounds like they've identified an issue where some chips (due to manufacturing variability?) lose signal integrity under certain conditions, and eventually cause a memory-access issue or similar? This testing stuff makes it sound like they're going through their inventory until they find a "silicon lottery" chip and shipping that out to everyone who complains. Not sure if I'm happy or sad about that.

Running kill-ryzen.sh right now. Really hitting some high temps. But, system is running like a champ - no real hiccup in the main firefox, even with a load of 16.xx/17.xx.. (Ryzen 1700, I think stock right now).

Edit: dang - failed 3884 seconds in.. Time to think about an RMA, I guess..
Edit2: got amazon to give me a 1-time exception, and return it, even though ~3 weeks past return time. WTF - prices went up $30-80 across the board and more in the meantime.. May be time to eat the cost of the MB and just get a 7700k and be happy... AMD you had your chance..
Edit3: Ouch - it's just an 80% refund (paid $270 + tax). Not sure I want to lose that much.. May have to deal with AMD after all.