AMD's processors don't appear to have the exact same attack surface for the Spectre and Meltdown speculative execution vulnerabilities that Intel's chips have, but the systems with red beating hearts still need updates. Recently, a Microsoft Windows patch for the forementioned exploits caused boot failures on some AMD machines. After a temporary stoppage, Microsoft has now resumed pushing the update.
The update in question is KB4056892. It was suspended on all AMD systems on January 3 after reports of machines going into an unbootable state after application of the patch. The company determined that the problem was limited only to the oldest AMD Athlon X2-era systems capable of running Windows 10. The software company resumed updating newer AMD systems yesterday and says it will have a fix in place for older systems still not receiving the patch within a week.
According to Paul Lilly at PC Gamer, Microsoft blames errors in AMD's documentation rather than anything within its patch. The software giant reportedly said that "[it] has received reports of some AMD devices getting into an unbootable state after installation of recent Windows operating system security updates. After investigating, Microsoft determined that some AMD chipsets do not conform to the documentation previously provided to Microsoft to develop the Windows operating system mitigations to protect against the chipset vulnerabilities known as Spectre and Meltdown."
For its part, AMD responded to The Verge earlier this week, stating that it's aware of the issue and that both companies "have been working on an update to resolve the issue and expect it to begin rolling out again for those impacted shortly."
Whatever the cause, most AMD-powered systems are now receiving updates and those that aren't should be getting them soon.
Hm, makes me curious how patch behaves on Pentium III…
I still have one running (last I checked it was at least !) in a PC Chips mATX board. Its the very rare PIII 100fsb as well.
The financial media’s presentation of this issue has, in the last 12 hours, become downright popcorn worthy. It’s almost like Intel is paying them or something. I think the most egregious I’ve seen is by Bloomberg (Ian King). With headlines like “AMD changes stance. Admits to chip vulnerabilities” – which is [s<]as close to[/s<] outright lying [s<]as you can get[/s<]. Lisa Sue was interviewed on CNBC earlier this week and never claimed that AMD was immune to Spectre, only that there was a "near zero" chance for one variant affecting users. Whereas now, Sue, for sake of simplicity, seems to be dropping the qualifiers and just stating that AMD is susceptible to Spectre. Then there is the CNBC Bloomberg interview with Ian King claiming that AMD is in big trouble with the insinuation that a large proportion of systems were affected by the MS patch. But then again, it is Wall Street, so I suppose I shouldn't be surprised.
The media is typically not a shining bastion of intelligence and rational thinking, and this situation is definitely a good example, but AMD does share blame on this. They (and their fanboys) immediately played the marketing card without waiting for all the facts, and the press ran with it by strongly implying that Meltdown+Spectre was only an Intel problem. Now that is backfiring as Spectre and its less straightforward mitigation gets more attention and pundits try to cover their butts by blaming someone else.
Hopefully the more cautious approach that AMD is now taking will keep them clear of the security hype cycle, because Spectre and its follow ons are going to be a big problem for everyone for a very long time.
It’s simple. The security issue as a whole was first discovered on Intel systems. So it makes sense that they were first targeted by the media. Then Intel responded saying that all modern CPUs were vulnerable. Then AMD responded initially saying that, at the time, they thought there was a “near zero” chance that AMD users would be effected. Shortly thereafter, they updated at admitted susceptibility to one variant of Spectre. Then days later, the media decides to go after AMD insinuating that they suddenly got caught with their pants down by “changing stance” and also insinuating that the MS patch was bricking a preponderance of AMD systems, when it was only a couple of old chipsets…
While you can criticize the choice to state “near zero chance” , I’m really not sure how AMD could have been more clear. They certainly didn’t seem to be hiding anything. Not that I blame Intel, as most CPUs are susceptible to Spectre to one degree, but their statement that this “issue” affects all CPUs (when Meltdown does not) was the most causally accurate statement of any made by either company.
The point though, is that the media really can’t have its cake and eat it: they can’t both report on something they can barely comprehend (these speculative caching security issues) while at the same time playing fast and loose with the truth to serve some bias or agenda.
The fact that Intel is doing a better job of managing the press shouldn’t be a surprise either.
Hooray for AMD processors!
Boo for AMD’s documentation!
[url<]http://www.amd.com/en/corporate/speculative-execution[/url<] Not surprising. Fundamentally, basic Intel cpu circuit designs are much older than Ryzen, and predate FX, even--so AMD hardware not being vulnerable is only logical. However, I think the threat to Intel cpus is grossly overblown, as are the fixes overblown in terms of performance penalties the fixes exact on Intel hardware--reminds me very much of the Y2k nonsense that had planes dropping from the sky, nuclear power plants imploding, etc....;) It's not quite that level of mindless panic, though. But it seems close in intensity but hopefully a much shorter duration! In the attached AMD bulletin, it states that no actual exploits of these vulnerabilities exist in the public domain--and so what we have is yet-another-proof-of-concept scare mongering. Once this general info hits the Internet, however, it pretty much forces the cpu makers to respond. Microsoft, however, seems inordinately good at bricking its own hybrid tablet hardware with both bios and OS patches...;) To see that they implemented a fix for old Phenoms without *testing* the fix on a Phenom system is hardly surprising. But I can believe what they say about Win7x64--and sort of believe it for Win8.1--I've been running Win10x64 since 10/2014 up to the present (I am still in the Insider's beta test program)--and I moved from dual booting with 8.1 to going exclusively to Win10x64 4/2015 and have never looked back. I actually thought WIn7x64 was a bit of a kludge compared with 8./1--and definitely think it's a kludge compared with the latest 1709 version of Win10x64. I'm on build 17063 at the moment. Fast and sweet--no compatibility problems--seems far more backwards-compatible than Win7x64--but I remember thinking the same thing about 8/.1 in relation to Win7x64. ...;) Edit: Wow...crowd around here has really gotten brainy, I see...;) I put the link to the official AMD statement up--before I added my text--and people gave *the link* negative votes??? Wtf? You didn't believe it--didn't like what it said? Jeez. Rough crowd...;)
[quote<]Fundamentally, basic Intel cpu circuit designs are much older than Ryzen, and predate FX, even--so AMD hardware not being vulnerable is only logical.[/quote<] It's a stretch to say that AMD hardware isn't vulnerable. Variant 1 is being mitigated in software, so AMD is just as vulnerable as Intel until the affected software is patched. Their response to Variant 2 is somewhat misleading; the exploit [i<]has[/i<] been demonstrated on AMD hardware, it's just not exposed in current default OS kernel configurations. Their response to Variant 3 doesn't include any equivocating, so I'm inclined to take them at their word on that one until proven otherwise. So while I agree the downvotes on the link were a bit hasty, your clarification doesn't help your case, and actually hurts it IMO.
I really haven’t seen the w2k panic… and it can be a rather harsh performance hit but mostly on databases. Which is great for server vendors especially those pushing private clouds. Older hardware is hit worse than newer hardware… especially those that already didn’t have performance headroom to lose.
But lets not use broad sweeping statements… lets just look at the facts.
[url<]https://www.servethehome.com/red-hat-outlines-meltdown-spectre-patch-performance-impacts/[/url<] The more system calls you do... the bigger the performance loss is. Some workloads are not affected and many users will never see an impact. Businesses will be hurt far worse than users. I think the "panic" level is appropriate especially for those that assumed things connected to the internet were secure-able. I expect AMD epyc market share to get a nice boost from this.
[quote<]I expect AMD epyc market share to get a nice boost from this.[/quote<] Oh it will get a temporary boost, but once it hits [b<]100%!![/b<] it won't be able to go any higher!
Never mind.
I feel like if you are running win10 on an athlon x2 chip, this patch is probably far from your first and far from the last of your problems. I am surprised those chips can support fast enough (and high enough capacity) RAM to functionally boot win10. I have one of these chips lying around at home inside a DFI board and it still works great; but that doesn’t mean I would dream of running anything newer on it than Windows XP MCE, and I certainly wouldn’t plug the thing into my network.
The patch affected Windows 7 too.
And there were Athlon X2 chips for Socket AM2, which supported up to 8GB of DDR2 RAM.
True, but I still cant imagine running Win 7 on IDE HDD with DDR1/DDR2 memory. I would think XP would still run more smoothly, but like I said, I haven’t tried updating x2 past XP for those exact reasons.
Edit: Actually now that I recall I think the higher end boards at that time did have Sata it was just gen 1 and very limited bandwidth. I suppose you could get it workable with an SSD but I would still worry about memory capacity and constantly hitting your pagefile.
FWIW, Win7 on an X2 with 4GB RAM and an SSD is still a semi-reasonable basic desktop. Most day-to-day desktop tasks aren’t especially limited by CPU power or RAM bandwidth.
Yep I forgot that Sata 1 was part of the very last generation of DDR1 boards. It was there, just hardly utilized yet because the cost of Sata based HDD at the time and the limited performance gains for the average user to go from ide hdd to sata hdd.
You’re also forgetting that the K8 X2 cores lived on into the early years of the Socket AM2 era.
That is true. I built my first PC with an X2 on socket 939 and a BFG7950 and it ran so great I didn’t even think about building a new machine until Sandybridge came out. I might have picked up a tiger direct once in that 8 year window to order another stick of ram and that was about it.
Until a couple months ago I had Win10 on a C2D (“pentium”) with 4GB of RAM and SSD, it was perfectly fine as a second box. But then I got a faster machine, so thats that.
IDE? I jumped to SATA in 2004ish when I bought my first Raptor and built an Athlon XP 2600 based system.
Lol whut? Socket AM2 was introduced in mid-2006, well after SATA drives were standard fare, and wasn’t superceded until IIRC 2009.
It’s perfectly plausible for there to be AM2 systems with 2 GB DDR2 and a supported install of Windows 7 still out there.
Also plausible is that quite a few of those were “upgraded” to Windows 10 by the almost-automatic unhelpful mass rollout of GWX.
[quote<]It's perfectly plausible for there to be AM2 systems with 2 GB DDR2 and a supported install of Windows 7 still out there.[/quote<] Replace "2 GB" with "4 GB" and that's my wife's desktop.
Re your edit: Yeah, the Socket 754/939 boards that had SATA were SATA Gen 1 (which was slow and buggy). But this got a lot better with Socket AM2, which supported SATA 2 and semi-reasonable (even by current standards) amounts of RAM.
Ah, that makes sense. I kind of missed the AM2 era due to that occurring in my reckless college years and me breaking the budget on my 939 build. It didn’t help any that ddr2/sata2 was released literally weeks after I finished my ddr1/sata1 build. When i saw that, I sort of rage quit my tiger direct subscription and buried myself under some bottles of Jaggermiester like any respectable young adult would do.
Heh… if you’d been reading TR back then you would’ve known what new tech was in the pipeline! 😉
Edit: …and I most certainly did [u<]not[/u<] miss the AM2 era, as that corresponded to a burst of system building activity for myself and family. Now I've got a lot of aging AM2 stuff sitting around in the parts pile or pulling duty in non-demanding systems (e.g. the above mentioned wife's desktop, and my file server).
So so very true. Alas I knew nothing of TR back then and most of my webbrowsing time went to Thotbot and alluctv
Jägermeister translates to Huntmaster or Master of the Hunt.
In the old world (Europe), schnapss is part of hunting tradition.
My socket 939 Opteron 170 on a ASUS A8N32-SLI Deluxe has Sata 2.
Athlon 64. The first desktop CPU to support 64 bit operating systems, 4GB+ of ram, which also has low latency due to architectural differences in early ddr vs later ddr versions.
Meanwhile, there are many atom tablets and compute stick devices that have 1-2 GB of low power ram, 16-32gb of flash storage, use slow integrated graphics, and come preinstalled with windows 10. Early atoms also didn’t even support 64 bit.
Yeah, the Athlon 64 is definitely too underpowered to functionally boot windows 10. /sarcasm.
You can totally still use a Athlon 64 x2 today, even with modern video cards and sata ssds, since most cards should be backwards compatible. It would be a bit bottlenecked, but should still work. The Opteron chips also had more cache than the lower end desktop chips, and there are a variety of boards that support the 64 chips with different levels of hardware support.
max of 3.2ghz and 8gb of ram is plenty for win10. And as long as the video is dx11 or better full UI is enabled as well.
K8 can easily run Windows 10 without any issues (The OS itself is rather easy on memory and is the slimmest member of the Longhorn dynasty). The real issue is the lack of platform support.
Post-Vista drivers for K8 platforms and hardware made in that area are scare at best. Core 2-era stuff is barely supported under Windows 10. Windows 7 EOL will likely be the final nail for Core-2 platform’s coffin.
I have a Turion 2 laptop that runs win10 far better than you’d ever think it would.
If it wasn’t for the hard drive…
[quote<]the problem was limited only to the oldest AMD Athlon X2-era systems capable of running Windows 10[/quote<] Bring on the pain!
The “Windows 10” bit is somewhat misleading, as the buggy patch was also released for Windows 7.
AMD and MS both share blame for this one. AMD for providing incorrect/incomplete docs, and MS for not testing the patch on all supported platforms before rolling it out.
FWIW my wife’s desktop got hit by this…
You know, Raven Ridge is awesome (and apparently wasn’t bricked).
JUST LEAVING THAT OUT THERE.
He said desktop, not netbook.
1. A desktop that old would be outperformed by a notebook.
2. New desktop RyZen APUs are just around the corner too.
1. Agreed. So what? I bought her a notebook a little while back; she almost never used it, and my son eventually co-opted it since it was basically sitting unused. (She also still doesn’t have a smartphone, by choice.)
2. If it becomes clear that her desktop needs a CPU upgrade, I have a Phenom 9550 that’s probably compatible with her existing mobo (option 1); I also have a spare AM3+ mobo, FX-8320, and 8GB of DDR3 RAM (option 2). Sure, neither of those would be as sexy as a brand new build with modern parts, but they would cost me exactly $0.
2. …and if she’s happy-enough with her current desktop’s performance, spending $0 is the best route to take, I think.
Indeed. Even the FX-8320 would be massive overkill for what she uses that PC for, and that’ll probably be what I use when I eventually do the next full upgrade of her system. Even today, for a “basic PC” an FX-8320 would be pretty darned “future proof”, assuming it is paired with a reasonable GPU and Microsoft doesn’t drop support for it.
Heck, I’m still on an FX-8350, and I’m grasping at straws trying to justify a Ryzen upgrade! The next upgrade I’ve got on deck is actually an LSI SAS/SATA controller for the existing FX-8350 rig, so I can ditch the flaky off-brand PCIe SATA cards I’m using to support all the drives I’ve got in there (plus an eSATA port or two)…
[quote<]CPU upgrade, I have a Phenom 9550[/quote<] Wow! I mean...just wow.
Even a decade+ old CPU is fine if all you need is basic desktop PC functionality. All complaints about her PC being slow stopped after I swapped the HDD for a SSD a couple of years ago.
Plus, if I built her a new PC now it would get Windows 10 on it, and that would just piss her off, because she really hates it when things on her PC change. We’ll start worrying about it when Windows 7 gets closer to final EOL.
While AMD deserves a bit of blame, there is one company that pushed out a “forced update patch” to an untested system.
Microsoft should NOT be bricking people’s computers, especially if said computer is still on their supported CPU list and the update is automatic / opt-out. It [b<]might[/b<] be fine to brick someone's computer on an opt-in experimental patch system a-la Windows Insider (still kinda a crappy thing to do), but bricking a mainline "stable" Windows installation? That's a [b<]major[/b<] problem on Microsoft's part. And that doesn't speak well for their testing team.
I don’t disagree with that; it felt a bit like they were trying to throw AMD under the bus. But AFAIK AMD has not denied their role in the screwup either.
FWIW Ubuntu apparently bricked a bunch of systems too, with their first attempt at a Meltdown/Spectre kernel patch… seems like everybody is rushing out patches with insufficient testing.
So… Microsoft’s fault for not testing patches… AMD’s fault for not making complete documentation. Given both companies statements… AMD did provide lacking docs…given they didn’t deny anything and said we are working with them. That’s as close to a sorry as you will get.
Having worked on OS products that required regression testing on old hardware platforms, it’s more than likely MS’s regression testing lab simply doesn’t have a test PC for every possible version of processor older versions of Windows still run on. Almost certain, in fact.
This is the type of scenario you simply cannot plan for and stay within your department’s assigned budget. You are forced to choose what you can best determine to be a representative sample, and hope that’s good enough. You especially don’t replace the most out-of-date hardware in your lab if it dies on you.
IMO, for an OS, there’s no excuse for not testing on every platform you claim to still support. Especially for a company as big as Microsoft.
We are talking about 13 year of CPUs and chipsets for Windows 10 (anything that supports NX can run Windows 10) and Windows 7 technically still supports CPUs as old as Pentium III Katmai. (And it is actually possible to get Window 7 on Pentium II)
That’s brutal number of combinations on support.
She’s still on an Athlon 64? Time to step her up to the FX, I say.
If you’re stepping UP to the FX, you should skip a step.
As in, all AMD anything from prior to Ryzen should be burned off the earth.
I knew someone would fall for it. 😉
I don’t know man. I’ve a home server sporting a PhII on an 880G mATX mobo w/IGP and DDR3-1600 ECC RAM. For what I use it for, it performs just fine. Even sits at 1.0V@800 MHz most of the time.
I think it idles at around 45-50W or so with 6 HDDs and 2 SSDs (w/one 2-port SATA add-in card). Not sure if I could do better w/RyZen tbh (though an R5-1600 would be plenty tempting for this kind of build).
Why not? I have a spare 8320. 😉
That’s exactly what I’ve been wanting you to say. 🙂
[quote<]After investigating, Microsoft determined that some AMD chipsets do not conform to the documentation previously provided to Microsoft to develop the Windows operating system mitigations to protect against the chipset vulnerabilities known as Spectre and Meltdown.[/quote<] Microsoft is like the by-the-book type we all hate while AMD is busy [url=https://www.youtube.com/watch?v=vdHBsWXaHN8<]buzzing the tower[/url<].
AMD’s not Goose?
I’m just saying that AMD might be dangerous, but it can fly with me any time!
There’s a joke about ‘great balls of fire’ in here somewhere…
I’m afraid I just lost that lovin’ feeling…
Which one do you want writing critical software that can bring down your computer?
So, releasing things that work to specification instead of breaking is bad, now?
Which I guess means the i7/Vega G-series is [url=https://i.imgur.com/l07MBOl.gif<]this[/url<]?