The computer and electronics super geek thread

[kajnrig]

Did the update happen automatically? What happens during the repair loop? Do you see anything on-screen? Is it a repair loop or an update loop? Former = problematic, latter = maybe not. Updates sometimes cause my PC to restart several times. Since you're going from 7 to 10, it might just be going through a whoooooole bunch of these.

A couple things you  might try to do:

1 - Go straight to the BIOS and change the boot hierarchy, put USB or CD-Rom as first (whichever you're using to boot externally from). (Maybe you already did this, but the post didn't specifically mention it so I'm going to assume otherwise). Assuming you can boot to the external medium, you might be able to repair Windows that way. Or if it's a Linux distro, you can use that to at least transfer your files to an external drive before...

2 - Fresh install Windows 10. Move the hard drive to a different computer/external enclosure/use Live USB distros to get any important data off of it, then format the drive and insert it back into the system and install a fresh copy of Windows 10. Searching up all the drivers to install afterwards will be a pain, but Dell's site SHOULD provide a comprehensive set of drivers for the main hardware.

EDIT:

adding quote so people don't have to skip back a page

41 minutes ago, kalvasflam said:

Ok, I need to get some help here, I have an older PC (2012 Dell XPS tower) and it went through a Windows 10 update recently (unwanted), and now, my computer is stuck in the repair loop, and can never get clear.  I've scoured the web, and still can't find a solution, when I use a boot media, the stupid thing doesn't want let me repair the unit.  I think the only thing left to do is a full reinstall.   But the damn thing doesn't let me reinstall from the boot media, and I don't know how to do it from the hard drive since it just goes into this idiot repair loop.  

 

Anybody have any ideas?

 

(full disclosure, I upgraded from win 7 to win 10, yes, I was a moron)  

 

Edited January 4, 2018 by kajnrig

[kalvasflam]

13 minutes ago, kajnrig said:

Did the update happen automatically? What happens during the repair loop? Do you see anything on-screen? Is it a repair loop or an update loop? Former = problematic, latter = maybe not. Updates sometimes cause my PC to restart several times. Since you're going from 7 to 10, it might just be going through a whoooooole bunch of these.

A couple things you  might try to do:

1 - Go straight to the BIOS and change the boot hierarchy, put USB or CD-Rom as first (whichever you're using to boot externally from). (Maybe you already did this, but the post didn't specifically mention it so I'm going to assume otherwise). Assuming you can boot to the external medium, you might be able to repair Windows that way. Or if it's a Linux distro, you can use that to at least transfer your files to an external drive before...

2 - Fresh install Windows 10. Move the hard drive to a different computer/external enclosure/use Live USB distros to get any important data off of it, then format the drive and insert it back into the system and install a fresh copy of Windows 10. Searching up all the drivers to install afterwards will be a pain, but Dell's site SHOULD provide a comprehensive set of drivers for the main hardware.

EDIT:

adding quote so people don't have to skip back a page

 

My bad, yep, tried #1, and booted direct from a USB.  The repair didn't help, kept coming back with a system can't be repaired message.  I figure the fresh install is the other option, I guess the good thing is that I already backed up all the data.  I guess I need o look at the Dell site for the drivers... geez, knew I should've just bought a Mac.

 

If anyone else has any suggestions, I'm all ears.  Thanks

[azrael]

Try recreating the boot media to do the clean install. The media may be non-bootable so when the system sees it, it won't boot.

------------------------

Rant: This crap with Intel...Maybe it is time I jumped back on the AMD-wagon after a 15-year hiatus.

[kajnrig]

47 minutes ago, azrael said:

Try recreating the boot media to do the clean install. The media may be non-bootable so when the system sees it, it won't boot.

------------------------

Rant: This crap with Intel...Maybe it is time I jumped back on the AMD-wagon after a 15-year hiatus.

A lone BBC article said that the issue also plagued AMD chips, but I'm 99% sure it's a misunderstanding on a probably non-techy journo's part.

That said, maybe wait a couple months if you can for at least the refresh lineup before switching. The Ryzen refresh will almost assuredly offer slightly better performance, maybe better IPC/clocks as well. I'm debating whether to switch on that "Ryzen+" or wait for Ryzen 2 next year.

[azrael]

23 minutes ago, kajnrig said:

A lone BBC article said that the issue also plagued AMD chips, but I'm 99% sure it's a misunderstanding on a probably non-techy journo's part.

That said, maybe wait a couple months if you can for at least the refresh lineup before switching. The Ryzen refresh will almost assuredly offer slightly better performance, maybe better IPC/clocks as well. I'm debating whether to switch on that "Ryzen+" or wait for Ryzen 2 next year.

By the time I actually put money down, I'm sure Ryzen updates will be out.
AMD hasn't clearly stated which generation is truly affected. Ryzen CPUs are vulnerable to the Spectre variants but can be patched at the OS level and are unaffected by Meltdown. AMD didn't clearly state if FX and their older CPUs are affected by both Spectre and Meltdown whereas all Intel CPUs since, at least, Ivy Bridge are affected by both exploits.

[kajnrig]

What is different about Spectre as compared to Meltdown? I thought Meltdown and Spectre both were referencing the same Intel-specific memory leak issue. Apparently ARM CPUs also suffer from the Spectre vulnerability?

[azrael]

Meltdown is Intel-specific, but affects some ARM CPUs. Spectre affects all manufacturers. Both vulnerabilities affect the same speculative execution process but Spectre requires more user involvement. Meltdown requires less user interaction and essentially, brute-forces its way. Spectre and Meltdown can be mitigated with a software patch to the OS but Meltdown requires a fix to the microcode on top of the OS patch. Both still requires a user to actually run the malicious code.

[kajnrig]

4 minutes ago, azrael said:

Meltdown is Intel-specific, but affects some ARM CPUs. Spectre affects all manufacturers. Both vulnerabilities affect the same speculative execution process but Spectre requires more user involvement. Meltdown requires less user interaction and essentially, brute-forces its way. Spectre and Meltdown can be mitigated with a software patch to the OS but Meltdown requires a fix to the microcode on top of the OS patch. Both still requires a user to actually run the malicious code.

Just as I saw this post, I also came across this video from Gamers Nexus:

 

[JB0]

AMD parts are not immune to Meltdown, it is just harder to pull off.

Basically, rumor and hearsay took hold before any public knowledge was available, and now both companies are using it to take potshots.

 

 

That said, I'm still mad at Intel over the management engine omnihack, and generally abusive business practices. 

 

 

 

Edit: relevant part of the whitepaper:

6.4 Limitations on ARM and AMD

We also tried to reproduce the Meltdown bug on several
ARM and AMD CPUs. However, we did not manage
to successfully leak kernel memory with the attack de-
scribed in Section 5, neither on ARM nor on AMD. The
reasons for this can be manifold. First of all, our im-
plementation might simply be too slow and a more opti-
mized version might succeed. For instance, a more shal-
low out-of-order execution pipeline could tip the race
condition towards against the data leakage. Similarly,
if the processor lacks certain features, e.g., no re-order
buffer, our current implementation might not be able to
leak data. However, for both ARM and AMD, the toy
example as described in Section 3 works reliably, indi-
cating that out-of-order execution generally occurs and
instructions past illegal memory accesses are also per-
formed.

Edited January 5, 2018 by JB0

[mikeszekely]

8 minutes ago, JB0 said:

That said, I'm still mad at Intel over the management engine omnihack, and generally abusive business practices. 

That said, my next CPU is still probably going to be Intel, because performance.

[JB0]

29 minutes ago, mikeszekely said:

That said, my next CPU is still probably going to be Intel, because performance.

Zen's trading blows in that field. And in terrs of performance per dollar, Intel's still losing after aggressive price cuts and spec upgrades across the board. We're probably gonna see more of that this year, as the gap's been hurting them.

 

This isn't sledgehammer vs P4, but it is still an exciting time.

[jenius]

I'm running a threadripper on one computer and coffee lake on another, both are gravy.

[kajnrig]

3 minutes ago, jenius said:

I'm running a threadripper on one computer and coffee lake on another, both are gravy.

1950X or 1920X? No Core i9 7980XE build?

I'm really interested in seeing AMD's APUs next year, both desktop and laptop varieties. The gaming reviews of the mobile Ryzen 5 2500u in the HP Envy X360 are really impressive given the minuscule thermal profile it has to work with. I foresee some small ITX builds over the course of this year for the kids, and maybe some affordable light gaming laptops as well.

This is definitely an exciting time.

[JB0]

4 hours ago, jenius said:

I'm running a threadripper on one computer and coffee lake on another, both are gravy.

Tangentally, Threadripper is the best name for a processor since Sledgehammer.

[jenius]

I got a 1950X around Black Friday for less than $800. I'm trying to move up to having a 4K computing existence and I didn't want all that encoding to kill me. I overbuilt my wife a computer which is secretly my back-up computer in case things with AMD didn't go smoothly. I can say the AMD machine is an encoding monster (though my point of reference is my previous SandyBridge E machine). Other than the amazingly fast encoding times, the AMD machine has had lots more issues than the wife's computer. Of course, during that same time, my wife's computer has just searched the Internet for recipes while my computer has been living in several different Adobe Apps and compiling/encoding. I also OC'ed the Threadripper but left my wife's computer at stock speeds so that might also account for some of my crashes and oddities along the way. Example of an oddity: both machines have the same NVME SSDs for boots. The Intel machine boots as you sneeze. The Threadripper takes what feels like approximately the same time my old SandyBridge took. I can go in BIOS and enable "quick boot" or something, and maybe I should, but I haven't bothered looking into it yet. Both computers are brand new motherboards so I may need to do a BIOS update at some point also. RAM-wise the Threadripper is using 3200CL17 while the Intel machine is using something faster but I doubt that means much. The wife's computer is an adorable ITX rig, Metis Plus. She's thrilled and I like knowing that if my computer ever has problems I can go fire hers up and keep working. 

[technoblue]

The CPU exploit is a combination of how modern processors access kernel memory and a CPU optimization known as speculative execution. In short, the intent of speculative execution isn't a negative one. It is supposed to provide additional concurrency for process intensive tasks at the hardware level when system resources are available and enhance the overall performance of CPUs. Here's the Wiki article for a quick reference:

https://en.wikipedia.org/wiki/Speculative_execution

Unfortunately for Intel, and to a lesser degree others like ARM, modern hardware designs have allowed this hole for hackers to use speculative execution to access kernel memory. For other CPUs, like legacy SPARC CPUs used on Solaris machines, kernel memory is not accessible to the user. Again, this is due to differences in hardware design. I wouldn't be surprised if AMD is not affected much, if at all, since they use a different design process from Intel. 

The only way to "fix" the issue for those CPUs that are affected is to 1) redesign the hardware or 2) apply a patch at the OS level, but the second option comes with a performance cost and who know how long it will take for the first option to make it to retail shelves. What's more worrisome for me, more so than the politics of which CPU is better, is how many other optimizations have created similar security holes at the hardware level. It's a scary thought when one starts to go down that road, because these are not simple fixes that can be applied with firmware updates and OS patches are merely a band-aid.

Edit: clarification

Edited January 6, 2018 by technoblue

[JB0]

Actually, older Sparcs are immune because Sun/Oracle was late to adopt out-of-order execution. A Sparc box sold after 2011(seriously, Oracle?) is probably vulnerable to Spectre(Fujitsu Sparcs adopted OOE in the 90s, but are less common to my knowledge). IBM's Power series is already confirmed vulnerable, and that extends to PowerPC. Any processor using out-of-order execution should be assumed vulnerable until proven otherwise. Don't confuse "researchers only tested Intel, AMD, and ARM" to mean that only those processors are affected.

Meltdown does require some architectual features, but as it works on both x86 and ARM processors, it is clear that it isn't an Intel-exclusive attack(though Intel's design makes it easier).

 

And to be clear, the attacks are a way AROUND how modern processors handle memory, not enabled by it.

[technoblue]

14 minutes ago, JB0 said:

And to be clear, the attacks are a way AROUND how modern processors handle memory, not enabled by it.

Yes, attacking private memory space using an otherwise benign CPU optimization.

For my part, I thought the SPARC/Solaris platform was too old to use any kind of speculative execution optimization, but I haven't really followed what Oracle did with SPARC. I always thought Oracle was into SUN more for Java than the hardware. Anyway, I do agree that all modern CPUs will need to be tested. This sort of news is a boon for verification and EDA. I just wouldn't be surprised if AMD isn't talking out of its arse. That's all I meant by it. 

The good news, since these hacks effect how the hardware and software (OS) interact with one another, is that the available OS patches should resolve the immediate security issue for most consumers. Personally, I'll be happy with hardware improvements too, especially if they do reduce the performance penalty caused by the OS patches. So far, though, the only early benchmarks I've seen are for LINUX and Windows.

[JB0]

Yeah, Oracle kept updating Solaris and Sparc for quite a while, even if they were late to a few parties.

Though they seem to have given up, after basically killing the line through ineptitude(appartently they decided to focus exclusively on the top-end and let everyone else take over the rest of the market... which predictably ended with the x86 servers making a rapid charge towards the niche they intended to service). I gather they fired their entire Sparc development team and almost all of the Solaris team late last year.

Fujitsu hasn't given up on Sparc yet, but... the architecture's days are clearly numbered and we can easily score it as yet another x86 fatality. Unless someone notices OpenSparc and decides to make "fully open" laptops. There's a niche for that, especially after the Intel Management Engine fiasco last year.

 

For what it i worth, I did a lot of looking around at the stuff after picking up a discarded Sun Ultra10 to play with, and that's most of why I have fresh knowledge of it.

[technoblue]

An Ultra10? Sweet. It's been 8 or 9 years since I last worked with any UltraSPARC-based workstation, which is why my knowledge is fuzzy.

Concerning the other topic, I noticed that this article from rasberrypi.org has been making the rounds:

https://www.raspberrypi.org/blog/why-raspberry-pi-isnt-vulnerable-to-spectre-or-meltdown/

It's a good read, especially if you are interested in some background abstraction of scalar, superscalar, and out-of-order processors, as well as hardware optimizations like cache and speculative execution. I also appreciated the simple example near the end which illustrates how these side-channel attacks are able to access data from restricted areas. 

[JB0]

Yeah, the Pi article is probably the closest we'll get to a layman's explanation. I was surprised they managed to make it as comprehensible as they did.

[kajnrig]

Gamers Nexus on some of the myths/misinformation surrounding Meltdown and Spectre:

 

[nhyone]

It is to Intel's benefit to muddy the waters as much as possible.

Meltdown is a pure Intel bug and it is very easily exploitable. A program could sniff memory as if there were no protection at alI. It makes a mockery of Hypervisors. Think cloud computing and servers. They definitely need to patch their systems for a 5-30% performance loss. For consumers? Less need, but it does mean that any website that you visit could theoretically hack you to some degree (e.g. get your password in memory).

Spectre is a general attack against modern CPU design, hence it affects almost all processors. Those not affected are slow-pokes. 

It is much harder and slower to pull off. I won't worry about it.

[David Hingtgen]

So, time to upgrade my PC.  But also the worst time EVER to be in the market for new graphics card because literally every card in North America has been bought up by bitcoin miners causing a massive supply/price change.  It's insane.  I should have bought a 1070 months ago.   Now, I'm seriously looking at used 970's.  (I have a 960, but no way can I afford to go up to a 1070 now or any time soon).   

Any chance anyone here is looking to upgrade from a 970/1060, and interested in selling their old card to me?   

[mikeszekely]

5 hours ago, David Hingtgen said:

So, time to upgrade my PC.  But also the worst time EVER to be in the market for new graphics card because literally every card in North America has been bought up by bitcoin miners causing a massive supply/price change.  It's insane.  I should have bought a 1070 months ago.   Now, I'm seriously looking at used 970's.  (I have a 960, but no way can I afford to go up to a 1070 now or any time soon).   

Any chance anyone here is looking to upgrade from a 970/1060, and interested in selling their old card to me?   

Not just graphics cards, there's a RAM shortage or something.  It's like double what it should cost.

I do have a GTX 970, and I'd sell it to you, but I didn't figure the GTX 10-series to be enough of an upgrade.  I'm waiting for a new Nvidia chipset before I upgrade.

[azrael]

Good thing I grabbed a 1070 before things got all cryptocurrency-crazy.

Dave, PM me. I got a 970 for you.

[kajnrig]

Unfortunately, all I have is a RX 480 8GB. Perfect if you wanted to get into this crypto craze, but that and the 1060 would hardly be enough of a step up from a 960, I'd imagine.

$500 if you want it.  What do you mean, that's a ripoff? It's $50 cheaper than the one on Newegg AND it has twice the VRAM! That's practically a steal!

[barurutor]

Actually the RX480/RX580 should be around R9 290x/390x level performance for less power draw, which makes it competitive with the GTX970 and a fair bit more powerful than the GTX960.

[David Hingtgen]

Well, thanks to azrael I now have a 970 on the way.  Should tide me over until 1070/1080 becomes more affordable.

[Mommar]

Seriously, the ridiculous price of PC components is because of bitcoin bs???

Edited February 7, 2018 by Mommar

[azrael]

4 minutes ago, David Hingtgen said:

 Should tide me over until 1070/1080 becomes more affordable.

That might be awhile.

[jenius]

The Bitcoin crash that's been happening may help eventually.

[JB0]

1 hour ago, Mommar said:

Seriously, the ridiculous price of PC components is because of bitcoin bs???

Ethereum, technically. But yes.

[kajnrig]

On 2/7/2018 at 11:49 AM, JB0 said:

Ethereum, technically. But yes.

Actually, suspicion is landing more on DRAM prices nowadays than cryptomining. Since all memory modules (DRAM and VRAM) are made by only three companies, either they're artificially raising prices (which has been criminally investigated in the past, I think...?) or it's getting harder to source/produce memory.

In other news, the Ryzen 2200G and 2400G APUs came out today, and all indications are that they're extremely good entry-level gaming chips. Some driver/compatibility optimization needs doing, of course, but they're definitely compelling products. I'm all but certain I'll use these in a couple simple builds with the nephews to play Overwatch on. Time to go parts hunting.

[mikeszekely]

53 minutes ago, kajnrig said:

Actually, suspicion is landing more on DRAM prices nowadays than cryptomining. Since all memory modules (DRAM and VRAM) are made by only three companies, either they're artificially raising prices (which has been criminally investigated in the past, I think...?) or it's getting harder to source/produce memory.

In other news, the Ryzen 2200G and 2400G APUs came out today, and all indications are that they're extremely good entry-level gaming chips. Some driver/compatibility optimization needs doing, of course, but they're definitely compelling products. I'm all but certain I'll use these in a couple simple builds with the nephews to play Overwatch on. Time to go parts hunting.

Yeah, I heard around 30fps on Ultra for Overwatch without a dedicated GPU.  I kind of want to be impressed, except I know that my next build will (like my current aging PC) have a dedicated GPU, almost certainly from Nvidia, and the new Ryzen APUs probably aren't marketed at someone like me.

On the other hand, I'd love to see them in a Surface-esque 2-in-1.