More Intel leaks.. this one is not good though

[Taf the Ghost]

3 minutes ago, M.Yurizaki said:

I don't really feel AMD has something that instills confidence past 2020. Especially when their lead processor architects left. Maybe they have someone brewing up the next big thing, but it's an uncertainty. At least Intel has more or less created a pattern for itself and stuck with it.

 

But I guess I should stop here before I incur the wrath of people wanting to throw molotovs at me for being an apologetic.

AMD's GPU side would be more of a concern with Raja gone, in theory. CPU side, Papermaster is still there. Keller wasn't responsible for Zen's design, though he was important for getting it through the design process. (He may be more responsible for moving AMD to a modular design approach, though, but he doesn't need to be there to continue with it.)

[TenThousand]

52 minutes ago, Princess Cadence said:

Thing is getting so serious with all the speculation though that now I seen a thread about what CPU to get for high end gaming and people said Ryzen instead of the i7 8700k... even when the main article on the subject already goes like:

What an intense day this has been.

That benchmark is fundamentally flawed. It uses a gpu/driver architecture that's been known to not perform very well on that os. Not to mention the settings used. It's more of a gpu test than it is a cpu one.

 

The change would be more apparent if it were tested with a better gpu, lower settings, or a worse cpu. Let's be real here, not that many people are using an 8700k. You can't seriously expect people to go buy one when a) there's no supply b) it's too expensive. All so the performance deficit won't be as noticeable.

[Mira Yurizaki]

1 minute ago, Taf the Ghost said:

AMD's GPU side would be more of a concern with Raja gone, in theory. CPU side, Papermaster is still there. Keller wasn't responsible for Zen's design, though he was important for getting it through the design process. (He may be more responsible for moving AMD to a modular design approach, though, but he doesn't need to be there to continue with it.)

According to the interwebs, they credit Keller as the lead designer along with Michael Clark. Papermaster is a C-level executive who wouldn't have anything to do with the design.

 

But since nobody seems to have mention Clark in any detail (or I glossed over it), that eases the concern a bit.

[cj09beira]

1 minute ago, Taf the Ghost said:

AMD's GPU side would be more of a concern with Raja gone, in theory. CPU side, Papermaster is still there. Keller wasn't responsible for Zen's design, though he was important for getting it through the design process. (He may be more responsible for moving AMD to a modular design approach, though, but he doesn't need to be there to continue with it.)

they seem to know very well what their weakness are, as the only thing we know about navi is that its a scalable gpu, which is the main problem with gcn, ao as long as they make it work, it should be good, and they also need to reduce voltage it doesn't help their case at all

[Princess Luna]

1 minute ago, TenThousand said:

.

I just meant that there are people rather having fun exaggerating the true extension of the these "performance losses"... for the instance of saying a Ryzen 5 1600 is going to be better at high refresh rate gaming than an i7 8700k because the i7 will lose 30% of gaming performance is completely ludicrous... yet there's plenty locking their mind around this already.

[AresKrieger]

Yeah whatever doesn't affect me directly as for indirect affect it will likely be minimal its not like security has been working in general anyway, if it accepts use input it can be hacked if it is connected to the internet it can be hacked remotely, there is no information left to be leaked on me if you are willing to look for it I'd bet given the sheer number of security breaches.

5 minutes ago, Princess Cadence said:

yet there's plenty locking their mind around this already.

No really this is more so a scenario of if you hated intel yell that they suck at the top of your lungs, if you liked them or indifferent you want more info, ie no one changed their minds

[Taf the Ghost]

8 minutes ago, cj09beira said:

they seem to know very well what their weakness are, as the only thing we know about navi is that its a scalable gpu, which is the main problem with gcn, ao as long as they make it work, it should be good, and they also need to reduce voltage it doesn't help their case at all

I'm going to be a contrarian and say that Raja leaving probably helps AMD's GPU side of things. Two of RX Vega's core features for gaming simply don't work/aren't on. I can appreciate that AMD is moving their GPU side of things to having a mid-level line and a Compute line, Vega just didn't meet almost anything it needed to do in the Consumer space.

[AresKrieger]

8 minutes ago, gabrielcarvfer said:

Seems to be implemented since 2012, so probably Ivy Bridge.

This is what we need actual info, the other things I'm trying to find (but failing hard because flu) is if this is a remote hole or if this required a hole already present it seems to be the latter but info is hard to read with a fever and dizzyness

[AnonymousGuy]

I'm going to bet right now that the fix only has a small effect on one benchmark.  None of this "up to 30%" bullshit that is being reported to inflate the drama.

 

Getting annoyed with "news" in general trying to make everything sound like it's a fucking tragedy.

[johnukguy]

1 minute ago, AresKrieger said:

This is what we need actual info, the other things I'm trying to find (but failing hard because flu) is if this is a remote hole or if this required a hole already present it seems to be the latter but info is hard to read with a fever and dizzyness

Stay hydrated. That's kind of more important right now.

[Princess Luna]

3 minutes ago, Arokhantos said:

If performance go's down i will expect class action lawsuit cos you no longer get advertised speeds.

Technically the "speed" is the very same it has always been, your CPU is just doing extra work to achieve the same result it did before in order to be "safe".

 

Intel is not getting sued.

[Mira Yurizaki]

5 minutes ago, Arokhantos said:

If performance go's down i will expect class action lawsuit cos you no longer get advertised speeds.

Which is baseless. It's a voluntary software change that needs to affect the performance. If I install Windows 7 on a system with an affected processor, the performance isn't going to change.

 

This is unlike Apple's situation where they more or less forced people to update iOS which had performance degrading mechanisms.

 

EDIT: Also Intel never claimed absolute performance benchmarks. Only relative ones with a huge fat disclaimer basically saying "we got these numbers from our labs, your millage may vary"

[cj09beira]

3 minutes ago, Taf the Ghost said:

I'm going to be a contrarian and say that Raja leaving probably helps AMD's GPU side of things. Two of RX Vega's core features for gaming simply don't work/aren't on. I can appreciate that AMD is moving their GPU side of things to having a mid-level line and a Compute line, Vega just didn't meet almost anything it needed to do in the Consumer space.

the question is why, my guess is that its driver related as i dont believe that they could fuck up most of the new features in one go. hopefully we get said features with vega 20, 

amd also needs a double precision card as the last one is very old now, but if amd manages to make scalable gpus most of their problems dissapear (until they need to scale each die pass 64rops and 64 cus )

 

[ItsMitch]

1 minute ago, Princess Cadence said:

Intel is not getting sued.

 

How can you be so sure? I've seen people sued for less and a decent enough lawyer can probs try for Negligence

[AresKrieger]

Just now, AnonymousGuy said:

I'm going to bet right now that the fix only has a small effect on one benchmark.  None of this "up to 30%" bullshit that is being reported to inflate the drama.

even if it does it's not like this hotpatch is the permanant solution, its a reactionary plug the hole with a towel to prevent the boat from sinking, well get more info in time I'm sure that effects will be minimal once intel makes a concrete fix and if not then people can complain with legitmacy

[Shakaza]

10 minutes ago, Arokhantos said:

If performance go's down i will expect class action lawsuit cos you no longer get advertised speeds.

But Intel is not liable for the reduction in speed because they couldn't possibly have predicted that this bug would occur, or the performance hit caused by the fix. It would be extremely unfair to Intel for them to get sued over this (a tragedy, eh? I know how much we all love Intel. ). Now, if it could be proven that Intel chose to conceal this bug from the public in order to gain every advantage in speed possible...well, you might have a case there, but I don't think that happened, and even if it did, good luck provin' it, mate!

[DeadlyTitan]

18 hours ago, SC2Mitch said:

Any word on what families will be affected, Skylake, etc?

Probably the families that have bought intel chips for Christmas. 

[emosun]

Is there a list of which cpu's are affected?

I know it said all within last ten years , but is that basically just every 64bit cpu and 32 bit isn't affected or what? is it basically everything netburst or everything after netburst or what?

[Princess Luna]

5 minutes ago, SC2Mitch said:

How can you be so sure? I've seen people sued for less and a decent enough lawyer can probs try for Negligence

Oh I mean allow me to correct myself, it can get sued, I just really believe no court will give victory to the suing part.

[Mira Yurizaki]

6 minutes ago, SC2Mitch said:

How can you be so sure? I've seen people sued for less and a decent enough lawyer can probs try for Negligence

If Intel can be sued successfully for this then maybe we should file class action lawsuits for anyone who's had a major security bug. Which is effectively every major player.

 

And all that means is the lawyers win. Class action lawsuits are dumb and we should be petitioning the FTC or the equivalent to fine the company instead.

[AresKrieger]

2 minutes ago, DeadlyTitan said:

Probably the families that have bought intel chips for Christmas. 

It's actually a non issue for most consumers in this scenario (at least directly) the issue is more so a scenario like this bank A stores login data on a server, someone gained access to it and then used these exploit to make client passwords legible.

 

If someone gains access to your home PC your likely screwed regardless of this bug except in some extreme circumstances

[leadeater]

8 minutes ago, SC2Mitch said:

How can you be so sure? I've seen people sued for less and a decent enough lawyer can probs try for Negligence

Probably because it's not Intel themselves causing the performance loss, its the kernel patch. So do we sue Microsoft for slowing down Intel processors?

[ItsMitch]

Just now, leadeater said:

Probably because it's not Intel themselves causing the performance loss, its the kernel patch. So do we sue Microsoft for slowing down Intel processors?

Sure why not, been an amusing year so far. 

[Taf the Ghost]

9 minutes ago, cj09beira said:

the question is why, my guess is that its driver related as i dont believe that they could fuck up most of the new features in one go. hopefully we get said features with vega 20, 

amd also needs a double precision card as the last one is very old now, but if amd manages to make scalable gpus most of their problems dissapear (until they need to scale each die pass 64rops and 64 cus )

 

Not sure why Vega has the issue with the Fast Path or new shaders, but there was an old rumor of an error at the design level that wasn't fixable. Though HBCC wasn't turned on, in driver, until about October, so the drivers were way behind. Vega is a major transition for AMD, but that doesn't mean it went well.

 

The MCM approach is simply what is needed in GPUs. They're getting small enough that running a cluster of 150 mm² GPUs will just work better. Getting them to work together, in an opaque manner, is the key issue.

[Taf the Ghost]

https://www.phoronix.com/forums/forum/phoronix/latest-phoronix-articles/998707-initial-benchmarks-of-the-performance-impact-resulting-from-linux-s-x86-security-changes/page4

 

Quote

It seems NVidia users can't run the patched kernel. The patch changes a previously non-GPL symbol to a GPL one, and the nvidia module build fails:

FATAL: modpost: GPL-incompatible module nvidia.ko uses GPL-only symbol 'cpu_tlbstate'

Good times 

This would be on Linux, so it almost assuredly won't effect Windows. However, this is going to be just one more headache invovled.