So the security patch slows the chip down anywhere from 5-30%. That is big for some people. I wonder if Intel will be sued for this ? I also wonder if Intel is still laughing and thumbing their nose at AMD because of Intel being #1 and AMD #2 ? God Bless


http://money.cnn.com/2018/01/0...-security/index.html

Apple should have stuck with RISC technology.

Apple says the iPhones are effected also.
those are ARM RISC

It's not just Intel. It apparently affects Intel, AMD, and ARM processors. It's just the risk is greatest to Intel chips.

You're supposed to replace the chip to completely eliminate the threat. Replace them with what, exactly? An abacus?

quote:

Originally posted by whanson_wi:
You're supposed to replace the chip to completely eliminate the threat. Replace them with what, exactly? An abacus?

quote:

Originally posted by whanson_wi:
You're supposed to replace the chip to completely eliminate the threat. Replace them with what, exactly? An abacus?

A newly made processor without the flaw, once they're designed and released.

quote:

Apple should have stuck with RISC technology.

What do you think is behind the CISC microcode? It's all RISC at the gate level.

I just read a story that said updating the software is enough to fix it.

quote:

Originally posted by Ackks:
I just read a story that said updating the software is enough to fix it.

I think the software fix is what causes the slow down in performance. The hardware flaw creates a security hole. To plug the hole, the software patch resolves the security hole at the expense of performance.

Conspiracy hat = ON

Was this really a flaw....or gubmint back door??

Better yet.....the replacement "FIX" chip contains the real gubmint back door!

quote:

Originally posted by ScreamingCockatoo:
Apple should have stuck with RISC technology.

It isn’t just Apple products that are affected.

quote:

Originally posted by RogueJSK:
It's not just Intel. It apparently affects Intel, AMD, and ARM processors. It's just the risk is greatest to Intel chips.

From what I've read the threats (Meltdown Spectre is what I believe they are called) have not been observed in real life - only theoretical.
I could be wrong but that is what I've gathered.
Plus a computer must already be vulnerable enough to have malicious "code" installed whatever that might be > THEN it is possible for the attack to succeed.
If all of that is so then this has been quite over-hyped.

quote:

Originally posted by Jimineer:
I think the software fix is what causes the slow down in performance. The hardware flaw creates a security hole. To plug the hole, the software patch resolves the security hole at the expense of performance.

I don't have much of a choice unless I want to replace the chip in every desktop and laptop I own.

I read that one hole involves a form of "Speculative prefetch" that loads predicted passwords and account info into the processor memory in anticipation of it's need. Negating this process could potentially slow the thru put down I suppose, but 30% seems a bit much to me.

RMD

quote:

Originally posted by smschulz:

quote:

Originally posted by RogueJSK:
It's not just Intel. It apparently affects Intel, AMD, and ARM processors. It's just the risk is greatest to Intel chips.

From what I've read the threats (Meltdown Spectre is what I believe they are called) have not been observed in real life - only theoretical.
I could be wrong but that is what I've gathered.
Plus a computer must already be vulnerable enough to have malicious "code" installed whatever that might be > THEN it is possible for the attack to succeed.
If all of that is so then this has been quite over-hyped.

I think the new bug was discovered by researchers fairly recently. So potential attackers were unaware of the security hole - until the leak of the bug got out before software could be patched.
From what I’m reading most of us won’t see the slowdown and the problem will impact large server type applications.

One does have to wonder if some researcher stumbled upon a backdoor.

quote:

Originally posted by Ackks:

quote:

Originally posted by Jimineer:
I think the software fix is what causes the slow down in performance. The hardware flaw creates a security hole. To plug the hole, the software patch resolves the security hole at the expense of performance.

I don't have much of a choice unless I want to replace the chip in every desktop and laptop I own.

Chances are you'll need to change your motherboard. Doubt they'll manufacture chips that will work with old motherboards.

quote:

Originally posted by rduckwor:
I read that one hole involves a form of "Speculative prefetch" that loads predicted passwords and account info into the processor memory in anticipation of it's need. Negating this process could potentially slow the thru put down I suppose, but 30% seems a bit much to me.

RMD

Speculative prefetch and speculative execution aren't about passwords and account info. Basically any time the processor has to decide a branch it loads the data for both branches and in some cases executes instructions for both branches before deciding which branch is the "right" one. It uses idle clock cycles in spaces between instruction executions on the different branches. In some highly performance-specific applications the performance hit could be WORSE than 30%.

EVERY processor that is remotely concerned with performance does speculative prefetch and speculative execution.

The exploits are about a way to get user programs to access OS data that has been prefetched into the cache in certain processors. Actually getting the right data loaded into the cache and then getting it out in a usable way is not trivial, though.

If the performance hit via security fix is true, business servers & data centers are going to be affected in a huge way!!

quote:

Originally posted by maladat:
It uses idle clock cycles

Years ago, I was on a consulting contract, supporting mainframe computers at a major customer's site.

During a meeting, somebody brought up the fact that certain usage reports showed more CUP time than wall clock time, and asked how that could be.

I was aware of the glitch in the reports -- somebody in the OS development group had made a math error when coding the calculations for the report, and the error had been caught and the fix was integrated in the next release of the OS.

Somehow, my mouth got ahead of my brain, and I started to explain the VTOS (Virtual Time Operating System) that they were running: the instruction fetch hardware looked ahead, realized that there were CPU cycles that were not going to be used in the middle of the night next week, so it used them as needed during current periods of heavy load.

I almost sold that concept.