OC3D Forums

OC3D Forums (https://forum.overclock3d.net/index.php)
-   OC3D News (https://forum.overclock3d.net/forumdisplay.php?f=47)
-   -   Intel CPUs hit by new "CacheOut" attack (https://forum.overclock3d.net/showthread.php?t=93825)

WYP 27-01-20 11:32 PM

Intel CPUs hit by new "CacheOut" attack
 
AMD processors are unaffected.

https://overclock3d.net/gfx/articles...181036228l.jpg

Read more about Intel processors being impacted by a new vulnerability called "CacheOut".

NeverBackDown 27-01-20 11:55 PM

I'll be honest I've lost count and the names of all the vulnerabilities. I should just switch to AMD at this point and not have to worry. Problem is Intel is still faster for me for my programming work... Bummer.

Dark NighT 28-01-20 12:45 AM

This is almost comical how many times intel keeps getting hit by new vulnerabilities, its like a bad joke and them sticking to 14nm++++++ doesn't help their cause.

Kleptobot 28-01-20 01:59 AM

I am by no means an expert in x86 architecture but it seems to me that this avalanche of exploits makes sense for several reasons:

1) This type of exploit, speculative execution exploits, had not previously been used. And so with the first instance of these attacks there has been massive interest in what other ones are possible

2) Intel had a larger share of the market and so offered a bigger target

3) Intel had a much more mature architecture, and had made massive IPC gains through speculative execution to improve the efficiency of their pipeline

4) this one is pure conjecture but maybe the ring bus architecture is more vulnerable to this type of exploit because of the way each core is connected to the cache

So if anything i would say there are only going to be more of these exploits, each with their own name :)

Gothmoth 28-01-20 09:42 AM

Quote:

Originally Posted by NeverBackDown (Post 1015541)
I'll be honest I've lost count and the names of all the vulnerabilities. I should just switch to AMD at this point and not have to worry. Problem is Intel is still faster for me for my programming work... Bummer.


are you sure?
is the compiler advantage still that great even when you consider buying at the same price point?


i do 3d rendering not coding but when compiling under linux the 3950x seems to blow intel out of the water.

in a linux compiling benchmark (3950x vs. 9980xe) i have seen recently, not one test was won by intel.


but i know when compiling it depends a lot on the benchmark (compiler used).
is are you depended on AVX512?

tgrech 28-01-20 11:11 AM

Quote:

Originally Posted by Gothmoth (Post 1015545)
are you sure?
is the compiler advantage still that great even when you consider buying at the same price point?


i do 3d rendering not coding but when compiling under linux the 3950x seems to blow intel out of the water.

in a linux compiling benchmark (3950x vs. 9980xe) i have seen recently, not one test was won by intel.


but i know when compiling it depends a lot on the benchmark (compiler used).
is are you depended on AVX512?

Yeah most compilation tasks in a typical stack are defined as embarrassingly parallel so the top end AMD parts should wipe the floor here, especially with modern or open source software/compilers.

My old FX8320 would even outperform my friends i7 6700K in many C/C++ compilation tasks.

But there's still a lot of old proprietary software out there still used heavily in some compilation stacks, Microsofts C++ linker was a point of contention for a while.

Quote:

Originally Posted by Kleptobot (Post 1015544)
I am by no means an expert in x86 architecture but it seems to me that this avalanche of exploits makes sense for several reasons:

1) This type of exploit, speculative execution exploits, had not previously been used. And so with the first instance of these attacks there has been massive interest in what other ones are possible

2) Intel had a larger share of the market and so offered a bigger target

3) Intel had a much more mature architecture, and had made massive IPC gains through speculative execution to improve the efficiency of their pipeline

4) this one is pure conjecture but maybe the ring bus architecture is more vulnerable to this type of exploit because of the way each core is connected to the cache

So if anything i would say there are only going to be more of these exploits, each with their own name :)

Points 1 and 2 hold a lot of truth, but regarding point 3), speculative execution itself has been around for decades and is widely used in everything from ARM processors up, and Intel's current performance lead is from process, they're behind AMD when you peg them clock for clock, so AMD are likely using speculative execution more widely and effectively (As the branch predictor is the key component when it comes to performance nowadays), however the maturity of their architecture does mean that it is very well studied and people have had much longer to find vulnerabilities in Intel's methods.

NeverBackDown 28-01-20 04:31 PM

Quote:

Originally Posted by Gothmoth (Post 1015545)
are you sure?
is the compiler advantage still that great even when you consider buying at the same price point?


i do 3d rendering not coding but when compiling under linux the 3950x seems to blow intel out of the water.

in a linux compiling benchmark (3950x vs. 9980xe) i have seen recently, not one test was won by intel.


but i know when compiling it depends a lot on the benchmark (compiler used).
is are you depended on AVX512?

It's not a large enough difference to warrant spending money to switch my system over and the downtime that comes with it. Afaik Intel is still faster for compiling though I haven't really looked all that much into it in a while


All times are GMT. The time now is 03:04 PM.

Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2021, vBulletin Solutions, Inc.