980ix AES hardware

[Rubycon]

Would this chip be ideal for hacking or what?!

There's an AES test in Everest and it's off the scale!

 

[SunnyD]

No. All it does is accelerate the encode/decode process for the data, not brute-force cracking of keys.

 

[eva2000]

indeed bloody fast

 

[Rubycon]

No. All it does is accelerate the encode/decode process for the data, not brute-force cracking of keys.

Well that makes sense. What other things could this be used for?

indeed bloody fast

23x200...nice. What cooling?

 

[Fox5]

The use case pitched for that has basically been "now it's feasible to encrypt your entire hard drive".

 

[Mark R]

Would this chip be ideal for hacking or what?!

There's an AES test in Everest and it's off the scale!

It would be better than existing CPUs.

It wouldn't be ideal though. I estimate that brute-forcing an AES key with an i980x would take around 1000 million million million million million million million million million million CPU-years.

 

[Cogman]

No. All it does is accelerate the encode/decode process for the data, not brute-force cracking of keys.

Depending on how the bruteforce is implemented, the decoding process is one of the slower ones, so this would be a boon to the hackers.

Though, it will still take an unworldly amount of time to actually break an AES encrypted file. Even going 20x faster.

 

[The-Noid]

Video cards are MUCH MUCH faster at bruteforcing.

 

[SunnyD]

Depending on how the bruteforce is implemented, the decoding process is one of the slower ones, so this would be a boon to the hackers.

Though, it will still take an unworldly amount of time to actually break an AES encrypted file. Even going 20x faster.

I suppose. But generally brute forcing an AES key is the stupidest way to do it. But I suppose it would help.

 

[Cogman]

I suppose. But generally brute forcing an AES key is the stupidest way to do it. But I suppose it would help.

There aren't exactly a whole lot of other choices. There are ways to speed up the brute forcing using a similar key attack, however, the decryption process is still the major time consumer.

 

[Cogman]

dp

 

[eva2000]

23x200...nice. What cooling?

EK Supreme LT + TFC 480 + MCP650 H20 in 28-33C room ambients

1.456v idle vcore and 1.408-1.416v load with loadline calibration disabled (vdroop as per intel spec).

just noticed that chart you posted at http://www.legitreviews.com/images/reviews/1245/sandra_crypto.jpg must have cryptographic and sha256 values swapped around by accident hehe

 

[Rubycon]

I suppose I should not be complaining about upper 70s to lower 80s with 211x21 1.4V Linpack temps - aircooled with VX.

The chart is from Sandra - never liked that software personally.

EK Supreme LT + TFC 480 + MCP650 H20 in 28-33C room ambients

1.456v idle vcore and 1.408-1.416v load with loadline calibration disabled (vdroop as per intel spec).

just noticed that chart you posted at http://www.legitreviews.com/images/reviews/1245/sandra_crypto.jpg must have cryptographic and sha256 values swapped around by accident hehe

 

[Voo]

Even if someone wanted to bruteforce AES (ahm well.. I hope he's got time), they wouldn't use a full x86 CPU, there would be better solutions for that - not cheap, but hey you'd also need a whole lot of 980s..

You'd also get a much better performance out of some GPUs - after all that stuff is probably one of the best non trivial usage scenarios for a SIMD architecture..

 

[Rubycon]

Even if someone wanted to bruteforce AES (ahm well.. I hope he's got time), they wouldn't use a full x86 CPU, there would be better solutions for that - not cheap, but hey you'd also need a whole lot of 980s..

You'd also get a much better performance out of some GPUs - after all that stuff is probably one of the best non trivial usage scenarios for a SIMD architecture..

Well technically the CPU has dedicated AES hardware so you're not using x86. Someone correct me if that's wrong but that's usually how hardware works.

And yes GPUs are very powerful both AMD and nVidia parts.

 

[Voo]

Well technically the CPU has dedicated AES hardware so you're not using x86. Someone correct me if that's wrong but that's usually how hardware works.

Well afaik Intel added those instructions to the x86 ISA, so you're using it, but just .1% of the whole transistor count probably

 

[Cogman]

Well afaik Intel added those instructions to the x86 ISA, so you're using it, but just .1% of the whole transistor count probably

Well, sort of.

The AES hardware is a part of the CPU, but that doesn't mean that it is part of the x86 instruction set. You could say that it is an extension of the x86 architecture, but not an actual part of it. They are careful to do that so you don't have each tech they have invented coming into public domain once their patent expires (which the x86 patent has expired).

In other words, AES, MMX, SSE1->4, 3dnow, ect, are not part of the x86 instruction set, rather they are extensions to the ISA. It has to be specifically programmed for, and programs that use it would not be considered strictly x86 (or i686 if you wish as i386 is pretty old now).