Intel Iris Pro 5200 and other current (Intel)APUs

[ocl]

Hi,
I understand that it's probably not of such an importance due to performance reasons, but is the Iris Pro 5200 GPU supported? If not, what performance do you estimate to expect of it (I mean the pure GPU computing performance, excluding the CPU)?

"Intel's OpenCL Beignet Project Is Gaining Ground":
http://www.phoronix.com/scan.php?page=ne...px=MTQzOTI

[radix]

probably about a GT650M

[ocl]

Ok, but since it's neither AMD, nor Nvidia, it's probably hard to say what the performance of this "GT650M" would be like (I mean an expected factor of like 3x more if it's going to be more AMD-like, not Nvidia-like), right?

And if it's "GT650M"-like fast, support one day would be nice, again, especially if the performance is going to be more AMD like (Nvidia has better Linux driver through).

[radix]

no speed is terrible, you could research this yourself.

[Rolf]

40 shaders at 0.2/0.3Ghz ?
It's a joke really.
It's more pathetic than GeForce 410M, which was the lousiest mobile Nvidia card in 2011.
Of course, if you like doing things the wrong way, sure, knock yourself out.

[ocl]

Ok, research (radix, actually Rolf needs to research this too ):
Intel 40 EUs -> 320 ALUs
GT650M has 384
1Intel shader != 1Nvidia shader

Rolf, 0.2GHz is the non-turbo modus, which actually saves battery life and is very good.

Anyway, maybe "The Team" could think about implementing it one day since the APU performance is increasing every year and not everyone can afford an extra 200€ card for hashing. These APUs are becoming quite performanceful I mention Intel because it supports open source drivers especially good.

[radix]

Until it becomes competitive theres really no reason to do this. The only reason that lower end cards are supported at all is because the driver package from AMD/nvidia support them so there is not much change needed.

[ocl]

Ok, I understand you guys are probably more info cracking/hacking than me and are more likely to buy a Desktop PC with a >200€ card just for hashing alone, not gaming. But think about how many Intel HD 4xxx/5xxx graphics (latest and fastest generation ATM) Desktop PCs and notebooks users are out there, where the users can't afford -not enough space- a >200€ card (e.g. because it's a notebook, Intel NUC or Gigabyte BRIX) or don't have a card at all.

My guess is if utilizing not only the CPU but also the GPU inside the CPU, the performance could easily double (or better if it's an Iris Pro 5200), which at least for so many would be nice to see. (Yes I know this is nowhere near a e.g. 7950)

Maybe it can't be done now because Intel's opencl drivers aren't ready jet, but one day. See link in first post.

Sad enough hashcat is not open source, imagine something happens to the developer, the development of hashcat would stop and it would stop working and disappear very quickly.

[epixoip]

1. Intel's libOpenCL only supports CPU on Linux. No GPU/APU support.

2. GPUs are not magic go-fast devices. The microarchitecture and ISA have to be well-suited for the task at hand. As it stands, Intel GPUs have very minuscule raw compute power, and their ISA is not optimal for password cracking. Most modern-day CPUs with XOP or AVX2 support will be faster than an Intel GPU.

3. oclHashcat is oriented toward advanced users and competitive users, not toward the casual user. If you aren't dedicated enough to buy a proper GPU, then Hashcat probably isn't for you.

4. Development would not stop if atom disappeared.

[ocl]

Alright, thanks for the clarification, epixoip.