[QUOTE=henryzz;431821]I suspect that this post could be amusing in 20+ years.[/QUOTE]
You mean "yet again." : ".....could [I] be[/I] amusing [I]yet again [/I]in 20+ years." :razz:

KNL is coming!!!

[url]http://hothardware.com/news/intels-72-core-xeon-phi-processor-now-commercially-available[/url]

Lowest price for CPU 2438$. :brian-e:

So waiting for first benchmark :rolleyes: Maybe it will make sense to buy one of them.

I'm trying to put it in the context of regular Intel CPUs, and I'm not sure I'm there.

Up to 3 DP TFLOPS sounds good. Or does it? My ancient R9 280X is roughly 1 DP TFLOP. How does a desktop CPU stack up? Taking a 6700k as example, I understand each core has two AVX units, each 256 bit (4xDP). 4 cores, at 4 GHz (put aside turbo and OC). Multiply together, I make that a peak rate of 128 DP GFLOPs. Someone sanity check that? That'll put a phi at around 24x the performance. Of course this doesn't take into consideration peak and actual throughput may differ.

[url]http://www.servethehome.com/intel-xeon-phi-knights-landing-generation-now-available/[/url]

Some extra info on this thing :bow::bow:

[QUOTE=mackerel;436603]Up to 3 DP TFLOPS sounds good. Or does it? My ancient R9 280X is roughly 1 DP TFLOP. How does a desktop CPU stack up? Taking a 6700k as example, I understand each core has two AVX units, each 256 bit (4xDP). 4 cores, at 4 GHz (put aside turbo and OC). Multiply together, I make that a peak rate of 128 DP GFLOPs. Someone sanity check that? That'll put a phi at around 24x the performance.[/QUOTE]
Your 6700K peaks at 256 DP GFLOPS: 4 GHz * 4 cores * 2 units * 4 DP/vector * 2 mul+add. You forgot the factor 2 for mul+add :smile:

[QUOTE=xtreme2k;436605][url]http://www.servethehome.com/intel-xeon-phi-knights-landing-generation-now-available/[/url]

Some extra info on this thing :bow::bow:[/QUOTE]

It sure sounds fun... even the "cheap" 7210 with 64 cores means you'd have 128 x AVX-512 units... even with the lower clock I think the combined power of that would be pretty amazing.

Now if we could only get our hands on a dev unit and the latest compilers, we could give George free reign on there to see what we can squeeze out of such a thing.

As-is, I wonder how something like Prime95/mprime would do by simply running it natively and watch it attempt to scale up to however many cores... we know that using the affinity scramble, it only goes up to 64 cores but maybe if you leave that part alone it can handle more?

Otherwise...what would happen? It just uses the first 64 cores and the rest are idle? Well... I wouldn't turn my nose up at 64 cores working together on a single exponent, and seeing how the scale factor per additional worker thread looks, with that fast memory tying things together. It'd be interesting.

I suppose that ideally, besides upping the # of cores it can handle, it might be worth making sure the work on each core will fit in whatever cache size, so the only time it's using the big memory is when aggregating all that work together for each iteration.

Plus, I found myself wondering how something like this would do at trial factoring, if the code were optimized for it like it is for GPUs...

Fun things ahead, at any rate.

I want one :)

I'd be happy with the base model. I'm curious.

I just grabbed a skylake xeon and it has better performance per watt than my titan black.

-- Craig

Interesting, when Intel or some vendors will start sale the KNL sustems? Intel said that they started shipping. But where is it? :confused2:

[QUOTE=Lorenzo;438103]Interesting, when Intel or some vendors will start sale the KNL sustems? Intel said that they started shipping. But where is it? :confused2:[/QUOTE]

I guess they're still in the pipeline, so to speak.

I just did a quick search and SuperMicro has some LGA 3647 motherboards listed:
[URL="https://www.supermicro.com/products/motherboard/Xeon_Phi/"]https://www.supermicro.com/products/motherboard/Xeon_Phi/[/URL]

No idea on when/price but one of them is an ATX form factor at least. Comes integrated in this workstation box: [URL="https://www.supermicro.com/products/system/tower/5038/SYS-5038K-I.cfm"]https://www.supermicro.com/products/system/tower/5038/SYS-5038K-I.cfm[/URL]

I keep waiting to see if [URL="https://www.pugetsystems.com"]Puget Systems[/URL] has anything in the KNL line available. They seem to offer the latest/greatest stuff, plus they're literally only a few miles from me. If I didn't already have a job, I'd want to work there because they seem like they have awesome toys. :smile:

[B]Madpoo[/B], Thank you for link! Nice to hear that KNL still alive :rolleyes:

Hehehe, I think we should wait ready systems not early than September (((

[QUOTE=Lorenzo;438119][B]Madpoo[/B], Thank you for link! Nice to hear that KNL still alive :rolleyes:

Hehehe, I think we should wait ready systems not early than September ((([/QUOTE]

Yeah, my guess is that when Intel says they're shipping, they may be going to government/universities that were promised some early units for their supercomputers, or something like that.

Or perhaps the companies that announced their server models are still working out some kinks and they're not really ready for sale yet.

Still, I think you can officially buy them from that Intel / Colfax "developer access program": [URL="http://dap.xeonphi.com/"]http://dap.xeonphi.com/[/URL]

They're just not as customizable as what we could get from 3rd parties like getting lower cost amount of RAM or disk options, etc. I mean, those things start at $4200 if you get one without any disks at all, so you're still paying for 6 x 16GB memory, and for me I'd be happy with 6 x 4GB or maybe 8GB on a dev platform and save some money.

Not sure CentOS would be my flavor of choice either but at least that could be changed unless there are some funky driver things that make it the only choice right now.

The inclusion of a "Parallel Studio XE Cluster Edition Named User License" is nice though...I don't know how much that would be separately but if you're planning to actually use something like this for dev work, I imagine you'd need it.

Whatever the case, it may be a while before I get my hands on one. :smile: Probably when HP comes out with a Proliant model, and I can work the #'s to show how awesome this would be for a virtual machine host. I'm imagining a cluster of a few of these compared to a larger amount of boring old Xeon E5 processors. :smile: A virtual host is definitely a good use of a CPU with many cores... things like SQL, IIS, and a lot of the other stuff I use, take good advantage of multiple threads.