Updated Knights Landing (KNL) Info.

[witeken]

TSMCs 16FF is not a real 16nm. TSMC themselves expect 15%.

That 15% is only for 16FF+.

 

[Arachnotronic]

Knights Landing looks pretty nice, and the ability to socket could help become a competitive advantage.

However, look at AMD and their FirePro W9100, 2.1 DP TFLOPS on 28nm. That's very impressive. Imagine a 16FF/+ W9100-successor. The potential could be 4-5 DP TFLOPS.

Not all FLOP/s are created equal. How many *usable* FLOP/s are there? That's where Knights Landing should excel and where NVIDIA's Tesla excels today over AMD's solutions.

 

[jdubs03]

Not all FLOP/s are created equal. How many *usable* FLOP/s are there? That's where Knights Landing should excel and where NVIDIA's Tesla excels today over AMD's solutions.

For sure, I agree, but that capability is still noteworthy. We don't really know the flops for Knights Landing, except 3+ DP FLOPS, so we could see 3.5. The transition from P1 to enhanced silvermont is very large, so it'll be interesting to see result.

 

[ams23]

Real-world measured efficiency can vary significantly from a basic manufacturer spec. And in the case of AMD, with newer Hawaii GPU's, they appear to be quoting peak TFLOPS throughput at maximum turbo frequencies and not at the GPU core clock frequency, which may be misleading depending on the workload and the throttling behavior.

 

[NTMBK]

This thing will also come as a standalone CPU! Was waiting for Haswell-E, but I think I'll wait for this instead since my FX8350 is still doing well.

What do you use your CPU for? On anything not using the full SIMD width this is going to be dog slow.

 

[xpea]

Not all FLOP/s are created equal. How many *usable* FLOP/s are there? That's where Knights Landing should excel and where NVIDIA's Tesla excels today over AMD's solutions.

and you right to say so ! paper sheet flops (or linpack or with any bench running in cache) are not very useful.
Current Xeon phi shows poor results in efficiency for many HPC workloads. One study here:
http://blog.xcelerit.com/intel-xeon-phi-vs-nvidia-tesla-gpu/

As we can see, Xeon Phi is destroyed by Kepler, so intel must increase their efficiency in KL if they want to be competitive with Nvidia Maxwell or Pascal...

 

[jhu]

What do you use your CPU for? On anything not using the full SIMD width this is going to be dog slow.

Image rendering