Hacker News Reader

Comment by dingo_bat | Hacker News Reader

> I wonder, what makes intel chips more expensive?

The fact that if you want a chip that powerful, there is only Intel. An example in GPUs are the NVIDIA Quaddro cards. If your frame needs more than 16GB memory to render, there is no other card available. So they sell the same chips packaged as a Quaddro at 5x the price of a Titan.

View on HN

I believe the biggest reason the workstation/enterprise GPU lines (AMD’s too) are priced so much higher than the gaming/consumer lines is because the “certifications” for CAD[1] and 3D graphics[2] software cost a king’s ransom plus require expensive testing and validation of every individual card before it’s sold.

[1] http://www.nvidia.com/object/quadro-design-and-manufacturing...

[2] http://www.nvidia.com/object/media-and-entertainment.html

> The fact that if you want a chip that powerful, there is only Intel.

How does, e.g., POWER8 compare? (And POWER9, soon?)

That said, POWER tends to be even more expensive than this.

> How does, e.g., POWER8 compare? (And POWER9, soon?)

POWER8 is not well suited for HPC workloads as their vector instructions (VMX/AltiVec) are only 128-bit wide vs Intel's latest Skylake Xeon CPUs have vector instructions (AVX512) that are 512-bit wide.

For non HPC workloads (e.g. integer arithmetic) the POWER8 CPUs were somehow competitive with Intel's CPUs when they were first released in 2015 but today AMD and Intel's CPUs are clearly faster.

From what I can tell CPU enthusiasts were not impressed by the POWER8 CPU architecture.

But for any particular version of "that powerful", the exclusivity doesn't last long. The burning need for hardware that didn't exist a couple years ago and will be mainstream in a couple more years isn't all that strong.

It is not so much a "chip that powerful" but a "chip that correct". Gaming GPUs are allowed errors because the rendered images are known to be visual output only, whereas an engineering GPU may be used for hard analysis, which cannot afford any rounding or other classes of the acceptable errors gaming GPUs are allowed.

Are you talking about double-precision floating point math? Or are you saying the gaming chips actually have issues?

It is just complete nonsense. In fact deep learning/AI chips used for neural networks usually have reduced floating point precision because they don't need it.

Deep learning is still a fairly new and small field, scientific computing workloads are usually mostly double-precision.

We don't need it because errors in any particular value aren't important - the calculations assume an amount of inherent error greater than that produced by the hardware.

nonsense

Do you have a source for those claims? It seems highly unlikely to me that certain types of cpu's or gpu's are allowed errors to cut costs. Rounding mechanics are complicated, but I'd say a bug is a bug. As mentioned in a comment by jackmott in this thread, "engineering purposes" also often actively reduce precision for increased performance.

Double precision floating point and the addition of ECC memory.

In the case of nVidia the quadro cards are the only ones sold as manufactured by the company and are incredibly high binned.

Lots of people on this thread are conflating a few things such as ML, but I work in an industry where we deal a lot with GPUs and its certainly a fact that flipped bits are significantly more common than many think. If you do any work that requires the output to be nonvolatile then you can run into issues.

Perhaps he is referring to ECC memory - high end pro model NVIDIA cards have it, and consumer cards don’t. Here is a paper showing that there are memory errors on the consumer non-ecc cards.

http://ieeexplore.ieee.org/document/5493404/?arnumber=549340...

Reply on news.ycombinator.com