Literally common sense since the beginning of wireless communications and coms in general.

All you have to look at is ASIC miners. Once they had them, they were 10x faster than GPUs easily and made GPUs useless for those algos. Something very similar can happen soon.

The fundamentals are different. Bitcoin mining is not intrinsically suited to acceleration on a GPU. It is a not-very-wide serial integer operation.

AI inference on the other hand is basically just very large floating point tensor matrix multiplication. What does an ASIC for matmul look like? A GPU.

No, GPUs are not a particularly ideal architecture for AI inference, it's just inference needs way more memory bandwidth than a general purpose CPU's memory hierarchy can handle.

> What does an ASIC for matmul look like?

A systolic array, and ultimately quite different than a GPU. This is why TPUs et all are a thing.

In general with a systolic array you get a quadratic speedup. For example with a 256x256 array, it takes 256 cycles to shift operations in and out, but in doing that you accomplish 65k MACs in 512 cycles for a speedup of 128x over serial.

The tensor cores in your NVIDIA chip are systolic arrays.

Not in the same way as a TPU, and again, the ISA and overall threading architecture matter.

Sorta? If that was the full story, TPU would not be a thing.

You are entirely mistaken. The TPU and GPU are organized very differently particularly with how the memory subsystem works.

In the big picture, TPUs are systolic arrays. They don't have threading, divergence, or similar.

GPUs in the big picture are SIMT, a hybrid of SIMD and multithreading where individual data streams in SIMD are relaxed to allow them to diverge somewhat.

Memory Wise the TPU can keep the partial products right in the array. Parameters and weights are held in large on die scratch memories, and backing those are streams coming from the HBM. The TPU acts as a single giant CISC coprocessor, and has much more predictable memory and communication patterns vs a GPU, which its design exploits for higher efficiency at inference vs GPUs.

So even if they use the word "tensor" and both have HBM based memory systems, how those are actually architected is very different.

TPUs are not fundamentally different or more efficient than NVIDIA hardware. They are just cutting out the middleman.

Asic for matmul is systolic array more or less

It’s going be pretty hard to dethrone Google when even Apple is going to Google for AI with another huge billion $ contract.

In fairness, that is for their model, not their brand. In this industry, the mind share and buy in OpenAI has is still second to none, even when Gemini 2.5 Pro served the vast majority of users far better than the OpenAI models, the latter MAU trounced all competitors.

We do not know for certain as of yet, but I’d be very surprised if Apple felt the need to loudly communicate to their user base that the upcoming Siri improvements aren’t Apple Intelligence. They won’t lie of course, but Googles brand in the space isn’t nearly as strong with the layperson, whereas they more than happily shouted their use of GPT from the rooftops initially.

This year.

Apple is in the business of contracting out new-hotness until their polish is done.

Then they put their name on the thing, pretend it’s new technology – and depending on the named feature, sometimes it is!

The new phones are faster though. I use tow different gen phones and you can def feel web pages load faster even just 1 gen up.

How is it Apple’s version of the notch when Apple invented that too. I actually perfer a notch over the island. My 13 pro is going strong, but the regular has me very intrigued.

If you look at the US healthcare system, it is 100% about fixing after the fact rather than prevention. Insurance would touch anything preventative.

Pretty effing hilarious the argument your diamond is worth less because it isn't full flaws and defects.

Could be the low power ecores being in use somehow? Meteor Lake has 3 types of core with 2 lpE cores in the SOC to try to turn off the Main P and E core tile. Lunar Lake removed the lpE cores and it does feel faster when surfing pages like reddit than my 12th gen and 5000hz laptops. I also tried the Ryzen AI and it is pretty close but 20% less battery life. They get a pretty crazy 15hr battery life now.

Its really iPhone 11 when things got crazy with deep fusion. The iPhone X actually has the most detailed screen. When i look at the same photos, somehow the X has better color and details than up iphone 14 max (i have every iphone till 14). If you look at raw image you will be blown away how little has changed since even iphone 4!

But at the end of the day its all about the photographer than the equipment. Just ask chefs. 20 students same kitchen, ingredients, recipe, 20 different tasting dish.

What really drove that in was when my art teacher took a shot of me with my camera and holy shit it was one of the best pics I've seen. He just has the eye at the end of the day.

I think we should give credit where credit it due in that the people in power in Israel is using the old holocaust as an excuse to commit another. The people in power do not 100% represent the people of a nation.