Recent content by DavidC1

You forget that SMT adds validation complexity. That increased difficulty not only causes potential delays but less focus on other parts of the core. If you need extra 1 month per generation then in 10 generations that's nearly a year of delay. Nevermind increasingly sophisticated hacking and...

Pentium 4 did not have enough execution resources to take full advantage of SMT, nevermind SMT4 which requires IBM's level of focus to get advantage of it. It's a 1-wide core helped by Trace Cache which was nowhere big enough to make up for the lack of issue width, double pumped but simple ALUs...

Everyone thinks that contra revenue was just Intel giving rebates for using their chip, misses the amount of what they were doing. If say you wanted to make a new company based on Intel's new Tablet offerings, you'd call up Intel and the easiest thing would be to rebrand their reference design...

~20-25%. Rest are all design and due to massive ARM share, all the top engineers going there. All the smart engineers go to where they get paid and are treated well. In 2000's that was Intel. In 2025 that is ARM vendors. Even ARM vendors have defections such as with Gerald Williams III and...

It is minor differences in SMT implementations that caused the difference originally. If you compare Sandy Bridge generation, they have lot more shared resources versus Ryzen's version of SMT where more resources are distributed and per thread. So Intel wanted 2-3% better ST performance. Intel...

I updated my post. Yes they are purely sold in big iron where they have very high margins. Everything they do is entirely optimized for Enterprise. It's not realistic for a PC chip that goes from 5W Tablet to 250W 6GHz enthusiast desktop.

That's because IBM actually puts lots of effort into gaining more from SMT, unlike AMD/Intel. AMD/Intel's implementations are basically barebone, and adds barely 5% transistors to a core, nevermind the whole chip. IBM's SMT added 25% extra transistors just for SMT. And those are careful...

It isn't so much the x86 that Intel is responsible for, but for all the other things they do, which is way more important. The support they give for the reference devices to be made on is seriously impressive. AMD barely does anything in that regard. They make CPUs and they make GPUs, that's...

The whole thing about artificially segmenting Atom to be 1/10 performance is what happens under management trained to be beancounters such as with Paul Otellini. Back then phone chips were selling $20 so from a beancounter viewpoint the profit margins were really no different than selling real...

It's just like the Core vs Netburst scenario. They called Core as combining Netburst with Pentium M, but it was a natural evolution of Pentium M. A total pivot with an entirely new architecture doesn't happen. Zen was new for AMD, but many fundamentals were shared with pre-Bulldozer chips and...

30% higher performance would put almost all of the titles in the desired 60 fps territory. That's what 4070 class performance is. Intel missed another opportunity. Enthusiasts drive sales, not average joes. They are ok with Intel HD Graphics.

Not sticking to their decisions. 60% higher perf is 4070 territory. With the rumors that its clocked higher 4070 Ti may have been possible. Q4 is 2 quarters from 9000 and 5000 competition. They would have been able to either price it higher for much better margin or be a perf/$ king like with...

Everything matters, down to the UI. The entirety of the internet is being optimized for tiny screen real estate. That's why there's complaints that it's slowing down for PC users such as the hamburger interface icon, needing extra steps to accommodate for 5-inch touch. Now tell me in light of...

I'll tell you why they got blindsided. My first Windows computer was Windows 95 with Plus! extension. It had a feature that could have been a predecessor to current mobile OSes. It was called something like Internet feature. When you enabled that, the icons went from double clicking to open it...

And you don't need "heavy branching" to make clustered decode work. Branching happens quite often enough that Tremont reaches 6-wide quite often even without the load balancer.