klelatti

36 ms·

511.
by klelatti 3y ago
> “My belief is in the next 5 to 10 years, RISC-V will take over all the data centers,” Keller told EE Times, Isn't this essentially impossible?
512.
by klelatti 3y ago
Amazing site. The fate of the South Bank studios makes me sad though. > The entire building including the 24-storey Kent House and all the studios were to be demolished, to be replaced with a 31-storey tower containing mostly luxury apar
513.
by klelatti 3y ago
Interesting. I sort of expected that for basic FP / SIMD but not for the full range - eg trig functions. Is that wrong?
514.
by klelatti 3y ago
Interesting - is there a reference for this?
515.
by klelatti 3y ago
Well all modern architectures have FP and SIMD so it’s not as though complex (microcoded) instructions have been banished.
516.
by klelatti 3y ago
> Otherwise, the intention was that bigger RISC-V cores would detect a conditional branch over a single instruction in the decoder and perform macro-op fusion into a conditional instruction. This seems like an overhead compared to actual
517.
by klelatti 3y ago
It may or may not be a dud but I’ve no idea how even a resounding failure could ‘destroy’ a company as profitable and with as successful a product as the iPhone.
518.
by klelatti 3y ago
I remember buying OmniOutliner on my first Mac just so I could try Kinkless. Can't believe that it's now 17 years ago!
519.
The big question of how small chips can get(ft.com)
2 points by klelatti 3y ago|0 comments
520.
by klelatti 3y ago
I did a rough calculation and came up with a current value of $80bn for Nvidia's offer for Arm based on the current Nvidia share price. An expensive regulatory loss for Masa given Arm will probably float for a lot less. Of course had t
521.
by klelatti 3y ago
This. The value of a stock should (in theory) be based on discounted cashflow with the emphasis on cash. A company can generate massive earnings but if it needs to invest huge amounts of capital to support those earnings then it should have
522.
Tenstorrent Partners with LG(morethanmoore.substack.com)
4 points by klelatti 3y ago|0 comments
523.
by klelatti 3y ago
> namely anybody can create FREELY a RISC-V microarchitecture AT WORLWIDE SCALE The problem with this argument is that it ignores the cost of creating the microarchitecture. It’s almost certainly cheaper to license an Arm A series core t
524.
by klelatti 3y ago
I’m interested to hear what you think the structure of the market for licensed RISC-V cores will be. How many vendors will there be and where will the revenue to support all these vendors will come from? At the moment licensing cores is not
525.
Britain's headlong slide from silicon supremacy to silicon irrelevance(edconway.substack.com)
6 points by klelatti 3y ago|3 comments
526.
by klelatti 3y ago
IBM, of course, did use (re-microcoded) 68000’s in a PC product - the XT/370. With a modified 8087 on a couple of cards these were used to emulate an S/370. https://en.wikipedia.org/wiki/PC-based_IBM_mainfram
527.
by klelatti 3y ago
The 68030 does look like a curious choice for a machine launched in 1988. Wikipedia says that they considered the Motorola 88000 but with a source that I've not been able to track down.
528.
by klelatti 3y ago
Also pretty implausible in 1990 say that Arm would be the highest performing surviving RISC architecture [1]. [1] yes, I know ARM64 looks nothing like the ARM ISA of 1990.
529.
Google “We have no moat, and neither does OpenAI”(semianalysis.com)
2455 points by klelatti 3y ago|1039 comments
530.
Intel Emerald Rapids Backtracks on Chiplets – Design, Performance and Cost(semianalysis.com)
1 points by klelatti 3y ago|0 comments
531.
Cloud Exit Pays Off in Performance Too(world.hey.com)
17 points by klelatti 3y ago|1 comments
532.
by klelatti 3y ago
Hi thanks so much for editing the comment. Much appreciated. Completely agree with your succinct summary of the what RISC is all about and the 801's place in the story. I've heard others call Cray's CDC's computers the f
533.
by klelatti 3y ago
Author here. Thanks for the comment and feedback. Really interesting point on MIPS vs RISC-I approaches. I was a bit puzzled by the IBM 801 comment as the second para of the previous post mentions the 801 and links to an earlier post that i
534.
by klelatti 3y ago
Largely agree but your wording is interesting - ‘generally considered’, ‘staying out of politics publicly’. I think that there has been a degree of constructive ambiguity in the past. No more and it’s 100% clear now that the King / Que
535.
by klelatti 3y ago
In my view the elephant in the room is that the monarchy in the UK has demonstrably failed to be a check or balance on a party with a majority in Parliament. Without a written constitution there is no limit on what a party with a majority i
536.
Microsoft in furious attack on UK as major gaming deal blocked(bbc.co.uk)
51 points by klelatti 4y ago|85 comments
537.
by klelatti 4y ago
Seems unlikely. 'RISC I: A Reduced Instruction Set VLSI Computer' appeared in May 81 https://dl.acm.org/doi/10.5555/800052.801895 (and of course the original RISC paper was even earlier in 1980) The iAPX
538.
by klelatti 4y ago
You can make these points without insulting the country. He’s not winning any friends with language like this.
539.
by klelatti 4y ago
And of course David Patterson did a take down of the performance of the 432 (which curiously doesn’t seem to mention RISC). https://dl.acm.org/doi/pdf/10.1145/641542.641545
540.
by klelatti 4y ago
From the FT: Activision, maker of the hit game Call of Duty, said the ruling “contradicts the ambitions of the UK to become an attractive country to build technology businesses”. It labelled the decision a “disservice to UK citizens, who fa

More ›