Intel is losing against AMD – Engadget | Hacker News Reader

Intel is losing against AMD – Engadget

engadget.com

45 pointsrbanffy25 comments

Will AMD be able to build a war chest to continue to compete after Zen? That is the question, or is it Athlon all over again?

It is really bad. Semiconductor technology wise they are falling behind which will make it hard to recover the investments in at least one generation. Heat distribution wise the AMD architecture is better. Yield wise the AMD approach is more effective. Product flexibility and ability to iterate AMD approach is better.

The biggest threat to Intel is economic. Each generation costs usually more. One generation funds the one after the next. Loose once and there is trouble, loose twice and it becomes existential.

One thing that is happening now as well is that Contract Manufacturers in Taiwan and China etc are all now doing AMD designs where previously they had not in a number of years. So this creates a newly refreshed foundation for AMD to be an alternative to Intel.

I wonder if/when Apple will be tempted to make the switch.

Apple is priming a switch to ARM. They don’t care about the amd64 game too much any more. Catalina is step 1 getting that done.

I thought Catalina just deprecated 32-bit apps?

iPadOS on the other hand is inching towards a desktop-like OS, ofcourse based on Apple's in-house ARM chips.

> iPadOS on the other hand is inching towards a desktop-like OS, ofcourse based on Apple's in-house ARM chips.

A desktop-like OS better includes lots of capabilities for development and debugging, e.g. attaching a debugger to a process, memory dumps, compiling and executing code etc. I do not believe that this fits Apple's plans and philosophy for iOS.

That's a very developer-centric view of what a "desktop-like" OS needs to have. Most users of desktop-like OSes don't care about these things. They're clearly not going after software developers for iOS, but that doesn't mean its not their plan to make it more desktop-like or compete with desktops for the non-pro markets.

I should add that I use my 9.7” iPad Pro for all my day to day computing needs except development.

that helps tremendously though if they plan on having an amd64 emulation layer, they don't need to bother in emulating 32 bit x86 (or the other legacy modes that x86 has).

I've been hearing this for about 10 years now. I don't think the Pro tower will be going ARM any time soon

3 years ago the question was more if they'll even build a Pro tower ever again. The tide has changed, but I wouldn't be surprised if it draws back again.

If AMD consistently outperforms Intel's high-end gear, I can easily see a Milan-based MacPro. Everything needed to trounce the current model is already in the EPYC Rome package.

They would have started their High TDP ARM CPU design at least 3 years ago, and had they done that, the NUVIA guys wouldn't have left Apple to start a company of their own design CPU for servers.

So I seriously doubt that is happening, not to mention it makes little financial sense.

Apple has been all in on Thunderbolt for so long that it wouldn't make sense for them to use start using AMD chips until USB 4 (which includes Thunderbolt) is available from AMD.

Intel already lost. It will be years before they can catch up to AMD assuming AMD does nothing.

Now Intel is in the rear view mirror and AMD has its sights on NVIDIA.

AMD stock is up over 130% this year.

Up over 130% in the last 12 months, but up ~20% in the last 13-14 since AMD took a nosedive in October 2018 with the rest of the semiconductor stocks.

What a come back; AMD was in the red for years and now this. Went with AMD + Nvidia in my current setup, and could as well go full AMD 5 or so years from now; My last AMD machine was with AMD K6-2 500mhz way back in 1999;

> Intel already lost.

That's a ridiculous thing to say about a company that has ~75% of the market share.

I really hope they try catching up to nVIDIA in terms of the development ecosystem for on-GPU computation. I hate being effectively-vendor-locked like that, due to the weakness of the competition - and this is coming from someone who likes, writes and publishes CUDA code.

That is the mainstream media narrative, may be AMD are beating Intel in price / performance or even in absolute performance, but the reality is Intel isn't hurt at all. They are still making record revenue and profits.

Intel is doing so well that even with the increased capacity they put in place they are still catching up to demands. And there was rumours about Intel outsourcing their CPU, I think that is unlikely but they are looking to outsource all of their Chipset to either TSMC or Samsung to further free up Capacity.

The press and response are all in flavours of AMD, but I would argue AMD is not doing a good enough job capturing more of it, especially in the Server and HyperScaler Segment. Even the optimistic research estimate AMD (only) to gain 10% Market shipment in the Server market by 2020. I would have expected the Zen 2 EPYC to be at least in the 10s if not 20s.

Most of the current Mobile Laptop shipment went with AMD merely because Intel could not supply them with enough chips. Not entirely sure how well this will play out once their Zen 2 APU arrives and Intel fix ( if they do ) their 10nm.

I just wish AMD could do better, having said that, despite all the dirty tactics used, I must give credit to Intel's marketing and sales people.

Could someone tell of their experience with AMD on Linux? I've read some old comments online about how there's some issues with like GCC.

I'm running a Threadripper 2950X in my primary workstation, Ubuntu 19.10. I have not encountered a single CPU-related issue with compilation in the 6 months I've been using this setup. In fact I've had less trouble overall than I've had on my MacOS laptop. IMO Linux has finally reached a point where it's solid enough for general purpose desktop use.

I have a ryzen 3600x with the x570 chipset and everything has worked flawlessly for the ~3 months since i got it. Im not into the mineuta of compiler optimizations so i cant speak to that, but there havent been any compilation failures.

The 1700X and 1800X had some issues with GCC failures under heavy load. Linux apparently uses more CPU resources faster than Windows does, and caused excess voltage drop in some scenarios, as I recall reading about it.

I think various sites online called it Performance Marginality.

That was only a problem with some of the first gen Ryzen / Zen 1 chips.

And I think it also involved the motherboard, because my 1700X never did it when I have my ASUS board set to boost with all-core overclock, but it did act a bit funny under normal settings.

With the current Zen 2 chips everything seems fine. I did have some weird RAM corruption errors with my 3900X when I had the RAM set to 3600 and I did 24 thread Rust compiles. However, that didn't happen with the release version BIOS, it was only after the ABA (I think? The one that fixed RDRAND) Agesa update that was supposed to fix the excess voltage levels and I think they went too far. Everything was perfect running the RAM at default 2400 speed.

Reply on news.ycombinator.com