Hacker News Reader

Comment by listless | Hacker News Reader

I find the “amazing, look what a woman did!” narrative to be incredibly insulting.

But congrats to Lisa Su as the person who turned AMD around. They have, IMO, utterly eclipsed Intel. I wouldn’t buy anything else.

View on HN

I'd love to know how much AMD's massive lead is due to AMD's designs and how much it is due to TSMC's 7nm node. Alder Lake on Intel 7 looks like it will largely close the gap, so I think that means AMDs designs are roughly on par with Intel's and the massive eclipsing is due to TSMC. Of course being on par with Intel's IP is still a huge accomplishment.

Most of their current lead is down to Zen which is architecturally superior to the current Intel Core architecture.

To be fair, as I understand it, feature density of Intel 7 nm is more like TSMC 5 nm, from what I have read. Also, I think they took a risk with switching to a chiplet design, and it paid off big for them. Not sure who was in charge of design was when they laid out that roadmap, but whoever it was, and Lisa Su and the upper management, get some kudos for that. Of course, it’s quite possible that move could have failed and we’d be laughing at them just like we did for the bulldozer architecture, so maybe they just got lucky :)

But I do agree, all these comparisons we’ve seen between Intel 14 and TSMC 7 have been… less than helpful for comparing technology. Good for end users making purchasing decisions, of course.

Off topic here, but it was probably Jim Keller who contributed to Intel doing chiplets. Looked like one of the few sane persons in the leadership when I worked there.

As an interested outsider (no connections or technical expertise), it seems like Keller’s tenure at a company very frequently coincides with very successful products being developed. Is that the view from inside the field as well? Is it seen as a bit of a coincidence, or is he really that good of a technical leader?

"we’d be laughing at them just like we did for the bulldozer architecture"

I remember buying the Phenom II X6 1100T day one (December 2010) because we were all waiting for AMD to rise like a phoenix and dethrone Intel's incredible Core series chips

10 months later (October 2011) Bulldozer finally launched, before crashing and burning on the launch pad almost immediately

It was a hard time being an AMD fan a decade ago :(

Hey, I too bought an 1100T before Bulldozer launched, and it gave me many years of great service.

Unfortunately, I needed to replace that server before Zen showed up, so the next iteration was (and is) an Intel chip. (Slightly tempted to upgrade it again ~soon, but not impressed by Intel's current "lower-power" Xeon designs, AMD's Epyc Embedded last had an iteration in Zen 1 and according to leaks won't see a new one until 2023(!) and Zen 4, and AArch64 is not bad but I haven't seen anyone selling a server board that was in the right spot between size/power/performance/price yet...)

>feature density of Intel 7 nm is more like TSMC 5 nm

yes but Intel announced 7nm delay while TSMC 5nm is in production and 3nm in risk production. Intel is behind.

Oh yeah, I didn’t mean to suggest Intel is ahead, I was more suggesting that chips built on Intel 7 nm and TSMC 7 nm were not really directly comparable on the technology side. Of course for an end user it doesn’t really matter what causes the performance, but we were talking more about comparing the architectures.

We will find out when Alder Lake launches. AMD have earned every penny they have so far, but Intel have new leadership, new designs (Zen took about 5 years to do from start to release, roughly, and 4 years after it's launch I think this is Intel's first proper competitor). Their new top of the line chip (12900k) looks like a bit of a monster.

Should be good for consumers for a while with all this new competition (P.S. Apple, please sell Raspberry Pi some M1s just to scare Intel and AMD)

I would argue AMD's designs aren't so much performance alone, but cheaper to manufacture. Effectively they brought the idea of breaking the chip into components, then putting them together later. That allowed for a cheaper and more robust design.

https://www.techpowerup.com/264271/amd-gives-itself-massive-...

As soon as I saw that back in 2016 I bought a ton of AMD stock. It was clearly a winner because they could have a lower cost design AND potentially better performance.

Price and feature availability had a lot to do with it. Up until Ryzen came out, a 4 core CPU was the largest core count available in an Intel "consumer" CPU. I had a 6 core Ivy Bridge CPU at one point (i7-4930k) and it cost me ~$550. The motherboard was another $300. Despite the premium, it was still considered a desktop CPU, so ECC was unavailable. And since it was a niche product, I was unable to find a replacement motherboard when I inadvertently damaged mine. Moreover, since the chipsets changed seemingly every new CPU release cycle, I couldn't upgrade either my CPU or my motherboard without replacing both.

AMD released the Ryzen 1700x, with 8 cores, for cheaper than what I paid for my 6 core, and something like half the price of Intel's 8 core offering. AMD made 4 core the baseline, while Intel was still selling 2 core i3s, and they made 6 cores the middle of the road. All of them had hardware threading enabled, whereas Intel had disabled hyper-threading in its i5 line.

AMD also committed to supporting the AM4 socket for multiple release cycles. With BIOS updates, motherboards that launched with with the 1XXX series can run the newest 5XXX series (subject to being able to power them). And the Ryzen motherboards came in two flavors: one that was very affordable (the BX50 series) and one that was more expensive, but had features enthusiasts want (the XY70 series).

I don't want to sound like I'm gushing, but AMD came out swinging with Ryzen. I'm probably misremembering, but looking back, I think the only advantage Intel really had at the time was single core performance (and for a niche audience, AVX512). Subsequent Ryzen releases have really closed the performance gap and there the TSMC process improvements help. AMD has garnered a lot of good will at this point. They shook up the stagnant CPU market with both features and price.

Are you sure X370 boards can run 5xxx Ryzen chips? I believe all of them top out at 3xxx series Ryzen. Still, that's a great run and much better than Intel has done for consumers.

I.m sure there were news that some companies add 5xxx support to x370 chipset. Not sure how many of them delivered anything.

IIRC, there were no motherboards at that time shipping with a large enough flash ROM to hold the microcode, etc. necessary to support every generation of CPU on the AM4 socket, so in practice supporting the Ryzen 5xxx processors would require a motherboard firmware update that removes support from some older processors—possibly the one you're using to prepare for the CPU upgrade. AMD decided that would be a support nightmare and instead opted to make CPU support somewhat loosely tied to chipset generation (beta BIOS versions excepted).

> With BIOS updates, motherboards that launched with with the 1XXX series can run the newest 5XXX series (subject to being able to power them).

It's socket compatible but not chipset compatible. X370 officially tops out at 3xxx series. Maybe you'll get lucky and find an unofficial bios hack for 5xxx but unlikely.

So to upgrade from a Ryzen 1xxx to a 5xxx in practice means a motherboard upgrade.

You cannot ignore that it’s unusual in her field. It isn’t easy being an outsider, and to rise up and do exceptionally well in that environment deserves accolades on many fronts.

I disagree. If a man were to accomplish something against the odds, that would just be a man being a man. Why is it any different when it’s a woman? This stuff is patronizing.

Because women in male dominated fields are typically not taken as seriously, need to do more than others for similar recognition, have few role models that look like them, often receive anti-women sentiment or even unwanted sexual advances along the way, are often told how they should or shouldn’t act in ways that their male colleagues don’t get, along with many other differences that men don’t have. It doesn’t have to be only women who receive such treatment — other minorities have their own experiences as such.

I’m curious what you even mean here as against the odds? If it’s a technical masterpiece, both a man and a women can equally be recognized by IEEE for that hard work. But if it’s a woman, than you also need to recognize there’s been many other hardships along the way in addition to the technical brilliance needed. It’s not patronizing at all — that would be if I were to suggest a woman isn’t as capable. Here it’s the opposite — I’m saying they’re just as capable, but society is setup in such a way to present many additional obstacles that men don’t have. To not recognize that is to ignore the realities of the world.

If you yourself are a woman or a minority, I’m sure you can appreciate that. If not, I’d encourage you to go and listen to the experiences of those who are — this is very well documented.

Reply on news.ycombinator.com