Live data from Hacker News

Nvidia buys $5B in Intel

tomshardware.com

481–490 of 638 posts

Re: Nvidia buys $5B in Intel

#481

The reason why Nvidia is buying now does not have to do anything with Arc or GPU competition. There are mainly two reasons. 1) This year, Intel, TSMC, and Samsung announced their latest factories' yields. Intel was the earliest, with 18A, while Samsung was the most recent. TSMC yieled above 60%, Intel below 60%, and Samsung around 50% (but Samsung's tech is basically a generation ahead and technically more precise),…

While #2 is true, there are a myriad of ways to do that without a press release.

Re: Nvidia buys $5B in Intel

#482
post #448

Nvidia's stake in Intel could have terrible consequences. First, it is in Nvidia's interest to kill Intel's Arc graphics, and that would be very bad because it is the only thing brighing GPU prices down for consumers. Second, the death of Intel graphics / Arc would be extremely bad for Linux, because Intel's approach to GPU drivers is the best for compatibility, wheras Nvidia is actively hostile to drivers on Linux.…

This misses the forest from the trees IMO: - The datacenter GPU market is 10x larger than the consumer GPU market for Nvidia (and it's still growing). Winning an extra few percentage points in consumer is not a priority anymore. - Nvidia doesn't have a CPU offering for the datacenter market and they were blocked from acquiring ARM. It's in their interest to have a friend on the CPU side. - Nvidia is fabless and has c…

> Nvidia doesn't have a CPU offering for the datacenter market and they were blocked from acquiring ARM. It's in their interest to have a friend on the CPU side.

Someone should tell nvidia that. They sure seem to think they have a datacenter CPU.

https://www.nvidia.com/en-us/data-center/grace-cpu-superchip...

Re: Nvidia buys $5B in Intel

#483

Earlier quoted context omitted.

I don't know what anybody would do with such a weak card. My RTX 5090 is about 10x faster (measured by FP32 TFLOPS) and I still don't find it to be fast enough. I can't imagine using something so slow for AI/ML. Only 2.2 tokens/sec on an 8B parameter Llama model? That's slower than someone typing. I get that it's a budget card, but budget cards are supposed to at least win on a pure price/performance ratio, even with…

I feel as though you are measuring tokens/s wrong, or have a serious bottleneck somewhere. On my i5-10210u (no dedicated graphics, at standard clock speeds), I get ~6 tokens/s on phi4-mini, a 4b model. That means my laptop CPU with a power draw of 15 watts, that was released 6 years ago, is performing better than a 5090. > The 5090 is 10x faster but only 6-8x the price I don't buy into this argument. A B580 can be bo…

You can get them for $2,000 now. One from Asus has been that price several times over the last few months. I got my PNY for 2200 or so.

Re: Nvidia buys $5B in Intel

#484
post #448

Nvidia's stake in Intel could have terrible consequences. First, it is in Nvidia's interest to kill Intel's Arc graphics, and that would be very bad because it is the only thing brighing GPU prices down for consumers. Second, the death of Intel graphics / Arc would be extremely bad for Linux, because Intel's approach to GPU drivers is the best for compatibility, wheras Nvidia is actively hostile to drivers on Linux.…

This misses the forest from the trees IMO: - The datacenter GPU market is 10x larger than the consumer GPU market for Nvidia (and it's still growing). Winning an extra few percentage points in consumer is not a priority anymore. - Nvidia doesn't have a CPU offering for the datacenter market and they were blocked from acquiring ARM. It's in their interest to have a friend on the CPU side. - Nvidia is fabless and has c…

>Nvidia is fabless and has concentrated supplier and geopolitical risk with TSMC.

East India Company has been conducting continental wars on its own. A modern company with $4T valuation and a country-GDP-size revenue and possessing key military technology of today and tomorrow wars - AI software and hardware, including robotics - can successfully wage such a continental war through a suitable proxy, say an oversized private military contractor (especially if it massively armed with drones and robots), and in particular is capable of defending an island like Taiwan. (or thinking backwards - an attack on Taiwan would cause a trillion or two drop in NVDA valuation. What options get on the table when there is a threat of a trillion dollar loss ... To compare - 20 years of Iraq cost 3 trillions, ie. 150B/year buy you a lot of military hardware and action, and efficient defense of Taiwan would cost much less than that.)

Re: Nvidia buys $5B in Intel

#485
post #454

Earlier quoted context omitted.

Conclusion: Buy AMD. Excellent Linux support with in-tree drivers. For 15 years! A bug is something which will be fixed. Nvidias GPUs are theoretically fast on initial benchmarks. But that’s mostly optimization by others for Nvidia? That’s it. Everything Nvidia has done is a pain. Closed-source drivers (old pain), out of tree-drivers (new pain), ignoring (or actively harming) Wayland (everyone handles implicit sync w…

> Excellent Linux support with in-tree drivers. For 15 years! Linux support has been excellent on AMD for less than 15 years though. It got really good around 10 years ago, not before.

ATI/AMD open source linux support has been blowing hot and cold for over 25 years now.

They were one of the first to actually support open source drivers, with the r128 and original radeon (r100) drivers. Then went radio silence for the next few years, though the community used that as a baseline to support the next few generations (r100 to r500).

Then they reemerged with actually providing documentation for their Radeon HD series (r600 and r700), and some development resources but limited - and often at odds with the community-run equivalents at the time (lots of parallel development with things like the "radeonhd" driver and disagreements on how much they should rely on their "atombios" card firmware).

That "moderate" level of involvement continued for years, releasing documentation and some initial code for the GCN cards, but it felt like beyond the initial code drops most of the continuing work was more community-run.

Then only relatively recently (the last ~10 years) have they started putting actual engineering effort into things again, with AMDGPU and the majority of mesa changes now being paid for by AMD (or Valve, which is "AMD by proxy" really as you can guarantee every $ they spend on an engineer is $ less they pay to AMD).

So hopefully that's a trend you can actually rely on now, but I've been watching too long to think that can't change on a dime.

Re: Nvidia buys $5B in Intel

#486
post #482
post #448

Earlier quoted context omitted.

This misses the forest from the trees IMO: - The datacenter GPU market is 10x larger than the consumer GPU market for Nvidia (and it's still growing). Winning an extra few percentage points in consumer is not a priority anymore. - Nvidia doesn't have a CPU offering for the datacenter market and they were blocked from acquiring ARM. It's in their interest to have a friend on the CPU side. - Nvidia is fabless and has c…

> Nvidia doesn't have a CPU offering for the datacenter market and they were blocked from acquiring ARM. It's in their interest to have a friend on the CPU side. Someone should tell nvidia that. They sure seem to think they have a datacenter CPU. https://www.nvidia.com/en-us/data-center/grace-cpu-superchip...

I wonder if this signal a lack of confidence in their CPU offerings going forward?

But there's always TSMC being a pretty hard bottleneck - maybe they just can't get enough (and can't charge close to their GPU offerings per wafer), and pairing with Intel themselves is preferable to just using Intel's Foundry services?

Re: Nvidia buys $5B in Intel

#487

There's a lot of concern in the comments here about what this means for ARC. The size of this investment while large isn't enough to warrant jeopardizing ARC though. Intel has a responsibility to all shareholders, and diminishing ARC would be a bad move for overall shareholder value. If Nvidia did try to exert any pressure to scrap ARC, that would be both a huge financial and geopolitical scandal. It's in the best in…

Could they use nvidia ip for the future arc line?

Re: Nvidia buys $5B in Intel

#488
post #203

Earlier quoted context omitted.

Apple did not want their x86 chips, they wanted their Xscale stuff. Apple went to Intel to get chips, the power envelope was appealing to Apple. Intel was the one to say no.

Right. But of course, intel was busy spinning off their Xscale business to Marvell. If they had seriously invested in it, they could have owned the coming mobile revolution. They did push hard on their UMPC x86 SoCs (Paulsbo and derivatives) to Sony, Nokia, etc. These were never competitive on heat or battery life.

> Paulsbo

You probably meant Poulsbo (US15W) chipset

Re: Nvidia buys $5B in Intel

#489

Nvidia's stake in Intel could have terrible consequences. First, it is in Nvidia's interest to kill Intel's Arc graphics, and that would be very bad because it is the only thing brighing GPU prices down for consumers. Second, the death of Intel graphics / Arc would be extremely bad for Linux, because Intel's approach to GPU drivers is the best for compatibility, wheras Nvidia is actively hostile to drivers on Linux.…

> Nvidia is actively hostile to drivers on Linux Nvidia is contributing to Nova, the new Nvidia driver for GSP based hardware. https://rust-for-linux.com/nova-gpu-driver Alexandre Courbot, an Nvidia dev, is comaintainer. https://www.phoronix.com/news/NOVA-Core-Co-Maintainer

Yeah I think Nvidia were hostile to Linux when they saw no value in it. Now it's where the machine learning is. It's the OS powering the whole AI hype train. Then there is also Steamdeck making Linux gaming not a complete write off anymore.

The days of Nvidia ignoring Linux is over.

Re: Nvidia buys $5B in Intel

#490

Nvidia's stake in Intel could have terrible consequences. First, it is in Nvidia's interest to kill Intel's Arc graphics, and that would be very bad because it is the only thing brighing GPU prices down for consumers. Second, the death of Intel graphics / Arc would be extremely bad for Linux, because Intel's approach to GPU drivers is the best for compatibility, wheras Nvidia is actively hostile to drivers on Linux.…

Conclusion: Buy AMD. Excellent Linux support with in-tree drivers. For 15 years! A bug is something which will be fixed. Nvidias GPUs are theoretically fast on initial benchmarks. But that’s mostly optimization by others for Nvidia? That’s it. Everything Nvidia has done is a pain. Closed-source drivers (old pain), out of tree-drivers (new pain), ignoring (or actively harming) Wayland (everyone handles implicit sync w…

> What we need? Reliability. And Linux support

Both of which NVidia does a lot better in practice! I'm all for open-source in-tree drivers, but in practice, 15 years on, AMD is still buggy on Linux, whereas NVidia works well (not just on Linux but on FreeBSD too).

> I don’t judge whether implicit sync or explicit are better.

Maybe you should.

Post reply on HN