DiabloD3

There will be no future Xe GPUs.

Druid, the 4th generation, has been shelved entirely, and all consumer DGPU products, possibly _all_ DGPU products, have been killed for Celestial.

Its likely the only products shipping with Celestial will be IGPUs for the next generation and a half, until Serpent Lake comes out with RTX graphics tiles from Nvidia, and then never another Xe product ever again.

A lot of games no longer run properly in Windows, and the only way to keep playing them is with Wine/Proton.

Windows performance is also often worse than Wine/Proton's.

GPU clusters have, largely, no actual value.

If anything, you might have to pay to have them disposed of, they don't really have any meaningful used eBay market outside of the randos that want to do high end extreme local inference in their basement.

Also, as for RAMmageddon, the inference SBCs that all of the AI bros bought don't have DIMMs, they're not even the right chip: its all GDDR and LPDDR. The only DDR DIMMs being consumed are for regular non-inference machines that help run the business and service infrastructure behind the scenes.

Neat, can't wait to see a llama.cpp PR for this.

OnePlus basically already went under a few years back.

They merged with Oppo, and ever since its just been rebranded Oppo phones... they're fine, they work, but they're not the magic OnePlus was.

Skip Samsung, they tend to not honor warranties and are ROM swap resistant.

Google is okay for now (I recently got a Pixel 10, its fine, it can run Graphene).

Next year Motorola flagships will also have official Graphene support.

Ironically, this is what people claim AI can do with a snap of the fingers.

Should be real simple if the HN AI echochamber is right, right?

I love how people say things like "extension spaghetti", as if all other non-standard APIs have the same problem: hardware gets new features that people want to use from that API, API gains extension to use that hardware feature.

CUDA is no different, in fact, often worse. Nvidia is bad at documenting which hardware does what things, and CUDA users often have to use third party tables to figure out what hardware can't do what and disappoint customers who unwisely invested into it.

Weird, most people have the exact opposite experience.

Having to deal with closed source opaque poorly documented stacks sucks.

Weird, since the most used open source inference engine is faster on Vulkan on platforms that offer multiple options, with the sole exception being Nvidia, due to poor Nvidia driver quality (which I am forced to assume is intentional, Nvidia wishes to maintain their moat after all).

Its easier to just get rid of your legacy code entirely and use Vulkan for compute, or have your compiler emit SPIR-V directly.

No reason to tie yourself to Nvidia's moat.

You'd have less problems with 27B, btw.

I was expecting Linda Hamilton, but apparently more than one set of twins were in there.

I still find it funny how the AI Bros bribed Trump, and he immediately turned around and cost them significant revenue.

As they say, qui cum canibus concumbunt cum pulicibus surgent.

Serpent Lake (and its siblings) cover both consumer desktop and laptop.

Arc's team was disproportionately effected by the layoffs. A lot of the major engineers that worked there changed their Linked in details to list Nvidia, AMD, etc.

As for DC and Workstation GPUs, they also seem to be canceled.

We'll see next year, I guess.

HN front page circa September through December of last year. Its part of the $5B investment into Nvidia, and the new CEO, Lip-Bu Tan, happily did whatever Nvidia told him to do.

Nova Lake (Series 4) is already too late in development to switch, so its shipping with Celestial, and Titan Lake (Series 5) after it will be essentially a Nova Lake refresh using a hybrid tile that is Celestial's graphics engine with Druid's media, memory, and bus controllers (since Druid development has been halted).

Serpent Lake (Series 6) will be the first that can ship with Nvidia tiles, which should launch in early 2028.

Unless something changes, there will be no DGPU Celestial products.

Unfortunately, the ink on the contract is already dry, so this is the end of Intel's GPU unit.

Yep, but the flip side is you can get two B70s for the price of a single 3090 (MSRP, obviously; 3090s used go for about the same as a new B70), and they're true 2 slot, so they can fit on x8/x8 consumer boards fine.

The side effect is Intel has fired their entire Arc team, including the driver team, as well have canceled all Celestial products (only low end combined GPU + IO die products will have Celestial, as its too late in production to change). All future Intel products will have Nvidia graphics tiles.

Llama.cpp is, of course, still trying to better support the existing Intel products, but its kinda hard when they might stop working any day due to driver breakage.

That is incorrect.

They both have GDDR6.

The B70 has 256 bit it bus at a clock speed of 2375mhz (608 GB/s), the 3090 has a 384 bit bus at a clock speed of 2438mhz (936 GB/s).

It isn't slower, it just has less channels, ie, it is less wide.

Sundaras aren't great, but are sufficient for this. Scarlett series is also good enough.

At 48khz, with a tone generator (not rando Youtube videos or anything), you should be able to clearly hear up to around 16khz (as in, can tell pitch of tone), and be able to tell a tone is being played at all up through 18khz, and hopefully up to 20khz.

The problem with RocksDB is Facebook uses it, and has a similar system that replicates data across the world... and still faces unrecoverable data loss because of RocksDB.

Yeaaaaaaaaaahhhh.... you might be a tiny bit deaf.

OTOH, we know nothing of your audio equipment nor how its setup.

Thats a weird way for Sony to announce the end of the Playstation

Interesting, they used to be the largest ZFS user.

Hard to Google for it without getting AI slop on it, but apparently they built their own stack in 2019.

Not sure I like their solution, "Meta-data is persisted in RocksDB databases using a proprietary KV store called QuarkDB." Unless QuarkDB has magically removed RocksDB's amazing ability to corrupt and lose data frequently, this whole thing sounds like a bad idea.

Also, their data is not stored on any one system (local XFS, Ceph, Lustre), a recipe for disaster.

In one pool, sure.

You can have more than one pool.

All using ZFS, too

I'm already playing 2.1, and have been since it came out like 3? 4? days ago.

Seeya sometime next decade.

As a reminder, the most dangerous job in Russia is "opposition leader".

MTP models share internal state with the main model, and also refer to parameters in the model.

They are more like a single model that has two separate attention head mechanisms.

They weren't the first to do MTP like this, and arguably did it wrong: the MTP heads are kept in a separate file and have to be welded in by the inference engine.

Qwen 3.6 shipped with working MTP first, and had working MTP in llama.cpp first.

Your father should only be looking at Kelly Blue Book values. What brand ends up retaining theirs year after year? Toyota.

He's right to buy American made when it comes to cars, we know what we're doing... he's just gonna have to buy a Toyota to get American.