The extra expense of the PCIe 5 circuitry would be balanced by having 1 less amplificator and one less set of Pads to the external world, while maintaning speed.

Wins all around.

The next Speed Bump will come with PCIe 6

Hope this, or something like it, gets standarized across the industry, like what happened with Dell's CAMM, that Del gave free of charge to JEDEC, so that CAMM2 could become an indsutry wide neutral standard.

Aside from Huawei and Cambricon, Deepseek also has Innosilicon, Moore Threads and Liusang as alternatives.

I guess that, for the R3 model, they will conduct small scale training trials to see if any of those work for them, to stay within chinese chips for training.

And it would not surprise me that they can achieve sway in to telling the "winner" what features they need emphasized in a future roadmap.

JM2C
YMMV

From InnoSilicon, to Huawei, to MooreThread, to the most recent entrant Lisuan.

Are these as good as nVIDIA's offerings? Nope.
Are these as good as AMD's offerings ? Nope.
Are these as good as Intel's Offerings? Nope.
Are these as good as Broadcomm's offerings? Nope

Are these alternatives good enough for the taks at hand (HPC and AI)? Well yes! Yes, they are good enough.

News for merd? Nope
Stuff that matters? Yes

For usoans at least.

Also, we bussiness nerds (Electronics engineer with an MBA here) find fascinationg how two failed mergers, one in '22 to "protect the company" and one in '23 blocked by the DoJ to "protect low fares" will lead to the biggest companies of the market getting Spirits' assets for pennies (or should it be nickels ;-) ) on the dollar

On Feb '22, frontier wanted to merge with Spirit, Spirit shareholders rejected the offer.
JetBlue Tried to acquire Spirit in '23 but the DoJ blocked the merger.

Now, the company will go bust, and the bigUns (American, delta, et al) will get all the assets from pennies on the dollar.

Great way to protect prices by the DoJ, and great way to protect your company shareholders...

Mergers sometimes are the lesser of two evils

Most people (including us techies) are lazy, and most people do not suffer from variable degrees of OCD as us terchies do, hence the need of autogrouping tabs.

Either by TNNs, hidden Markov Models, crowdsourcing or any other thechnique

This "feature" will not be available until about june 2026. By then it will either be polished or be removed...

All advantages

While that may be the case in advanced abrahamic economies, when seen worldwide, the bulk of vegetarians are so for religious reasons (see bhudists, tibetans and hindustanis among other groups)

In those cases, the religious implications of lab-grown meat are not as clear cut as for "fashion vegetarians"

Vegetarian burger chain company "neat burger", backed by the likes of lewis hamilton ans leo di caprio has entered liquidation.

https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fwww.planetf1.com%2Fnews%2F...

All this "vegetarian stuff that pretends to be meat" started with bhudist monks in SE asia. You see, they become monks in adulthood, but by then they already know the delicious taste of all sorts of meat. But since a monk has to be vegetarian, they developed recipes to mimic the shape, texture and flavour.

I've tasted that food many times, and is delicious, and i'd rather have that than an ultrqa-processeed burget patty

Keep up the good work

CoD 6 had anticheat.
SecureBoot and TPM 2.0 have been mandatory on OEM Machines since 2018
At most in Oct 2026 Win10 will be out of support from versions of the OS that gamers can legally get their hands on
And Every supported Win11 version WILL have TPM2.0 and SecureBoot

So, why code AND SUPPORT two versions of the anticheat for CoD7? One without TPM2.0 and secureboot that will last only a year, and one with them that will keep going on for a loooooong time. Is silly.

And is not only CoD7. The new Battlefront will also require TPM2.0 and Secure Boot (again, required on new OEM Machines since 2018).

Is Like getting mad because Quake required a math coprocessor, and you had an SX machine

No, we were not looking at the same machine. I stoped following apple closely after the switch to ARM, but should have known that they would neglect the Pro Tower, they have done it before...

Yes, the Decked up studio IS the best value, but that gives even more gravitas to my point. For training, nVIDIA, for inference, Apple.

"Gpt-oss-120B model will require about 80 GB of memory."

So either an nVIDIA A100 or a maxed M2 Mac Pro.

The current nVIDIA A100 costs 10,000.oo ~15,000.oo U$D, has 80GB of VRAM, and needs a server around it.
The decked out M2 Mac Pro cost a smidge less than U$D 10,000.oo* , but it has so much RAM (192GB), that you can run inference on 2 different models of similar size (one on the GPU and one on the NPU), with left over memory for the OS.

For training, nVIDIA is the right choice, but for inference? Apple hands down.

* And you get a Keyboard and a mouse included ;-)

Would this morality core make a better AI? Or just slow down the IA's tendency towards enslaving the Human race?

Who knows?