Vijesti, novosti i glasine o nadolazećim nVidia grafičkim karticama

[The Exiled]

nVidia Hopper might have huge 1000 mm² die monolithic design

Hopper is still sporting a classic monolithic die design despite previous rumors, and it appears that nVidia's performance targets have led to the creation of a monstrous, ~1000 mm² die package for the GH100 chip, which usually maxes out the complexity and performance that can be achieved on a particular manufacturing process. This is despite the fact that Hopper is also rumored to be manufactured under TSMC's 5nm technology, thus achieving higher transistor density and power efficiency compared to the 8nm Samsung process that nVidia is currently contracting. At the very least, it means that the final die will be bigger than the already enormous 826mm² of nVidia's GA100.

If this is indeed the case and nVidia isn't deploying a MCM (Multi-Chip Module) design on Hopper, which is designed for a market with increased profit margins, it likely means that less profitable consumer-oriented products from nVidia won't be featuring the technology either. MCM designs also make more sense in nVidia's HPC products, as they would enable higher theoretical performance when scaling - exactly what that market demands. Of course, nVidia could be looking to develop an MCM version of the GH100 still; but if that were to happen, the company could be looking to pair two of these chips together as another HPC product (rumored GH-102). ~2,000 mm² in a single GPU package, paired with increased density and architectural improvements might actually be what nVidia requires to achieve the 3x performance jump from the Ampere-based A100 the company is reportedly targeting.

Izvor: TechPowerUp

[The Exiled]

RTX 4000 rumors: September release, over 800W TDP

nVidia’s upcoming RTX 4000 series of graphics cards, codenamed Lovelace, are rumored to arrive this September. But in addition to availability and price concerns, you’ll probably need a new PSU for the high-end models, which are said to consume over 800W of power. We heard reports last year claiming the RTX 4000 cards would use some of the most power-hungry GPUs we’ve ever seen. That’s especially true of the AD102 that’s rumored to have a TGP of 800W or more. The latest rumors come from prolific leakers Kopite7kimi and Greymon55, both of whom say they’ve heard claims that the AD102 GPUs will have scarily high TGPs. The former thinks the RTX 4080 will carry a 450W TGP and the RTX 4080 Ti will be 600W. The RTX 4090, meanwhile, will come with a monstrous 800W rating. According to a report from Hardware Times, nVidia is prepared to pay dearly for the ability to meet growing demand for increasingly powerful graphics cards. Specifically, it may be offering up to $10 billion to TSMC for a significant chunk of its 5nm manufacturing capacity. Other companies tapping TSMC’s N5 process node are Bitmain, AMD, and Apple.

Izvor: TechSpot

[Lewis]

ono dvije grafe u boostu + CPU = 2kW, strasno

[The Exiled]

Iskreno, koliko god pretjerano zvučalo, ne bi me čudilo da budu tu negdje, jer već sad imaju 450W i 500W BIOS-e za RTX 3090 modele.

S ovom novom opet idu na ogroman monolitan čip, dok AMD već ima spreman i funkcionalan MCM, tj. chiplet dizajn za RDNA 3.

[Disco]

Ima video od od Toma iz MLID koji misli da to nije istina, i kao jedan od glavnih razloga navodi da bi to bilo nemoguće hladit ako srče 800+ wata. Pa to je imho apsolutno suluda potrošnja za igranje…uz skupi HW, trebat ce imat i za struju. Nadam se da nije istina, i nadam se da ce onda bar AMD imati nesto normalno.
E da Tom misli da su to mozda i serverske grafe ili ona cudesa za AI, ili je to isto, ne sjecam se sto je točno rekao

[The Exiled]

Autor Disco

Ima video od od Toma iz MLID koji misli da to nije istina, i kao jedan od glavnih razloga navodi da bi to bilo nemoguće hladit ako srče 800+ wata. Pa to je imho apsolutno suluda potrošnja za igranje…uz skupi HW, trebat ce imat i za struju. Nadam se da nije istina, i nadam se da ce onda bar AMD imati nesto normalno.
E da Tom misli da su to mozda i serverske grafe ili ona cudesa za AI, ili je to isto, ne sjecam se sto je točno rekao

Da, pogledal sam i sve to stoji, ali velim ništa me ne bi začudilo da slože takve potrošače, ne mora čak ni biti tih nagađanih 800W+, dosta je da bude iznad ovih 450W ili 500W koliko gutaju sadašnji high-end RTX modeli. HPC/AI/ML modeli nadolazećih kartica bi stvarno mogli i trošiti čuda tehnike uz HEP na brzom biranju, jer njima nije problem po defaultu služiti vodeno hlađenje. Kaj se AMD RDNA 3 tiče, sve ovo kaj nVidia priprema opet s monolitnim čipom, je baš zato, jer isto ko i Intel najlakše budu najboljiJi kad se potrošnja nafrlji do kraja, pa kud puklo da puklo. Tom se isto na to osvrnul da nVidia već sad zna da RDNA 3 MCM dizajn uz bolje performanse nudi i bolju učinkovitost, pa da onda odmah u startu našpanaju svoje čisto za svaki slučaj. Sve u svemu, budemo pričekali da vidimo kaj imaju spremno, ali kak god se okrene, nove Radeon i GeForce kartice budu trošile više, jedino kaj bi moglo ispasti da nVidia stvarno pretjera.

EDIT:

GPU giant nVidia is investigating a potential cyberattack

Izvor: BleepingComputer

[Eladio]

Nadam se da se jako varam, hm, moj savjet onima koji cekaju bi bio kupite grafu sad dok ne padne i Taiwan..
Brijem da nize cijene i bolju dobavljivost nego sto sad imamo, nećemo imati tako skoro

[Lewis]

Ja sam kupio jos dvije 3090 dok nije nestasica

[Eladio]

Nisi me bas shvatio.. geopoliticki nemir uz rat u Ukrajini i mozebitni novi ratovi (Kina/Taiwan) te sankcije koje idu u paketu uz rat ne ide na ruku IT trzistu... to je jedan novi faktor... naravno nisam baba vanga, ali samo nagadjam da tko očekuje još pad cijena, mogao bi se neugodno iznenaditi ako krenu nove nestašice i poskupljenja

[MadBoY]

Ukrajina proizvodi 70% neona, koji se koristi u proizvodnji chipova. Bez obzira kolike zalihe proizvođači imaju, ovo će im biti izgovor za dizanje cijena.

[The Exiled]

Hacking group threatens to leak nVidia GPU driver and firmware data, already selling GA102/104 LHR algorithm bypass

Izvor: Interneti

Što dalje ova situacija ide, to se viši čini da je nVidia interno vrtjela onaj fake LHR unlocker za koji se ispostavilo da je malware, pa sad imaju problem.

[p3r02d3r0]

Nvidia's DLSS source code leaked by hackers

The group has threatened to release more information unless Nvidia removes its Ethereum hash rate limiter.

[Disco]

Mah, ti hackeri se prijete da će pustit u opticaj nadolazeće specke novih grafa...nV se ''ukakala'' od potencijalne štete...