PC Ekspert Forum - Vijesti i novosti o Intel procesorima i čipsetima

PC Ekspert Forum (https://forum.pcekspert.com/index.php)

- Intel (https://forum.pcekspert.com/forumdisplay.php?f=8)

- - Vijesti i novosti o Intel procesorima i čipsetima (https://forum.pcekspert.com/showthread.php?t=49484)

Citiraj:

Autor Sinac (Post 3677582)

:goood: Jel se to meni čini ili Intelu kola idu nizbrdo? :D

Otelo se skroz van kontrole, nema više ni rekordnih zarada, ali oni i dalje lupaju PowerPoint slajdove ko Maks po diviziji.

I u ovoj zadnjoj najavi daljnjih odgoda i kašnjenja opet su više vremena potrošili na PowerPoint, nego na konkretne stvari.:):D

Bili bi sasvim OK kad bi se sami stavili na pravo mjesto i opet prodavali dobre procesor po 1100 kn poput i5 10400F, naravno uz solidne ploče od 750 kn.
Ovak nek crtaju i dalje, bude ih valjda spašavala USA vlada kao strateškog nacionalnog i vojno industrijskog chip proizvođača.
Bumo vidli kaj budućnost donosi, zadnji rekordni gubitak je zabrinjavajuć.

Najgore od svega je to da oni stvarno nisu imali nikakav plan ni program, a više od desetljeća stajanja na mjestu ih sad košta u svim segmentima, posebice u serverskom dijelu gdje su nekad imali 90%+ tržišta.

Intelovi problemi su počeli još na 22nm procesu, nastavili su se na 14nm proces i zacementirali na 10nm, ali u onih deset godina bez konkurencije nikoga nije bilo briga, niti je ikoga zanimalo - quad-core do ibera.

I sad brže-bolje sukladno najavama konkurencije, Intel također non-stop najavljuje od svega pomalo, ali uz relativno malu konkretnu realizaciju, dok AMD, nVidia, ARM, Samsung i TSMC idu svojim putem.

Steta, imaju dobre procesore, cak stovise, i onda si dozvole ovakve gluposti, vjerojatno je to posljedica njihovog bahacenja, "ne moze nam niko nista, jaci smo i od sudbine":D Sad polako dolazi na naplatu..

Slazem se da je 10400f bio odlican proc po jako dobroj cijeni. Samo da je u intelu vise pameti, bilo bi sve daleko bolje.

Za nadolazeći mobilni Mateor Lake i desktop Arrow Lake će sve osim CPU chipleta biti proizvedeno od strane TSMC-a.

IMHO isti taj TSMC će im za koju godinu odrađivati ama baš sve, dok će Intel kao i Apple, AMD i nVidia "samo" dizajnirati CPU i GPU.:)

Intel details PowerVia chipmaking tech: Backside Power performing well, on schedule for 2024:kafa:

Citiraj:

At next week’s annual VLSI Symposium, Intel will be presenting a trio of highly-anticipated papers about their progress with their upcoming PowerVia chip fabrication technology – the company’s in-development implementation of backside power delivery networks. Along with Intel’s RibbonFET technology for gate-all-around transistors, PowerVia and RibbonFET are slated to serve as Intel’s big one-two punch to the rest of the silicon lithography industry, introducing two major chip technologies together that Intel believes will vault them back into the fab leadership position. Combined, the two technologies are going to be the backbone of Intel’s “angstrom” era fab nodes, which will go into high volume manufacturing next year, making Intel’s progress with the new technologies a subject of great importance both inside and outside of the company – and one which Intel wants to address.

When it comes to Intel’s chip manufacturing technology, the stakes for Intel’s R&D groups couldn’t be any higher than they are right now. The long-time leader of the fab would has faulted – repeatedly – and is now in the middle of a multi-year effort to course-correct, not only to get back that leadership position which they’ve lost, but to break into the contract chip manufacturing business in a big way. As a result, while VLSI research papers do not normally attract a ton of outside attention, this year in particular is a big exception. With RibbonFET and PowerVia set to go into production next year, Intel is reaching the point where they’re wrapping up R&D work on the first generation of those technologies. Intel is now at the point where, to their peers in the VLSI industry, they can present their first findings around producing a complex logic test chip. And to Intel’s investors and other outsiders, Intel can show the first real evidence that their efforts to get back on track may very well be succeeding, giving Intel an opening to leap ahead of the competition that the company greatly needs.

To that end, at next week’s symposium Intel is going to be disclosing a great deal of information around their implementation of backside power delivery network technology, which they call PowerVia. Central to these papers is Blue Sky Creek, an Intel “product-like” logic test chip that implements backside power delivery on the EUV-enabled Intel 4 process technology. With Blue Sky Creek, Intel intends to demonstrate that not only do they have PowerVia working with a CPU on time for high volume manufacturing next year, but that the performance and manufacturing benefits of backside power delivery are everything Intel has promised. Suffice it to say, Intel is preparing for this year’s VLSI conference to be a very big moment for the company.

Ultimately, PowerVia is perhaps the single biggest make-or-break moment for Intel in terms of fully recovering momentum and potentially retaking leadership within the silicon lithography business. If Intel can deliver on its promises, the company is expecting to be at least two years ahead of TSMC and Samsung in deploying backside power delivery – and that means at least two years of reaping the cost and performance benefits of the technology. TSMC for its part is not expecting to deploy backside power until its N2P node in late 2026 or early 2027, while it remains unclear when Samsung will make their own transition. As for Intel, if everything continues to go to plan, Intel will begin high volume manufacturing with PowerVia in 2024, when the company is scheduled to bring both its Intel 20A and Intel 18A processes online. The first consumer processor to launch using the technology will be Intel’s Arrow Lake architecture, which will be a future generation Core product built on the 20A node.

Izvor: AnandTech

Malo pomalo i eto nas na 800W procesoru

https://www.pcgamer.com/intel-cpu-wi...nd-draws-800w/

To je ono kaj smo komentirali u vrijeme kad je Intel krenul sa svojim hibridnim big.LITTLE pristupom, tak da što lakše istovremeno dostignu AMD po pitanju jezgri, a opet da ne troši kao da nema sutra.

Sapphire Rapids ima samo veće punokrvne jezgre, zato je i potrošnja prikladna, ali i nadolazeći Raptor Lake-R budu nasvirali iznad 6.2GHz tak da bude i s tim porasla potrošnja uz zagrijavanje.

Za tih 13 kupaca uz PR da za performanse nema kompromisa, i više nego dobar, mogli su odmah zaokružiti na kW i amen :D

KaCmo kod PR-a :):D...

Citiraj:

Intel to launch new Core Processor branding for Meteor Lake: drop the i, add Ultra tier:kafa:

Citiraj:

As first hinted at by Intel back in late April, Intel is embarking on a journey to redefine its client processor branding, the biggest such shift in the previous 15 years of the company. This shift is due to begin in the second half of the year, when Intel will launch their highly anticipated Meteor Lake CPUs. Meteor Lake represents a significant leap forward for the company in regards to manufacturing, architecture, and design – and, it would seem, is prompting the need for a fresh product naming convention. The latest branding change, as previously mentioned, is the biggest in the last 15 years of Intel. While some naming schemes can be confusing and cluttered (ahem, Xeons), Intel is aiming for something that's still reasonably clear for the client processors. Which is great timing, because the hardware underpinning them is about to get fairly complex with the transition chiplet-based Meteor Lake and its mix-and-match of dies built on the Intel 4 process and built over at rival-supplier TSMC. First and foremost, under their new branding system Intel is doing away with leading generational branding.

https://i.postimg.cc/6TwgfZzb/Intel-...23-678x452.jpg

Izvor: AnandTech

šteta, meni se baš sviđa i3, i5, i7, i9

ak niš bar ću imat zadnju generaciju ''i'' procesora prije promjene nomenklature :D

Bolje da su išli na G kao gaming :D

to je već lođitek kapario

I tak, konačno su u Intelu sami sebi priznali ono kaj je očito posljednjih 5 godina, odnosno da im više ne cvate cveće u preduzeće, a i konkurencija je zajebana stvar.

Potencijalni spin-off Intelovih proizvodnih kapaciteta podsjeća na AMD-ovu GlobalFoundries situaciju otprije 15 godina, s tom razlikom da je AMD bio u puno gorem stanju.

Citiraj:

Intel Manufacturing gets its own profit and loss potentially gearing up for a spin-off:kafa:

Citiraj:

Intel has a plan to regain process leadership and redefine how it does business. Intel announced IDM 2.0, but the next step is interesting since it is effectively one of the key steps one would take if it was planning to spin off a business. Intel has two main challenges right now. First, it is seeing declining revenue, partly due to a cyclical market and partly due to missing product launch windows. Second, it is trying to ramp capacity in its fabs, while it is also utilizing TSMC for some leading-edge designs (further shifting capacity.) The second one really boils down to trying to ramp new costly processes. As such, the announcement today had a heavy cost focus. One of the biggest concepts was that Intel is transitioning to IDM 2.0. We have covered this multiple times, but the transition will change the way that Intel looks at its manufacturing organization. Intel is going to start treating its internal business units more like external fabless customers. That means, revenue from IFS as well as its internal customers will now have its own P&L.

In one of the more honest slides, Intel said it went from leadership in the 14nm Intel v. 16nm TSMC era, but lost leadership in the 10nm Intel/ 7nm TSMC era. Intel also says it will take until Intel 18A against TSMC 2nm at this rate to regain process leadership. Intel 3 and Intel 20A seem behind TSMC 3nm from Intel’s slide projections. Intel marks 10 nm (aka Intel 7) as significantly lagging behind TSMC 7 nm on time-to-market. TSMC 5 nm has already been in the market for over 3 quarters now, and it's only later in Q3 that Intel could release the first products based on the rival Intel 4 node ("Meteor Lake" compute tile). Intel 3 could similarly lag by a couple of quarters behind TSMC 3 nm, as could Intel 20A. It's at the 2 nm-class where Intel claims that if it executes the IFS and IDM 2.0 roadmap correctly, the Intel 18A foundry node should beat TSMC 2 nm-class nodes both technologically, and at time-to-market.

Izvor: ServeTheHome, Tom's Hardware i TechPowerUp

Tanka je granica između paranoje i realnosti.

P.L.A.N._i_P.R.O.G.R.A.M.:):D

Citiraj:

Intel's next-gen Arrow Lake CPUs were going to be the first to be built upon the 20A process node but those plans have allegedly changed as the company now focuses on using TSMC's 3nm node. There had been some rumors circulating that Intel might have dropped its 20A process node which was going to play a big role in the development of the next-gen Arrow Lake CPUs. The TSMC 3nm product roadmap also included N3P and N3P which offer higher performance and also lower power. Meanwhile, Intel 20A was looking to be a very strong competitor to TSMC's 3nm offerings but it looks like Arrow Lake might be missing out on that. We know from Intel's own presentation that the CPU "Compute Tile" was meant to utilize the Intel 20A process node while the GPU tile was going to be based upon the TSMC 3nm node.

A nije ovo nikako dobra vjest.
Još da krkne Samsungova proizvodnja i onda ostajemo samo na TSMCu, a Winnie Pooh bi ih mogao izbrisat sa lica zemlje ....tj to se baš više i ne spominje kao opcija, ali nikad se ne zna.
Ugl sve vodi prema nekakvom monopolu i napucavanju cijena. :(

Jebiga, situacija je alarmantna, dok TSMC i Samsung, tj. svi osim Intela na ovaj ili onaj način sve svoje najave konkretno i realiziraju. To je ono kaj smo pričali već nebrojeno puta, Intel već godinama koristi TSMC-ove usluge, prije i nego su službeno izašli u javnost s tim informacijama, da bi istovremeno uz korištenje TSMC-ovih kapaciteta, glorificirali vlastite, pa na temelju toga privlačili klijente, a na kraju budu i oni postali fabless kao Apple, nVidia i AMD. OK, možda se u Intelu stvarno dogovore sami sa sobom, ali sudeći prema ovome kaj rade zadnjih par godina - i dalje se vrte u vlastitoj magli obećanja, najava i odgoda.

EDIT:

Citiraj:

Intel’s 5 Nodes in 4 years

:chears:

Intel Raptor Lake Refresh Basic Specs Leaked

Ovaj put Intel piči dalje sa jezgrama pošto AMD ne mrda, i3 sa 6 jezgri, i5 sa 8....

https://www.techpowerup.com/img/Vpd0WFxV8iaIanDh.jpg

https://www.techpowerup.com/img/7YqMcrQ4QGxz2iC9.jpg

O potrošnji nećemo....još. ;)

A opet, ako cijene ne napreduju proporcionalno, to jest neki opipljivi napredak, good!

Samo high-end ponuda je besmislenija nego ikad, kome trebaju ti Cinebench akceleratori, xx600 je dovoljan za baš sve. Vjerojatno će i cache u tim 8p konfiguracijama biti isti?

i5 14600KF bu zvijer za igre. Bome i i3 bu ozbiljan cpu za igre, pogotovo ak ga postave sa niskom cijenom.

i3 bude sigurno koliko i i5-12400f.

Citiraj:

Autor Disco (Post 3685779)

Ovaj put Intel piči dalje sa jezgrama pošto AMD ne mrda, i3 sa 6 jezgri, i5 sa 8....

Nikak nije dobro, kak god se okrene.:) 2017. je jedna od najaktualnijih stvari bila priča o gomilanju besmislenih jezgri, jer nikome ne treba više od Intel quad-core ili hexa-core modela koji su u to vrijeme bili smatrani high-end gaming ponudom, da bi danas s ovim nadolazećim Raptor Lake-R vijestima taj isti broj jezgri bil smješten u početni rang nove Intel ponude. I sva sreća da su se stvari tak posložile, inače bi stvarno još uvijek bili na quad i hexa uređajima za sulude cifre. Makar tehnički Intel i dalje nudi osamerojezgrene modele, kak god se okrene i ove manje jezgre imaju svoj efekt, mada nije toliko izražen u svima bitnim igrama. Kaj se AMD-a tiče, već danas za Zen 4 nude 16-core CCD + SMT + AVX-512 i istu stvar imaju u pripremi za Zen 5, ali većina tih čuda tehnike umjesto za desktop ide u 128C/256T i 192C/384T EPYC primjerke i mobilne APU modele. Makar ko zna, možda slože kakvo 32C/64T Zen 5 čudovište bazirano na 16-core dual-CCD rješenju, pa ga nakrcaju popratnim 3D cacheom, ali opet za igre većina krajnjih kupaca od te skalamerije nema neke pretjerane koristi.

Pa ja se isto nadam da će AMD krenuti od 8 jezgri sa zen5. I da će njihove i Intelove cijene biti normalnije, a ne ovo što kolega gore misli da će i3 biti 200injak€, nije to to onda.

Zasad nemaju quad u zen4 generaciji, i tako bi trebalo ostati. Te CCD-ove su dosad valjda dobivali tako da su namjerno rezali ispravne čipove u zen2 i 3 generaciji.
Sve ovisi o konkurenciji. AMD nema killer instinct da istrijebi konkurenciju cijenama kad može i bojim se da ćemo gledati opet neku ravnotežu ako Intel ne proizvede neku dobru konkurenciju nove generacije. Cijene Zen4 su mogle odmah biti puno niže da su htjeli, ali ne, cijelo vrijeme su htjeli dizati profit margin, ne marketshare.
Ili nemaju količine jer serveri gutaju sve što se proizvodi, pa za desktop daju taman koliko imaju da se proda skupo, i to im najviše odgovara.

Citiraj:

Intel Arrow Lake-S to be 6% to 21% faster than Raptor Lake-S:kafa:

Citiraj:

Igor’sLAB recently obtained an internal presentation from Intel revealing performance projections for the upcoming Arrow Lake-S, the next-generation client platform. Although the platform’s launch is still some time away, the discussed data provides us with numerous details that would typically remain undisclosed for months. The projected data is compared to the upcoming ’14th Gen Core’ Raptor Lake-S Refresh, which consists of 8 Performance and 16 Efficient cores—similar to Arrow Lake-S. One notable finding is that Arrow Lake’s PL1 and PL2 may be configured at 250W, which is slightly lower than Raptor Lake’s LGA1700 platform by 3W. However, this difference in power is unlikely to yield significant performance variations. Arrow Lake-S, featuring an entirely new architecture, is projected to outperform the current lineup by 1% to 21%, depending on the benchmark used.

Izvor: Igor'sLAB

Da, tak ispada, mada su bila očekivanja i najave od kakvih 30% do 40% za single-thread i 50% do 75% za multi-thread, ovisno o tome koliko manji šparnih jezgri ubace, posebice za 8P + 32E kombinaciju.

Ukoliko je ovih ~20% stvarno Intelova interna procjena, onda nije baš nešto previše, obzirom da Arrow Lake u najboljem slučaju stiže tek za godinu dana, a realno krajem 2024. i početkom 2025.

Zen 5 konkurencija bude do onda već naveliko i naširoko dostupna, do te mjere da se Arrow Lake krajem 2024. i početkom 2025. nalazi okružen običnim i 3D cache Zen 5 varijantama.

Ak ovaj intel ne ponudi nis vrijednog spomena, odo ja na amd zen5:D
Haj end procesori, puno trose i griju su.. Ocekivao sam vise od arrow lake-a.

Uskoro mi nece bit jasno, kaj ce uopce intel ponudit kupcima, ako ce imat nekaj pametnog.

Amd dok se razmase s ovim 3d cache procesorima, tesko ce ih ko moci zaustavit, a odgovora na tu "prijetnju," je nepostojeci bar sto se tice intela.

Nažalost Intel se fino sjebal u godinama bez konkurencije, a Arrow Lake bi možda imal i veći efekt da nisu otkazali Meteor Lake za desktop, tak da budu sad imali poveću prazninu u odnosu na Raptor Lake-R.

Mobilni Meteor Lake bude ionak u svemu osim CPU chipleta odrađen od strane TSMC-a, a kak je krenulo Arrow Lake bude lako moguće kompletno proizveden u TSMC-ovim fabrikama.

Kaj se pak Intelove ponude općenito tiče, CPU divizija im je jedina koja nije upitna, jer sve ostalo su na ovaj ili onaj način otkazali, odgodili, sfušali, prodali ili kompletno ugasili.