Šta je novo?

nVidia Maxwell

Pa nije samo to, performace/watt je bitan i za high-end karte, najvise zbog Tesla/Quadro verzija koje idu u servere.
 
Poprilicno me podseca na situaciju Cypress vs. Barts... isti proces izrade, ali daleko optimizovaniji cip po svakom parametru.
 
Tako nesto, mada izgleda da je ovde veci fokus na potrosnju.
Ako GTX 750Ti ima performanse kao GTX 650Ti boost (110W) a trosi ~75W, sta ce tek biti kad je prebace u 22nm..
Vrlo moguce da ce nece imati slabiji GPU od ovog u Maxwell seriji.
 
Izgleda da od tog prebacivanja u 22nm nema nista u skorije vreme, kad ni slabije cipove ne mogu da prave u 22nm.
 
Ima li neko malo bolje informacije, kad se mogu ocekivati pravi Maxvell-i, znaci 20nm naslednik 680GTX recimo, naslednici Titana i 780Ti ce dosta kasnije verovatno.

Koliko se ja secam, sve je to bilo najavljeno za 2014tu, kakva je situacija, zna li iko ista?
 
Pouzdane informacije tesko da neko ima.

Mozda tek Q3/Q4, plus ako bude sledili Kepler strategiju, ocekuj prvo naslednika GTX 680/770 a tek kasnije naslednika GTX 780Ti/Titan.
 
AMD je ugazio prvi u 40nm pa u 28nm sigurno ce prvi krenuti i u 20nm proces tako da dok ne vidimo neke AMD karte cisto sumnjam da cemo vidjeti Nv u 20nm procesu, mozda se varam, ali nece jos dugo biti na trzistu karte radjene u 20nm procesu - nista ne ukaziva na to.
Ovo sto Nv izbaca maxvel dobra je stvar, bar iste performanse uz manju potrosnju kad vec nije spreman 20nm proces, dobro bi bilo da izbace maxvel u svim segmentima znaci nasljednik Gk110.
 
Guzva u TSMC-u, nema dovoljno kapaciteta, a sada ima daleko vise zainteresovanih za najnapredniji proces (ako je verovati pojedinim izvorima imaju i Apple i Qualcomm i MediaTek kao musterije za 20nm).
 
Izgleda da je malo drugacije nego sto je prvobitno iscurelo:

As it turns out, the 768/960 CUDA count as read by GPU-z was wrong. The application could not read the real number because the architecture has been changed compared to Kepler. GPU-Z assumed that each SMX holds 192 CUDA cores, which was not correct. The real number is 128. Full GM107 has 5 SMs with 128 CUDAs each, which gives 640 in total. This means that GTX 750 Ti has 640 CUDAs (5 SMX) and GTX 750 has 512 (4 SMX).

GM107 has a TDP of 60W

The GM107 will not even utilize the full power delivered by PCI-E connector (75W). While operating at default frequencies it won’t need any additional power source. Although manufacturers will still add the power connector, for the sake of stability or increasing the overclocking headroom.

There are two generations of Maxwell GPUs

It has been revealed by newest CUDA driver that NVIDIA will make two generations of Maxwell GPUs:

28nm: GM108, GM107
20nm: GM206, GM204, GM200

GM107 will replace GK107 with a performance of GeForce GTX 480

You should find this particularly interesting. While GM107 utilizes 4 times less power than Fermi GF100, it will offer the same performance (actually even slightly better).

The GM107 diagram

This is the first ever diagram of Maxwell GPU. There is one Graphics Processing Cluster. Full GM107 has 5 SMMs with 640 cores in total. Each SMM has 8 TMUs, so the total count is 40. The GPU has 16 ROPs and two memory controllers (both 64b).

Maxwell-GM107.png


What is new in Maxwell

Okay, so the SM has been redesigned, the layout has changed, but what exactly makes the Maxwell better? Well here’s the list:

Larger L2 cache.
This is the main difference between Kepler and Maxwell. Larger L2 cache will limit the queries to the GPU. GM107 L2 cache has 2MB. GK107′s cache has 256KB.
Workload balancing and complier-based scheduling has been improved.
The number of instructions per clock cycle has been increased.
SM has been redesigned into four processing blocks (as explained above).
Maxwell introduces even faster H.264 encoding and decoding with improved NVENC (which is used, for instance, in ShadowPlay).
New GC5 power state (low sleep state).

As opposed to previous leaks, the die size of GM107 is even smaller, not 156 but 148mm2. Compared to GK107 the density of CUDAs per mm2 has increased roughly by 30%. The density of transistors increased by 15%. Remember, this is all on the same fabrication process.

http://videocardz.com/49557/exclusive-nvidia-maxwell-gm107-architecture-unveiled

28a.jpg


28b.jpg


28c.jpg


28c.jpg


28d.jpg
 
Poslednja izmena:
Karta stvorena za male Steam mašine.
 
Ovde se već u ponudi na jednom sajtu nalazi Asusov GTX 750Ti za 190 evra. Mnogo je ako ovo čudo po performansama ne može da potuče GTX 660. Inaće mi se kartica baš sviđa, mala, hladna, treba mi baš ovako nešto da se rešim raspalog Radeona 5770. Valjda će uskoro i GM108.
 
Cena je losa.

Ali GTX 750Ti trosi kao GTX 650 a duplo je brzi. To se neki put ne postize ni prelaskom na novi proizvodi proces.
 
To je onda super. Tu negde blizu kao R7 260X, a duplo manja potrosnja. :)

Jos da non-Ti modeli budu oko 10-11k, u rangu R7 250/250X, i odlican posao su napravili.
 
Da li to dobro čitam, ili oni zaista sa Maxwellom i dalje furaju samo DX 11.0 feature level?
 
A sta treba da furaju? Super popularni DX 11.05.0x v3? Ili da idu odmah na DX15?
 
Posle Keplera koji je imao odličan perf/wat mislim da niko nije očekivao još jednu ovako dobru optimizaciju na istom proizvodnom procesu. Naravno, nVidia nebi bila nVidia da ne naplaćuje bonus za svoj brand. Za 10K RSD ova kartica bi bila prava mala bomba.

Nadam se da će potrošnja eksponencijalno rasti sa performansama pa da time dobijemo naslednika GTX680 u 130W segmentu :)
 
pojavile su se u firmi ne E ali cene... uhhhh, paprene!
 
Pa naravno kada misle da smo svi mi ovde ovce, da smo sisali vesla i ne znamo koliko kostaju te karte preko.
 
Poslednja izmena:
Slaba je u igrama za te pare.
 
Prema ovih nekoliko compute testova sa Anand-a:

http://www.anandtech.com/show/7764/the-nvidia-geforce-gtx-750-ti-and-gtx-750-review-maxwell/21

GTX 750Ti je jednak ili brzi od GTX 660 koji ima 50% vise shejdera sto znaci da se dosta poradilo i na ovome.

Dobar napredak Nvidia sad nema konkurenciju, amd-e karti nema na trzistu nvidia moze da zacjeni koliko oce karte prodace se, mislim da je visoka cjena, ali napisao sam sta je uzrok cim se pojave amd-e karte na policama(kad propadne BTc i ostale valute :D) i ovoj ce sletiti cjena.
 
Ako gledas cene kod nas, onda nije uzrok taj koji si pomenuo. Ima AMD karti koliko hoces u tom segmentu. Pa i R7 260X 2 GB je preskup kod nas, kosta ~17K
Treba cene da se malo slegnu, a i da izbace GTX 750Ti 1 GB, onda ce moci da se prica.
 
Dobro na domacem trzistu si upravu nije samo nestasica uzrok visokoj cjeni tu je onaj nesretni PDV zezno sve, vidio sam R7 260X 1GB i to asus za 15k sigurno kad izbace i GTX750Ti 1GB ice cjena do 15k, ali tu ce stici i R7 265 koja je opet brza od navedeni bas su do 150$ natrpali dosta karti.
 
Jos nesto zanimljivo sam nasao:

One change in the new SM will be especially consequential for certain customers—and possibly for the entire GPU market. Maxwell restores a key execution resource that was left out of Kepler: the barrel shifter. The absence of this hardware doesn't seem to have negative consequences for graphics, but it means Kepler isn't well-suited to the make-work algorithms used by Litecoin and other digital currencies. AMD's GCN architecture handles this work quite well, and Radeons are currently quite scarce in North America since coin miners have bought up all of the graphics cards. The barrel shifter returns in Maxwell, and Nvidia claims the GM107 can mine digital currencies quite nicely, especially given its focus on power efficiency.

http://techreport.com/review/26050/nvidia-geforce-gtx-750-ti-maxwell-graphics-processor

U kombinaciji sa malom potrosnjom moglo bi da bude interesantno...jos samo da se vidi koliko dobro kopaju.
 
Poslednja izmena:
35% bolje compute performanse sa manjom potrosnjom. Sjajne vesti za GPU raytracing:)
 
Nazad
Vrh Dno