Gráfica nVidia Maxwell (GTX750/GTX970 e GTX980 Series)

Hellgardia

Power Member
Para separar a discussão das actuais Kepler do outro tópico criei este para se ir postando o que se sabe das Maxwell.

Entretanto encontrei isto:
A short update. We recently released details about Maxwell coming to mobile and desktop very soon. We also said that the mobility flagship 880M will remain a re brand while some of the lower models will be upgraded to Maxwell architecture. Well Nvidia has just more or less confirmed the first such chip.
[h=1]Nvidia Updates Specs Chart of G820M adding DX11.2, Maxwell GM117 GPU instead of GF117?[/h]We already know Maxwell architecture is coming to low-mid end Geforce Mobility chips soon and the fact that Nvidia has silently updated the spec chart of the 820M more or less confirms the first lucky gpu. Here is the updated spec sheet of the 820M:

As you can see DX 11 has been updated directly to DX 11.2. The Geforce 820M should feature the GF117 Core that is based on the Fermi Architecture, but if this is true then it means that it will have the GM117 Core instead. We have also received confirmation of the same from our friends over at
Computerbase.de They also arrived at the same conclusion as us, that Nvidia has updated the 820M’s architecture from Fermi to Maxwell. Also the fact that Fermi GF117 was originally advertised at OpenGL 4.3 which is now OpenGL4.4 is a massive hint.



Read more: http://wccftech.com/geforce-820m-maxwell-mobility/#ixzz2qeVejYxx

http://wccftech.com/geforce-820m-maxwell-mobility/

Se tiverem algo que já se saiba postem aqui. Se acharem que é cedo demais e que não vale a pena existir a thread então, by all means, algum mod que apague. Caso contrário, fire away :P
 
Momento WTF:

NVIDIA Readies GeForce GTX 750 Ti Based on "Maxwell"

NVIDIA's next-generation GPU architecture, codenamed "Maxwell," will debut this February, with the unexpectedly positioned GeForce GTX 750. The card will launch on February 18, to be specific. Maxwell will introduce a host of new features for NVIDIA, beginning with Unified Virtual Memory. The feature lets the GPU and CPU share the same memory. Such a feature is already implemented on the current CUDA, but Maxwell could be designed to reduce overhead involved in getting the thing to work. The next big feature is that Maxwell GPUs will embed a 64-bit ARM CPU core based on NVIDIA's "Project Denver." This CPU core will allow the GPU to reduce dependency on the system's main processor in certain GPGPU scenarios. Pole-vaulting the CPU's authority in certain scenarios could work to improve performance

Getting back to the GeForce GTX 750 Ti, NVIDIA's aim is simple, to see how "Maxwell" performs on the existing, proven 28 nanometer silicon fab process, before scaling it up on the future 20 nm nodes, with bigger chips. Given its name, we expect it to be positioned in between the GTX 760 and the GTX 660 in terms of gaming performance, but we won't be surprised if it falls into an entirely different league with GPGPU. There are no specifications at hand.

fonte

Portanto temos então um rebrand ao contrário :-D
 
Depois pontos muito interessantes na arquitectura:

Unified Virtual Memory
Maxwell GPUs will embed a 64-bit ARM CPU

O primeiro vai permitir talvez uns bons aumentos de performance e redução de consumo de memória.
O segundo deve aumentar o poder GPGPU que bem precisam, dado que isso tem sido o mais fraco contra a AMD.

Parece-me bastante bem
 
Acho dificil de acreditar, principalmente na data de Fevereiro e no facto de incluir o core "Denver", que a Nvidia disse publicamente no CES só estar disponivel no segundo semestre (para o SoC K1, mas quer dizer, se já está pronto para ser usado numa GPU porque é que iriam atrasar a introdução no K1)...
 
Depois pontos muito interessantes na arquitectura:

Unified Virtual Memory
Maxwell GPUs will embed a 64-bit ARM CPU

O primeiro vai permitir talvez uns bons aumentos de performance e redução de consumo de memória.
O segundo deve aumentar o poder GPGPU que bem precisam, dado que isso tem sido o mais fraco contra a AMD.

Parece-me bastante bem

Sobre os dois pontos:
1. A memória unificada vai fazer muito pouco para aumentar performance. Isto serve apenas para simplificar um pouco a programação, em que o driver passa a tratar de cópias de memória entre a main memory e o GPU, ao invés de isso ser feito explicitamente em código. Aliás, já existe nas ultimas versões de CUDA, e não é algo tão elaborado como o HSA/hUMA da AMD (porque ai estamos a falar uma APU com acesso à mesma memória, aqui não).
2. Se há coisa que a NVidia não precisa é de aumentar o GPGPU. Eles deliberadamente capam as GeForce para proteger as Quadro e Tesla, onde têm uma performance em GPGPU muito superior à AMD (com CUDA). De qualquer forma, esse CPU ARM nunca terá grande efeito em GPGPU. É muito fraquito, e servirá basicamente para evitar algumas round-trip GPU-CPU-GPU, mantendo tudo no GPU, e isto, apesar de melhorar a performance, não será algo de estonteante.
 
Atenção que o CPU ARM não é um core denver completo mas sim um core baseado no denver.

O core vai servir sobretudo como unidade de DMA e de controlo de coerência da memória principal e da memória gráfica.
 
GeForce GTX 750 Ti "Maxwell" First Performance Numbers Out

Ahead of its rumored mid-February launch, members of Taiwanese tech forum Coolaler.com posted the first performance benchmark numbers of the card. Originally thought to be positioned between the previous-generation GeForce GTX 660 and current GTX 760, the GTX 750 Ti, according to these numbers is on average 10 to 15 percent slower than the GTX 660, which should put its performance somewhere in between the GTX 650 Ti Boost and the GTX 660.

Then again, the testers must be using some very early drivers, and performance figures of the GTX 750 Ti should get clearer as its mid-February launch date approaches. The GeForce GTX 750 Ti is an important GPU for NVIDIA, as it's the first to be based on its next-generation "Maxwell" GPU micro-architecture. NVIDIA is trying the architecture out on current 28 nm process, before launching bigger chips based on the next-generation 20 nm fab process.



http://www.techpowerup.com/197038/geforce-gtx-750-ti-maxwell-first-performance-numbers-out.html
 
Infelizmente sem se saber detalhes sobre quantidade de shaders, consumos e dimensões não se fica a saber muito sobre a eficiência da nova arquitectura.
 
Resta só referir que poderemos estar a falar de um GM107, nessas 750ti.

Se tivermos em atençao que nomalmente os G*107 equivalem a gpus entrada de gama, os numeros parecem-me bastante interessantes.
 
Here is the first picture of a couple of NVIDIA GM107 silicons in a tray, ahead of graphics card assembly. The packages appear to be as big as those of the GK106 from the previous generation, however, the die itself is estimated to be smaller, at roughly 156 mm², compared to the 221 mm² die of the GK106, and the 118 mm² of the GK107. The best part? All three chips are built on the same 28 nm silicon fab process. So what makes the GM107 die smaller than that of the GK106 despite having a similar feature-set? Narrower memory bus. The GM107 is said to feature a 128-bit wide GDDR5 memory interface, in comparison to the 192-bit wide interface of the GK106.

Apart from the 128-bit wide GDDR5 memory interface, the GM107 is said to feature a total of 960 CUDA cores, 80 TMUs, and 16 ROPs. The CUDA core count is identical to that of the GK106. The GM107 is built on NVIDIA's next-generation "Maxwell" GPU architecture. It will form the foundation of two SKUs, the GeForce GTX 750 Ti, and the GeForce GTX 750. The former features the full complement of 960 CUDA cores; while the latter isslightly cut down, and features just 768. The TDP of the GTX 750 Ti is approximated to be around 75 Watt. If true, the GTX 750 duo will set new standards on the performance-per-Watt metrics. NVIDIA is expected to launch both, later this month.


http://www.techpowerup.com/197617/nvidia-gm107-maxwell-silicon-pictured.html



EDIT: Entretanto vi que tinhas postado os benchmarks.
 
Última edição:
isto de ter um cpu ARM é interessante, será que jogos multiplayer grandes tipo battlefield vai ganhar com isto? especialmente para quem não tem quad cores?
 
ASUS GeForce GTX 750 series detailed, plus the real specs of GM107!

Full Maxwell GM107 GPU has 640 CUDA Cores!

First of all I want to share some new information. Maxwell GM107 GPU does not have 960 CUDA cores, neither does it have 768. The full chip has 640 cores. So you’re probably wondering why everyone (including me) said it has 960. It was because of the wrong GPU-Z reading. I have seen many materials, each claiming it’s 960, but as it turns out Maxwell architecture is completely redesigned, causing GPU-Z to read the wrong number. Even manufacturers were not aware of that and they released materials with inaccurate specs. I will cover this more closely in my next posts.

ASUS GeForce GTX 750 Ti (GTX750TI-OC-2GD5)

Now let’s talk about the GTX 750 Ti from ASUS. ASUS is preparing quite a few 750 series cards, but I couldn’t find mode detailed info about them, apart from the two in this post. The OC model comes with a Dual Fan cooling solution, which is said to provide better airflow and lower noise level.
The card is clocked at 1072 MHz with a boost at 1150 MHz. The memory runs at stock 5400 MHz. It has 2GB of GDDR5 memory and 128-bit interface. It does not support SLI. Card requires one 6-pin power connector.

ASUS GeForce GTX 750 Ti has 640 CUDA cores.

ASUS GeForce GTX 750 (GTX750-PHOC-1GD5)

The GTX 750 has 1GB of RAM and it does not have any power connector. The PHOC runs at 1059/1137 MHz, with a memory at 5010 MHz.
Card is equipped with HDMI, DVI-D, and VGA display output. It’s a single-slot design, but the cooler takes more space than one slot.

ASUS GeForce GTX 750 has 512 CUDA cores.

ASUS-GeForce-GTX-750-Ti-Chinese.png



NVIDIA Maxwell GM107 architecture unveiled

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).
  • GM107 has a die size of 148mm2

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.

Maxwell-GM107-706x620.png
 
Última edição:
Tuesday, 18 February 2014 09:51GTX 750 Ti launches today for $149

Written by Fuad Abazovic




nvidia.gif


Behold, Maxwell

Today is the big day for Nvidia. The company is set to introduce Maxwell, the new generation graphics architecture later today, but the first product is no flagship. Maxwell is power oriented, it will need much less power and at the same clock it will perform faster than Kepler or Fermi.We have heard that compared to Fermi, Maxwell performs at four times less power than Fermi, you can get high-end GTX 480 at one quarter of the wattage.
Pricewise the Geforce GTX 750 TI will sell for $149 and will compete with the Radeon HD R7 265. AMD launched this card to spoil Nvidia’s party as it should cost the same and perform on about the same level. The Geforce GTX 750 will sell for $119 and again AMD reacted by dropping the price of Radeon R7 260X to the exact same price point.
Maxwell will offer decent performance and has been described as “solid product” that will have good performance and low power consumption, but it won’t blow our mind, not yet.


The mind blowing Maxwell parts are coming later this year, in the form of 20nm GM200 products. They will offer a significant performance boost to the high end graphics market and they should dethrone Radeon R9 290X and Geforce GTX 780 TI cards. Since GM 200 parts are coming in the latter half of the year, AMD shouldn’t be far behind with its 20nm parts.

http://fudzilla.com/home/item/33971-gtx-750-ti-launches-today-for-$149


http://videocardz.com/49647/nvidia-geforce-gtx-750-ti-official-specifications-performance

GeForce GTX 750 (TI) specifications

GeForce GTX 750 Series Specifications
GeForce GTX 650 TiGeForce GTX 750 TiGeForce GTX 750
GPU28nm GK10628nm GM107-40028nm GM107-300
GPU Config768 : 64 : 16640 : 40 : 16512 : 32 : 16
GPU Clock928 MHz1020 / 1085 MHz1020 / 1085 MHz
Memory Clock1350 MHz1350 MHz1250 MHz
Video Memory1GB GDDR52GB or 1GB GDDR51GB GDDR5
Memory Bus128-bit128-bit128-bit
TDP110W60W55W
NVIDIA is preparing two cards, both based on GM107 GPU, the first Maxwell processor. Maxwell GM107 features 640 CUDA cores, 40 TMUs and 16 ROPs. This is also a configuration of GTX 750 TI. The full GPU has 5 SMMs (Streaming Multiprocessors Maxwell). The GTX 750 non-Ti will pack 4 SMMs, which means it has 512 CUDA cores and 32 TMUs. The rest is pretty much the same, although GTX 750 will offer even lower TDP of 55W (5W less than GTX 750 Ti), and less memory (although you will be able to find 750 Tis with 1GB too.
You could say that NVIDIA took the same path as AMD did with Hawaii processor. It’s still on the same fabrication process, but the architecture is improved and more power efficient. If you’re interested in the architectural improvements make sure to check my previous post. Below is the official specs sheet and the SMM in detail.

Some official pictures:
The card is one-and-half slot design. The reference board will be adopted by few manufactures, but don’t expect them to use the reference cooler. Both reference Maxwell models do not have any power connectors. Power is drawn directly from the interface, which is more than enough to provide enough current and voltage.


NVIDIA GeForce GTX 750 performance
GeForce GTX 750 performance is not breathtaking, but it shouldn’t be considered by the plain numbers and framerate comparisons to other cards, unless we consider how little energy it requires to compete with them.
Let me just start by saying there are no detailed benchmarks results for GTX 750 non-Ti (from official source). What we do have however, is one chart showing a comparison to its direct competitor — Radeon R7 260. As you can see in NVIDIA-picked titles it offers the same or considerably better performance. While it doesn’t look very special, we should mention that R7 260 consumes 95W and GTX 750 only 55W. For HTPC builders this is more than enough to choose GeForce card.

NVIDIA GeForce GTX 750 Ti performance
NVIDIA is really proud that they managed to reduce the power consumption almost 3 times compared to Fermi GTX 480 (the one you could fry eggs on).
The TDP has been dramatically reduced to 60W from Fermi’s 250W. GeForce GTX 480 has 384-bit interface so this comparison is nothing without clearly stated resolution (this chart is at 1080p).


Moving on to more detailed benchmarks. In AMD titles, like Tomb Raider or Battlefield 4 the difference is almost nonexistent. Other games follow the same path. We have a 5% (Tomb Raider) to 38% (COD) performance difference. The average is 20%, so it’s safe to assume that we are looking at 10-15% difference.
Let’s not forget Radeon R7 260X has 896 Stream Processors and 56 Texture Units, so it’s quite obvious that Maxwell is more powerful on architectural level, providing better performance with almost twice less power consumption.


Extreme overclocking

There is one key part that wasn’t covered yet. Because of extremely low TDP the clocks speeds suffered. NVIDIA had to choose the lowest possible clocks to obtain 60W. For that reason Maxwell GM107 GPU has extreme overclocking potential. In fact you will even find cards with 28% factory-overclocking.
Long story short, you will easily be able to obtain 1200 MHz clock, but only with the cards equipped with additional power connectors.
Release date

NVIDIA GeForce GTX 750 series will be released on February 18th. The prices remain unknown.
 
Back
Topo