free estadisticas Saltar al contenido

0

febrero 11, 2020

After the first official images of DG1, Intel's first architecture-based dedicated graphics card Xe, many are waiting to learn more about the technical characteristics of the company's new GPUs, solutions that will cover all market segments and that in the server segment could make their debut by the end of the year. In this regard, the portal DigitalTrends has published what seems to all intents and purposes an internal Intel slide, a very interesting document at first glance that seems to clarify some key features of these graphics cards.

Stressing that we cannot 100% confirm the authenticity of the slide, the data emerged refer to a solution in particular called Arctic Sound, a name we had already heard of in April 2018 and that in some way could indicate a product in an advanced stage of development.

According to what emerges from the following slide, Intel Xe graphics cards will use a multi-chip design (or better MCM – Multi Chip Module) with some modules called "Tile"; the number of Tiles should vary based on the price range and the scope of use / TDP (1,2 or 4), each tile instead should integrate 128 EU (remember that Intel DG1 offers 96 EU).

Arcitc Sound in particular will be equipped with an architecture of up to 4 Tiles for a total of 512EU, a graphics card so configured could offer a maximum of 4096 processing units (see Cuda Core / Stream Processor). However, the data that strikes us is the TDP, a good 400 / 500W for a solution that among other things will need one voltage of 48 volts, a feature that we can obviously find only on the server platform.

In essence, Intel could articulate its proposal in three basic solutions, four if we take into account the DG1 model which basically has a single EU 96 tile architecture.

  • Data Center: up to 4 tiles / 512 EU / 4096 cores / TDP 400-500W
  • High end / wokstation: 2 Tile / 256 EU / 2048 core / TPD 300W
  • Mid-range: 1 Tile / 128EU / 1024 core / TDP 150W
  • Low end (Intel DG1): 1 Tile / 96 EU / 768 core / TDP 75W

Rate this post

0

febrero 11, 2020

NVIDIA updates its catalog of graphics cards for notebooks with models GeForce MX350 is GeForce MX330, two economic and low-consumption solutions that update the GeForce MX200 series saddle models introduced about a year ago. The manufacturer, who has avoided issuing an official release, as usual does not provide particular technical details on the new GPUs, except for the type of memory used, i.e. the now tested GDDR5.

Thanks to colleagues from notebookcheck.net however, we are able to give you more information about the specifications of these cards which, as expected, will still use the Pascal GPU architecture with a 16nm production process.

In details, NVIDIA GeForce MX350 is equipped with a Pascal GP107 GPU with 640 Cuda Core, or the one we find on the most famous GeForce GTX 1050 in the notebook variant. The card has a 25W TDP and uses 7 GHz GDDR5 memories on a 64-bit bus, the GPU frequency is instead set at 1354 with Boost at 1468 MHz.

GeForce MX330 instead it receives the same Pascal GPU108 GPU that NVIDIA used on the previous GeForce MX250; we are talking about a 384 Cuda Core chip that is flanked by 7 GHz GDDR5 memory, always on a 64-bit bus. In this case, the standard 25W variant with a base / boost frequency of 1531/1594 MHz and a low consumption version (TDP 12W) which drops in frequency to 746/936 MHz is provided.

The new NVIDIA GeForce MX350 and MX330 should be available shortly on new generation notebooks or in any case destined to reach the market in the coming months; at the moment we do not know if the reference price for OEMs will change compared to the previous generation of GeForce MX cards.

Rate this post